timaeus/pile-pubmed_central · Datasets at Hugging Face

text stringlengths 12 2.52M	meta dict
Data sets that are subject to our analysis are available online. Data set B: <http://www.fc.up.pt/addi/ph2%20database.html>, Data Set A: <https://doi.org/10.17026/dans-zue-zz2y>. Because of concerns related to patient privacy and permissions set by the clinic that collected the data, some access restrictions apply to Data Set A. Introduction {#sec001} ============ Melanoma, the neoplasm of the pigment cells of the skin, is still a challenge both for clinicians and CAD (Computer Aided Diagnosis) specialists. Observations with the naked eye and with instruments, especially with popular optical or even with digital dermoscopes require long-term experience which is hard to achieve not only for general practitioners but also for dermatologists. Diagnosis of mature melanoma moles, due to asymmetry, variety in colors or border irregularity, may not be difficult, which is quite the contrary to early melanoma lesions that lack those indications. Effective treatment of melanoma i.e. a high (\>95%) 5 or 10-year survival rate consists in early detection and resection of the malignant skin lesion \[[@pone.0211318.ref001]\]. When not excited at an early stage, melanoma penetrates deep from epidermis to the skin and finally transfers to the lymph nodes and other internal organs by metastasis. At this stage the mortality rate is extremely high and especially for at least a decade has become a medical problem. This problem refers to all countries but particularly these where melanoma morbidity rate is elevated. Statistics says that women got melanoma moles on the legs and men on the back. Crucial for improving the patient survival rate is early detection and for that the only reliable method is biopsy. For obvious reasons (surgical complications, ANS-Atypical Nevus Syndrome, economic reasons) excision cannot be a diagnostic tool or standard mass treatment. The key role is precise detection of the tumor. Clinical diagnosis of early or micro-melanoma is extremely difficult even for experts. A differentiation between pigment and non-pigment skin lesions may be a challenging task but classification of different forms of pigment (melanocytic) lesions can be very complex. From the clinical point of view three forms of pigment lesions can be misdiagnosed: benign nevus, dysplastic (atypical) nevus and melanoma. The dysplastic nevus may be a precursor to malignancy and often exhibits features visually identical or close to melanoma \[[@pone.0211318.ref002]\]. On that account melanoma CAD systems are very popular and of key importance. They help in the early diagnosis of the skin lesions and support non-invasive methods in dermatology. The CAD methods implement machine learning paradigms and are based on data sets of (histopatology or expert ground truth) known cases. Unfortunately such data sets are sparse and with limited statistics of events and perhaps limited or unknown quality (e.g. image resolution, compression). Recognition of melanoma with such computer systems can be divided into: i) direct machine learning (ML) to recognize melanoma from its features, and ii) ML of specific skin lesion patterns/structures that are indicative of melanoma (lesion border, blue veil, geometric asymmetry etc.). The main feature types classified in the melanoma recognition systems include color and/or geometry-based features, histogram-like features, wavelet-based features and other less common (e.g. Gabor etc.). Literature on the melanoma CAD is ample and deal with all the steps: accuracy of clinical and histopathological examinations (i.e. the ground truth), dermoscopy and digital acquisition of lesion images, preprocessing (removal of artifacts, filtering), and finally pattern recognition and machine learning of melanoma. The majority of articles refer to dermoscopy and/or digital imaging of melanoma, since spectral, trans-illumination, ultrasonography or tomography systems, although promising, are not mass skin diagnostic systems. In 2013 Masood et al. \[[@pone.0211318.ref003]\] described the state of the art of the melanoma classification of dermoscopic images with statistics, plus they compared and discussed the results and conditions that affected the analyzed techniques. Later on we refer to this work to scrutinize our results. The latest review on computational methods and their applications as well as trends for automatic melanoma diagnosis was published in 2016 \[[@pone.0211318.ref004]\]. It was focused on feature extraction, feature selection, classification algorithms and evaluation procedures. Overview of statistics and results from the melanoma CAD and segmentation issues can be also found in \[[@pone.0211318.ref005], [@pone.0211318.ref006]\]. The main achievement of this work is first quantification of the C-SVM classification performance of melanoma (in terms of AUC) over a broad range of wavelet bases and SVM kernels. This article is a continuation of our latest research on wavelet-based features of melanoma, derived from dermoscopy images, and classified with various ML paradigms \[[@pone.0211318.ref007], [@pone.0211318.ref008]\]. Our motivation is driven by a systematic search for resolution invariant features and learning methods contributing to the computer pattern recognition of cutaneous melanoma. Motivation {#sec002} ---------- The first and still important approaches to the melanoma detection were segmentation techniques \[[@pone.0211318.ref004]--[@pone.0211318.ref006], [@pone.0211318.ref009]\]. Since dermoscopy images can be taken under different conditions, such techniques can suffer from unstable image illumination, different optical magnification, different skin complexions and presence of artifacts (hairs, reflections, bubbles of immersion fluid etc.). Under such circumstances successful pattern recognition is still possible, but may be limited in performance or strongly sensitive to datasets. Identification methods based on spatial and frequency information found in the skin texture can be a promising alternative. Application of the wavelet- and wavelet-packet- transform to human skin data was first applied in \[[@pone.0211318.ref010]\]. Patwardhan et al. \[[@pone.0211318.ref011]\] studied wavelet packets to decompose the sub-bands of the pigmented skin texture, which proved to derive features sensitive to the class of the dermoscopy image \[[@pone.0211318.ref007], [@pone.0211318.ref008], [@pone.0211318.ref012]\] and then be subject to appropriate machine learning methods. Some WPT attempts are also shown here \[[@pone.0211318.ref013]--[@pone.0211318.ref015]\]. Most of the melanoma wavelet-based classifiers quoted in the literature used only a single wavelet base (usually Daubechies 3 or 4 \[[@pone.0211318.ref011], [@pone.0211318.ref016]\]) to build classification models. In our latest research all popular and well defined wavelets are used. We showed that some selected wavelet bases prove to be more efficient and robust for the machine learning of dermoscopic images than the others. They also keep high classification efficiency in the downgraded image resolutions, that is exhibit resolution invariance. In \[[@pone.0211318.ref007]\] this was analyzed in the ensemble of different model types, and in \[[@pone.0211318.ref008]\] in simple back-propagated neural classifier setups. In this work we specifically study SVM classifiers of the wavelet-based features for the melanoma detection to analyze how wavelet bases can affect the quality of the margin-maximization learning paradigm. The choice of SVM is due to the inherent advantages of the method: i) the ability to generalize well as it maximizes the margin between the classes, ii) no local minima, and iii) flexibility of kernel selection within a unified architecture. Although in this article we want to study and compare with the previous works and literature the SVM classifiers (our work is the first quantification of the C-SVM classification performance of melanoma over a broad range of wavelet bases and SVM kernels), our primary objective is to contribute to optimal feature extraction methods. We believe that adequate wavelet base(s) representation can both yield high classification efficiency and be robust and efficient in worse or reduced resolution environments. For that objective we study one by one different learning paradigms, SVM included. Computational (memory, time) aspects of any machine learning experiment (not particularly SVM) are important factors which should be carefully planned. This is not only the algorithm convergence and regularness that matters, but if the developed classifier is to be used e.g. to support medical diagnosis (in CAD this is the ultimate goal), practical aspects play a role. Computer parts and systems constantly evolve. Nowadays personal smartphones with ARM-based processors are widely used as medical diagnostic tools, the melanoma CAD included (e.g. \[[@pone.0211318.ref017]\], see \[[@pone.0211318.ref018]\] for a review). Since (usually) ML algorithms for image recognition demonstrate high complexity, and small hand-held devices have limited processing power and memory (and the battery capacity), it is of great importance both to: efficiently probe the parameter space of the solution, and to look for features and learning paradigms that preserve high efficiency also in lower image resolutions. In this work our objective is to select the best wavelet bases in terms of AUC classification performance of the SVM binary classifiers of melanoma/dysplastic lesions and analyze how those wavelet bases perform with the reduced image resolutions. Materials and methods {#sec003} ===================== Experimental data {#sec004} ----------------- Data sets that are subject to our analysis are available on-line: 1. Data set A: <https://doi.org/10.17026/dans-zue-zz2y>, 2. Data set B: <http://www.fc.up.pt/addi/ph2%20database.html>. Because of concerns related to patient privacy and permissions set by the clinic that collected the data, some access restrictions apply to data set A. Data set A was collected in Poland from anonymous dermoscopy patients in 2012-2013 \[[@pone.0211318.ref019]\]. There were 102 malignant melanoma (M) and 83 dysplastic nevus (D) cases (altogether 185 images). Most (not all) of the JPEG images were annotated with some patients data like: sex, age, and location of the lesion (this kind of information is not used in this study). All of the images referred to lesions that were resected and examined under microscope in a histopathological lab. All the histopathological examinations were performed by one experienced team of pathologists, which could prevent from contingent false positives in the sample (we investigated this problem earlier and found few misdiagnosed cases by inexperienced specialists). This is important, since the histopathological examination is the ground truth which affects assignment of the class labels. The images were collected with resolution of 2272x1704 by Minolta Z5 digital camera with an extra dermoscopy extension and then JPEG-compressed as 3x8-bit RGB components. No magnification details were available. On the dermoscopic images no pre-processing tasks took place. This was due to the lack of rough artifacts (black borders, hairs, droplets of immersion fluid, etc.), but also to eliminate any bias on the final wavelet base selection. To be capable of performing the wavelet transform, the RGB representation (three integers per pixel) had to be translated into an indexed representation. This was done by linear, monotonic color mapping. Data set A was used to prepare two derived data sets of the same size by (recursively) averaging neighbor pixel values in 2x2 blocks: Images in the data set A2 had resolution of 1136x852. Images in the data set A4 had resolution of 568x426. Since each iteration of the wavelet decomposition downscales the input image by a factor of 2 both in rows and columns, to allow for three wavelet iterations the 568x426 set was padded with zeros (rows 427 and 428). Data set B was downloaded from the public database PH^2^ of the Automatic computer-based Diagnosis system for Dermoscopy Images (ADDI) in Portugal \[[@pone.0211318.ref020]\]. The PH^2^ database contains 200 dermoscopic images of melanocytic lesions (80 common nevi, 80 atypical nevi, and 40 melanomas) with medical annotations of expert dermatologists. The annotations refer to the dermoscopic criteria (Asymmetry: 0 = Fully Symmetry, 1 = Asymmetry in One Axis, 2-Fully Asymmetry; Pigment Network: AT = Atypical, T = Typical; Dots/Globules: A = Absent, AT = Atypical, T = Typical; Streaks: A = Absent, P = Present; Regression Areas: A = Absent, P = Present; Blue Whitish Veil: A = Absent, P = Present; Colors: 1 = White, 2 = Red, 3 = Light-Brown, 4 = Dark-Brown, 5 = Blue-Gray, 6 = Black), and clinical/histological diagnosis (0 = Common Nevus, 1 = Atypical Nevus, 2 = Melanoma). For most of the images binary masks from lesion segmentation is available. The PH^2^ images were taken during follow-up examinations at the Dermatology Service of Hospital Pedro Hispano (Matosinhos, Portugal) through Tuebinger Mole Analyzer with magnification of 20x. They are 3x8-bit RGB color images with a resolution of 768x560 pixels. The image size was cropped centrally to 760x552 pixels for the sake of the aforementioned wavelet recursive transformation. Since data set A contained only melanoma and displastic (atypical) lesions, we limited data set B to similar (not identical) medical cases. Finally data set B had the following structure: 40 melanomas (33 from clinial + histopathological, 7 lesions from only clinical examination), 77 common/dysplastic nevi (5 from clinial + histopathological, 72 lesions from only clinical examination). Altogether in B 113 lesion images were analyzed. In order to directly compare the efficiency of the SVM classifiers between data set A and B, the latter was additionally resized to resolutions A/A2/A4 and a series of experiments with the mixed data set A + B was done. Clinical statistics show that the incidence rate of melanoma is up to 5% (on average) within all 'suspicious' pigment lesions \[[@pone.0211318.ref021]\]. This means that the melanoma class is the minority class and is under-represented compared to the benign class. This status holds in the clinic, whereas in the melanoma CAD scientists attempt either to balance the number of cases from both classes (data set A is about this status) or to introduce sub- or over-sampling techniques \[[@pone.0211318.ref022]\] (this is required in B). In this work we took for both the data sets A and B the whole statistics for the minority class (melanoma) and experimented with different sub-sampling ratios \[[@pone.0211318.ref023]\] of the benign lesions. As in \[[@pone.0211318.ref008]\] for the data set A no major change in the classification performance (\<5% for B) was observed. This estimation allows us to conclude that the data imbalance problem in our experiment does not impact noticeably. In fact, in B the class ratio is 1:2, which is not a dramatic imbalance. Disregarding the way how imbalance is removed/controlled in this or (generally) other ML experiments, the problem in the clinic still holds, because a question arises how majority cases are picked out for the data set. It follows from our experience that usually those images are selected which visually fits in very well with the set of benign (dysplatic) cases. This may, however, seriously impact 'production' classifiers. We conclude that bias comes preferably from the 'clinical' source and not our data statistics or procedures. Wavelet features {#sec005} ---------------- Skin is texture that manifests repetitive patterns, pigment network and different structures (globules, streaks, etc.). This reality can be analyzed in a range of frequencies and spatial scales, which goes beyond pure Fourier analysis. In this section we introduce wavelet features that are used in this work. The Discrete Wavelet Transform (DWT) decomposes a signal into a coarse (average) signal and a signal of details. This can be interpreted, after Mallat \[[@pone.0211318.ref024]\], in terms of lowpass (L) and highpass filters (H). The subsequent levels of DWT operate recursively on the lowpass (scaling) part of the output. Products of the transform are downsampled (decimated) by two at each level. In the wavelet packet transform (WPT), the filtering operations are also applied to the signal of details. This results in a substantial amount of subbands in the output and makes the time (space) vs. frequency analysis much finer than in DWT. Multiresolution analysis with WPT can reveal scale-based features of the signal, which provides more precise information than other signal analysis techniques. Our working hypothesis is that WPT suits well to determine distinctive features characterizing the class of the dermoscopy image. Images are two-dimensional signals so one iteration of the Mallat filtering algorithm produces four sub-signals which can be considered as LL, LH, HL and HH filters after 1D wavelet transform on the rows and then on the columns. Those filters can be interpreted as decimated sub-images. We used three iterations of WPT. Since all wavelet filters are downsampled (by a factor of 2) after each run of the wavelet transform (in rows and columns separately), four subimages were produced. In [Fig 1](#pone.0211318.g001){ref-type="fig"} a summary image with all the used stages of decomposition is shown. It is evident that all channels of decomposition are used (WPT). ![The wavelet packet transform (WPT) decomposition.](pone.0211318.g001){#pone.0211318.g001} In one iteration four filters were produced, so in three turns we had 1 + 4 + 16 = 21 different transformation branches. In each branch we produced a list of twelve features. As features we took simple measures based on the energy content of the (sub)images \[[@pone.0211318.ref011], [@pone.0211318.ref012]\]: 1. e~i~---energies, 2. e~i~/e~max~---maximum energy ratios, 3. e~i~/Σe~k~, k ≠ i---fractional energy ratios, where energy e~i~, i = 1, 2, 3, 4 was defined as a sum of absolute values of the pixels. Our feature vector was composed of 21 × 12 = 252 components. Such feature extraction was carried out in different wavelet bases: orthogonal wavelets: Haar, Daubechies, Symlets, Coiflets, bi-orthogonal wavelets, and reverse bi-orthogonal wavelets. Those wavelet bases possess certain characteristics which however is not discussed in this article. SVM learning {#sec006} ------------ The Support Vector Machine (SVM) learning method finds a hyperplane that separates data points from two distinct classes in the way that it maximizes the margin between the two classes \[[@pone.0211318.ref025], [@pone.0211318.ref026]\]. The mathematical formulation of the problem \[[@pone.0211318.ref027]\] involves finding β and b such that for all the data points (x~i~, y~i~) it holds: $$\begin{array}{r} {f\left( x \right) = x^{\prime}\beta + b = 0,\; y_{i}f\left( x_{i} \right) \geq 1} \\ \end{array}$$ where the latter equality is accomplished for those data points that are (called) support vectors (y~i~ f(x~i~) = 1). In a case the two classes are not separable by a hyperplane SVM introduces a soft margin i.e. it releases the condition y~i~ f(x~i~) ≥ 1 − ξ~i~. The optimization problem formulated by Lagrangian (in the simplest form of the L^1^ norm): $$\begin{array}{r} {L = \frac{1}{2}\beta^{\prime}\beta + C\sum\limits_{i}\xi_{i} - \sum\limits_{i}\alpha_{i}\left( y_{i}\left( x_{i}^{\prime}\beta + b \right) - \left( 1 - \xi_{i} \right) \right) - \sum\limits_{i}\mu_{i}\xi_{i}} \\ \end{array}$$ with the bounds: $$\begin{array}{r} {\beta = \sum\limits_{i}\alpha_{i}y_{i}x_{i},\;\sum\limits_{i}\alpha_{i}y_{i} = 0,\;\alpha_{i} = C - \mu_{i},\;\alpha_{i},\mu_{i},\xi_{i} \geq 0} \\ \end{array}$$ is usually translated into an expression with the α coefficients only: $$\begin{array}{r} {max_{\alpha}\sum\limits_{i}\alpha_{i} - \frac{1}{2}\sum\limits_{i}\sum\limits_{j}\alpha_{i}\alpha_{j}y_{i}y_{j}x_{i}^{\prime}x_{j}} \\ \end{array}$$ with ∑~i~ α~i~ y~i~ = 0 and 0 ≤ α~i~ ≤ C, C being the regularization parameter. This concept is very successful in machine learning and data mining applications \[[@pone.0211318.ref028]\], especially when non-linear transformations of the primary data space (kernels) are introduced. The appropriate choice of kernel can map primary features to a larger dimensional space where the classes are (better) separable. Four kernels are well known and widely used: the linear kernel $x_{i}^{T}x_{j}$, which is a linear combination of features, the polynomial kernel $\left( x_{i}^{T}x_{j} + r \right)^{p}$ with the parameters: p-degree of polynomial and r-offset, the gaussian (radial basis) kernel exp(−γ \* \\|x~i~ − x~j~\\|^2^), γ \> 0, and sigmoid (MLP) kernel. Quadratic programming which is required to solve the optimization problem defined above is broadly implemented in various statistical toolboxes and professional computer libraries e.g. in \[[@pone.0211318.ref029]\]. In this study we take advantage of both the LIBSVM library \[[@pone.0211318.ref029]\] and Matlab Statistical Toolbox \[[@pone.0211318.ref030]\]. We take into account the binary SVM classifiers with the linear, polynomial, and gaussian kernel. The SVM paradigm, as compared to standard back-propagated neural networks, does not have hidden parameters, and is a global approach not prone to stuck at local minima. Different bounds set on the margin can affect the (upper bound of) generalization error and can control overfitting. Bayesian optimization {#sec007} --------------------- The objective of every learning method is to minimize a cost function, which is usually the cross-validated loss. In the learning paradigm exploited in this work, i.e. C-SVM, we want to search for the best parameters of the kernel and the domain size C, as well as the number of support vectors in the resulting model. One framework to minimize a real-valued objective function within a certain box is Bayesian Optimization (BO) \[[@pone.0211318.ref031]\]. BO is a global optimization technique (starting points are of no importance) but there are cases where it does not always secure accurate results. If no a priori knowledge about parameters is available, the widest finite range should be covered. BO algorithm is time-consuming due to large computational burden every iteration, therefore for the sake of experimental requirements it should stop if the cross-validated error rate (objective function) drops below a certain level or (at least) after reaching a fixed number of iterations, or maximum run time. BO introduces the so called acquisition function, that determines which points of the space should be covered. 'Goodness' of the point is evaluated upon the 'amount' of improvement it provides for the optimization. There can be different methods to achieve this and our model of choice is based on the posterior distribution function Q. If at x~best~ the lowest value of the posterior mean is μ~Q~(x~best~) then the 'expected improvement' is: $$\begin{array}{r} {EI\left( x,Q \right) = E_{Q}\left\lbrack max\left( 0,\mu_{Q}\left( x_{best} \right) - f\left( x \right) \right) \right\rbrack} \\ \end{array}$$ To escape from local minima of the objective function, in order to balance sampling at points with high and low efficiency of the model, the acquisition function checks after each iteration the standard deviation of the posterior distribution function $\sigma_{Q}^{2}\left( x \right)$. If it drops below a defined level, the area is recognized as 'overexploited' and the algorithm modifies the kernel to raise the variance σ~Q~. This offsets the solution to a new not yet examined point. Such strategy both covers the whole space and concentrates on the best feasible points. Results and discussion {#sec008} ====================== Our setup was coded first with LIBSVM \[[@pone.0211318.ref029]\] and then recoded to Matlab \[[@pone.0211318.ref030]\]. The experiments ran on a 4-core i7 workstation with 32GB RAM. We cross-validated (10-fold CV) our SVM classifiers on the standardized (N(0, 1)) predictor vectors. Search for the best subset of features is usually a tradeoff between reducing the bias and computational cost, and preserving the classification performance. In this work we did not reduce our feature vector and kept all the wavelet features in the analysis. Our objective was to finally compare different learning methods, so we had to remove any bias coming from feature selection methods, which can affect the analyzed learning paradigms in a different way. This was discussed in \[[@pone.0211318.ref032]\] where feature selection algorithms (CFS, PCA, GSFS) reduced the complexity of the classification, but performance was highly dependent upon the classifier. Analysis and discussion of our results is divided into several stages: 1. Analysis of the absolute SVM classification performance for all the kernel functions for data set A (A/A2/A4) and data set B (original resolution), which includes Figs [2](#pone.0211318.g002){ref-type="fig"}--[5](#pone.0211318.g005){ref-type="fig"}. Our objective is to analyze an overall classification performance of the kernel functions. In [Fig 6](#pone.0211318.g006){ref-type="fig"} we analyze average AUC levels for the gaussian, linear, and polynomial kernels. We also present classification performance for all the kernel functions for data set B resized to resolutions A/A2/A4 (Figs [7](#pone.0211318.g007){ref-type="fig"}--[9](#pone.0211318.g009){ref-type="fig"}). Again, in [Fig 10](#pone.0211318.g010){ref-type="fig"} we analyze average AUC levels for the three kernels. 2. Analysis of the classification performance for a fixed kernel function for three image resolutions for data set A (Figs [11](#pone.0211318.g011){ref-type="fig"}--[13](#pone.0211318.g013){ref-type="fig"}) plus for the combined data set A + B(resized) (Figs [14](#pone.0211318.g014){ref-type="fig"}--[16](#pone.0211318.g016){ref-type="fig"}). Our objective here is to pick up those wavelet bases that show outstanding classification performance for individual data sets with specific image resolutions. Our second objective is to select those wavelet bases that preserve high classification efficiency through the presented spectrum of resolutions (resolution invariance). Analysis of the joint data set A + B cast light both on compatibility of the data sets and stability of the learning algorithm (independence on the data set). In [Fig 17](#pone.0211318.g017){ref-type="fig"} we show the number of support vectors for the three kernel functions used in the SVM classifier of data set A. This is also a test for quality and stability of features built upon certain wavelet numbers. 3. Analysis of our SVM results against our previous experiments with the wavelet features: ensembling (Figs [18](#pone.0211318.g018){ref-type="fig"}--[20](#pone.0211318.g020){ref-type="fig"}) and feed-forward neural networks classifiers (Figs [21](#pone.0211318.g021){ref-type="fig"}--[23](#pone.0211318.g023){ref-type="fig"}). Our objective is to show the property of the resolution invariant wavelet features for melanoma detection in the background of other tested classification paradigms. 4. Analysis of our SVM results with help of the melanoma CAD summary articles (Masood \[[@pone.0211318.ref003]\], Oliveira \[[@pone.0211318.ref004]\]) and some individual studies from the literature. ![AUC as a function of the wavelet number (see text) for data set A for the three SVM kernels optimized by the Bayesian search.\ Resolution is (2272x1704). To easily follow the trend the points belonging to the same kernel function are connected with a line.](pone.0211318.g002){#pone.0211318.g002} ![AUC as a function of the wavelet number for data set A for the three SVM kernels optimized by the Bayesian search.\ Resolution is A2(1136x852).](pone.0211318.g003){#pone.0211318.g003} ![AUC as a function of the wavelet number for data set A for the three SVM kernels optimized by the Bayesian search.\ Resolution is A4(568x426).](pone.0211318.g004){#pone.0211318.g004} ![AUC as a function of the wavelet number for data set B (original resolution 760x552) for the three SVM kernels optimized by the Bayesian search.](pone.0211318.g005){#pone.0211318.g005} ![Average AUC levels for the three kernel functions, for data set A (resolutions A, A2, A4).](pone.0211318.g006){#pone.0211318.g006} ![AUC as a function of the wavelet number for data set B for the three SVM kernels optimized by the Bayesian search.\ Resolution is resized to (2272x1704).](pone.0211318.g007){#pone.0211318.g007} ![AUC as a function of the wavelet number for data set B for the three SVM kernels optimized by the Bayesian search.\ Resolution is resized to (1136x852).](pone.0211318.g008){#pone.0211318.g008} ![AUC as a function of the wavelet number for data set B for the three SVM kernels optimized by the Bayesian search.\ Resolution is resized to (568x426).](pone.0211318.g009){#pone.0211318.g009} ![*Average AUC levels for the three kernel functions, for data set B* resized to resolutions:** A, A2, A4.](pone.0211318.g010){#pone.0211318.g010} ![AUC for the SVM gaussian kernel as a function of the wavelet number for the data set A (resolutions: A, A2, A4).](pone.0211318.g011){#pone.0211318.g011} ![AUC for the SVM polynomial kernel as a function of the wavelet number for the data set A (resolutions: A, A2, A4).](pone.0211318.g012){#pone.0211318.g012} ![AUC for the SVM linear kernel as a function of the wavelet number for the data set A (resolutions: A, A2, A4).](pone.0211318.g013){#pone.0211318.g013} ![AUC for the SVM gaussian kernel as a function of the wavelet number for the combined data set A + B (B is resized to A, A2, A4).](pone.0211318.g014){#pone.0211318.g014} ![AUC for the SVM polynomial kernel as a function of the wavelet number for the combined data set A + B (B is resized to A, A2, A4).](pone.0211318.g015){#pone.0211318.g015} ![AUC for the SVM linear kernel as a function of the wavelet number for the data set A + B (B is resized to A, A2, A4).](pone.0211318.g016){#pone.0211318.g016} ![Number of support vectors for the three kernel functions used in the SVM classifier of data set A.](pone.0211318.g017){#pone.0211318.g017} ![AUC for the ensemble learning optimized by the 'recall' measure as a function of the wavelet number for data set A (resolutions: A, A2, A4).](pone.0211318.g018){#pone.0211318.g018} ![AUC for the ensembling learning as a function of the wavelet number for data set B resized to resolutions A, A2, A4.](pone.0211318.g019){#pone.0211318.g019} ![AUC for the ensembling learning optimized by the 'recall' measure as a function of the wavelet number for data set A + B (B is resized to A, A2, A4).](pone.0211318.g020){#pone.0211318.g020} ![AUC for the neural feed-forward classifier as a function of the topology and learning function (see text for the details).\ Data set is A in the native A and reduced resolutions A2 and A4.](pone.0211318.g021){#pone.0211318.g021} ![AUC for the neural feed-forward classifier as a function of the topology and learning function (see text for the details).\ Data base is B in the native resolution 760x552.](pone.0211318.g022){#pone.0211318.g022} ![AUC for the neural feed-forward classifier as a function of the topology and learning function.\ Data base is A + B (B is resized to the native and reduced resolutions of the data base A).](pone.0211318.g023){#pone.0211318.g023} Ad.1 {#sec009} ---- Full results of our experiments on data set A/A2/A4 are presented in Figs [2](#pone.0211318.g002){ref-type="fig"}--[4](#pone.0211318.g004){ref-type="fig"} for the three kernel functions and the three resolutions. In these and the follow-up figures the wavelet number denotes: 1 = Haar, 2-11 = Daubechies (db1-db10), 12-18 = symlets (sym2-sym8), 19-23 = coiflets (coif1-coif5), 24-38 = bi-orthogonal wavelets (Bior1.1, 1.3, 1.5, 2.2, 2.4, 2.6, 2.8, 3.1, 3.3, 3.5, 3.7, 3.9, 4.4, 5.5, 6.8), and 39-53 = reverse bi-orthogonal wavelets (Rbio1.1, 1.3, 1.5, 2.2, 2.4, 2.6, 2.8, 3.1, 3.3, 3.5, 3.7, 3.9, 4.4, 5.5, 6.8). The way how images, so 2D signals, are analized by the filter bank is another factor. In this work 1Dx1D filtering (separately along rows and columns) was performed, which followed Mallat \[[@pone.0211318.ref024]\] (in other approaches 2D filtering elements are potentially possible). For the SVM optimization we used the Sequential Minimal Optimizer \[[@pone.0211318.ref031]\] and initiated the alpha coefficients with zeros. To check for the optimization convergence the gradient tolerance was set for 10^−3^. The AUC values were averaged over 10-fold cross-validation (10 turns, 9 training subsets and 1 hold-out testing subset in each turn). The optimizatoin space that was checked included: - kernel functions: gaussian, polynomial, linear, - parameters of the kernel functions: for 'gaussian' it is the inverse variance parameter (γ), for 'polynomial' it is the order of the polynomial (d), - the maximum penalty imposed on margin-violating data points (C). We scanned the above-mentioned model parameters in a log step between 1e-3 and 1e6. The maximum number of iterations was fixed for 60 and the run time was left infinite. Figs [2](#pone.0211318.g002){ref-type="fig"}--[4](#pone.0211318.g004){ref-type="fig"} show AUC (the Area Under the ROC Curve) vs. the wavelet number for data sets A(2272x1704), A2(1136x852), and A4(568x426) for the binary SVM classifiers of melanoma vs. displastic lesion. This is a Bayesian search over an available space of the gaussian (marker is square), linear (triangle), and polynomial (circle) kernel. Clearly visible is an inphase sinusoidal fluctuation of AUC through all the resolutions. Since all the kernel functions fluctuate over the wavelet number in a similar way (the trend, not the absolute level) this must be the feature set that is responsible for the phenomenon, not the learning paradigm. Indeed, variability in the classification performance comes both from the individual wavelet base, and is also present within each wavelet family. The latter is obvious since wavelets from one wavelet family differ in scale and offset. The skin texture apparently 'prefers' some scales. The wavelets that were used in the analysis have different properties like smoothness, symmetry, or support. The wavelet base complies with some model constraints and has a diverse impact on the texture classification. Haar, Daubechies, symlets, and coiflets are orthogonal wavelets. Orthogonal wavelets are energy preserving and non-redundant, which follows the formula X^2^ = A^2^ + D^2^ (X-image, A-reconstructed image of approximation, and D-reconstructed image of details). After Mallat \[[@pone.0211318.ref024]\] wavelets can be represented by a filter bank. In this case this is a pair of a single scaling function and a wavelet. The bi-orthogonal wavelet filters, on the other hand, generate one scaling function and a wavelet for decomposition, and another pair for reconstruction. They have the property of perfect reconstruction i.e. X = A + D and a linear phase, which makes them more symmetric and compact (these properties are useful in signal analysis). For the sake of symmetry we can divide the selected wavelet bases into: symmetric functions (Haar and (reverse)-biorthogonal wavelets), nearly symmetric functions (symlets, coiflets), and asymetric functions (Daubechies). Taking into account this variety of properties we conclude that they are individual wavelets, not wavelet families that make the SVM classification of melanoma dermoscopy images more (or less) efficient, although certain wavelet families have a higher number of 'efficient' wavelets than the others. Performance of the SVM classification on data set B for the three kernels is shown in [Fig 5](#pone.0211318.g005){ref-type="fig"}. Images from this data set have rather low resolution but (unexpectedly) they yield (average) AUC values about 10% higher. Two factors may play a role: Compression: Data set A is a collection of JPEG images compressed by a lossy algorithm. Rough estimation on the information content of an image from data set A is possible from its JPEG file size (assumed we skip all the headers from different stages of the compression). We have the following (average) file sizes: 1.35MB (A, 2272x1704), 0.59MB (A2, 1136x852), and 0.17MB (A4, 568x426). This yields: 2.79 bit/pix, 4.89 bit/pix, and 5.45 bit/pix from the highest to the poorest resolution, respectively. Data set B, in turn, has (standardized) resolution of 760x552 and the average file size is 1.33MB. This yields information capacity of about 24.26 bit/pix. Such an extremely high information content is due to its lossless file format, which is BMP. Even if we compare data set B only to the lowest resolution data set A4, the factor between them is about five. We know that the JPEG algorithm has a quantization stage which 'flattens' pixels values based on the JPEG quality/compression ratio. This may affect the image structure which translates to decreased sensitivity for patterns in the compressed image. Quantitative estimations between data sets A and B are not possible as far as (pre)processing steps in B are unknown. Class content: The existing melanoma pattern recognition methods (a brief comparison is done later in this article) are rather 'advanced' systems so are 'costly' in terms of feature extraction/selection and classification. This reflects the 'clinical' difficulty in discriminating melanoma from displastic (anomal) skin lesions. Data set A has two distinct classes: melanoma and dysplastic nevus while data set B, in turn, has class melanoma and a mixture of dysplastic (1/3) and common (2/3) nevus. While the benign class in B (common + dysplastic nevus) must be more widely distributed in the feature space than dysplastic cases alone (A), the 'distance' between the melanoma and this (mixed) class in B should be larger i.e. more efficient for classification, than in A. Those two factors may be, though qualitative, but reasonable explanation for the difference in the average AUC level between the data sets A and B. Since there are a few variables in our SVM experiments (the wavelet bases, different resolutions, and finally the three kernel functions) in order to interpret the results and draw straight-forward conclusions we will limit the number of factors and answer the question which kernel function produces the best average classification performance in terms of AUC. [Fig 6](#pone.0211318.g006){ref-type="fig"} has three sections in which we show an average level of AUC produced by the SVM classifiers with the appropriate kernel functions: gaussian (dotted line), polynomial (dash-dotted line), and linear (solid line) for the nominal A and two downgraded resolutions A2 and A4. The polynomial kernel function keeps its average level in the reduced image resolutions while the liner one looses classification efficiency considerably. An interesting finding is that the gaussian kernel rises continuedly towards lower resolutions. Beyond this point (568x426), when we go further with the resolution reduction the classification performance drops rapidly to a level of about 70%. For the polynomial and linear kernel we can see a drop in performance for the intermediate resolution (1136x852) and then they recover loss and even outrun a little bit in 568x426. This behavior is independent of the absolute AUC level, this happens also for the linear kernel whose performance beyond the nominal 2272x1704 is poor. In order to check whether those average levels are data set specific (A) or general, we resized images from the data set B to the resolutions of the data set A. This can be found in Figs [7](#pone.0211318.g007){ref-type="fig"}--[9](#pone.0211318.g009){ref-type="fig"}. Based on the numerical results of Figs [7](#pone.0211318.g007){ref-type="fig"}--[9](#pone.0211318.g009){ref-type="fig"} we could depict in [Fig 10](#pone.0211318.g010){ref-type="fig"} an average level of AUC produced by the SVM classifiers with the same kernel functions (gaussian, polynomial, and linear) for data set B resized to resultions A, A2, and A4. Again, the linear kernel has a decreased (average) level of AUC along with the reduced resolution. As in A the gaussian kernel performs better for (568x426), but slightly worse than the polynomial kernel for (1136x852). Taking into account statistical aspects of learning and possible performance drops for (1136x852) apparent in A, both Figs [6](#pone.0211318.g006){ref-type="fig"} and [10](#pone.0211318.g010){ref-type="fig"} look similar. From this analysis we can conclude the following: 1. kernel functions play a key role in the performance of the SVM classification. The poorest linear kernel (dot product of the feature vector components) cannot appropriately map features-related (noisy) classes of melanoma and benign cases. The polynomial kernel, which can translate the primary feature vector into more dimensional space (degree of the polynomial is a parameter of the model) performs better, but the gaussian kernel outperforms all the other. This is because it provides the feature vector mapping in a more nonlinear fashion through introducing cluster-like subspaces. Apparently this is suitable for the SVM classifiers for the melanoma detection. 2. wavelet features classified by SVM with different kernel functions have 'preferential' resolutions where they reach the most optimal classification efficiency. Our analysis on two data sets (A, B) show that for the gaussian kernel the most efficient space for the SVM classifier is this of 568x426. This observation is important since can affect the computational aspects of the melanoma classification. Low(er)-resolution image analysis would be potentially resource-saving without (noticeable) degradation of the classification performance. Ad.2 {#sec010} ---- In stage 1 we performed an analysis of the average SVM classification performance for all the kernel functions for two data sets: A and B. We concluded that tha gaussian kernel outperforms the polynomial and linear kernel especially towards lower image resolutions and that both data sets yield compatible results. In stage 2 we focus on the gaussian kernel function and analyze the absolute classification performance (in terms of AUC) for the three image resolutions (2272x1704, 1136x852, 568x426) in A and the combined data set A + B. Our first objective here is to pick up the wavelet bases that show outstanding classification performance for the individual data sets with specific image resolutions. The second objective is to select those wavelet bases that preserve high classification efficiency through the presented spectrum of resolutions (resolution invariance). Analysis of the joint data set A + B should prove both compatibility of the data sets and stability of the learning algorithm (independence on the data set). Although our analysis is focused on the gaussian kernel (A: [Fig 11](#pone.0211318.g011){ref-type="fig"}, A + B: [Fig 14](#pone.0211318.g014){ref-type="fig"}), we present results for the other two kernel functions as well (A: Figs [12](#pone.0211318.g012){ref-type="fig"} and [13](#pone.0211318.g013){ref-type="fig"}, A + B: Figs [15](#pone.0211318.g015){ref-type="fig"} and [16](#pone.0211318.g016){ref-type="fig"}). This should cast light on robustness and stability of the wavelet features. Below we present for the individual kernel functions some consolidated figures with the three image resolutions. In each resolution we select a set (we arbitrarily choose n = 6) of wavelet bases that show an outstanding AUC classification performance. In [Table 1](#pone.0211318.t001){ref-type="table"} we call it 'Condition 1'. In 'Condition 2' we require that the wavelet number has the highest AUC. 10.1371/journal.pone.0211318.t001 ###### Wavelet numbers collected for the kernel functions (gaussian, polynomial, linear) satisfying some conditions for the AUC values (numbers in parentheses have equal AUC values). Condition 1: outstanding SVM classification efficiency. Condition 2: highest classification efficiency. Condition 3: high classification efficiency in all the resolutions (order of priority in A). Condition 4: high classification efficiency in all the resolutions (order of priority in A2). Condition 5: high classification efficiency in all the resolutions (order of priority in A4). ![](pone.0211318.t001){#pone.0211318.t001g} Data Set Resolution Condition 1 Condition 2 Condition 3 Condition 4 Condition 5 ------------ ------------ ----------------------------- ------------- ------------------------- ------------------------ ------------------------- gaussian A 2272x1704 (46 48) 26 (39 2 23) 46 46 48 26 2 39 12 A2 1136x852 26 43 (2 39 46 3) 26 26 2 46 39 19 30 A4 568x426 (46 26) 25 47 48 42 46 46 26 48 2 19 12 B 760x552 3 (8 10 13 15 26) 3 \- \- \- A + B 2272x1704 41 (46 48) 45 40 23 41 41 46 48 40 6 9 A2 + B2 1136x852 24 (26 2) 39 25 42 24 24 2 26 25 12 47 A4 + B4 568x426 26 (46 25) 12 (47 48) 26 26 46 25 12 48 2 polynomial A 2272x1704 48 46 41 39 12 2 48 46 39 2 25 26 19 A2 1136x852 26 43 46 2 24 19 26 26 46 2 25 19 39 A4 568x426 26 46 25 47 48 42 26 26 46 2 19 39 25 B 760x552 26 (40 44 10 3 8) 26 \- \- \- A + B 2272x1704 46 45 41 48 53 15 46 46 24 9 26 20 25 A2 + B2 1136x852 2 39 26 24 12 25 2 24 26 25 19 2 9 A4 + B4 568x426 46 25 42 49 26 24 46 46 25 24 26 2 9 linear A 2272x1704 46 48 41 23 26 2 46 46 26 12 2 39 19 A2 1136x852 46 2 26 25 39 1 46 46 2 26 39 19 18 A4 568x426 46 26 47 25 14 12 46 46 26 19 18 12 39 B 760x552 (40 8) (25 37 47 13) 40 \- \- \- A + B 2272x1704 41 46 45 48 40 14 41 46 48 41 26 15 25 A2 + B2 1136x852 47 12 52 46 25 2 47 46 24 26 3 2 39 A4 + B4 568x426 46 25 26 43 48 1 46 46 26 2 3 48 43 As a next point we analyze high-AUC wavelet numbers almost constant in all the three resolutions. Our 'algorithm' to choose such cases is: 1. step 1: for i ∈ {wavelet numbers} calculate SUM(i) = ∑~j=1,2,3~ AUC~i~(resolution~j~) 2. step 2: choose n (n = 6) consecutive numbers: $i = \text{argmax}_{i}SUM_{i}$ Conditions 3,4,5 in [Table 1](#pone.0211318.t001){ref-type="table"} select wavelet numbers that show high efficiency in all the resolutions (assumed a data set and a kernel function). In such a set the order of priority of the wavelet numbers depends on the resolution preference of the analyzed dermoscopy images. In 'Condition 3' preferential is resolution 2272x1704, in 'Condition 4' 1136x852, and in 'Condition 5' 568x426. [Table 1](#pone.0211318.t001){ref-type="table"} shows the following: 1. Two most frequent wavelet numbers are marked: 46-in bold face, 26-in italic. 2. Wavelet numbers producing high AUC ('Condition 1), the best one included ('Condition 2'), belong for the most part to the Bi-orthogonal (24-38) and Reverse Bi-orthogonal (39-53) wavelet families, 3. wavelet numbers exhibiting good resolution invariance (keeping high AUC results in the three resolutions) come out in all the kernel functions ('Condition 3,4,5'). Some auxiliar information about how different wavelets 'fit' to the classifier can also be found in [Fig 17](#pone.0211318.g017){ref-type="fig"}. 'Good' wavelets from [Table 1](#pone.0211318.t001){ref-type="table"} (from the classification performance point of view) have a (preferably) small number of the support vectors. They are also less dispersed both between different kernel functions and different image resolutions. The perfect case waveletnumber = 46 shows very small values of the support vectors and almost perfect independence of the kernel function. This can be attributed to good quality and stability of features built upon wavelet number 46. Collective behavior of all the wavelet bases measured both in AUC and in the number of support vectors ([Fig 17](#pone.0211318.g017){ref-type="fig"}) casts light on how the SVM classifier 'prefers' the studied resolutions. It seems that the nominal resolution 2272x1704 and 568x426 are comparable in terms of the classification performance and stability, but 1136x852 is apparently worse (in [Fig 17](#pone.0211318.g017){ref-type="fig"} see the dispersion in the number of support vectors between the kernel functions). After analysis of [Table 1](#pone.0211318.t001){ref-type="table"} we come to some conclusions regarding the SVM classification of the dermoscopy images (class melanoma vs. class displastic/common nevus) with the presented wavelet features. First, we can see that efficient SVM learning of dermoscopy images is possible with some selected wavelet bases. Most of the high-performance wavelet bases belong to the symmetrical, bi-orthogonal or reverse bi-orthogonal wavelet families. Second, there are some wavelet numbers which keep the SVM classification performance on an almost stable level through a range of resolutions. The two wavelet bases that show best performance in all the studied resolutions are: 46 (RBio3.1), and 26 (Bior1.5). Ad.3 {#sec011} ---- In this section we compare our SVM results for data set A and A + B with the previous experiments with data set A concerning ensembling \[[@pone.0211318.ref007]\] and artificial neural networks \[[@pone.0211318.ref008]\]. Ensemble learning consists in learning of single weak and diverse learners \[[@pone.0211318.ref033]\]. To build an ensemble of models we started from an empty ensemble and added step-by-step the best model in the so called Out-of-Train procedure (OOT) (after Breiman's Out-Of-Bag technique \[[@pone.0211318.ref034]\]). Several models were trained on the training set and these models were compared by evaluating the prediction errors on the validation set. The best models became ensemble members until the ensemble got the desired size. Six model families were used for the training: LDA, Kernel Ridge Regression with a polynomial kernel k(x, x′) = (a + x.x′)^b^, MLP trained with the RPROP descent (with the changeable number of nodes), Perceptron trained with a second order gradient descent, C4.5 decision trees, and Matlab data trees (dtree). The wavelet features were learnt by an ensemble of models in this way, that the ensemble optimized (one by one) seven different quality measures: accuracy, precision, F1-score, fp rate, specificity, ber and recall. The final model was tested on a separate unseen set of data and the area under the ROC curve (AUC) was calculated. In those experiments we measured AUC and its error bar to draw some conclusions about stability of the models. Accuracy and recall were concluded to have the highest AUC levels plus modest error bars. In [Fig 18](#pone.0211318.g018){ref-type="fig"} we plot classification performance of the ensemble built on data set A by optimizing the recall quality measure. As we can see there are similar fluctuations of AUC over the wavelet number as in SVM ([Fig 11](#pone.0211318.g011){ref-type="fig"}) and the average level of AUC in ensembling is higher than this in SVM by about 6%. This difference in performance between ensembling and SVM should be attributed both to the different learning paradigms (ensembles may be more 'flexible' for optimization as far as high dimensional feature space is concerned), and (less) to the Bayesian optimization process, which is a statistical procedure. The most optimal wavelet base in our ensembling experiments (\[[@pone.0211318.ref007]\]) was reported wavelet number 46 i.e. reverse bi-orthogonal wavelet base Rbio3.1 and 26 i.e. bi-orthogonal wavelet base Bior1.5. The mixture set A + B ([Fig 20](#pone.0211318.g020){ref-type="fig"}) yields similar results, except for more wavelet numbers that satisfy the (almost) constant performance in all the three resolutions i.e. 46, 26, 24, 25, 39. We remember that those wavelet numbers are present in [Table 1](#pone.0211318.t001){ref-type="table"}. Artificial neural networks (ANN), in turn, is a black-box approach to the knowledge acquisition but can model complex relationships between inputs and outputs due to the nonlinear processing capabilities of its constituent neurons. ANN are widely used as classification systems for the melanoma CAD. In \[[@pone.0211318.ref008]\] we briefly commented on the existing works on ANN and contributed to the automatic classification of melanoma from dermoscopy images by ANN. In our study on data set A we used 252 input neurons that represented the same set of wavelet features as in this work, extracted for one wavelet base, which was RBio3.1 (wavelet number = 46). There are no no methodical studies in the literature how the ANN structure (hidden layers) affect the melanoma classification performance. After careful considerations and computational attempts on both CPU- and GPU-based parallel computing platforms we arbitrarily limited ourselves to a few combinations of hidden neurons grouped into one or two hidden layers: \[10\], \[20\], \[10-10\], \[20-10\], \[20-20\], which is not a considerably worse setup compared to the literature. The validation data was used to optimize (by mean square error) and update the weights in the backpropagation phase where all the layers had sigmoid activation functions. The testing set was used to calculate AUC on the unknown subset of dermoscopy images from data set A. We used several training algorithms with 'standard' base parameters: Levenberg-Marquardt μ = 0.001 (L1), Bayesian Regularization μ = 0.005 (L2), Broyden-Fletcher-Goldfarb-Shanno (L3), Conjugate Gradient with Powell-Beale restarts (L4), Fletcher-Powell Conjugate Gradient (L5), Polak-Ribiére Conjugate Gradient (L6), Gradient Descent learn\_rate = 0.01 (L7), Gradient Descent with Adaptive Learning learn\_rate = 0.01 (L8), Gradient Descent with Momentum learn\_rate = 0.01, momentum = 0.9 (L9), Variable Learning Rate Gradient Descent learn\_rate = 0.01, momentum = 0.9 (LA), One Step Secant (LB), Resilient backpropagation learn\_rate = 0.01, Δ = 0.07 (LC), and Scaled Conjugate Gradient (LD). Our objective was to find the best performing ANN for the classification of melanoma dermoscopy images under the assumption that only those aforementioned limited topologies are taken into account (starting with \[10\] hidden neuron on one hidden layer up to \[20 20\] hidden neurons on two layers). Beside classification performance our experiments had to show the computational efficiency (in terms of learning epochs required to gain a defined error level). For that we grouped results of our ANN setups in terms of pairs (number\_ofepochs, setups\_finished). As a necessary efficiency condition for further analysis we took the median of the number of epochs for the finished setups, which was 20 epochs. In [Fig 21](#pone.0211318.g021){ref-type="fig"} we plot numerical results of AUC for the resolutions 2272x1704, 1136x852, 568x426, for the five setups of the hidden layers and for the thirteen different back-propagation learning algorithms assumed that the number of epochs is below 20. Absence of L7-LA proves that methods based on (variations of) gradient decent converges very slowly (maximum number of epochs was sometimes even above 1000). Among the remaining cases there are no single winners in our optimization space, different combinations of the learning algorithms and topologies reach high AUC values. Each topology has a preferable learning algorithm assumed a resolution, and in reverse each learning algorithm has the most efficient topology. In any case one can optimally select a triple (resolution, learning\_alg, hidden\_layers) to reach a high AUC and match the experimental setup. In search for resolution invariance we scan through all the resolutions and see that especially L1 (Levenberg-Marquardt) shows extreme robustness for almost all the topologies e.g. (A, L1, \[10 10\]) = 0.97, (A2, L1, \[20\]) = 1.0, and (A4, L1, \[20\]) = 0.98. L6 (Polak-Ribiére Conjugate Gradient) proves to be the second best (A, L6, \[10 10\]) = 0.96, (A2, L6, \[20 10\]) = 0.99, and (A4, L6, \[10\]) = 0.93. The map of high AUC levels through all the learning functions shows us that images of 1136x852 pixels are the most suitable for classification with ANN (for SVM the preferable resolution was 568x426). Higher absolute results (although more sparse in the optimization space) are reached for data set B ([Fig 22](#pone.0211318.g022){ref-type="fig"}). The same applies to [Fig 23](#pone.0211318.g023){ref-type="fig"} for the joint data set A + B. Our ANN experiments for one particular wavelet base (wavelet number 46) agree qualitatively with the SVM results and the ensembling results in high classification efficiency and that the wavelet feature (46) is resolution-proof. Unfortunately no more wavelet bases were studied with ANN due to considerable computational burden. Ad.4 {#sec012} ---- A detailed/comprehensive comparison between various methods on the classification of melanoma and benign lesions cannot be done. While the reviews show the techniques and results of the published works, they cannot explain why and to which extent the classification performance varies with the data set, and with the method (the pre-processing step, feature extraction and selection included). This is because not enough information about all the crucial steps of the data collection and analysis are available from the papers, which was criticized by the reviews' authors. Sometimes even the presented methods have apparent drawbacks. In fact most of the results falling to the same machine learning paradigm seem to be comparable. This is because we usually compare some absolute performance measures i.e. levels (assumed the same measures) with no variance/dispersion of the results. There are only few works applying a method or the whole pattern recognition system to more than one data set. When different features are extracted/used, comparing performance of even identical machine learning methods is controversial. There is a lot of melanoma decision support systems, each having its own paradigm, a feature set, and machine learning approaches. Here we want to show our results against the background of existing experiments, based on the review of Masood \[[@pone.0211318.ref003]\], and more contemporary Oliveira \[[@pone.0211318.ref004]\] (crucial bibliography can be found there). Image acqusition, pre-processing, segmentation, feature extraction (and selection), and finally classification are fundamental steps in the non-invasive computer diagnosis of melanoma. Dermoscopic images, we use in this work, are the predominant part of the existing visual material recording skin lesions. This technique was briefly characterized in the introduction. The most important factor that makes this technique most useful in the clinical diagnosis is illumination and magnification of the moles. There are also drawbacks because artefacts such as hair or light reflections and liquid bubbles get magnified as well. For all that dermoscopy remains the most sensitive visual examinations that outperforms plain macroscopic images. Some more advanced techniques like trans-illumination, laser-based methods, ultrasonography or spectral systems are not widely used and are not significant in the mass melanoma screening. Melanoma image recognition can be reliable when trained on robust and comprehensive data. Unfortunately the only few existing standard data sets that can be used for training and testing in various melanoma CAD systems are relatively fresh and even if some statistics amount to some thousands of images, usually only a fraction of cases can be used. This is due to the variety of classes (non-pigment/pigment lesions, dermal/junctional/compound nevi, typical/dysplastic lesions, melanoma/other skin cancer cases etc.). For the melanoma binary classifiers, which have clinical significance, only melanoma and dysplastic cases should be selected. The most important existing ML experiments for melanoma usually operate on proprietary small data sets where the melanoma class has 20, 40, sometimes up to 100 cases. They are usually unavailable in public and those experiments are not compared against the standard data sets. Data set A, which we use, is a medium-size data set for melanoma classification, and its advantage is high resolution, which is not the case of most existing data sets. High resolution images allow for experiments with reduced, downgraded resolutions. In this context data set A is unique. To check performance of the trained SVM model one of the aforementioned public data sets was chosen---data set B described in the section 'Experimental Data'. Prior to segmentation some pre-processing steps can be done, if required. They can remove artifacts and enhance contrast. As it was explained in Motivation data sets A and B are taken without any methods enhancing image quality. Segmentation is an important step to extract regions of interest (ROI) in the image. Lesion border detection methods play an important role and this step has its own vast methodology and classification approaches (SVM included). As it is clearly visible from the literature, segmentation is rather unstable method, which can yield quite good results when fined tuned to the image local conditions. In fact even different images from the same data set can have slightly different magnification and/or illumination parameters (not to mention about the different artifacts), different skin complexions etc. so adjustment to one global (data set wide) parameter set is hard. Feature extraction methods for melanoma discrimination can be based on diverse grounds. They include clinical approaches (based on asymmetry, border, color, diameter, evolution or elevation of features), widely known as ABCD(E), 7-Point Checklist, Menzies etc., pattern analysis methods, where specific global or local patterns are visible in the dermoscopy image (homogeneous, starburst, globular, etc.), or various shape or color representation/variation features. Finally texture based features may be used, which determine some statistical texture descriptors e.g. the gray-level co-occurence matrix, (dis)similarity, entropy, momenta etc. Some popular methods to extract those aforementioned features is: thresholding, color discrimination, discontinuity-based segmentation (active contour), region-based segmentation (split and merge, morphological flooding), soft computing and fuzzy logic included. Other feature extraction methods found in the literature include filter-based approaches: Gabor (dyadic) filters, Fourier power spectrum, Gaussian derivative kernels or multichannel filtering. Some authors raise objections that a lot of different features are used to feed sophisticated classifiers of melanoma, but there is little discussion about the real meaning of those features and most of the studies do not report the details of their feature extraction procedures. Wavelet features which are exploited in this work probe the whole area of the image in various scales and frequency subsets, and are sensitive filters of localized frequencies in the skin texture. When this filtering method is done recursively through all the decomposition branches, the wavelet packet transform (WPT) is applied. Identification of malignant melanoma by wavelet analysis is less artifacts sensitive, independent of 'visual' conditions of the image thus less error-prone. References on experiments with the wavelet features are collected in the Introduction plus a brief summary is done in \[[@pone.0211318.ref003]\] and \[[@pone.0211318.ref004]\]. The wavelets features eliminate the segmentation step so are a promising alternative to most of the feature sets on the market for the melanoma detection. Results of our previous experiments and this work show that selected wavelet features are classification efficient and demonstrate robustness in a chosen spectrum of classifiers (ensembling, ANN, SVM). Feature selection is used to eliminate redundant, irrelevant or noisy features in order to reduce classifier complexity for better generalization and to make the learning process more time-, memory-, and storage-effective. From the known studies very few report the details of their feature selection procedures. The available ones treat feature selection as an optimization problem and apply greedy, heuristic, or genetic algorithms or some customized approaches. In our method we do not use any feature selection procedures, which was explained in the Motivation. The classification step to discriminate melanoma lesions vs. dysplastic or other lesions include different learning paradigms. This is usually divided into families of learning methods. They are: artificial neural networks (MLP, RBF), decision trees (CART, C4.5, etc.), Bayesyan networks, support vector machines (linear, polynomial and gaussian kernel) and ensembles (random forest, bagging, boosting, etc.). All those approaches are widely represented in the literature and summerized in the reviews. Below we restrict to SVM-based results. A brief theory to SVM methods were presented in the Introduction. This is a linear classifier, which can adapt to non-linearity of the model through kernel functions. Despite their inherent advantages, there exist relatively few studies investigating the utility of SVMs in melanoma recognition. SVMs are used as melanoma classifiers in two scenarios: - as an indirect classifier identifying one of the clinical (dermoscopic) features of melanoma (dots, streaks etc.), thus a feature extraction method, and - as a global image classifier based on various melanoma features (geometry, colors, wavelet filters etc.). The first function of SVM is represented in the five main articles. Celebi et al. \[[@pone.0211318.ref021]\] took a methodological approach and first determined the lesion border, then they extracted lesion shape and color/texture features. The feature set was optimized by various feature selection algorithms and the optimal feature subset size was ranked by an SVM classifier with the RBF kernel. Experiments on 564 images (15.6% melanoma, 84.4% benign) yielded a specificity of 92.34% and a sensitivity of 93.33% with a feature set of 18 out of all 437 features. The appropriate values of the kernel parameters, C (cost/penalty) and gamma (kernel width), were determined on a grid (C ∈ \[2^−5^, 2^15^\], and γ ∈ \[2^−15^, 2^3^\]). After the grid-search, the SVM classifier was trained with the optimal parameters (C, γ) = (2.0, 0.125). In \[[@pone.0211318.ref035]\] texture, border-based, and geometrical features were extracted from 289 dermoscopy images (114 malignant, 175 benign). The texture features were derived from a wavelet tranform, the border features were derived from spatial and frequency factors of the lesion border model, and the geometry features are derived from shape indexes. The most optimized features were selected by the gain-ratio method and the SVM classifier with the RGB kernel and optimized by SMO. Apart from SVM also random forest, logistic regression, and hidden naive Bayes models were tested. For the 23 features accuracy was 91.26% and ROC was 0.937. The authors concluded that the texture with (fewer) border and geometry features outperform pure texture information only. Based on color symmetry and texture analysis Abbas et al. \[[@pone.0211318.ref036]\] developed a system classifying the melanoma tumor patterns (reticular, globular, cobblestone, homogeneous, starburst, parallel ridge, multicomponent pattern). Such multicomponent patterns are well analyzed by learning algorithms that can assign each input pattern to multiple class labels simultaneously. In this work ML-SVM (RBF kernel), ML-kNN and multi-label ranking (AdaBoost.MC) were used, the latter being the best classifier for the problem. For the ML-SVM they reported classification sensitivity of 89.28%, specificity of 93.75% and AUC of 98.6% (as quoted by \[[@pone.0211318.ref004]\]) and concluded that the developed pattern classifier based on color--texture features agrees with dermatologists' perception). Mirzaalian \[[@pone.0211318.ref037]\] investigated visual streaks as one of the most important dermoscopic criteria for the diagnosis of malignant melanoma. The streaks in 99 dermoscopic images were identified by quaternion tubularness, which is sensitive to the radial components of streaks. Presence or absence of the regular and irregular streaks (measured by some flux-based descriptors for different number of bands, K ∈ \[5, 13\], and thicknesses Δ ∈ \[2, 8\]) was validated by the SVM classifier (RBF kernel), yielding AUC of 93% in the best model. Maglogiannis \[[@pone.0211318.ref038]\] was segmenting and counting dark dots and globules from dermoscopy images (108 benign + 100 melanoma lesions, 632 × 387 pixels) based on inverse non-linear diffusion. The optimal set of the dot features plus some region-based descriptors were classified (among others) by SVM with a polynomial kernel (degree = 5) in the malignant-non malignant lesion setup. For the best feature subset they achieved sensitivity of 88.46%, specificity of 92.31% and accuracy of 90.38%. The SVM classifier was the main supervised discriminator in the works mentioned below. They are widely commented in the reviews \[[@pone.0211318.ref003], [@pone.0211318.ref032], [@pone.0211318.ref039]\]. In \[[@pone.0211318.ref039]\] the data set contained 1619 lesion images: 600 common nevi, 144 dysplastic nevi, and 65 melanoma for the training set, and 690 common nevi, 80 dysplastic nevi, and 40 melanoma for the test set, all in resolution of 752x582 pixels. The authors analyzed basic, shape and color features (alltogether 107 features) with different normalization conditions and concluded that on both dichotomous (common nevi vs. dysplastic nevi + melanoma; melnoma vs. common + displastic nevi) and trichotomous tasks (correctly distinguishing all three classes) the ANNs performed similarly as logistic regression and SVMs, and better than the k-nearest neighbors and decision trees. For the dichotomous classification the optimal polynomial kernel was linear (C = 100) and yielded AUC of 0.92, and the optimal Gaussian RB kernel had inverse variance γ = 10^−4^ and C = 100 and yielded AUC of 0.97. \[[@pone.0211318.ref032]\] reviewed the state of the art of the visual features used for skin lesion classification in the stages: segmentation, border detection and color/texture processing. They also compared the performance of several classifiers and concluded that the SVM classifier (RBF kernel) seems to achieve higher performance in terms of sensitivity and specificity, followed by ADWAT and CART algorithms. Work \[[@pone.0211318.ref003]\] is a comparative assessment of skin cancer diagnostic models. They critically examined practices, problems, deficiencies and prospects from the image acquisition to classification of dermoscopic images. Some techniques are commented on the conditions that affect their feasibility. Single works that are referred to in the above mentioned reviews are briefly commented on below. Amico et al. \[[@pone.0211318.ref040]\] constructed various asymmetry measures (the so called Size Functions) to discriminate melanocytic lesions by SVM (kernel function was a third degree polynomial), which was implemented at a clinical level. They gained in the best tests sensitivity of 96.8% and specificity of 87.2%. \[[@pone.0211318.ref041]\] examined the melanoma discrimination capacity of some skin specialists (31 melanomas + 103 nevi) and the automatic data analysis for the melanoma early detection system (ADAM, 42 melanomas + 435 nevi). The ADAM system showed a slightly higher sensitivity (84%) and a lower specificity (72%), compared with the physicians. In \[[@pone.0211318.ref042]\] a local thresholding algorithm was proposed to extract separation, border, texture and color based features. Those features were used to construct a classifier based on SVM (malignant melanoma versus dysplastic nevus). The best accuracy of the RBF kernel was 91.84% (sensitivity = 91.30%, specificity = 91.87%). \[[@pone.0211318.ref043]\] developed a decision support system for the dermoscopic images by combining (by the Bayes theory) outputs from different classifiers (SVM, GML, kNN). On a collection of 358 dermoscopic images they used local color and texture-related features to achieve accuracy of 76%. In \[[@pone.0211318.ref044]\] a diagnostic system for dermatologists based on SVM models of melanoma was built. They investigated 14 geometry and color features. They tested four distinct kernels: polynomial, sigmoid, RBF and the k-MOD decreasing. The best SVM model with the k-MOD decreasing kernel function got 89% sensitivity and AUC = 76% using a set of 199 digital dermoscopic images (101 melanomas, 98 dysplastic). Work \[[@pone.0211318.ref045]\] aimed at determining the best system for the skin lesion classification out of one global (one of them was SVM, RBF kernel) and one bag-of-features classifier based on local features. The other objective was to compare the role of color and texture features in lesion classification and determine which set of features is more discriminative. They concluded that color features outperform texture features when used alone and that both methods achieve very good results on 176 dermoscopy images, i.e. sensitivity of 92% and specificity of 72% for the SVM global method against sensitivity = 100% and specificity = 75% for local methods. A classification system for four types of skin cancers: melanoma, basal cell carcinoma (BCC), actinic keratosis (AK), squamous cell carcinoma (SCC) was reported in \[[@pone.0211318.ref046]\]. The GLCM based texture features ware extracted from each of the four classes and given as input to a multi-class SVM. The accuracy of classification to the one of the four skin cancer classes was 81.43%. Masood et al. \[[@pone.0211318.ref047]\] analyzed 168 images (112 melanoma and 56 benign) rescaled to a resolution of 720x472. The lesion area was segmented using the Histogram based Fuzzy C Mean algorithm for Level Set initialization (H-FCM-LS). 45 features (15 GLCM, 5 GTDM-Gray Tone Difference Matrix, 15 FMI-WPT, and 10 Autoregressive) were extracted for each image The FMI-WPT features (Fuzzy Mutual Information based Wavelet Packet Transform (FMI-WPT) were extracted to level 3 and then some fuzzy sets were constructed. The number of features that maximized the classification performance was evaluated using fuzzy-set entropy based criterions. The constructed feature sets are used separately as well as in different combinations for feeding the classifier. Finally, Self-advisable Support Vector Machine (SA-SVM) was used for classification. SA-VM (both linear and kernel based) uses information generated from misclassified data in the training phase and thus, improves performance by transferring more information from training phase to the test phase. The results obtained by SA-SVM were significantly better than the results of traditional SVM. The SA-SVM diagnostic system achieved an overall accuracy of 90%, with sensitivity 91% and specificity 89%. In \[[@pone.0211318.ref049]\] features based on asymmetry, border irregularity, color variations and diameter were calculated from an illumination compensated segmented image after noise removal by iterative dilation. The SVM classifier was optimized by Sequential Minimal Optimization (SMO) for the parameters that varied in the ranges: C ∈ 0.1, 0.2, ..., 5 for soft margin and σ ∈ \[1, 11\]. The achieved sensitivity and specificity for the different sets of training and test data elements were respectively 87% and 94% (1-fold), and 90% and 75% (10-fold CV). Amerald et al. \[[@pone.0211318.ref050]\] proposed the high-level intuitive features (HLIFs) to model the ABCD criteria commonly used by dermatologists. They experimented with various data sets concluding that concatenating the proposed HLIFs with some low-level features increased classification accuracy. The SVM classifier with the linear kernel on the best data set achieved sensitivity 92.52%, specificity 96.22%, and accuracy 96.64%. Choudhury et al. \[[@pone.0211318.ref051]\] proposed a multilayer decomposition based textural and color features for SCC, BCC, melanoma and actinic keratosis. The normalized GLCM and histogram of oriented gradients (HOG) were used as textural feature descriptors, and were extracted from different layers of the image than color histograms. The base and detailed layers were decomposed by the weighted least squares (WLS) edge-preserving decomposition. These features were fed to multi-class SVM (MSVM) and extreme learning machine (ELM) for classification. An average accuracy of 94.18% for MSVM was better than 90.5% with ELM. Alquran et al. \[[@pone.0211318.ref052]\] collected asymmetry, border, color and diameter (ABCD) features extracted using the GLCM method after segmentation using thresholding. Those features were selected by principal component analysis (PCA), than the Total Dermoscopy Score (TDS)then the SVM classification was done. The achieved classification accuracy was 92.1%. The most recent work on SVM classified malicious and benign skin lesions by Ashtami \[[@pone.0211318.ref053]\] proposed new features characterizing border irregularities on both complete and incomplete lesions. Those features plus color and texture features (GLCM) were classified by a Support Vector Machine (SVM) model on two dermoscopy databases with images of two human races: caucasian an xanthous race. Their results, sensitivity 97.82%, specificity 75%, and accuracy 88.46% are slightly better than for their ANN classifier. There are few direct SVM systems based on wavelet features (and they are not fully comparable with our experiments): Surowka et al. \[[@pone.0211318.ref014]\] tested performance of several machine learning paradigms and methods: neural networks, support vector machines, and Attributional Calculus, applied to dermatoscopic images of potentially malignant pigmented lesions. The features were obtained using the multiresolution wavelet-based decomposition of the image. The SVM AUC was 0.937. In \[[@pone.0211318.ref048]\] 255 features were extracted from Wavelet Packet Transform (WPT) and reduced by Particle Swarm Optimization (PSO) to optimize the SVM classifier. In the pre-processing step the Wiener2, Gabor, median filtering and histogram equalization was introduced to the images. Each image was further segmented by edge detection and thresholding. All the tasks yielded classification sensitivity of 94.1%, specificity of 80.22%, and accuracy of 87.13% Takruri et al. \[[@pone.0211318.ref016]\] analyzed 448 digital and dermoscopic images (benign and malignant) from two sources and segmented them by k-means clustering. They derived wavelet (db4), curvelet and color based features both from grayscale and color original images which resulted in a sensitivity of 86.4% and 76.9% and specificity of 88.1% and 85.4% for the wavelets and curvelets respectively. The obtained results were discussed to be comparable to those obtained by dermatologists. In [Table 2](#pone.0211318.t002){ref-type="table"} we show our SVM numerical results (best wavelet numbers in each data set plus the parameters) against all the quoted literature results (Tables [3](#pone.0211318.t003){ref-type="table"} and [4](#pone.0211318.t004){ref-type="table"}) for the SVM learning paradigm of non-wavelet and wavelet features. We conclude that our results are reliable and compatible to the rest of the works. Unfortunately, due to too few melanoma classification systems utilizing both the wavelet features and the SVM classifiers plus unique conditions of the existing experiments, and last but not least, lack of details from the groups, makes a full quantitative comparison still unfeasible. 10.1371/journal.pone.0211318.t002 ###### Results of the melanoma SVM learning (this work). Legend: dn = dysplastic nevus, cn = common nevus. RBF kernel parameter: γ, polynomial kernel parameter: degree; \#features = 252, partition: 10-fold CV. Resolutions: A = A + B = 2272 x 1704, B = 760 x 552, A2 = A2 + B2 = 1136 x 852, A4 = A4 + B4 = 568 x 426. ![](pone.0211318.t002){#pone.0211318.t002g} Data set \#images \#class A (melanoma) \#class B wavelet number wavelet base kernel param. C AUC \#supp.vect. ------------ ---------- ---------------------- --------------- ---------------- -------------- --------------- ---------- ------- -------------- gaussian A 185 102 83 dn 46 Rbio3.1 4.68e-09 9.81e+06 0.870 72 A2 185 102 83 dn 26 Bior1.5 2.41e-04 9.69e+06 0.859 81 A4 185 102 83 dn 46 Rbio3.1 3.72e-06 4.76e+04 0.859 60 B 113 40 73 (cn + dn) 3 Db2 1.61e-03 1.54e+06 0.932 68 A + B 298 142 156 (cn + dn) 41 Rbio1.5 6.98e-04 6.96e+02 0.881 120 A2 + B2 298 142 156 (cn + dn) 24 Bior1.1 4.25e-03 2.36e+04 0.877 177 A4 + B4 298 142 156 (cn + dn) 26 Bior1.5 5.03e-03 1.64e+04 0.854 218 polynomial A 185 102 83 dn 48 Rbio3.5 3 9.53e+06 0.870 73 A2 185 102 83 dn 26 Bior1.5 3 5.78e+05 0.854 83 A4 185 102 83 dn 26 Bior1.5 4 6.50e+05 0.859 102 B 113 40 73 (cn + dn) 26 Bior1.5 3 1.53e+00 0.940 50 A + B 298 142 156 (cn + dn) 46 Rbio3.1 6 7.21e+03 0.868 92 A2 + B2 298 142 156 (cn + dn) 2 Db1 4 2.56e+05 0.871 125 A4 + B4 298 142 156 (cn + dn) 46 Rbio3.1 2 2.33e-01 0.868 138 linear A 185 102 83 dn 46 Rbio3.1 \- 2.63e+04 0.870 73 A2 185 102 83 dn 46 Rbio3.1 \- 5.48e+01 0.843 76 A4 185 102 83 dn 46 Rbio3.1 \- 8.28e+03 0.854 62 B 113 40 73 (cn + dn) 40 Rbio1.3 \- 5.58e+00 0.923 41 A + B 298 142 156 (cn + dn) 41 Rbio1.5 \- 9.39e+06 0.871 106 A2 + B2 298 142 156 (cn + dn) 47 Rbio3.3 \- 1.09e-05 0.858 118 A4 + B4 298 142 156 (cn + dn) 46 Rbio3.1 \- 1.16e-05 0.854 123 10.1371/journal.pone.0211318.t003 ###### The SVM-based melanoma machine learning experiments. Legend: n = nevus, cn = common nevus, dn = dysplastic nevus, otherwise = benign; parameters for the RBF kernel: (C, γ); parameter for the polynomial kernel: (degree).?/blank space = data unavailable. ![](pone.0211318.t003){#pone.0211318.t003g} ----------------------------------------------------------------------------------------------------------------------------------------------------------------- Ref\ \#images \#class A\ \#class B \#class C partition \#features\ kernel\ sens spec acc AUC (resolution) melanoma (best) (params) ----------------------------- ---------- ------------ ------------ ----------- -------------- ------------- ------------------- -------- -------- ------- ------- \[[@pone.0211318.ref039]\]\ 1619 65/40 600/690 cn 144/80 dn train/test 107 poly (1)\ 84,5%\ 88,5%\ 92%\ (752x582) RBF (100, 10^−4^) 92,1% 95% 97% \[[@pone.0211318.ref040]\]\ 977 25/25 500/427 n train/test poly (3) 96,4% 87,2% (magnif 16x) \[[@pone.0211318.ref041]\]\ 477 22/20/\ 218/217/\ train/test/\ 84% 72% (magnif 16x) 31 of 42 103 of 435 test doctors \[[@pone.0211318.ref042]\] 1041 69 972 n RBF (?, 7) 91.3% 91.9% 91.9% \[[@pone.0211318.ref014]\]\ 39 19 20 dn 231(10) RBF\ 97.4% (2272x1704) (512, 0.000244) \[[@pone.0211318.ref021]\] 564 88 476 10-fold CV 437(18) RBF (2, 0.125) 93.3% 92.3% \[[@pone.0211318.ref043]\] 358 134 106 118 dn RBF 75.7% \[[@pone.0211318.ref044]\] 199 101 98 dn 130train/\ 14 89% 64% 76% 69test \[[@pone.0211318.ref035]\] 289 114 175 \(23\) RBF 91.3% 93.7% \[[@pone.0211318.ref037]\]\ 99 RBF 93% (768x512) \[[@pone.0211318.ref045]\] 176 RBF 92% 72% \[[@pone.0211318.ref036]\]\ 350 80/20%\ RBF (1,?) 89.3% 93.8% 98.6% (768x512) train/test ----------------------------------------------------------------------------------------------------------------------------------------------------------------- 10.1371/journal.pone.0211318.t004 ###### The SVM-based melanoma machine learning experiments (contd). Legend: dn = dysplastic nevus, otherwise = benign; parameters for the RBF kernel: (C, γ); BCC = Basal Cell Carcinoma, SCC = Squamos Cell Carcinoma, a.ker. = actinis keratosis. blank space = data unavailable. ![](pone.0211318.t004){#pone.0211318.t004g} -------------------------------------------------------------------------------------------------------------------------------------------------------------------- Ref\ \#images \#class A\ \#class B \#class C partition \#features\ kernel\ sens spec acc AUC (resolution) melanoma (best) (params) ----------------------------- ---------- ------------ -------------- ----------- ------------ ------------- ------------------------ -------- -------- ------- ----- \[[@pone.0211318.ref038]\]\ 208 100 108 poly (5) 88.5% 92.3% 90.4% (632×387) \[[@pone.0211318.ref046]\]\ 359 77 BCC(84)\ 81.4% (150x112) a.ker.(101)\ SCC(101) \[[@pone.0211318.ref047]\]\ 168 112 56 84 train\ 45 91% 89% 90% (720x472) 84 test \[[@pone.0211318.ref048]\] 79 50 29 64 train/\ 255 RBF\ 94.1% 80.2% 87.1% 15 test\ (3525.0051, 0.0084732) 15 test \[[@pone.0211318.ref016]\] 448 93 + 142 121 + 92 0.8train/\ 52\ RBF 86.4%\ 88.1%\ 0.2test 55(25) 76.9% 85.4% \[[@pone.0211318.ref049]\] 146 108 38 1-fold CV\ RBF 87%\ 94%\ 10-fold CV 90% 75% \[[@pone.0211318.ref050]\] LOO linear 92.5% 96.2% 96.6% \[[@pone.0211318.ref051]\] 94.2% \[[@pone.0211318.ref052]\] 92.1% \[[@pone.0211318.ref053]\] 151 97.8% 75% 88.5% -------------------------------------------------------------------------------------------------------------------------------------------------------------------- Conclusion {#sec013} ========== Computer aided diagnostic systems are common. For the early detection of cutaneous melanoma they play a crucial role to support clinics and general practitioners. Our work contributes to this effort in the field of feature extraction. Appropriate features project the information from a (dermoscopy) image into a space where classes (malignant and benign lesions) are well separated. This step is critical for the performance of the classifier and the learning procedure. In this work we extracted features from the WPT wavelet transform for the sake of efficient SVM classification. We analyzed 53 wavelet bases to select those that are efficient for the SVM learning, and to probe the SVM parameter space. Our experiments were performed on two different data sets of unequal resolution, and biased with different illumination, representation, and compression. The SVM paradigm involves optimization of a convex cost function, therefore it is a 'global' classifier, less prone to overfitting. It is flexible in a high dimension feature space through an appropriate choice of kernel, is robust and simple. We tested the linear, polynomial and gaussian kernel and searched (Bayesian optimization) the kernel parameters (C, γ, degree of polynomial). From this analysis we concluded that: 1. Kernel functions play a key role in the performance of the SVM classification, the best kernel was the gaussian function. 2. Wavelet features classified by SVM with different kernel functions have 'preferential' resolutions where they reach the most optimal classification efficiency. Analysis on two data sets (A, B) shows that for the gaussian kernel such most efficient resolution is 568x426. For the melanoma classification lower image resolution would be potentially resource-saving without (noticeable) degradation of the classification performance. 3. We selected wavelet bases that performed the best for a certain resolution and SVM kernel function, and also such that keep high classification efficiency towards downgraded image resolutions (Rbio and Bior wavelet families, particularly Rbio3.1 and Bior1.5). Our findings are compatible with our previous machine learning experiments on the melanoma discrimination (ensembling and artificial neural networks) where high classification accuracy and stable behavior over a range of resolutions was observed as well. Regarding our SVM classification of melanoma against dysplastic nevus, our classification performance is in line with other research groups, however a detailed comparison with other experiments with different data sets, selection criteria, different pre-processing steps and finally different features is problematic. General outlook is that further analyses on various (large) public data sets of dermoscopic skin lesion images should be done to probe new learning paradigms and yield quantitative results contributing to the melanoma feature extraction. [^1]: Competing Interests:The authors have declared that no competing interests exist.	{ "pile_set_name": "PubMed Central" }
Introduction ============ Heat shock protein 90 (HSP90), a 90 kDa ATP-dependent molecular chaperone, is the most abundant intracellular protein in mammalian cells and is essential for protein folding, assembly, and degradation processes.[@b1-ott-8-1849],[@b2-ott-8-1849] It is overexpressed in response to a variety of physiological and environmental insults, allowing cells to survive potentially lethal conditions due to its cytoprotective functions.[@b3-ott-8-1849]--[@b5-ott-8-1849] In this regard, HSP90 has recently received much attention due to its overexpression in certain types of cancer,[@b6-ott-8-1849]--[@b9-ott-8-1849] where it is associated with a poor prognosis and contributes to resistance to chemotherapy and radiation.[@b10-ott-8-1849] Unlike other molecular chaperones that are involved in the primary folding of nascent polypeptides, HSP90 uses repeated cycles of client protein binding, hydrolysis of ATP, and interaction with its co-chaperones such as HSP70, Cdc37, HOP, p23, and Aha1 to control the stability and activity of hundreds of client proteins.[@b11-ott-8-1849],[@b12-ott-8-1849] Importantly, the majority of these client proteins are involved in critical signaling pathways necessary for cellular proliferation, cell cycle progression, and apoptosis, including steroid hormone receptors, kinases, transcriptions factors, human epidermal growth factor receptor 2 (HER2), Akt, mutant BRAF, and mutant p53 among others.[@b12-ott-8-1849]--[@b16-ott-8-1849] Consistent with this, HSP90 inhibition results in the simultaneous degradation of many of these clients (via the ubiquitin-proteasome pathway), leading to cell-specific growth arrest, apoptosis of cancer cells, and antitumor activity in preclinical models.[@b3-ott-8-1849],[@b17-ott-8-1849] Thus, HSP90 has evolved as an important molecular target in cancer therapy, and 16 different first- and second-generation HSP90 inhibitors have entered clinical testing ([Table 1](#t1-ott-8-1849){ref-type="table"}). The prototype HSP90 inhibitor geldanamycin provided proof-of-concept for HSP90 inhibition; however, geldanamycin and its derivatives (17-allylamino-17-demethoxygeldanamycin \[17-AAG\] and 17-dimethylaminoethylamino-17-demethoxygeldanamycin \[17-DMAG\]) could not be fully developed due to a number of safety and pharmacological limitations. Consequent efforts using a variety of different chemical scaffolds have led to the development of highly potent, second-generation, small molecule HSP90 inhibitors with improved pharmacological properties and safety profiles.[@b18-ott-8-1849] This review will summarize information on the pharmacology, preclinical activity, and current clinical development of ganetespib, a novel small molecule, second-generation HSP90 inhibitor developed by Synta Pharmaceuticals. Pharmacology and preclinical data ================================= The HSP90 chaperone consists of an amino (N) terminal, a carboxy (C) terminal, and a middle (M) domain. The N-terminal contains the ATP-binding pocket, and both the N and C termini have a drug-binding site. The M domain mostly participates in forming active ATPase and acts as a docking site for clients and co-chaperones.[@b19-ott-8-1849],[@b20-ott-8-1849] Similar to many other HSP90 inhibitors, ganetespib also acts by binding to the N-domain ATP-binding pocket of HSP90.[@b21-ott-8-1849] The first-generation inhibitors were a semi-synthetic derivative of the natural product, geldanamycin, and included 17-AAG (17-Allyl-17-Demethoxygeldanamycin \[Tanespimycin\]), 17-DMAG (17-desmethoxy-17-N,N-dimethylaminoethylaminogeldanamycin \[alvespimycin\]), IPI-504/17-allylamino-17-demethoxygeldanamycin hydroquinone hydrochloride (Retaspimycin), and IPI-493/17-desmethoxy-17-amino geldanamycin ([Table 1](#t1-ott-8-1849){ref-type="table"}).[@b22-ott-8-1849] While the first-generation inhibitors provided proof-of-concept, they failed to advance further clinically due to various issues including poor pharmacological properties, adverse toxicity profiles, or suboptimal efficacy. The most clinically significant off-target toxicity with the geldanamycin derivatives was hepatotoxicity, which has largely been attributed to the presence of a benzoquinone moiety.[@b22-ott-8-1849] Although 17-DMAG lacked the solubility issues observed with 17-AAG, the toxicities reported with 17-DMAG, including liver, ocular, and cardiac toxicities, ultimately led to the cessation of its further clinical development.[@b23-ott-8-1849],[@b24-ott-8-1849] IPI-504 was designed to be more water soluble than both 17-AAG and 17-DMAG. Although the clinical activity of IPI-504 was promising in both non-small cell lung cancer (NSCLC)[@b25-ott-8-1849] and gastrointestinal stromal tumors (GIST),[@b26-ott-8-1849] there were treatment-related deaths due to hepatotoxicity in the Phase III RING (Retaspimycin in GIST) trial of this HSP90 inhibitor.[@b27-ott-8-1849] This prompted mandatory dose reductions of IPI-504 in the ongoing HER2-positive metastatic breast cancer trial, which ultimately may have led to the lack of efficacy seen with this inhibitor in that trial.[@b27-ott-8-1849] In contrast, the second-generation HSP90 inhibitors generally encompass the resorcinol moiety of radicicol or are purine derivatives; the exception to this classification is the SNX-522 agent, which does not fit among these designations ([Table 1](#t1-ott-8-1849){ref-type="table"}).[@b12-ott-8-1849],[@b23-ott-8-1849] Notably, this group of inhibitors lack the hepatoxicity noted with the first-generation inhibitors due to the absence of the benzoquinone moiety. Ganetespib (\[5-\[2,4-Dihydroxy-5-(1-methylethyl)phenyl\]-4-(1-methyl-1H-indol-5-yl)-2,4-dihydro-\[1,2,4\]triazol-3-one\]) is a resorcinol-based, non-geldanamycin, synthetic small molecule second-generation HSP90 inhibitor (molecular weight =364.4), similar to NYP-AUY922, AT-13387, and KW-2478. In differentiation to the other HSP90 inhibitors within this subgroup, ganetespib contains a triazolone moiety ([Figure 1](#f1-ott-8-1849){ref-type="fig"}, [Table 1](#t1-ott-8-1849){ref-type="table"}).[@b28-ott-8-1849] In vitro, ganetespib exhibits potent cytotoxicity in a wide variety of hematological and solid tumor cell lines, including those that express mutant kinases (including BCR-ABL, FLT3, c-KIT, EGFR, and B-RAF) that confer resistance to small molecule tyrosine kinase inhibitors. The half maximum inhibitory concentration values calculated for ganetespib are in the low nanomolar range, and ganetespib is at least 20-fold more potent than 17-AAG.[@b28-ott-8-1849] Interestingly, ganetespib not only induces rapid degradation of known HSP90 client proteins but also exhibits sustained activity even with short exposure times.[@b29-ott-8-1849] The cytotoxicity of ganetespib in these cell lines is predominantly mediated via an irreversible commitment to apoptosis, likely following growth arrest and effects on the cell cycle.[@b29-ott-8-1849] Similarly, ganetespib has potent antitumor activity in vivo as demonstrated by significant growth inhibition and/or regression in solid tumor and hematological xenograft models.[@b29-ott-8-1849] Evaluation of the microregional activity of ganetespib demonstrates efficient distribution throughout the tumor tissue, including the hypoxic regions \>150 µm from the microvasculature, resulting in sustained inhibition of proliferation and induction of apoptosis throughout the tumors.[@b28-ott-8-1849] Ganetespib also exhibits preferential tumor retention compared with normal tissues with a half-life of over 58 hours in NSCLC xenograft models vs 3 hours in the plasma and 5--6 hours in normal liver and lung tissues.[@b30-ott-8-1849] With regards to its pharmacological profile, as discussed earlier, ganetespib lacks the dose-limiting hepatotoxicity reported with the geldanamycin analogs. Additionally, ganetespib lacks the ocular toxicity that has been reported with NVP-AUY922 and SNX-522. The mechanism of this ocular toxicity is hypothesized to be related to photoreceptor degeneration due to retinal drug distribution and retention, which has been seen with NVP-AUY922 exposure in rodent models. In contrast, ganetespib does not accumulate in the rat eye and is rapidly eliminated from the retinal tissues.[@b31-ott-8-1849],[@b32-ott-8-1849] Taken together, these promising results suggest that ganetespib may be more potent in its antitumor activity compared with first-generation inhibitors and has an optimal safety profile that predicts for a superior therapeutic index. This has provided a compelling rationale to further develop this agent clinically. Single-agent Phase I/II trials ============================== An open-label dose-escalation study of single-agent ganetespib given intravenously in advanced solid tumors established the maximum tolerated dose at 216 mg/m^2^ given weekly for 3 out of 4 weeks. The recommended Phase 2 dose was hence established as 200 mg/m^2^. Dose-limiting toxicities (DLTs) included one grade 3 amylase elevation at 150 mg/m^2^, one grade 3 diarrhea, and two cases of asthenia (grade 3 and grade 4) at 259 mg/m^2^. Overall, the most common adverse events (AEs) were gastrointestinal in nature and fatigue, predominantly of grade 1 and 2 severity, and easily manageable.[@b33-ott-8-1849] Partial response (PR) was noted in a patient with metastatic colorectal cancer, and disease stabilization (SD) was noted in 23 of the 44 evaluable patients, including 1 patient each with NSCLC and GIST, whose tumors harbored the BRAF G469A and PDGFRA^D842V^ exon 18 mutations, respectively. The most common AEs in another Phase I trial of ganetespib in advanced solid tumors, administered twice weekly for 3 out of 4 weeks, were also diarrhea and fatigue. Two DLTs of elevated transaminases were noted at 10 mg/m^2^ and 144 mg/m^2^, respectively.[@b34-ott-8-1849] In this trial, a durable PR was noted in a patient with metastatic melanoma, and durable SD was reported in two patients with NSCLC. Two Phase I trials have also been conducted with single-agent ganetespib in patients with hematologic malignancies. These studies determined the recommended Phase 2 dose to be 200 mg/m^2^ for the once weekly and 90 mg/m^2^ for the twice weekly regimens, respectively. Again, the most common AEs in these studies were also diarrhea and fatigue. With regards to DLTs, there were elevated liver enzymes in one patient in the once weekly trial, and hyperbilirubinemia, hyponatremia, QTc prolongation, and transaminitis for the twice weekly trial.[@b23-ott-8-1849],[@b35-ott-8-1849],[@b36-ott-8-1849] Notably, there were no reports of severe hepatotoxicity, ocular toxicity, or cardiac toxicity in these patients. Since many HSP90 client proteins are mutated, overexpressed and/or chimeric kinases important for tumor growth and survival, including HER2 in breast cancer, mutant EGFR or anaplastic lymphoma kinase (ALK) in lung cancer, mutant KIT in GIST, and mutated BRAF in melanoma, a number of Phase II trials of single-agent ganetespib in selected tumor types that express such oncoproteins have been undertaken to exploit the concept of "oncogene addiction". Supported by preclinical evidence that ganetespib has potent antitumor activity in different breast cancer subtypes, a Phase II trial of single-agent ganetespib at 200 mg/m^2^ intravenously (IV) weekly for 3 out of 4 weeks was conducted in patients with unselected metastatic breast cancer.[@b32-ott-8-1849] The most common toxicities in this trial were of grade 1 and 2 severity and included diarrhea, fatigue, nausea, and hypersensitivity reactions. While the study did not meet its prespecified criteria for overall response rate in this heavily pretreated group of patients, clinical activity was notable in patients with trastuzumab-refractory HER2+ and triple negative breast cancer (TNBC). Specifically, two patients achieved a PR, and four of the seven patients who achieved SD had estrogen receptor positive (ER+)/HER2+ disease.[@b32-ott-8-1849] Unlike HER2+ breast cancer or hormone receptor-positive breast cancer, TNBC lacks a unique molecular alteration that can be therapeutically targeted. Yet, clinical activity has been noted with ganetespib in this subset of patients with evidence of tumor shrinkage, specifically in patients with lung metastases.[@b32-ott-8-1849],[@b37-ott-8-1849] This latter observation is in line with findings from syngeneic mouse models of spontaneous and experimental metastases, where ganetespib suppressed lung colonization and tumor growth.[@b37-ott-8-1849] There are many molecular alterations for TNBC that have been identified but are not yet validated as therapeutic targets. Several of these abnormal proteins are known HSP90 client proteins, such as EGFR, hypoxia-inducible factor 1 (HIF-1α), and KIT among others, thus underscoring the therapeutic potential for ganetespib even in this aggressive subtype of breast cancer.[@b38-ott-8-1849] Another Phase II trial evaluated the activity and tolerability of ganetespib administered at 200 mg/m^2^ IV for 3 out of 4 weeks in previously treated patients with NSCLC enrolled in three cohorts: cohort A (mutant EGFR), cohort B (mutant KRAS), and cohort C (no EGFR or KRAS mutations). Of the 66 patients enrolled to cohort C, four had PR with all four cases being crizotinib-naive and harboring the ALK gene rearrangement. As expected, the most common AEs were diarrhea, fatigue, nausea, and anorexia.[@b39-ott-8-1849] A Phase II trial in patients with metastatic ocular melanoma is also evaluating progression-free survival with single-agent ganetespib administered at the same dosing schedule ([ClinicalTrials.gov](http://ClinicalTrials.gov) Identifier: NCT01200238). Another pilot window-of-opportunity study is evaluating changes in the expression of biomarkers due to ganetespib, which is administered twice weekly for 2 weeks prior to surgery in patients with head and neck cancers ([ClinicalTrials.gov](http://ClinicalTrials.gov) Identifier: NCT02334319). The biology behind combination trials ===================================== Contrary to expectations based on the presence of mutant KIT, an HSP90 client, and therefore amenable to HSP90 inhibitor therapy, a Phase II trial of single-agent weekly ganetespib reported limited efficacy in patients with GIST (12/23 evaluable patients had SD: 4 SD ≥16 weeks, 8 SD ≥8 weeks).[@b40-ott-8-1849] This negative result was considered to be related to a possible lack of a sufficient duration of inhibition of this client or its downstream pathways, which was observed both in preclinical models and in patient biopsies done as a part of this Phase II trial.[@b40-ott-8-1849] The limited success of HSP90 inhibitors in GIST and other client-protein-driven patient populations remains poorly understood but has certainly led to the emergence of the concept of rationally designed combination studies to enhance efficacy. These strategies include combinations with cytotoxics, other targeted therapies, and radiation and are briefly discussed in the following. HSP90 protects cells under conditions of stress, and, therefore, HSP90 inhibitors have the ability to sensitize cells to the toxic effects of chemotherapy and radiation therapy.[@b22-ott-8-1849] In support of this, preclinical data from various cell lines and xenograft models suggest additive or synergistic antitumor activity when HSP90 inhibitors are combined with various systemic cytotoxics including anthracyclines and taxanes.[@b22-ott-8-1849],[@b41-ott-8-1849],[@b42-ott-8-1849] The synergistic benefit observed by the addition of HSP90 inhibitors to taxanes is likely multifactorial, resulting from increased cytotoxicity and apoptosis, Akt inactivation and sensitization of the tumor cells to induction of apoptosis by a taxane, loss of pro-survival signaling, and exacerbation of mitotic catastrophe.[@b12-ott-8-1849],[@b41-ott-8-1849],[@b43-ott-8-1849],[@b44-ott-8-1849] Indeed, the combination of ganetespib with paclitaxel or docetaxel enhances antitumor growth and is synergistic in TNBC, NSCLC, and ovarian cancer models.[@b37-ott-8-1849],[@b45-ott-8-1849],[@b46-ott-8-1849] Most importantly, this combination has a nonoverlapping toxicity profile. Ganetespib can also potentiate the cytotoxic activity of doxorubicin via enhanced DNA damage and mitotic arrest and thus can confer superior efficacy to doxorubicin-containing regimens.[@b37-ott-8-1849] Approximately 15%--25% of lung adenocarcinomas have tumor-associated KRAS mutations.[@b47-ott-8-1849] These mutations are negative predictors of response to currently available EGFR tyrosine kinase inhibitors, and these patients have unfavorable clinical outcomes.[@b48-ott-8-1849] Currently, there are no approved anti-KRAS-directed therapies. While KRAS by itself is not a known client protein, its downstream effector pathways, namely, the PI3K/AKT/mTOR and the RAF/MEK/ERK pathways, are shown to be sensitive to HSP90 inhibition. In fact, the combination of ganetespib with a dual PI3K/mTOR inhibitor demonstrated superior antitumor efficacy in a xenograft model, supporting further investigation of this dual-targeted approach.[@b49-ott-8-1849] On the other hand, ALK is a sensitive client of HSP90 inhibition, and ganetespib can overcome acquired resistance to the ALK inhibitor, crizotinib in ALK + NSCLC, both in xenograft models and in patients.[@b50-ott-8-1849] Furthermore, the L1196M mutant form of EML4-ALK, considered to be an acquired mutation conferring resistance to crizotinib, continues to be sensitive to other structurally different ALK tyrosine kinase inhibitors and HSP90 inhibitors like 17AAG and ganetespib in both cell lines and the clinical setting.[@b50-ott-8-1849],[@b51-ott-8-1849] Similarly, mutant BRAF is a very sensitive HSP90 client, and ganetespib was more potent than single-agent vemurafenib in BRAF-driven melanoma cell lines.[@b52-ott-8-1849] Since BRAF mutations lead to dysregulation of the RAF/MEK/ERK signaling axis, a number of trials are evaluating the combination of BRAF inhibitors and MEK inhibitors not only to improve efficacy but also to overcome acquired resistance. Along the same lines, the combination of ganetespib with vemurafenib or selective MEK inhibitors has shown synergistic activity in xenograft models. In a preclinical study, ganetespib in combination with the MEK inhibitor TAK-733 also caused significant tumor regressions in vemurafenib-resistant xenograft tumors, thus providing a rationale for this combination to treat tumors that have developed acquired resistance to vemurafenib.[@b52-ott-8-1849] HSP90 inhibitors can also enhance tumor cell sensitivity to radiation. HSP90 clients such as Akt and ErbB2 are thought to be associated with radioresponse, which in turn can protect against radiation-induced cell death. Degradation of these proteins by HSP90 inhibition therefore enhances tumor cell death in many cell lines and xenograft models.[@b53-ott-8-1849] This combination can also increase apoptosis and enhance G2 arrest and hence is being evaluated in tumors such as rectal cancer.[@b54-ott-8-1849] HSP90 inhibitors may also have a significant role to play in hormone-receptor resistant breast cancer based on the role of HSP90 in regulating the post-translational folding of the estrogen and progesterone receptors.[@b55-ott-8-1849] In preclinical hormone receptor positive breast cancer models, ganetespib reverses endocrine resistance and reduces heterogeneity in the disease control achieved by hormonal therapies. This work has formed the basis for evaluating the combination of ganetespib with endocrine therapies such as fulvestrant, a complete estrogen receptor antagonist ([Table 2](#t2-ott-8-1849){ref-type="table"}). Bortezemib was the first therapeutic proteasome inhibitor that was approved in the United States by the Food and Drug Administration for the treatment of multiple myeloma and mantle cell lymphoma. Preclinical findings suggest that the combination of HSP90 inhibition with bortezemib can enhance bortezemib-triggered apoptosis and induce a prolonged intracellular accumulation of ubiquinated proteins credited to the synergistic suppression of chymotryptic activity of the 20S proteasome.[@b56-ott-8-1849],[@b57-ott-8-1849] This has been the basis for combination trials of various HSP90 inhibitors, (including the ongoing ganetespib studies) with bortezemib in patients with multiple myeloma.[@b58-ott-8-1849],[@b59-ott-8-1849] Together, these compelling preclinical studies have formed the rationale for many Phase I/II clinical trials in combinations with ganetespib as a way to optimize efficacy and delay or limit acquired resistance. Combination Phase I/II trials ============================= [Table 2](#t2-ott-8-1849){ref-type="table"} lists several ongoing combination trials of ganetespib with cytotoxic agents such as taxanes and doxorubicin, radiation, fulvestrant, and other targeted agents such as sirolimus, crizotinib, Ziv-Aflibercept, and bortezemib for several tumor types. Some of these combination studies have reported results and are discussed in detail here. The combination of docetaxel and ganetespib is in advanced clinical testing in NSCLC based on the results from the Phase IIb GALAXY-I (Study of Ganetespib + Docetaxel in Advanced NSCLC) trial.[@b60-ott-8-1849] This Phase II trial randomized 252 stage III/IV NSCLC patients to docetaxel alone (75 mg/m^2^ on day 1 of a 3 week cycle; n=127) or the same dose of docetaxel and ganetespib (150 mg/m^2^ on days 1 and 15; n=125). There was a nonsignificant improvement in overall survival (OS) in the ganetespib arm vs the control arm (9.8 months vs 7.4 months; hazard ratio \[HR\] =0.82; P=0.082); there was also a trend toward improved PFS (4.5 months for ganetespib vs 3.2 months for the control group; HR =0.84; P=0.038). Importantly, in a prespecified group of patients enrolled in the trial more than 6 months after diagnosis (n=176), there was a significant 4.3 month OS advantage with ganetespib compared with the control of docetaxel alone (10.7 months vs 6.4 months; HR= 0.61; P=0.0093); PFS was also significantly improved in the treatment arm (5.4 months vs 3.4 months; HR= 0.61; P=0.0041). For the combination, the most common AE across all grades was diarrhea (48%), which was well managed with the use of over-the-counter antidiarrheals. The most frequently reported grade 3--4 AEs for ganetespib vs the control group were neutropenia (37% vs 38%), febrile neutropenia (11% vs 2%), and anemia (8% vs 2%).[@b60-ott-8-1849] The Phase III GALAXY-2 trial will compare second-line ganetespib and docetaxel vs docetaxel alone in advanced adenocarcinoma of the lung of those who are more than 6 months past a diagnosis of advanced disease. The target enrollment is 500 patients, and the primary endpoint is OS. Key secondary endpoints include OS in three subpopulations (mKRAS, elevated lactate dehydrogenase, and lactate dehydrogenase-5), PFS, overall response rate, and duration of response. Tumor tissue and blood samples will be collected for planned correlative studies ([ClinicalTrials.gov](http://ClinicalTrials.gov) Identifier: NCT01798485). Preliminary results from a Phase I trial of ganetespib plus paclitaxel and trastuzumab, a humanized monoclonal antibody against HER2, were recently presented at the San Antonio Breast Cancer Symposium.[@b61-ott-8-1849] There were no DLTs or grade 3 toxicities attributable to ganetespib. In this group of patients heavily pretreated with prior pertuzumab and ado-trastuzumab emtansine, one of six evaluable patients achieved a PR and four others had SD for a CBR of 60%. This trial is ongoing.[@b61-ott-8-1849] Given the exciting preclinical data of doxorubicin and ganetespib in xenograft models, this combination is being studied to determine a signal of efficacy in relapsed/refractory NSCLC ([Table 2](#t2-ott-8-1849){ref-type="table"}). Limitations and future directions ================================= As is the case with all targeted therapies, there is a pressing need to develop companion diagnostics or predictive biomarkers to better select patients who might derive the most benefit from ganetespib therapy. Due to the prolonged tumor retention of this inhibitor, it is evident that to optimize treatment, target modulation within the tumor itself is necessary. This will help expand the role of ganetespib and this class of agents beyond the tumors that are addicted to a sensitive client protein. Non-invasive \[18F\]-Fluorodeoxyglucose Positron Emission Tomography imaging has most recently been utilized as a means for measuring early response to ganetespib therapy; however, the utility of this approach has yet to be validated in large prospective trials.[@b12-ott-8-1849],[@b22-ott-8-1849],[@b23-ott-8-1849] Ongoing studies such as the window of opportunity trial in head and neck cancer, a Phase I trial in rectal cancer, and the ongoing Phase III GALAXY-2 trial in NSCLC are collecting tumor tissues for biomarker analyses that may also yield valuable information and guide further development of this agent. Conclusion ========== Altogether, our knowledge of the HSP90-complex machinery continues to grow. While as a class, HSP90 inhibitors have certainly emerged as an exciting multifaceted anticancer strategy, there are no currently approved HSP90 inhibitors for any cancer indication. The improved safety profile and superior efficacy of ganetespib compared with the other HSP90 inhibitors is attributed to its small molecular weight, increased lipophilicity, and absence of the benzoquinone moiety, allowing for a greater therapeutic index. Clinically, ganetespib has been shown to be very well tolerated with a notable lack of significant cardiac, liver, and ocular toxicity, which have been the limiting factors for further development of many of the other HSP90 inhibitors. While efficacy has been promising in NSCLC and HER2-positive metastatic breast cancer, many Phase I/II clinical trials are utilizing combinatorial approaches with other therapeutic agents to further improve the outcomes with ganetespib and also to expand its role in other disease types. Data from the ongoing Phase III trial in NSCLC along with the concurrent development of predictive biomarkers in other key studies have the potential to make a significant impact on the subsequent applications of ganetespib therapy and perhaps may lead to a path of regulatory approval for this promising HSP90 inhibitor. Komal Jhaveri would like to thank the Terri Brodeur Breast Cancer Foundation and all the patients and their families who participated in the clinical trials. Disclosure The authors report no conflicts of interest in this work. ![Chemical structure of ganetespib.](ott-8-1849Fig1){#f1-ott-8-1849} ###### First- and second-generation HSP90 inhibitors HSP90 inhibitors Class Pharmaceutical company ---------------------------------------- ------------------------- ---------------------------------------- First-generation HSP90 inhibitors Tanespimycin (17-AAG, KOS-953) Geldanamycin derivative Kosan Biosciences/Bristol-Myers-Squibb Alvespimycin (17-DMAG) Geldanamycin derivative Kosan Biosciences/Bristol-Myers-Squibb Retaspimycin (IPI-504) Geldanamycin derivative Infinity Pharmaceuticals IPI-493 Geldanamycin derivative Infinity Pharmaceuticals Second-generation HSP90 inhibitors CNF2024/BIIB 021 Purine Biogen Idec MPC-3100 Purine Myriad Pharmaceuticals/Myrexis Debio 0932 (CUDC-305) Purine-like DebioPharm PU-H71 Purine Samus Therapeutics Ganetespib (STA-9090) Resorcinol--Triazole Synta Pharmaceuticals NVP-AUY922 (VER-52269) Resorcinol--Isoxazole Novartis NVP-HSP990 Not reported Novartis KW-2478 Resorcinol Kyowa Hakko Kirin Pharma AT13387 Resorcinol Astex SNX-5422 Indazol-4-one Serenex/Pfizer DS-2248 Not reported Daiichi Sankyo Inc XL888 Not reported Exelixis Abbreviation: HSP90, heat shock protein 90. ###### Ongoing combination trials of ganetespib with other therapeutic agents ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Phase Disease Combination Dose and schedule Primary endpoint [ClinicalTrials.gov](http://ClinicalTrials.gov) Identifier --------------------- --------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------ -------------------------------- ------------------------------------------------------------ I Multiple myeloma Ganetespib +/−bortezemib Ganetespib IV days 1, 4, 8, 11 every 3 weeks\ MTD NCT01485835 Bortezomib IV or subcutaneous days 1, 4, 8, 11 every 3 weeks I/II Phase 1 includes multiple sarcoma subtypes and Phase 2 MPNST Ganetespib plus sirolimus Ganetespib 150 mg/m^2^ IV on days 1, 8, and 15 IV over 1 hour\ Toxicity/clinical benefit NCT02008877 Sirolimus 4 mg taken orally once daily on a continuous dosing schedule I Rectal cancer Ganetespib, capecitabine, and radiation Ganetespib 60 mg/m^2^ once weekly ×2 weeks prior to starting XRT and then on days 1, 8, 15 for cycle 1 and then on days 29 and 36 for cycle 2\ Response rate NCT01554969 Capecitabine at 825 mg/m^2^ twice daily for the entire duration of radiation therapy I HER2+ metastatic breast cancer Ganetespib, paclitaxel, and trastuzumab Ganetespib IV over 1 hour on days 1, 8, and 15\ MTD NCT02060253 Trastuzumab IV over 30 minutes, and paclitaxel IV over 1 hour on days 1, 8, 15, and 22 II Neoadjuvant breast cancer (I-SPY 2) Ganetespib + paclitaxel Ganetespib 150 mg/m^2^ once weekly for 3 out of 4 weeks\ pCR NCT01042379 Paclitaxel: 80 mg/m^2^ IV weekly ×12 weeks Randomized Phase II HR+ breast cancer Arm A: Fulvestrant\ Fulvestrant I M on day 1 and 15 of cycle 1, day 1 of cycle 2 and each subsequent cycle\ PFS NCT01560416 Arm B: Fulvestrant + ganetespib\ Ganetespib IV over 1 hour on days 1, 8, and 15 of each cycle Arm C: Crossover from Arm A to Arm B at progression I ALK positive lung cancers Crizotinib and ganetespib Ganetespib IV over 1 hour on days 1 and 8 of a 21-day cycle\ MTD and PFS NCT01579994 Crizotinib 250 mg orally twice daily continuously I/II Solid tumors/refractory small-cell lung cancer Ganetespib + doxorubicin Ganetespib 100 mg/m^2^ or 150 mg/m^2^ IV on days 1 and 8 of a 21-day cycle\ MTD/ORR NCT02261805 Doxorubicin 50 mg/m^2^ IV on day 1 of a 21-day cycle Phase I/II Malignant pleural mesothelioma (MESO-02) For Phase II: Ganetespib + cisplatin/pemetrexed Or Ganetespib + carboplatin/pemetrexed Ganetespib IV on day 1 and day 15 of each cycle\ DLT, MTD/PFS NCT01590160 Cisplatin 75 mg/m^2^, day 1 every 21 days Pemetrexed 500 mg/m^2^, day 1 every 21 days\ Carboplatin AUC5, day 1 every 21 days Pemetrexed 500 mg/m^2^, day 1 every 21 days I/randomized II Metastatic, p53-mutant, platinum-resistant ovarian cancer (GANNET53) For Phase II: Arm A: ganetespib plus paclitaxel\ Ganetespib IV once weekly for 3 out of 4 weeks\ PFS NCT02012192 Arm B: paclitaxel Paclitaxel 80 mg/m^2^, given IV once weekly for 3 out of 4 weeks I/II Recurrent ovarian, fallopian tube, or primary peritoneal cancer Paclitaxel and ganetespib Paclitaxel and ganetespib IV over 1 hour on days 1, 8, and 15 DLT, RP2D, PFS at 6 months, RR NCT01962948 I Refractory gastrointestinal carcinomas, non-squamous non-small- cell lung carcinomas, urothelial carcinomas, and sarcomas Ganetespib and Ziv-Aflibercept Ganetespib IV weekly on days 1, 8, and 15 of a 28-day cycle\ MTD, safety NCT02192541 Ziv-Aflibercept IV every 2 weeks, on days 1 and 15, of a 28-day cycle ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Abbreviations: MTD, maximum tolerated dose; IV, intravenous; MPNST, malignant peripheral nerve sheath tumors; HER2+, human epidermal growth factor receptor 2 positive; DLT, dose limiting toxicity; RP2D, recommended Phase 2 dose; PFS, progression-free survival; RR, response rate; HR+, hormone-receptor positive; IM, intramuscular; ORR, objective response rate; pCR, pathologic complete response; ALK, anaplastic lymphoma kinase.	{ "pile_set_name": "PubMed Central" }
Introduction {#s1} ============ In most epithelial tissues, the Na^+^-K^+^-ATPase is located at the basolateral membrane and provides the driving force for a variety of Na^+^-dependent transport processes across the plasma membrane, which are critical for sustaining animal homeostasis (Schuurmans-Stekhoven and Bonting, [@B40]; Skou and Esmann, [@B42]). In marine teleost fish, the coordinated physiology of gill and intestinal epithelia is responsible for the mobilization of ingested seawater toward the submucosal vascular beds, and the extrusion of the excess of NaCl from ingested seawater, leading to a net balance of water intake (Evans, [@B21]; Evans et al., [@B22]). The key molecular components in the machinery for the elimination of NaCl against a dramatic electrochemical gradient reside in a specialized cell type, namely chloride cells, which are located in the branchial epithelia (Evans, [@B21]; Evans et al., [@B22]). Despite the complexity of iono- and osmoregulatory processes in marine fish, involved mechanisms are ultimately dependent on the exergonic activity of the Na^+^-K^+^-ATPase (Evans, [@B21]; Evans et al., [@B22]). Previous studies performed in our laboratory have disclosed differences in the biochemical properties of the Na^+^-K^+^-ATPase activities along the intestinal tract of the gilthead seabream, Sparus aurata (Díaz et al., [@B13]; Almansa et al., [@B1]; Dópido et al., [@B16]). On the basis of the differential sensitivities to ouabain, calcium and to ionic strength, it was concluded that α-subunit of Na^+^-K^+^-ATPase expression varied along the intestinal tract with α~1~ being ubiquitously expressed, but α~3~-subunit only observed in distal regions (Almansa et al., [@B1]). Later on, we found that differences in thermodynamic properties existed between intestinal segments from gilthead seabream and that these differences were likely due to different phospholipid microenvironment rather than to differential expression of α-subunits (Almansa et al., [@B2]). The influence of the lipid microenvironment composition on the thermodynamic and kinetic properties of the Na^+^-K^+^-ATPase is pervasive and has been recognized for long (Wheeler and Whittam, [@B49]; Klingenberg, [@B30]; Brasitus, [@B6]; Cornelius and Skou, [@B11]; Yeagle et al., [@B51]; Muriana et al., [@B34]; Gerbi et al., [@B25], [@B26]; Ventrella et al., [@B47]). From these early studies it soon became clear that the degree of unsaturation of membrane phospholipids, as well as membrane cholesterol contents, played modulatory roles on the kinetic features of the Na^+^-K^+^-ATPase, which were initially interpreted as secondary to differences in membrane fluidity. However, the precise molecular association between degree of phospholipid unsaturation, type of phospholipid, cholesterol-to-phospholipid relationships, on the membrane biophysics (and geometry) and dynamics of membrane-bound integral proteins is just starting to be unraveled. Studies using membrane models have demonstrated direct interactions between Na^+^-K^+^-ATPase α-subunits and specific membrane phospholipids and fatty acids, and that there exit interaction sites at the hydrophobic surfaces of both α-subunits and regulatory FXYD protein in the buried helixes exposed to the phospholipid bilayer (Arora et al., [@B3]; Cohen et al., [@B8]; Esmann and Marsh, [@B20]). At present, there exist solid evidence for direct and specific interactions of different phospholipids and cholesterol which affect both the stability and molecular activity of the Na^+^-K^+^-ATPase, with essential roles in physiological regulation linked to membrane lipid composition (Gerbi et al., [@B25], [@B26]; Yeagle et al., [@B51]; Crockett and Hazel, [@B12]; Else and Wu, [@B18]; Cornelius, [@B9]; Almansa et al., [@B2]; Esmann and Marsh, [@B20]; Cornelius et al., [@B10]). Unlike model membranes, where lipid composition is set to specific compositions, usually containing few molecular lipid species, membranes from living cells display an enormous biochemical complexity with \~10,000 different molecular species, likely depending on the organism and cell type, which, in turn, are subjected to continuous remodeling in response to intracellular and extracellular signals (Dowhan, [@B17]; Ernst et al., [@B19]). Important (and abundant) lipid molecules tightly associated to membrane physical properties in living cells are polyunsaturated fatty acids (LCPUFA). These fatty acids generally esterify glycerol backbone at sn-2 position in membrane phospholipids. Unlike saturates and monoenes, which can be synthesized by most vertebrates, including marine fish, LCPUFA cannot be produced (or are produced in very limited amounts) and their needs are mostly covered by their food intake. Amongst most important LCPUFA fatty acids, those belonging to the n-3 series (n-3 LCPUFA) and n-6 series (n-6 LCPUFA) are considered essential (Sargent et al., [@B39]; Díaz and Marín, [@B14]; Spector and Kim, [@B43]), and their deficiency are associated with a number of pathophysiological conditions and with the failure in adaptive responses (Bell et al., [@B4]; Gerbi et al., [@B25], [@B26], [@B24]; Sargent et al., [@B39]; Bogdanov et al., [@B5]; Russo, [@B38]; Díaz and Marín, [@B14]; Matsunari et al., [@B33]). In the present study we have aimed to determine the influence of n-3 LCPUFA-deficient diets in the lipid composition of isolated intestinal and branchial epithelial cells of the gilthead seabream in vivo, and the extent to which such lipid modifications impact the thermodynamic properties of Na^+^-K^+^-ATPase from epithelial cells. The outcomes indicate that branchial epithelium is much more resistant to diet induced-LCPUFA depletion, in terms of phospholipid remodeling, than the intestinal counterparts. To the best of our knowledge, this is the first study demonstrating that the lipid composition of epithelial cells differs depending on their histological origin. Also, the results disclose severe changes in the thermodynamic behavior of the Na^+^-K^+^-ATPase that correlate with altered lipid environment-protein interactions in vivo. Materials and methods {#s2} ===================== Animals, diets, and cell preparation ------------------------------------ Gilthead seabream (S. aurata) (average weight 400 g) were initially reared at the National Institute of Oceanography of Tenerife (Spain) in seawater (35°/~oo~) at 20°C and fed commercial fish pellets (CONTROL), containing 2.07% n-3 LCPUFA (dry weight basis, DWB). A subset of specimens were then reared in the Atlantic Ocean (Los Gigantes, Southwest coast of Tenerife, Spain) in parallel tanks containing 10 individuals each, and fed once a day with an amount of pellets equivalent to 1% their biomass of the control diet (CONTROL) or a LCPUFA-deficient diet (DEFICIENT) for 6 months. The experimental diet containing no n-3 LCPUFA and very low amounts of n-3 PUFA was based on olive oil as lipid source, and obtained from Stirling University\'s Fish Nutrition facility (UK). Detailed composition of diets is shown in Table [1](#T1){ref-type="table"}. After decapitation, intestinal segments were isolated from pyloric caeca and from the proximal and distal portions of the intestine as described previously (Almansa et al., [@B1]). For branchial epithelia, animals were first perfused through continuous ventricular injection of Ringer physiological solution to remove blood cells from gills. All tissues were rinsed in ice-cold Ringer solution, and submitted to epithelial cells isolation following the procedures described in detail in Dópido et al. ([@B16]). Isolated cells were immediately placed in the homogenization solution containing 50 mM sucrose, 20 mM TRIS, 1 mM EDTA, 1 mM of the protease inhibitor, phenylmethylsulfonyl fluoride (PMSF), pH 7.5 (adjusted with TRIS/HCl) and kept at 4°C. Once homogenized, samples were stored in 1 ml aliquots at −80°C until analysis. All experimental manipulations were carried out following the procedures approved by the ethics committee (Comité de Ética de la Investigación y de Bienestar Animal: CEIBA) from Universidad de La Laguna. ###### Composition of control and PUFA-deficient diets. Control Deficient ---------------------- ------------- --------------- Protein 40 38 Lipids 21 19 Celulose 3 5 Humidity 9 10 Ash 7 9 Data in % Weight Fatty acids Control Deficient 14:0 6.8 tr 16:0 18.9 11.8 16:1 7.6 1.4 18:0 3.7 2.9 18:1 12.7 72.1 18:2 n-6 4.3 9.6 18:3 n-3 1.4 0.6 18:4 n-3 2.9 -- 20:1 2.4 0.4 20:2 n-6 0.2 tr 20:4 n-6 0.7 tr 20:4 n-3 0.7 tr 20:5 n-3 12.6 tr 22:1 2.2 0.4 22:5 n-6 0.4 tr 22:5 n-3 1.4 tr 22:6 n-3 13.2 tr Monoenes 26.3 74.3 Saturates 30.7 14.7 n-3 PUFA 32.2 0.9 n-6 PUFA 5.9 9.8 n-3 LCPUFA 27.9 tr n-6 LCPUFA 1.1 tr n-3/n-6 5.5 0.1 Data in % area. PUFA: polyunsaturated fatty acids (≥2 double bonds). LCPUFA: Long-chain polyunsaturated fatty acids (≥18 carbons and ≥2 double bonds). ATPase assays ------------- Na^+^-K^+^-ATPase activities were measured in triplicate as the difference in inorganic phosphate (Pi) production from ATP in the presence or absence of 1 mM ouabain, under steady-state conditions, as described previously (Ventrella et al., [@B48], [@B47]; Díaz et al., [@B13]; Almansa et al., [@B1]). Briefly, 25 μl of protein suspension, containing 50--100 μg protein, were added to test tubes containing 1 ml of incubation media (25 mM HEPES, 200 mM NaCl, 10 mM KCl, 5 mM MgCl2, and adjusted to pH 7.4 with TRIS) and allowed to preincubate for 5 min in a temperature-controlled water-jacketed chamber set at the desired temperature. The reaction was started by the addition of 50 μl (5 mM final concentration) of vanadate-free ATP and incubated for 10 min at different temperatures ranging from 4 to 50°C under continuous agitation. The amount of Pi produced was determined by the method of Forbush ([@B23]). Accordingly, at the end of the incubation period, the reaction was stopped by adding 1 ml of an ice-cold solution containing 2.8% ascorbic acid, 0.48%, ammonium heptamolybdate, 2.8% SDS and 0.48 M HCl, and the tubes were placed at 4°C for 10 min. Afterwards, the reaction was developed by incorporating 1.5 ml of 2% sodium citrate, 2% sodium-m-arsenite and 2% acetic, incubated at 37°C for another 10 min, and the absorbance read at 705 nm. Na~2~HPO~4~ was used as Pi standard for calibration curves. Corrections for unspecific ATP hydrolysis were made by measuring the amount of Pi liberated in the absence of protein samples at each temperature tested. Specific Na^+^-K^+^-ATPase activities were expressed as μmol Pi/mg prot.hr. Lipid analysis -------------- Gill and intestinal epithelial homogenates were also submitted to lipid analyses. For these purposes, cell collections were homogenized in chloroform/methanol (2:1 v/v) containing 0.01% of butylated hydroxytoluene as antioxidant, and then the organic solvent was evaporated under a stream of nitrogen. Lipid contents were determined gravimetrically. Lipid extracts were re-dissolved in chloroform/methanol (2:1 v/v) and stored in 1 mL glass vials in a nitrogen atmosphere free of O~2~ at −20°C until analyses. Lipid classes were separated by one-dimensional double development high performance thin layer chromatography (HPTLC) as described elsewhere (Olsen and Henderson, [@B35]; Martín et al., [@B32]) and quantified by scanning densitometry using a Shimadzu CS-9001PC dual wavelength flying spot scanner. The polar lipid fraction (PL) was separated from neutral lipids by silica sep-pak cartridges and then subjected to acid-catalyzed transmethylation to yield fatty acid methyl esters (FAME) (Christie, [@B7]). FAME were separated and quantified using a Shimadzu GC-14A gas chromatograph equipped with a flame ionization detector, an integrator and a fused silica capillary column Supelcowax TM 10 (30 m × 0.32 mm I.D.). Individual FAME were identified by reference to a multi-standard mixture which included authentic standards from marine fish (Supelco, Bellefonte, USA) and further confirmation of identity carried out by mass spectrometry when necessary. Statistics and calculations --------------------------- All results are expressed as means ± SEM for, at least, four different determinations. Experimental data were submitted to one-way ANOVA followed by Tukey\'s test or to Kruskall-Wallis analysis followed by Mann-Withney U test, were appropriate. Correlation and determination coefficients were obtained by Pearson\'s approach. Multivariate analyses were performed using Principal Component Analyses (PCA). Statistical calculations were performed using SPSS (v.15.0 SPSS Inc., Chicago). Estimations of regression equations (lineal and exponential) and related parameters were performed by non-linear regression analysis tools using Sigma Plot software (Jandel Scientific, San Rafael, CA). A p-value below 0.05 was considered to achieve statistical significance. Materials --------- Ouabain, sodium dithionite, vanadate-free Na~2~ATP, HEPES, EDTA, TRIS, type IV Collagenase and PMSF were purchased from Sigma-Aldrich (Biosigma, Spain). Dimethylsulfoxide and HPTLC plates were obtained from Merck (Germany). Silica sep-pak cartridges were supplied by Millipore (Milford, MA). All reagents were analytical grade. Results {#s3} ======= Lipid profiles in isolated gill and intestinal epithelial cells from control fish --------------------------------------------------------------------------------- Lipid composition of gill epithelia and isolated enterocytes obtained from animals fed control diets are shown in Table [2](#T2){ref-type="table"}. It can be observed that neutral and polar total lipid contents as well as individual lipid classes are rather homogeneous between intestinal segments. Only phospahtidylglicerol (PG, lowest in posterior intestine) and free fatty acids (FFA, lowest in anterior intestine) significantly differ between regions. However, it turns out that lipid profiles in epithelial cells from gill origin notably differ from those of enterocytes. Thus, gill epithelial cells contain lower amounts of phosphatidylcholine (PC), phosphatidylinositol (PI), phospahtidylglycerol (PG), phosphatidylserine (PS), phosphatidylethanolamine (PE), and as a consequence, total polar lipids (TPL), as well as higher levels of cholesterol (CHO), sterol esters (SE), free fatty acids (FFA) and resultant total neutral lipids (TNL) compared to enterocytes (Table [2](#T2){ref-type="table"}). ###### Lipid classes and total polar lipid fatty acid composition of isolated epithelia from gilthead seabream reared under standard conditions. Pyloric caeca Anterior intestine Posterior intestine Gill epithelia ------------------------ ------------------- ------------------------ ------------------------- -------------------- LIPID CLASSES LPC 0.27 ± 0.46 0.18 ± 0.21 0.27 ± 0.65 0.20 ± 0.29 SM 1.85 ± 0.56 2.08 ± 0.70 2.10 ± 0.86 1.04 ± 0.44 PC 19.38 ± 2.46a 17.49 ± 2.13a 17.77 ± 1.87a 4.21 ± 1.50b PS 3.36 ± 0.80a 3.23 ± 0.56a 2.81 ± 0.89ab 1.08 ± 0.46b PI 4.66 ± 0.57a 4.30 ± 0.77a 3.72 ± 0.70a 1.03 ± 0.21b PG 5.31 ± 0.58a 4.31 ± 0.95a 3.11 ± 0.78b 1.02 ± 0.77c PE 12.13 ± 2.14a 10.97 ± 1.86a 10.27 ± 2.54a 3.03 ± 2.10b TPL 46.96 ± 6.48a 42.56 ± 5.46a 42.02 ± 6.56a 11.61 ± 4.91b MAG 1.93 ± 1.14b 1.92 ± 1.12b 1.41 ± 1.24b 7.39 ± 3.82a DAG 0.00 ± 0.00 0.00 ± 0.00 0.27 ± 0.67 0.52 ± 0.74 CHO 18.95 ± 2.24b 18.42 ± 1.64b 19.46 ± 0.57b 25.83 ± 2.11a FFA 1.78 ± 1.83b 5.33 ± 3.64a 7.39 ± 4.53a 14.44 ± 6.08a TAG 23.24 ± 5.16b 24.48 ± 2.50b 23.16 ± 4.14b 33.46 ± 1.24a SE 5.57 ± 1.66 5.61 ± 2.81 6.06 ± 2.00 3.22 ± 0.36 TNL 53.04 ± 6.48b 57.44 ± 5.46b 57.98 ± 6.56b 84.86 ± 6.72a LT (μg/mgprot) 554 ± 60a 375 ± 49b 345 ± 53b 560 ± 80a FATTY ACIDS 14:0 1.07 ± 0.53b 0.44 ± 0.24b 0.32 ± 0.21b 2.05 ± 0.11a 16:0 21.99 ± 1.38 21.34 ± 1.09 21.05 ± 0.92 22.08 ± 0.75 16:1 0.77 ± 0.34b 0.45 ± 0.17b 0.57 ± 0.22b 3.73 ± 0.64a 18:0 10.31 ± 0.79b 11.04 ± 0.84ab 12.11 ± 0.23a 8.56 ± 0.45b 18:1 n-9 9.84 ± 0.62b 10.70 ± 0.27b 10.10 ± 0.87b 14.85 ± 0.21a 18:2 n-6 2.49 ± 0.05 2.49 ± 0.03 3.21 ± 0.40 2.85 ± 0.01 18:3 n-3 0.00 ± 0.00b 0.00 ± 0.00b 0.00 ± 0.00b 0.17 ± 0.00a 18:4 n-3 0.09 ± 0.08 0.12 ± 0.10 0.13 ± 0.11 0.17 ± 0.02 20:1 1.41 ± 0.14b 1.98 ± 0.40b 1.49 ± 0.13b 0.45 ± 0.16a 20:2 n-6 0.36 ± 0.07 0.30 ± 0.03 0.20 ± 0.23 0.00 ± 0.00 20:4 n-6 2.59 ± 0.27b 2.41 ± 0.18b 2.62 ± 0.25b 5.10 ± 0.43a 20:4 n-3 0.48 ± 0.04a 0.53 ± 0.06a 0.54 ± 0.03a 0.39 ± 0.05b 20:5 n-3 6.15 ± 1.55b 6.32 ± 1.94b 6.14 ± 1.13b 9.15 ± 0.38a 22:1 0.25 ± 0.05a 0.40 ± 0.27a 0.30 ± 0.16a 0.00 ± 0.00b 22:5 n-6 0.46 ± 0.04 0.56 ± 0.07 0.60 ± 0.05 0.57 ± 0.10 22:5 n-3 2.25 ± 0.36a 2.47 ± 0.23a 2.55 ± 0.21 2.53 ± 0.32 22:6 n-3 34.13 ± 2.17a 35.04 ± 0.79a 32.70 ± 1.50a 21.69 ± 0.33b TOTALS AND INDEXES Monoenes 12.29 ± 0.90b 13.53 ± 0.53b 12.47 ± 0.58b 20.19 ± 0.21a Saturates 33.69 ± 1.71 33.35 ± 1.72 34.03 ± 1.14 34.06 ± 0.67 n-6 6.03 ± 0.38b 5.98 ± 0.09b 6.86 ± 0.53b 8.85 ± 0.57a n-3 LCPUFA 43.03 ± 1.93a 44.38 ± 1.67a 41.95 ± 1.33a 33.76 ± 1.08b 18:1/n-3 LCPUFA 0.16 ± 0.00b 0.17 ± 0.01b 0.17 ± 0.01b 0.44 ± 0.01a UI 248.96 256.74 244.40 194.35 Different letters in the same row state for statistical differences with p-values below 0.05. LPC, Lyso-phosphatidylcholine; SM, Sphingolipids; PC, Phosphatidylcholine; PS, Phosphatidylserine; PI, Phosphatidylinositol; PG, Phosphatidylglycerol; PE, Phosphatidylethanolamine; TPL, Total Polar Lipids; MAG, Monoacylcerides; DAG, Diacylcerides; CHO, Cholesterol; TAG, Triacylcerides; FFA, Free fatty acids; SE, Sterol esters; TNL, Total Neutral Lipids; TL, Total Lipids; UI, Unsaturation Index. With the exception of 18:1/n-3 LCPUFA ratio, UI and TL, all data are expressed as % of total lipid. TL is expressed as μg lipids/mg total protein. Regarding fatty acids from polar lipids, significant differences between enterocytes were observed for stearic acid (18:0), linoleic acid (18:2n-6) and docosapentaenoic acid (DPA, 22:5n-6), which were higher in posterior intestine. However, when compared with gill epithelium significant differences in most fatty acids were observed. Thus, levels of palmitoleic acid (16:1), linoleic acid (18:2n6), linolenic acid (18:3n-3), and total monoenoic and total n-6 fatty acids were higher than in enterocytes, while contents of eicosatetraenoic acid (20:4n-3), eicosapentaenoic acid (20:5n-3), docosahexaenoic acid (22:6n-3) and total n-3 LCPUFA were notably lower than in the enterocytes within the same animals. Consequently, the 18:1/n-3 LCPUFA relationship, was significantly higher in gill epithelial cells. The results presented so far reveal that lipid profiles in gill epithelia substantially differ from that in enterocytes. This was further confirmed by multivariate analyses using the principal component analyses (PCA). The results of the factor scores plot shown in Figure [1](#F1){ref-type="fig"} indicate that the three populations of enterocytes were clustered together while the gill population was clearly segregated (Figure [1A](#F1){ref-type="fig"}). The variables explaining most variance (overall variance 76.90%) were 18:1n-9, 20:4n-6 (ARA, arachidonic acid), linolenic acid (18:3n-3), myristic acid (14:0), which were positively related to PC1, and DHA, stearic acid (18:0) and 20:1 which were negatively related to PC1 (Figure [1A](#F1){ref-type="fig"}). PC2 was positively correlated to three n-6 fatty acids (18:2n-6 or linoleic acid, 20:3n-6, and 22:5n-6) and negatively to palmitic acid (16:0). Indeed, differences in PC2, allowed a certain differentiation between enterocytes, with caecal and posterior enterocytes appearing as opposite clouds, and anterior enterocytes positioned between them (Figure [1B](#F1){ref-type="fig"}). ![Multivariate analyses using PCA of fatty acids from total polar lipids in gilthead seabream epithelia. (A) Scatterplot for Factor scores 1 and 2, and grouped to identify different epithelial sources in control animals. (B) Factor loadings for Principal Components 1 and 2 (PC1 and PC2) obtained for fatty acids from TPL in control animals. The fraction of total variance explained by each principal component is indicated in parentheses. (C) Factor loadings for Principal Components 1 and 2 (PC1 and PC2) obtained for fatty acids from TPL in enterocytes from Control and LCPUFA-deficient animals. As in (B), the fraction of total variance explained by each principal component is indicated in parentheses. (D,E) Scatterplots for Factor scores 1 and 2 grouped to identify different diets (D), and enterocyte sources (E). For details see corresponding results section.](fphys-07-00589-g0001){#F1} Effects of n-3 PUFA-deficient diets on lipid profiles of isolated gill and intestinal epithelial cells ------------------------------------------------------------------------------------------------------ In the next experiment we aimed to determine the impact of dietary n-3 PUFA deficiency in the lipid composition of the four populations of epithelial cells. The results in Table [3](#T3){ref-type="table"} indicate that TPL were reduced by the treatment in enterocytes (at least in caecal and anterior enterocytes). This effect was due to general reductions of PC, PS, PI, and PE in deficient animals although differences only scarcely reached statistical significance. However, none of these trends were observed in gill epithelia. Regarding total neutral lipids, a generalized increase in the contents of SE, triacylglycerides (TAG) and TNL was observed in enterocytes from deficient animals irrespective of the intestinal origin. As in the case of TPL, none of these effects were observed in branchial cells. ###### Effects of experimental diets on lipid classes and total polar lipid fatty acid composition of isolated epithelia. Pyloric Caeca Anterior Intestine Posterior Intestine Gill epithelia ------------------------ ------------------- ----------------------------------------------- ------------------------- ----------------------------------------------- -------------- ----------------------------------------------- -------------- ---------------------------------------------- LIPID CLASSES LPC 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 1.86 ± 0.12 0.68 ± 0.29[^\^](#TN1){ref-type="table-fn"} SM 1.18 ± 0.36 1.08 ± 1.01 0.67 ± 0.06 0.61 ± 0.14 0.80 ± 0.04 0.41 ± 0.21[^\^](#TN1){ref-type="table-fn"} 1.52 ± 0.29 1.02 ± 0.38 PC 16.21 ± 5.36 10.31 ± 2.75 16.61 ± 0.10 11.30 ± 1.89[^\^](#TN1){ref-type="table-fn"} 13.61 ± 3.50 11.40 ± 3.23 8.23 ± 0.70 8.71 ± 3.06 PS 2.12 ± 0.25 1.58 ± 0.31 2.24 ± 0.49 2.04 ± 0.10 3.07 ± 1.00 2.04 ± 0.34 3.16 ± 0.31 3.79 ± 0.38 PI 4.50 ± 1.18 3.32 ± 0.70 3.82 ± 0.12 2.85 ± 0.73 3.28 ± 0.64 3.36 ± 0.61 2.49 ± 0.24 2.13 ± 0.41 PG 5.06 ± 0.51 3.89 ± 0.67 4.59 ± 0.02 3.49 ± 1.05 2.27 ± 1.64 3.29 ± 0.65 2.54 ± 0.112 2.44 ± 0.23 PE 12.14 ± 1.57 7.43 ± 1.04[^\^](#TN1){ref-type="table-fn"} 13.12 ± 0.30 10.97 ± 1.93 12.39 ± 2.51 13.30 ± 1.42 6.81 ± 0.43 6.11 ± 1.80 TPL 41.22 ± 8.18 27.63 ± 4.20[^\^](#TN1){ref-type="table-fn"} 41.07 ± 0.21 31.28 ± 5.21[^\^](#TN1){ref-type="table-fn"} 35.44 ± 7.40 33.84 ± 5.02 26.64 ± 0.46 24.88 ± 5.57 MAG 1.10 ± 0.82 1.32 ± 1.01 0.56 ± 0.05 0.15 ± 0.11[^\^](#TN1){ref-type="table-fn"} 1.05 ± 0.43 0.69 ± 0.08 1.33 ± 0.44 1.38 ± 2.23 DAG 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.02 ± 0.03 0.00 ± 0.00 0.00 ± 0.00 0.35 ± 0.34 0.28 ± 0.66 CHO 28.90 ± 3.38 28.25 ± 3.18 21.37 ± 0.53 20.77 ± 1.62 26.46 ± 1.72 28.45 ± 0.84 28.52 ± 1.02 27.38 ± 3.78 FFA 8.72 ± 5.87 5.26 ± 2.75 12.66 ± 1.35 10.20 ± 3.03 17.25 ± 8.64 9.22 ± 1.23 10.21 ± 0.79 10.72 ± 3.20 TAG 17.18 ± 5.57 31.89 ± 5.62[^\^](#TN1){ref-type="table-fn"} 21.06 ± 1.45 33.51 ± 3.00[^\^](#TN1){ref-type="table-fn"} 16.64 ± 2.21 23.79 ± 4.27[^\^](#TN1){ref-type="table-fn"} 21.78 ± 1.33 21.07 ± 3.65 SE 2.85 ± 0.08 5.62 ± 0.93[^\^](#TN1){ref-type="table-fn"} 2.44 ± 0.13 4.03 ± 0.83[^\^](#TN1){ref-type="table-fn"} 2.91 ± 1.41 3.82 ± 0.67 11.13 ± 0.31 11.29 ± 0.70 TNL 58.77 ± 8.18 72.36 ± 4.20[^\^](#TN1){ref-type="table-fn"} 58.10 ± 0.63 68.71 ± 5.21[^\^](#TN1){ref-type="table-fn"} 65.04 ± 7.29 65.98 ± 5.03 73.35 ± 0.46 72.12 ± 5.57 PHO/CHO 1.43 0.98 1.92 1.51 1.34 1.19 0.87 0.88 LT (μg/mgprot) 767 ± 90 1090 ± 65[^\^](#TN1){ref-type="table-fn"} 865 ± 60 1230 ± 180[^\^](#TN1){ref-type="table-fn"} 761 ± 90 998 ± 190[^\^](#TN1){ref-type="table-fn"} 520 ± 20 518 ± 70 FATTY ACIDS* 14:0 0.42 ± 0.12 0.60 ± 0.24 1.45 ± 1.45 0.44 ± 0.28 0.59 ± 0.08 0.41 ± 0.20 0.96 ± 0.39 0.68 ± 0.23 16:0 16.88 ± 0.55 14.59 ± 2.57 18.61 ± 5.12 12.10 ± 2.68 17.36 ± 1.19 11.63 ± 0.67[^\^](#TN1){ref-type="table-fn"} 22.85 ± 3.92 22.65 ± 5.23 16:1 1.60 ± 0.32 1.78 ± 0.24 2.19 ± 1.03 1.19 ± 0.48 1.61 ± 0.23 1.18 ± 0.23 3.76 ± 0.50 2.19 ± 0.16[^\^](#TN1){ref-type="table-fn"} 18:0 11.22 ± 1.25 9.88 ± 1.80 13.11 ± 0.64 10.63 ± 2.72 13.47 ± 1.18 10.15 ± 1.69[^\^](#TN1){ref-type="table-fn"} 11.81 ± 2.01 12.11 ± 4.14 18:1n-9 9.07 ± 2.85 21.41 ± 6.72[^\^](#TN1){ref-type="table-fn"} 9.98 ± 3.34 23.84 ± 9.39[^\^](#TN1){ref-type="table-fn"} 6.86 ± 1.52 22.92 ± 5.35[^\^](#TN1){ref-type="table-fn"} 14.29 ± 3.22 20.51 ± 4.55 18:1n-7 2.73 ± 0.17 2.71 ± 0.10 2.76 ± 0.11 2.72 ± 0.20 2.31 ± 0.22 2.60 ± 0.20 3.06 ± 0.19 2.74 ± 0.30 18:1n-5 0.10 ± 0.09 0.11 ± 0.10 0.00 ± 0.00 0.00 ± 0.00 0.06 ± 0.06 0.06 ± 0.11 0.00 ± 0.00 0.09 ± 0.16 18:2n-6 5.29 ± 0.28 7.43 ± 1.97 4.87 ± 0.78 8.32 ± 3.42 4.95 ± 0.68 7.37 ± 1.50 5.81 ± 0.58 6.87 ± 2.92 18:3n-3 0.32 ± 0.06 0.24 ± 0.04[^\^](#TN1){ref-type="table-fn"} 0.44 ± 0.07 0.34 ± 0.05[^\^](#TN1){ref-type="table-fn"} 0.30 ± 0.08 0.27 ± 0.07 0.38 ± 0.04 0.11 ± 0.18[^\^](#TN1){ref-type="table-fn"} 18:4n-3 0.16 ± 0.08 0.20 ± 0.17 0.17 ± 0.07 0.10 ± 0.11 0.07 ± 0.06 0.14 ± 0.02 0.06 ± 0.11 0.05 ± 0.09 20:1 0.72 ± 0.06 1.05 ± 0.17 0.99 ± 0.13 0.92 ± 0.80 0.77 ± 0.06 1.52 ± 0.28[^\^](#TN1){ref-type="table-fn"} 1.47 ± 0.65 1.23 ± 0.24 20:2n-6 0.47 ± 0.04 0.53 ± 0.14 0.43 ± 0.09 0.58 ± 0.24 0.20 ± 0.34 0.28 ± 0.49 0.49 ± 0.07 0.64 ± 0.16 20:4n-6 2.68 ± 0.48 1.33 ± 1.18 1.56 ± 1.36 1.48 ± 0.42 2.16 ± 0.15 1.52 ± 0.21[^\^](#TN1){ref-type="table-fn"} 2.11 ± 0.96 1.78 ± 0.62 20:4n-3 0.63 ± 0.04 0.54 ± 0.16 0.60 ± 0.26 0.50 ± 0.24 0.56 ± 0.18 0.25 ± 0.26 0.10 ± 0.17 0.08 ± 0.13 20:5n-3 7.71 ± 0.65 6.10 ± 3.00 6.62 ± 2.28 5.92 ± 2.75 4.84 ± 1.56 4.37 ± 1.96 2.88 ± 1.08 1.37 ± 1.20 22:1 0.05 ± 0.04 0.10 ± 0.09 0.09 ± 0.09 0.14 ± 0.14 0.02 ± 0.04 0.05 ± 0.08 0.41 ± 0.72 0.00 ± 0.00 22:5n-6 0.70 ± 0.05 0.46 ± 0.15 0.79 ± 0.11 0.68 ± 0.40 0.96 ± 0.10 0.57 ± 0.22[^\^](#TN1){ref-type="table-fn"} 0.55 ± 0.14 0.24 ± 0.21[^\^](#TN1){ref-type="table-fn"} 22:5n-3 3.08 ± 0.50 2.32 ± 0.54 2.31 ± 0.74 2.26 ± 0.54 3.19 ± 0.51 2.50 ± 0.53 1.31 ± 0.32 1.23 ± 0.22 22:6n-3 27.12 ± 1.09 18.18 ± 2.10[^\^](#TN1){ref-type="table-fn"} 23.57 ± 3.39 18.17 ± 3.07[^\^](#TN1){ref-type="table-fn"} 28.83 ± 1.50 21.72 ± 2.32[^\^](#TN1){ref-type="table-fn"} 13.46 ± 4.99 11.87 ± 2.96 TOTALS AND INDEXES Monoenes 15.73 ± 2.42 28.17 ± 6.45[^\^](#TN1){ref-type="table-fn"} 17.31 ± 2.67 29.86 ± 9.76[^\^](#TN1){ref-type="table-fn"} 12.93 ± 0.98 29.55 ± 5.13[^\^](#TN1){ref-type="table-fn"} 25.54 ± 5.32 29.79 ± 4.23 Saturates 30.02 ± 1.25 26.56 ± 4.66 35.17 ± 7.79 24.39 ± 6.12 33.11 ± 2.76 23.37 ± 2.60[^\^](#TN1){ref-type="table-fn"} 37.46 ± 6.44 36.58 ± 9.60 n-3 39.20 ± 0.45 27.74 ± 5.67[^\^](#TN1){ref-type="table-fn"} 33.88 ± 6.20 27.32 ± 6.13 38.05 ± 3.35 29.60 ± 5.06 18.19 ± 6.56 14.71 ± 3.83 n-6 9.46 ± 0.29 10.40 ± 1.24 7.81 ± 2.14 11.20 ± 3.05 8.60 ± 0.46 10.14 ± 1.54 9.16 ± 1.70 9.82 ± 3.82 n-3LCPUFA 38.61 ± 0.43 27.17 ± 5.62[^\^](#TN1){ref-type="table-fn"} 33.10 ± 6.27 26.85 ± 6.13 37.51 ± 3.31 28.92 ± 4.74 17.75 ± 6.49 14.55 ± 3.56 n-3/n-6 4.15 ± 0.17 2.73 ± 0.89 4.44 ± 0.56 2.70 ± 1.42 4.44 ± 0.51 3.02 ± 0.97 1.95 ± 0.34 1.56 ± 0.29 18:1/n-3LCPUFA 0.23 ± 0.07 0.85 ± 0.43 0.30 ± 0.08 0.98 ± 0.61 0.18 ± 0.03 0.83 ± 0.34[^\^](#TN1){ref-type="table-fn"} 0.88 ± 0.35 1.42 ± 0.19[^\^](#TN1){ref-type="table-fn"} UI 261.62 206.09 227.69 209.64 252.62 220.28 152.06 137.98 p \< 0.05 compared to controls. Abbreviations as in Table [2](#T2){ref-type="table"}. Profiles of fatty acids in polar lipids shown in Table [3](#T3){ref-type="table"}, also indicate severe changes in enterocytes from deficient animals. Thus, a dramatic reduction of DHA occurs in all enterocyte preparations, reflecting the composition of deficient diets. Furthermore, the depletion of DHA in enterocytes was accompanied by pervasive increase in oleic acid (18:1n-9), linoleic acid and total monoenoic fatty acids. Strikingly, none of these changes appear to affect gill epithelium, as only few minor fatty acids were significantly changed. Given that main differences between enterocytes from LCPUFA deficient animals were observed for fatty acids, we next performed multivariate analyses using PCA. The results illustrated on Figure [1C](#F1){ref-type="fig"} revealed that two principal components explained 64.1% of overall variance, with PC1 being positively related to saturates (mainly 14:0, 16:0, and 18:0) and negatively to polyunsaturated fatty acids (mainly 20:4n-6, 20:5n-3, and 22:6n-3). Interestingly, PC2 was positively related to 18:1n-9, but PC1 was insensitive to this fatty acid, despite large differences between groups. Plotted factor scores revealed that deficiency-induced changes in fatty acids composition of phospholipids was so dramatic that two clusters corresponding to control and deficient enterocytes were plainly segregated (Figure [1D](#F1){ref-type="fig"}). Further, when factor scores were plotted for each intestinal segment, it was observed that irrespective of the intestinal origin, the three groups of enterocytes are represented in both CONTROL and DEFICIENT clouds (Figures [1D,E](#F1){ref-type="fig"}). Thermodynamic properties of Na^+^-K^+^-ATPase in isolated enterocytes and branchial cells ----------------------------------------------------------------------------------------- We next examined the temperature-dependence curves for the Na^+^-K^+^-ATPase reaction rates from the three groups of enterocytes and branchial cells under control conditions. Arrhenius plots for the Na^+^-K^+^-ATPase from intestinal and branchial preparations along with the Eyring plots (in the insets) are shown in Figure [2](#F2){ref-type="fig"}. Arrhenius plots indicate that the overall reactions of ATP hydrolysis for the Na^+^-K^+^-ATPase activities from all sources proceeded with temperature discontinuity points (Td) and two activation energies, i.e., above (Ea~1~) and below (Ea~2~) the breaking temperature. The calculated values for Td were lowest for posterior enterocytes (15.6°C) and highest for gill epithelia (22.3°C), being Td for pyloric caeca (17.2°C) and anterior intestine (16.4°C) enterocytes very close each together. Regarding Ea~1~ it is noticeable that values of enterocytes were similar between segments (in the range 12.33 kcal/mol -- 15.58 kcal/mol), but notably lower than the value obtained in gill epithelia (17.98 kcal/mol). Conversely, Ea~2~ were similar between epithelial cells from intestine and gills (in the range 5.46 kcal/mol--7.72 kcal/mol). Of note, it can also be observed that maximal activities are attained in the range 30--35°C, well beyond the standard rearing (and environmental) temperature of this species, and that increasing the assay temperature above 45°C (I~t~, inactivation temperature) leads to a pronounced decrease in ATPase activity, which has been interpreted as result of progressive enzyme denaturation and/or destabilization of membrane microenvironment as thermal stress increases. This thermosensitive behavior of Na^+^-K^+^-ATPase is common between ectotherms (Else and Wu, [@B18]). ![Thermodynamic features of Na^+^-K^+^-ATPase from gilthead seabream epithelia reared under standard conditions. Illustrated correspond to Arrhenius plots (Log\[activity\] vs. 1/T) and Eyring plots (insets, Ln\[k/T\] vs. 1/T) for the hydrolysis of ATP by Na^+^-K^+^-ATPase in each epithelial preparation. Td: Arrhenius breaking (or discontinuity) temperature. Ea~1~ and Ea~2~: activation energies below and above the point of discontinuity, respectively. I~t~: discontinuity at inactivation temperature. Each point corresponds to the average of four different experiments.](fphys-07-00589-g0002){#F2} By applying the transition states theory using Eyring plots we estimated activation enthalpy (ΔH^‡^) and activation entropy (ΔS^‡^). Results are illustrated in the insets of Figure [2](#F2){ref-type="fig"}. From the slopes and intercepts of ln(k/T) vs. 10^3^/T plots below and above Td, it was calculated the activation enthalpy and entropy values. In general, lower enthalpy and entropy values are observed above Td. In enterocytes, ΔH^‡^ values were similar between sections and ranged between 11.78 kcal/mol and 12.50 kcal/ mol below Td, and between 4.92 kcal/mol and 7.77 kcal/mol above Td. These values were considerably higher for the case of gill epithelia, which were 18.53 kcal/mol and 10.06 kcal/mol as calculated below and above Td, respectively. Regarding activation entropy, highest entropy values were observed for gill epithelia both above (36.86 kcal/°K/mol) and below (65.95 kcal/°K/mol) Td, while in enterocytes, ΔS^‡^ remained similar and ranging between 48.49 kcal/°K/mol (caeca) and 46.01 kcal/°K/mol (anterior intestine) below Td, and 31.78 kcal/°K/mol (anterior intestine) and 22.07 kcal/°K/mol (caeca) above Td. Effects of alterations in lipid profiles on the thermodynamic properties of Na^+^-K^+^-ATPase --------------------------------------------------------------------------------------------- We next assessed the potential effects of lipid alterations induced by n-3 PUFA deficient diets on the thermodynamic properties of Na^+^-K^+^-ATPase from intestinal and branchial epithelial cells. The results are illustrated in Figure [3](#F3){ref-type="fig"}. The first obvious differences when compared to control cells were the disappearance of Td and the increase in activation energy values in the enzyme from enterocytes. Indeed, Ea~1~ values ranged from 50.13 kcal/mol in caecal enterocytes to 24.9 kcal/mol in the case of anterior enterocytes. The second most dramatic change was the increased thermosensitivity displayed by the enzyme from enterocytes, which started to inactivate (I~t~) around 26°C, nearly 20° lower than in control cells. Consequently, Ea~2~ could not be estimated in enterocyte populations. These deleterious effects of n-3 PUFA deficiency in enterocytes are indicative of membrane instability and protein inactivation as a consequence of environment destabilization. Strikingly, gill epithelium demonstrates, once again, a considerable resistance to LCPUFA depletion and its effects were much less prominent in Td, Ea~1~, and Ea~2~, though slight differences compared to control animals were detected for Td and Ea~2~, but not for Ea~1~. ![Effects of PUFA-deficient diets on thermodynamic properties of Na^+^-K^+^-ATPase from gilthead seabream enterocytes from anterior intestine, posterior intestine, pyloric caeca, as well as from gill epithelium. Illustrated correspond to Arrhenius and Eyring plots for the hydrolysis of ATP by Na^+^-K^+^-ATPase in each epithelial preparation as indicated in Figure [2](#F2){ref-type="fig"}. Each point corresponds to the average of four different experiments.](fphys-07-00589-g0003){#F3} Changes in activation enthalpy (ΔH^‡^) and activation entropy (ΔS^‡^) for Na^+^-K^+^-ATPase in enterocytes below (I~t~) were dramatic, with a generalized increase in ΔH^‡^ (24.36 kcal/mol, 33.14 kcal/mol and 44.61 kcal/mol in anterior, posterior and caecal enterocytes, respectively) and ΔS^‡^ (84.26 kcal/°K/mol, 113.94 kcal/°K/mol and 171.27 kcal/°K/mol in anterior, posterior and caecal enterocytes, respectively). However, for gill epithelia, ΔH^‡^ values remained similar to controls both above (10.06 kcal/mol) and below (18.53 kcal/mol) Td. Likewise, ΔS^‡^ changes closely resembled those of controls with values of 36.68 kcal/°K/mol and 65.95 kcal/°K/mol, above and below discontinuity point, respectively. As n-3 LCPUFA deficiency affected only fatty acid composition of membrane phospholipids, but not lipid classes, levels of most relevant fatty acids were compared with the Arrhenius plot parameters (Td, Ea~1~, Ea~2~) and activation enthalpy (ΔH^‡^) and entropy (ΔS^‡^) for enterocytes belonging to the three regions as well as for gill epithelia (Figure [4](#F4){ref-type="fig"}). Initially, we obtained the Pearson\'s correlation coefficient matrixes for all fatty acids and ratios which were differentially affected by diets against thermodynamic parameters. First, we found significant relationships between levels of 18:1n-9 (Td = 9.46 + 0.83^\^\[18:1n-9\], r* = 0.96, p \< 0.001), total monoenes (Td = 6.76 + 0.63^\^\[monoenes\], r* = 0.96, p \< 0.001), DHA (Td = 32.72 − 0.61^\^\[DHA\], r* = 0.93, p \< 0.05) or n-3 LCPUFA (Td = 3.04 − 0.38^\^\[n-3 LCPUFA\], r* = 0.91, p \< 0.05), and Td (Figure [4A](#F4){ref-type="fig"}, left panel). Noticeably, it was observed an opposite influence of monoenes (positively related) and n-3 polyunsaturated fatty acids (negatively related) and Td. The positive effect of monoenes was also related, on one side, to saturated fatty acids as revealed by the negative relationships between saturates/18:1n-9 (Td = 34.48 − 8.012^\^\[saturates/18:1n-9\], r* = 0.85, p \< 0.05) and Td, and to n-3 LCPUFA, on the other side, as indicated by the negative correlation between n-3 LCPUFA/monoenes (Td = 26.30 − 3.93^\^\[n-3 LCPUFA/monoenes\], r* = 0.88 p \< 0.05) and Td (Figure [4A](#F4){ref-type="fig"}, middle panel). More precisely, this later relationship could be defined as a very significant negative exponential relationship between the ratio DHA/18:1n-9 (Td = 14.69 + 20.66^\^e^−0.81\\[DHA/18:1n-9\]^, r = 0.96, p \< 0.005) and Td (Figure [4A](#F4){ref-type="fig"}, right panel). Since Td disappeared in n-3 LCPUFA deficient enterocytes (but not in gills) these epithelia were excluded from these regression analyses. ![Relationships between thermodynamic signatures of Na^+^-K^+^-ATPase from gilthead seabream epithelia and their lipid features. (A) Effects of individual, grouped or ratios fatty acids on Arrhenius (discontinuity) breakpoints (Td). (B) Effects of lipid parameters on activation energies below Td or I~t~: (Ea~1~). (C) Effects of lipid ratios and selected fatty acids on activation enthalpy (ΔH^‡^) and activation enthalpy (ΔS^‡^).](fphys-07-00589-g0004){#F4} On the other hand, Ea~1~ was only related to total monoenes and particularly to 18:1n-9 (Ea~1~ = 14.43 + 0.086^\^\[1.2^\[18:1n-9\]^\], r* = 0.91, p \< 0.005) levels, but not obvious relationships were observed for saturated or polyunsaturated fatty acids individually (Figure [4B](#F4){ref-type="fig"}, left panel). However, these later two lipid groups appear to substantially affect Ea~1,~since their ratios to 18:1n-9 provided good estimations of Ea~1~(Figure [4B](#F4){ref-type="fig"}, right panel),revealing in both cases very significant exponential decays. Finally, we assessed the relationships between activation enthalpy (ΔH^‡^) and activation entropy (ΔS^‡^) as dependent variables and lipid parameters. We found that enthalpy of activation was closely related to the DHA/18:1n-9 ratio (ΔH^‡^ = 13.17 − 198.7^\^e^−3.17\\[DHA/18:1n-9\]^, r = 0.92, p \< 0.01) and to saturates/18:1n-9 (ΔH^‡^ = 11.43--94.57^\^e^−2.11\\[saturates/18:1n-9\]^, r = 0.98, p \< 0.001) (Figure [4C](#F4){ref-type="fig"}, left panel). Enthalpy (ΔH^‡^) of activation was also positively (and exponentially) related to monoenes, in particular to 18:1n-9 (ΔH^‡^ = 11.63 + 0.28^\^e^−0.17\\[/18:1n-9\]^, r = 0.96, p \< 0.005) but not to n-3 LCPUFA or DHA (Figure [4C](#F4){ref-type="fig"}, middle panel). In the case of activation entropy (ΔS^‡^), two relevant relationships were observed, i.e., for 18:1n-9 (ΔS^‡^ = 46.22 − 0.57^\^e^−0.18^\^\[18:1n-9\]^, r = 0.96, p \< 0.005) (shown in Figure [4C](#F4){ref-type="fig"}, right panel), and for the ratio DHA/18:1n-9 ratio (ΔS^‡^ = 50.38 + 778.31^\^e^−3.46\\[DHA/18:1n-9\]^, r = 0.92, p \< 0.01) (not shown). Discussion {#s4} ========== In the present study, we have assessed the potential involvement of membrane lipid profiles and thermodynamic properties of the Na^+^-K^+^-ATPase in four populations of epithelial cells in the gilthead seabream. First, we observed that fatty acid composition of membrane phospholipids slightly differ between enterocytes isolated from the different intestinal segments. However, most dramatic differences were observed between enterocytes and branchial epithelial cells. To the best of our knowledge this is the first study convincingly demonstrating such heterogeneities in membrane lipid composition of epithelial cells. Highlighting main differences, we observed that, under control conditions, enterocytes are more enriched in long-chain polyunsaturated fatty acids (mainly DHA), total polar lipids, PC, PE, PS, PI, and PG, compared to gill epithelial cells. Conversely, branchial cells contain higher levels of total neutral lipids (mainly CHO, FFA, MAG, and TAG), and 20-carbons polyunsaturated fatty acids (ARA and EPA), as well as monoenes. These features allowed a neat discrimination between epithelial cells in the multivariate analyses performed here, which allowed us to define a differential lipid fingerprint between epithelial cells of intestinal and branchial origins. As the biochemical structure of membrane bilayer is finely tuned to cope with the specific cellular functions, it can be speculated that in gill epithelia, membrane phospholipids are more adequate to provide an efficient permeability transmembrane barrier against severe osmo-ionic gradients (Wood and Shuttleworth, [@B50]; Evans et al., [@B22]). Another conclusion that can be extrapolated from these data, is that as membrane surface is greater in enterocytes than in branchial cells (due to the presence of an extensive brush border), it would be expected a higher level of TPL. On the other hand, epithelial branchial cells exhibit higher levels of FFA, MAG, and TAG, which is in consonance with the high density of mitochondria and higher metabolic rates, especially in chloride cells (Wood and Shuttleworth, [@B50]; Evans et al., [@B22]), which accomplish the major task of NaCl extrusion against a very large electrochemical gradient. The second goal of our study was the assessment of the extent to which lipid profiles in epithelial cells might be affected by dietary modifications. Our dietary manipulation consisted in providing an n-3 PUFA-deficient diet which contained similar amount of total lipids compared to control standard diets, but changed proportions of PUFA, monoenoic and saturated fatty acids. The results showed that deficient diets have a profound impact on lipid profiles in enterocytes. With the sole exception of TAG which were significantly increased in all intestinal segments studied, lipid classes remained unaffected. However, most important changes were observed in fatty acid composition of membrane phospholipids. Thus, levels of total monoenes (mainly 18:1n-9) were significantly increased, while DHA, total n-3 LCPUFA and 18:3n-3, were reduced. All these findings revealed a direct influence of diet composition on the enterocyte membrane composition. However, it is worthwhile mentioning that the increase in 18:1n-9 in membrane phospholipids may also be part of a homeoviscous strategy to keep membrane fluidity in conditions of LCPUFA depletion. Strikingly, fatty acid composition of branchial cells from the same animals remained very similar between control and deficient diets, with only small changes in some minor fatty acids. These observations are very relevant from the homeostatic point of view, and pinpoints to the existence of powerful lipostatic mechanisms responsible for the biochemical stability of branchial cell membranes (including chloride cells, mucous cells, epithelial respiratory cells and pillar cells). This indeed leads us to hypothesize that the extremely relevant physiological functions of branchial cells (Wood and Shuttleworth, [@B50]; Evans et al., [@B22]) are tightly linked to a precise lipid composition in cell membranes. The next goal in the present study was the assessment of how changes in lipid profiles of cell membranes might affect the thermodynamic properties of epithelial Na^+^-K^+^-ATPase. Results from Arrhenius plots indicate that the overall reactions of ATP hydrolysis for the Na^+^-K^+^-ATPase activities from all sources followed convex behaviors (Truhlar and Kohen, [@B46]) with temperature discontinuity points (Td) and two activation energies, above (Ea~1~) and below (Ea~2~) the discontinuity, with Ea~1~\>Ea~2~. Slight differences in these parameters were detected for control enterocytes, but values for control gill epithelia were substantially different, being Td and Ea~1~ higher than in enterocytes. Similar results have been observed in mucosal scrapings (Almansa et al., [@B2]) and branchial microsomal fractions from gilthead seabream (Ventrella et al., [@B48]) and in the sea bass Dicentrarchus labrax (Ventrella et al., [@B47]) reared under similar conditions. Thermodynamically, the lower Td values obtained for enterocytes imply lower values for the thermotropic transitions of protein-lipid and protein-protein interactions and, coherently, a temperature reduction in the activation energies-shift point (Truhlar and Kohen, [@B46]). Accordingly, the enzyme from pyloric caeca would display lower activation energy above 17.20°C, which would allow an increase in the enzyme activity at lower temperatures (close to Td), compared with anterior and posterior enterocytes, whose shift points were clearly lower. Conversely, the enzyme from gills would require substantially higher temperatures (above 22.26°C) to benefit from a lower activation energy (as inferred from Ea~2~). This apparent unfavorable condition may not be compensated by the ΔH^‡^ values (both above and below Td) found in gill Na^+^-K^+^-ATPase. These assertions have been demonstrated experimentally in enzyme activity assays performed under standard conditions, where branchial Na^+^-K^+^-ATPase specific activity measured at 20°C was found \~50% lower than in averaged enterocytes. Also, these results are in line with lower activation enthalpy for enterocytes compared to gill epithelia (Table [4](#T4){ref-type="table"}). Thus, it seems that there exist environmental restrictions leading to restricted degrees of freedom (i.e., translational, rotational, and vibrational) in the transition state for the hydrolysis of ATP by Na^+^-K^+^-ATPase in branchial membranes. This is perhaps the price must be paid in gill epithelia to cope with extreme osmotic and ionic gradients far from equilibrium. It is known that the Na^+^-K^+^-ATPase reaction cycle includes two major conformational intermediates, namely E1 (with high affinity for ATP) to E2 (with phosphatase activity) (Schuurmans-Stekhoven and Bonting, [@B40]; Skou and Esmann, [@B42]; Kaplan, [@B29]), and that the transition between them occurs with a large enthalpy and entropy change. Moreover, changes in membrane order (reduction) favors the E1 conformation and ATP binding (i.e., by PUFA or ethanol) while the increase in membrane order by reduced temperature or PUFA depletion, favors the E2 conformation with ATP hydrolysis and K^+^ binding (Swann, [@B45]; Skou and Esmann, [@B42]; Kaplan, [@B29]). Thus, the higher absolute values of ΔH^‡^ (and ΔS^‡^) of activation for the hydrolysis of ATP in gill Na^+^-K^+^-ATPase suggest more ordered membranes, where the stabilization forces required for the transition of E1 to E2 conformations for hydrolysis are higher than for enterocytes. Indeed, our analyses of membrane lipid composition (with significantly lower contents of total n-3 LCPUFA in gill epithelia) and the lower membrane unsaturation index found in gill cell membranes (194.35) compared to enterocytes\' (ranging 244.4--256.7) support the thermodynamic observations and may well explain the observed differences between intestinal and branchial epithelia. ###### Summary of thermodynamic parameters calculated for epithelial cells from control individuals and animals fed LCPUFA deficient diets. **Ea~1~ (kcal/mol) Ea~2~ (kcal/mol) Td (°C) ΔH^‡^ \< Td (kcal/mol) ΔH^‡^ \>Td (kcal/mol) ΔS^‡^ \< Td (kcal/°K/mol) ΔS^‡^ \> Td (kcal/°K/mol) ---------------------- ---------------------------------------- ------------------------ ---------------------------------------- ---------------------------------------- ------------------------------- ----------------------------------------- ----------------------------------- CONTROLS Gill 17.98 7.72 22.26 18.53 10.06 65.96 36.87 Pyloric caeca 13.14 5.46 17.20 12.59 4.92 48.50 22.07 Anterior intestine 12.33 6.48 16.42 11.78 7.00 46.01 25.87 Posterior intestine 12.75 8.28 15.58 12.20 7.77 47.13 31.78 LCPUFA DEFICIENT** Gill 17.40 4.40 26.24 16.84 3.83 59.41 15.92 Pyloric caeca 50.13[^\#^](#TN2){ref-type="table-fn"} 26.09[^\#^](#TN2){ref-type="table-fn"} 49.61[^\#^](#TN2){ref-type="table-fn"} 171.27[^\#^](#TN2){ref-type="table-fn"} Anterior intestine 24.91[^\#^](#TN2){ref-type="table-fn"} 27.22[^\#^](#TN2){ref-type="table-fn"} 24.36[^\#^](#TN2){ref-type="table-fn"} 84.26[^\#^](#TN2){ref-type="table-fn"} Posterior intestine 33.68[^\#^](#TN2){ref-type="table-fn"} 25.77[^\#^](#TN2){ref-type="table-fn"} 33.14[^\#^](#TN2){ref-type="table-fn"} 113.93[^\#^](#TN2){ref-type="table-fn"} Values were obtained below I~t~. The relationships between epithelial membrane lipids and thermodynamic properties of Na^+^-K^+^-ATPase was further demonstrated in the n-3 PUFA deficiency experiments. This maneuver dramatically affected lipid profiles in enterocytes. As mentioned before, the main effects being a considerable increase in oleic acid and a substantial DHA depletion. Paralleling these findings, Td disappeared and Ea~1~ increased by 2--3.8-fold compared to control animals. LCPUFA depletions also augmented the thermosensitivity of intestinal Na^+^-K^+^-ATPase and reduced the values for inactivation temperature (I~t~). Noticeably, none of these parameters were significantly affected in gill epithelia, which reinforce the hypothesis that branchial cells are endowed with a highly efficient lipostatic mechanism ensuring the biochemical stability of membrane lipids (and lipid-lipid and lipid-protein interactions), ultimately giving support to the splendid physiological complexity of fish gills. Severe changes in absolute values of ΔH^‡^ (and ΔS^‡^) of activation for the hydrolysis of ATP were observed in enterocytes (but again not in gill epithelia), which increased by 2--4-fold for ΔH^‡^ and 1.8--3.5 fold for ΔS^‡^ below inactivation temperature (I~t~). These observations suggest more ordered membranes and higher stabilization forces for the transition E1 to E2 conformations. In agreement, unsaturation indexes in enterocytes (but not in branchial cells) were reduced by LCPUFA depletion by 14% (index values ranging 261-227 in controls, and 206-227 in LCPUFA deficient cells). The deleterious effects of increased Ea~1~, ΔH^‡^ and ΔS^‡^ were readily demonstrated in the specific activity assays (as measured at 20°C, the rearing temperature of seabream). Indeed, unlike branchial preparations, in enterocytes\' homogenates, Na^+^-K^+^-ATPase specific activity fell by average 82.8% compared to controls. In order to explore more deeply the individual relationships between specific membrane lipids and lipid ratios, as independent variables, and thermodynamic parameters, we used non-linear regression analyses. We found that Td was significantly negatively related to DHA and positively to 18:1n-9 (and monoenes). Expectedly, the ratio n-3 LCPUFA/monoenes (and more precisely the ratio DHA/18:1n-9) was negatively related to both Td, following a negative exponential relationship. Overall, these data indicates that the relative amounts of 18:1n9 and DHA in membrane phospholipids are critical in determining the discontinuity point and the transition Ea~1~↔Ea~2~ and pinpointed that small changes in DHA and/or 18:1n-9, have a great impact in setting Td as per their exponential relationship. Moreover, the same ratio (DHA/18:1n-9) was also found to be a significant predictor of Ea~1~ and ΔH^‡^ values, observing in both cases negative exponential equations describing their relationships. In these cases, both total monoenes and 18:1n-9 (but not saturates or DHA alone) were significantly related to Ea~1~, ΔH^‡^ and ΔS^‡^ but following exponential relationships, which indicates a very relevant role of 18:1n-9. Overall, these data indicate the relevant effects of DHA and monoenoic fatty acids (and their relative proportions) in setting thermodynamic traits for Na^+^-K^+^-ATPase, at least for ATP hydrolysis. Finally, current available information based on reconstitution and crystallographic studies may provide, at least in part, a molecular interpretation in support of our observations. Thus, recent studies using purified detergent-soluble recombinant α/β or α/β/FXYD Na^+^-K^+^-ATPase complexes, have revealed three separate functional effects of phospholipids and cholesterol that are exerted at separate binding sites for phophatidylserine/cholesterol (site A), polyunsaturated phosphatidylethanolamine (site B), and saturated PC or sphingomyelin/cholesterol (site C) (Cornelius et al., [@B10]; Habeck et al., [@B27]). These binding sites likely correspond to three lipid-binding pockets identified in the crystal structures of the Na^+^-K^+^-ATPase (Shinoda et al., [@B41]; Laursen et al., [@B31]). The three sites appear to have different effects on Na^+^-K^+^-ATPase complexes, being site A stabilizing, site B stimulatory and site C inhibitory (Cornelius et al., [@B10]; Habeck et al., [@B27]). Therefore, it is expected that direct and specific interactions of different phospholipids and cholesterol within the protein molecular framework, will determine both the stability and molecular activity of Na^+^-K^+^-ATPase, eventually giving membrane lipid composition an essential role in its physiological regulation. As it has been mentioned before, the activating site is associated to polyunsaturated fatty acid binding, especially when esterifying the sn-2 position of neutral lipids such phosphatidylethanolamine or phosphatidylcholine. Indeed, these two phospholipids represent by average 66% of epithelial TPL (Tables [2](#T2){ref-type="table"}, [3](#T3){ref-type="table"}), and DHA is, by far, the most abundant fatty acid in epithelial polar lipids. Recently, the stimulation of Na^+^-K^+^-ATPase activity of the purified human α1β1 or α1β1FXYD1 complexes by neutral PUFA PC or PE has been reported, being the stimulation structurally selective for neutral phospholipids (Haviv et al., [@B28]). The structural selectivity for the neutral phospholipid and asymmetric saturated plus PUFA fatty acyl chain structure is actually a strong indication for a specific interaction with Na^+^-K^+^-ATPase. Indeed, molecular modeling of E1\~P and E2·P conformations bound to any of these polyunsaturated phospholipids suggests that their specific binding may facilitate the E1P↔E2P conformational transition (mainly by lowering the activation energy) (Cornelius et al., [@B10]). In addition, DHA (and other LCPUFA) impose a general physical effect on the bilayer physicochemical state as it has conformational properties that keep highly structured but fluid membrane bilayer capable to accommodate rapid protein conformational changes (Rabinovich and Ripatti, [@B36]; Stillwell and Wassall, [@B44]; Diaz et al., [@B15]). Presumably, adjustment of DHA levels within membrane phospholipids (mainly phosphoglycerides) would accomplish effective physicochemical changes over a wide temperature range, ensuring the adaptation of cell membranes to environmental fluctuations and metabolic activity (Sargent et al., [@B39]). The latter being considered an important mechanism involved in thermal acclimation in ectothermic organisms (Raynard and Cossins, [@B37]; Else and Wu, [@B18]; Ernst et al., [@B19]). Interestingly, the stimulatory effects of polyunsaturated PE or PC seem to be independent of cholesterol and the FXYD protein (Cornelius et al., [@B10]). This is relevant because cholesterol is quite abundant in epithelial lipid profiles (the second most abundant neutral lipid after TAG, and accounting for \~40% of TNL), and because cholesterol has been long recognized to be essential for optimal Na^+^-K^+^-ATPase activity (Crockett and Hazel, [@B12]; Cornelius, [@B9]; Cornelius et al., [@B10]). The effects of cholesterol on Na^+^-K^+^-ATPase are associated to its ability to impose a high degree of conformational order on the phospholipids acyl chains thereby stabilizing the liquid-ordered lipid phase, which is dominated by changes in activation entropy (Cornelius et al., [@B10]), but also to direct interaction with site A (Shinoda et al., [@B41]; Habeck et al., [@B27]). However, as we observed no changes in cholesterol contents, neither between tissues nor between diets (Table [3](#T3){ref-type="table"}), observed differences in activation ΔH^‡^ and ΔS^‡^ cannot be explained by distinct membrane cholesterol. Finally, we wish to introduce a word of caution on the extent to which our results may be interpreted. Indeed, although we have empirically studied the thermodynamic parameters of the epithelial Na^+^-K^+^-ATPase from intestinal and branchial origins, and correlated them with the membrane phospholipid composition, it is not possible to provide molecular details on the real enzyme-membrane interactions nor giving precise insights into the physical mechanisms underlying the observed responses. In conclusion, the results illustrated in the present study reveal that subtle differences in the lipid matrix of the lipid microenvironment surrounding the Na^+^-K^+^-ATPase may explain the heterogeneous thermodynamic features of the Na^+^-K^+^-ATPase in epithelia from the gilthead seabream reared under control conditions. In enterocyte preparations, these effects were exacerbated by induction of membrane n-3 LCPUFA deficiency. Noteworthy, epithelial cells from gill origin were notably resistant to diet-induced modifications of membrane lipid composition and, consequently, to alterations in the thermodynamic features of the Na^+^-K^+^-ATPase. We conclude that n-3 LCPUFA and 18:1n-9 (together with cholesterol) are critical elements for the fine tuning of the Na^+^-K^+^-ATPase activity within the context of the epithelial plasma membrane, which underlies the essential role of the Na^+^-K^+^-ATPase in the physiological regulation of osmo/ionoregulatory tasks in euryhaline teleost fish. Author contributions {#s5} ==================== RD and CR performed the lipid analyses and specific ATPase activity experiments. TG was in charge of temperature-dependence experiments. MD designed the study, analyzed the data (together with RD) and wrote the manuscript. All authors reviewed the drafted manuscript. Conflict of interest statement ------------------------------ The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. We are indebted to Lupe Acosta for invaluable assistance during the experimental stage of the study. We are grateful to Biosearch Life-Puleva Biotech for continuous support to our project at the Laboratory of Membrane Physiology and Biophysics. This research was partly supported by research grant SAF2014-52582-R from MINECO (Spain). [^1]: Edited by: Angel Nadal, Universidad Miguel Hernández de Elche, Spain [^2]: Reviewed by: Pasquale Stano, University of Salento, Italy; Angel Catala, National University of La Plata, Argentina [^3]: This article was submitted to Membrane Physiology and Membrane Biophysics, a section of the journal Frontiers in Physiology	{ "pile_set_name": "PubMed Central" }
Review of the literature {#sec0001} ======================== Background {#sec0002} ---------- Tuberculosis (TB) is an airborne infectious disease that most commonly affects the lungs. The causative organism is the obligate aerobic, acid-fast bacillus Mycobacterium tuberculosis [@bib0001], [@bib0002], [@bib0003]. There are numerous manifestations of extrapulmonary tuberculosis, many involving the eye [@bib0004]. This review will focus primarily on the intraocular forms of TB. A previous review summarized orbital TB [@bib0005]. Literature search strategy {#sec0003} -------------------------- The literature was reviewed using a PubMed search with both Medical Subject Headings (MeSH) and keywords. MeSH terms included tuberculosis, ocular tuberculosis, eye infections and visual acuity. Keywords included eye, intraocular, periocular, ocular, uveitis, sclerouveitis, panuveitis, choroiditis, retina, retinal, tuberculosis, and "ocular tuberculosis." Results were limited to available peer-reviewed, English-language journals published between 1930 and 2015. All papers were reviewed, (including single case reports), to determine if they should be referenced in this review. Epidemiology {#sec0004} ------------ The Centers for Disease Control and Prevention (CDC) estimates that one third of the world\'s population is infected with TB, but only ten percent of infected persons develop clinical manifestations of the disease [@bib0006]. Of the ten percent with detectable disease, sixteen to twenty-seven percent have extrapulmonary TB involvement, which includes those with intraocular findings [@bib0007]. The incidence of intraocular TB has been reported to range from 1.4 to 18 percent [@bib0004], [@bib0008], [@bib0009], [@bib0010], [@bib0011], [@bib0012], [@bib0013]. Age over forty, female gender, and HIV infection increase the risk of extrapulmonary TB, and individuals with HIV also have an increased risk of ocular TB [@bib0004]. Intraocular involvement {#sec0005} ----------------------- Hematogenous spread is the primary mechanism by which TB affects the eye [@bib0014], [@bib0015], [@bib0016]. However, direct local extension and hypersensitivity responses from infection elsewhere in the body can also result in intraocular findings [@bib0014]. Intraocular TB often affects the ciliary body and choroid due to the high regional oxygen tension of these tissues, and uveitis, especially posterior uveitis, is the most common form of intraocular TB [@bib0006], [@bib0014], [@bib0016]. Regardless of the clinical presentation, multiple recurrences of inflammation despite treatment should increase the level of suspicion for intraocular TB in a patient with TB risk factors. The wide variety of ways in which TB can affect the intraocular tissues are described below and summarized in [Table 1](#tbl0001){ref-type="table"} with a corresponding diagram outlining the relevant structures of the eye in [Fig. 1](#fig0001){ref-type="fig"}.Table 1Clinical presentations of intraocular tuberculosis.Table 1TissuePossible presentationsAnterior uvea/ Pars planaGranulomatous anterior uveitisIris nodulesIris atrophyIntermediate uveitis  LensCataract  ChoroidTuberclesTuberculomasAbscessesChoroiditis  RetinaMacular edemaIntra- or preretinal hemorrhageRetinitisVasculitisNeovascularizationNeuroretinitisEales disease  Optic nerveOptic neuritisRetrobulbar neuritisPapillitisPapilledemaTubercles  GlobePanuveitisEndophthalmitisPanophthalmitisGlobe ruptureFig. 1Anatomical diagram of the eye. The various anatomical structures of the eye that can be affected by tuberculosis are labeled. Labels correspond to the structures listed in [Table 1](#tbl0001){ref-type="table"}Fig 1 ### Anterior uveitis {#sec0006} TB can cause a granulomatous uveitis with iris and angle granulomas, mutton-fat keratic precipitates, posterior synechiae, and occasionally hypopyon [@bib0017], [@bib0018], [@bib0019], [@bib0020]. A pigmented hypopyon, iris nodules, and iris atrophy have also been reported [@bib0014], [@bib0021], [@bib0022]. Cataract can develop as a result of ongoing inflammation and steroid treatment, and extensive synechiae can lead to angle closure glaucoma [@bib0014]. Of patients with TB-related uveitis, anterior uveitis has been reported in 12 to 36 percent of the cases; these patients are more likely to have broad-based posterior synechiae and less likely to have filiform synechiae than patients with uveitis unrelated to TB [@bib0006], [@bib0023], [@bib0024]. ### Posterior and panuveitis {#sec0007} Panuveitis has been reported in 11 to 20 percent of patients with TB uveitis ([Fig. 2](#fig0002){ref-type="fig"}), whereas posterior uveitis accounts for 35 to 42 percent of intraocular TB [@bib0006], [@bib0023]. Multifocal choroiditis is the most common manifestation of posterior segment involvement ([Fig. 3](#fig0003){ref-type="fig"}) [@bib0006], [@bib0014], [@bib0015], [@bib0023]. Retinitis usually occurs in the setting of concomitant choroiditis rather than as an isolated syndrome [@bib0006]. TB-associated posterior uveitis can also take the form of serpiginous-like choroiditis ([Figs. 4](#fig0004){ref-type="fig"} and [5](#fig0005){ref-type="fig"}), which is hypothesized to be a hypersensitivity reaction that progresses relentlessly despite steroid treatment. [@bib0006], [@bib0025]. Serpinginous-like choroiditis may be an important marker for TB even in patients residing in non-endemic regions [@bib0026], [@bib0027]. In contrast to classic serpiginous choroiditis, the lesions in TB-associated serpiginous-like choroiditis are more pigmented, more likely to be multifocal, and often arise outside the peripapillary region [@bib0028]. However, both TB-associated choroiditis and true serpiginous choroiditis can be difficult to treat. Patients with serpiginous-like choroiditis are typically from TB-endemic regions and/or of Asian Indian ethnicity [@bib0029]. In one report, serpiginous-like choroiditis was found in 21.5 percent of patients with TB uveitis in Tunisia, North Africa, and another report found that patients with TB-associated uveitis are more likely to have serpiginous choroiditis than patients with uveitis unrelated to TB [@bib0023], [@bib0024]. Since chorioretinal inflammation may breach Bruchs membrane, all types of chorioretinal scars should be monitored for the development of neovascularization [@bib0014].Fig. 2Panuveitis secondary to tuberculosis. Optical coherence tomography of the right eye (A) reveals subfoveal fluid with intraretinal cysts and diffuse thickening. There is mild diffuse macular leakage on fluorescein angiography (B) with more distinct areas of leakage in the central and temporal macula. Indocyanine green demonstrates a vague area of hypercyanescence in the superior macula (C). The left eye looked similar.Fig 2Fig. 3Multifocal choroiditis secondary to tuberculosis. Montage color fundus photos of the right and left eyes (A and B) demonstrate multifocal punched out lesions in both eyes. The lesions are hyperfluorescent on mid-phase fluorescein angiography (C and D). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)Fig 3Fig. 4Serpiginous-like choroiditis secondary to tuberculosis. Montage color fundus photos of the right and left eyes (A and B) demonstrate extensive retinal pigment epithelial changes, atrophy, and scarring in both eyes. The lesions are hypofluorescent centrally with hyperfluorescent borders which do not leak on fluorescein angiography (C). The lesions are hypocyanescent on indocyanine green (D). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)Fig 4Fig. 5Serpiginous-like choroiditis secondary to tuberculosis. Montage color fundus photos of the right and left eyes (A and B) demonstrate multiple white, irregular lesions and mottling of the retinal pigment epithelium in both eyes as well as a peripapillary serpiginous-like lesion in the left eye. The lesions are hyperfluorescent on fluorescein angiography (C) and hypocyanescent on indocyanine green (D) as shown for the left eye. Optical coherence tomography reveals macular thinning with ratty outer segment changes in the right eye (E); the left eye has an epiretinal membrane with peripapillary subretinal material corresponding to a fibrotic scar as well as elevation of the inferotemporal macula due to a large choroidal lesion (F). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)Fig 5 ### Tubercles, tuberculomas, and subretinal abscesses {#sec0008} Another manifestation of TB-associated posterior uveitis is the tubercle, which is a small nodule that is usually located in the posterior pole. Tubercles may be unilateral or bilateral, typically lack associated vitreitis, and are usually multifocal in miliary TB [@bib0014], [@bib0030]. Tubercles are strongly associated with meningitis [@bib0031]. Large choroidal tuberculomas measuring up to 14 mm are encountered less commonly [@bib0032], [@bib0033]. Tuberculomas are yellow subretinal masses that mimic tumors and often have associated exudative retinal detachments [@bib0006]. They can occur in immunocompetent patients, those with disseminated TB, and those with central nervous system TB [@bib0034], [@bib0035], [@bib0036], [@bib0037], [@bib0038]. Sub-retinal abscesses occur as a result of liquefaction necrosis in caseating granulomas and can develop in patients with disseminated TB [@bib0014], [@bib0039]. Vitreitis and retinal hemorrhages can also be associated with these abscesses [@bib0006]. ### Endophthalmitis and panophthalmitis {#sec0009} If sub-retinal abscesses go untreated, they can rupture and result in endophthalmitis [@bib0015], [@bib0040], [@bib0041]. TB endophthalmitis is exceedingly rare, with only 18 cases published worldwide [@bib0042]. Panophthalmitis, which includes scleral involvement, can also occur, leading to globe rupture or scleral calcification in advanced cases [@bib0014], [@bib0043]. ### Retinal vasculitis and other retinovasculopathies {#sec0010} Retinal vasculitis, which is more common with TB-associated intraocular inflammation than non-TB associated uveitis, typically presents as a periphlebitis and very rarely involves the arterioles in intraocular TB [@bib0024], [@bib0044]. Periphlebitis is typically accompanied by vitreitis and is the second most common presentation of intraocular TB [@bib0045]. In a patient with TB risk factors, TB should be highly suspected in the setting of exudative hemorrhagic periphlebitis [@bib0046] Ischemic central retinal vein occlusion has also been reported in the setting of retinal vasculitis due to TB [@bib0047]. Eales disease, first described by Henry Eales in 1882, is characterized by peripheral capillary nonperfusion, neovascularization, recurrent vitreous hemorrhages, periphlebitis, and intraocular fibrovascular proliferation in a quiet eye ([Fig. 6](#fig0006){ref-type="fig"}). This entity typically affects otherwise healthy men in their third to fourth decade of life from TB-endemic countries. Associated systemic symptoms have been reported to include epistaxis, peripheral circulation disorders, headache, and constipation [@bib0048], [@bib0049], [@bib0050]. M. tuberculosis DNA has been detected in patients with Eales disease, but it remains unclear whether or not TB is directly causative of this entity, or if it is a separate retinal hypersensitivity reaction [@bib0019], [@bib0046], [@bib0051], [@bib0052], [@bib0053].Fig. 6Eales disease. Montage color fundus photo of the right eye (A) reveals a temporal frond of neovascularization in a patient with a positive Mantoux skin test. The corresponding fluorescein angiogram (B) shows leakage from the area of neovascularization with adjacent peripheral capillary nonperfusion. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)Fig 6 ### Neuroretinitis and optic neuropathy {#sec0011} Neuroretinitis typically occurs secondary to choroidal spread of a peripapillary infection [@bib0006], whereas optic nerve infiltration and subsequent optic neuropathy can occur secondary to hematogenous spread or choroidal extension [@bib0015]. Optic neuropathy can also be the result of a hypersensitivity response without direct infection of the nerve [@bib0014]. Manifestations of TB optic nerve involvement can also include papillitis, papilledema, optic neuritis, retrobulbar optic neuritis, and optic nerve tubercles [@bib0051], [@bib0054], [@bib0055]. ### Immunocompromised patients {#sec0012} Ocular involvement of TB is both more common and more severe in immunocompromised patients. There have been reports of extensive choroidal infiltrates and nodules, severe vitreitis, endophthalmitis, choroidal tuberculomas, choroiditis, chorioretinitis, and bilateral optic neuritis in the immunocompromised with TB [@bib0056], [@bib0057], [@bib0058], [@bib0059], [@bib0060], [@bib0061]. Immune recovery uveitis occurs in HIV positive patients with TB who have reconstitution of the immune system after starting antiretroviral therapy [@bib0046]. The findings in immune recovery uveitis can vary widely and include anterior uveitis, hypopyon, vitreitis, papillitis, panuveitis, retinal or optic disc neovascularization, retinal detachment, cystoid macular edema, epiretinal membrane formation, vitreomacular traction, and macular hole [@bib0062]. One reported case was so severe that it resulted in globe rupture [@bib0063]. It should be noted that immune recovery uveitis can also occur in treated HIV patients who do not have TB. Ophthalmic imaging characteristics of intraocular TB {#sec0013} ---------------------------------------------------- ### Fluorescein and indocyanine green angiography {#sec0014} On fluorescein angiography, active choroidal tubercles show early hypofluorescence and later hyperfluorescence [@bib0006]. Tuberculomas show early hyperfluorescence with late pooling in areas of exudative retinal detachment [@bib0006]. Serpiginous-like choroiditis demonstrates early hypofluorescence with late hyperfluorescence and leakage of the active borders of the lesion [@bib0015]. Inactive serpiginous-like choroiditis lesions have staining borders and less hyperfluorescent centers. Retinal vasculitis from TB will demonstrate leakage surrounding the involved vessels as well as peripheral capillary nonperfusion [@bib0006]. Indocyanine green angiography of choroidal lesions demonstrates early hypocyanescence followed by late hypercyanescence surrounding a central zone of hypocyanescence [@bib0015]. ### Optical coherence tomography {#sec0015} Optical coherence tomography (OCT) may be helpful in detecting and quantifying macular edema and other pathology such as chorioretinal lesions [@bib0014]. While a significant degree of intraocular inflammation could result in decreased image resolution, the OCT may still provide valuable information when clinical examination by slit lamp and ophthalmoscopy is very difficult [@bib0064]. ### Fundus autofluorescence {#sec0016} Fundus autofluorescence can be particularly useful in cases of serpiginous-like choroiditis [@bib0065]. Serpiginous-like choroiditis progresses from an ill-defined area of hyperautofluorescence in the acute stage, to a well-defined hypoautofluorescent halo surrounding an area of hyperautofluorescence in the subacute stage, and finally a uniformly hypoautofluorescent lesion after resolution [@bib0065]. ### Ocular ultrasound {#sec0017} B-scan ultrasonography is also useful when severe inflammation or cataract limits the view of the fundus. While mass-like tuberculomas can mimic tumors, A-scan shows low to medium internal reflectivity, which may help differentiate the tuberculoma from other entities [@bib0003], [@bib0015]. Diagnosis {#sec0018} --------- The myriad of intraocular TB manifestations can mimic other diseases, making diagnosis challenging. Delays in diagnosis of intraocular TB can lead to permanent vision loss and even loss of the affected eye(s) [@bib0066]. Without microbiologic confirmation, which can be challenging to obtain, a diagnosis of intraocular TB may be empiric; therefore it is necessary to maintain a high index of suspicion aided by a targeted review of systems and directed laboratory testing [@bib0014], [@bib0067], [@bib0068]. Response to anti-tuberculosis treatment (ATT) may be the only definitive evidence of intraocular TB, but patients should not be committed to long courses of systemic ATT without first undergoing a thorough evaluation to rule out other entities such as syphilis and toxoplasmosis that may have similar findings [@bib0006], [@bib0069]. ### Mantoux skin test {#sec0019} A Mantoux skin test involves the intradermal injection of a purified protein derivative followed by examination for cutaneous induration 48 to 72 hours after the injection [@bib0070]. A positive result is caused by a delayed-type hypersensitivity reaction and is defined as 1) induration measuring greater than 5 mm in HIV-positive patients; 2) induration greater than 10 mm in high-risk persons, including healthcare workers, nursing home patients, and those living in endemic areas; 3) induration greater than 15 mm in all other patients [@bib0006]. Cutaneous hypersensitivity has been shown to correlate directly with ocular hypersensitivity in rabbits, indicating this test may be useful in diagnosing ocular disease [@bib0071]. Unfortunately, Mantoux test interpretation is subjective, and the test has low sensitivity and specificity [@bib0006]. Anergy to the protein can lead to false negative results, especially in immunocompromised patients [@bib0006]. False positive reactions may be encountered in patients who have been exposed to different species of mycobacteria, those who have received multiple Mantoux tests in the past, or those who have previously received Bacillus Calmette-Guerin (BCG) vaccinations [@bib0006], [@bib0014], [@bib0072]. However, induration greater than 10 mm should not be attributed to prior BCG vaccination according to the United States Preventative Services Task Force [@bib0073]. ### Interferon gamma testing {#sec0020} The QuantiFERON TB-Gold test (QFTG: Qiagen, Germantown, Maryland, USA) is a blood test designed to specifically detect TB. One advantage over the Mantoux test is the fact that the QuantiFERON blood test does not require a return office visit for interpretation [@bib0074]. The blood test quantifies interferon gamma release by using M. tuberculosis antigens to stimulate sensitized T cells from TB-infected patients. Theoretically this test should not yield false positive results due to exposure to other mycobacteria or prior BCG vaccination [@bib0074], [@bib0075]. However, the false positive rate has been reported to be as high as 41% in healthcare workers and 80.5% in HIV-positive patients at low risk for TB [@bib0076], [@bib0077], [@bib0078], [@bib0079]. Nevertheless, while the QuantiFERON test may have the most utility in patients who have been vaccinated with BCG or for those who are unlikely to return for skin test reading, the CDC recommends the QuantiFERON test as a screening tool that may be used in place of a Mantoux skin test in all situations were a Mantoux test would be recommended [@bib0080]. ### Chest x-ray and computerized tomography {#sec0021} Chest x-ray or computerized tomography of the chest demonstrating cavities, consolidation, or lymph node enlargement can provide supporting evidence for TB [@bib0009]. Any positive chest imaging finding should prompt sputum evaluation to distinguish between healed TB lesions and active disease [@bib0081]. Ocular TB can occur in the absence of pulmonary TB, so a normal chest radiograph does not exclude the diagnosis of intraocular TB [@bib0003], [@bib0082]. ### Acid-fast bacteria and histology {#sec0022} The diagnosis of intraocular TB can be proven if acid-fast bacteria are demonstrated via microscopy or Lowenstein-Jensen media culture of intraocular fluids or tissue; however, the yield of this diagnostic test is typically low due to a paucity of organisms [@bib0014], [@bib0015]. Necrotizing granulomas with positively staining acid-fast bacilli by Ziehl-Neelsen stain are also compatible with TB [@bib0009]. ### Polymerase chain reaction {#sec0023} Polymerase chain reaction (PCR) amplification of mycobacterial DNA can be utilized in the evaluation of suspected intraocular TB, as the ocular specimen volume or size is often too small to detect the mycobacterium by microscopy or culture [@bib0083], [@bib0084]. PCR has the potential to be an important tool in the diagnosis of intraocular TB, and has proven useful in vitreous samples from patients with serpiginous-like choroiditis [@bib0083], [@bib0085], [@bib0086]. However, testing of aqueous and vitreous humor may be falsely negative due to a relative paucity of organisms in TB-associated uveitis [@bib0087]. Therefore, while it can be useful, positive M. tuberculosis PCR is not required to make the diagnosis of intraocular TB [@bib0015], [@bib0088], [@bib0089]. Treatment {#sec0024} --------- ### Medical treatment {#sec0025} Because systemic treatment is required to treat ocular TB, patients should be co-managed by an ophthalmologist and an infectious disease specialist. Medical ATT for ocular TB utilizes the same quadruple-drug regimen used to treat pulmonary TB. The CDC-recommended regimen consists of isoniazid, rifampin, pyrazinamide, and ethambutol for a total of 2 months, followed by an additional 4 to 7 months of dual therapy with isoniazid and rifampin [@bib0090]. Extended courses of therapy may be necessary due to multi-drug resistance and/or the typically slow and sometimes relapsing nature of ocular TB [@bib0014], [@bib0066]. Rifabutin, fluoroquinolones, interferon gamma, and linezolid may be utilized in cases of multi-drug resistance or patient intolerance to any of the four recommended agents [@bib0006]. Infectious disease specialists may choose to treat latent TB with isoniazid and/or rifampin rather than quadruple therapy, although it has become increasingly common to conclude that intraocular TB requires quadruple drug therapy [@bib0090]. Pyridoxine (vitamin B6) supplementation is necessary for patients taking isoniazid in order to prevent peripheral neuropathy [@bib0091]. ### Ocular side effects of medical therapy {#sec0026} A baseline eye exam must be performed prior to starting ATT, and patients should be monitored for medication side effects, which can sometimes be confused with worsening ocular TB [@bib0006]. Isoniazid and ethambutol can cause optic neuropathy, which presents as decreased visual acuity, decreased color vision, or cecocentral scotomas [@bib0092]. Ethambutol can also cause optic neuritis characterized by red-green dyschromatopsia and disc edema, which can lead to optic atrophy [@bib0006]. The offending medication should be promptly discontinued at the first sign of vision-threatening side effects, as these symptoms may be reversible, with vision often returning to normal 10 to 15 weeks after drug cessation [@bib0006]. Vitamin B12 can also occasionally help vision recovery, but in some cases vision loss may be permanent [@bib0006]. Rifabutin can cause anterior, intermediate, or panuveitis that is typically treatable with corticosteroids but may also necessitate discontinuation of the medication [@bib0003]. ### Use of steroids {#sec0027} While choroidal tubercles and tuberculomas often respond well to ATT alone, many other intraocular manifestations of TB require corticosteroids plus ATT to prevent damage to the ocular tissues secondary to the inflammatory response [@bib0015], [@bib0093]. Corticosteroid treatment can be critical for certain manifestations of ocular TB, such as choroiditis, which can actually undergo paradoxical worsening upon initiation of ATT [@bib0006]. Topical and locally injected corticosteroids are important in the management of anterior and intermediate uveitis, and systemic forms are an important component in retinal vasculitis treatment [@bib0003], [@bib0006], [@bib0017]. It is critical, however, that corticosteroids are never used in the absence of ATT, as this not only leads to more frequent recurrences of ocular inflammatory symptoms but can also cause significant worsening and potentially dissemination of the infection [@bib0006], [@bib0094]. ### Surgical treatment {#sec0028} Typically, ocular TB can be managed with medical therapy alone. However, in cases where retinal neovascularization develops, laser photocoagulation may be necessary to treat areas of retinal capillary nonperfusion and decrease the stimulation of neovascular growth [@bib0006]. Pars plana vitrectomy has also been reported as an adjunctive therapy for TB endophthalmitis, and full thickness eye wall resection with pars plana vitrectomy has been used to treat a tuberculous granuloma [@bib0042], [@bib0095]. Occasionally, patients may also require cataract surgery; however, this procedure should be postponed until all intraocular inflammation has been completely controlled for at least 3 months [@bib0014]. Elective surgical procedures such as cataract surgery should be delayed even longer in young or noncompliant patients and in those with severe ocular damage from intraocular TB [@bib0014]. Conclusion {#sec0029} ========== Intraocular TB is a difficult diagnosis, as it can mimic many other more common etiologies of intraocular inflammation or uveitis. A high index of suspicion for TB and cooperation with infectious disease specialists are paramount, because timely diagnosis and treatment may prevent irreversible vision loss [@bib0096]. Ann M. Farrell for assistance with literature search. Funding: Supported by Mayo Foundation, Rochester, MN; Research to Prevent Blindness, New York, NY; and the VitreoRetinal Surgery Foundation.	{ "pile_set_name": "PubMed Central" }
Introduction {#s1} ============ Dengue (DEN) and chikungunya (CHIK) are two mosquito-borne viral infections transmitted by mosquitoes of the genus Aedes. Dengue viruses (DENV-1, -2, -3, -4) belonging to the Flaviviridae family, Flavivirus genus [@pntd.0000706-Lindenbach1] are of primarily concern as they are responsible of the most important arboviral disease widely distributed in the tropical world [@pntd.0000706-Gubler1]. Dengue infection may be unapparent or induce an undifferentiated febrile illness, a classic dengue fever (DF), or a dengue hemorrhagic fever (DHF). The highest prevalence of DEN is observed in South-East Asia and South Americas with approximately 50--100 million cases and 25 000 deaths per year [@pntd.0000706-Gubler2]. The transmission is mainly ensured by the highly anthropophilic Aedes aegypti in urban areas [@pntd.0000706-Gubler3]. However, Aedes albopictus may act as a secondary vector in rural areas and even as the main vector when Ae. aegypti is not present or too scarce as observed in some localities in China, Japan, Hawaii, and Seychelles [@pntd.0000706-Gratz1]. Ae. albopictus was indeed the only vector in the recent dengue outbreaks observed in the Indian Ocean on La Reunion Island [@pntd.0000706-Pierre1] and Madagascar [@pntd.0000706-Ratsitorahina1]. Chikungunya virus (CHIKV), first isolated in Tanzania in 1953 [@pntd.0000706-Ross1], belongs to the Togaviridae family, Alphavirus genus [@pntd.0000706-Kuhn1] and is endemic to Africa, India and South-East Asia. In Africa, the virus is maintained within a sylvatic cycle with wild mosquitoes (Aedes furcifer, Aedes luteocephalus, Aedes taylori, Aedes africanus) feeding preferentially on primates [@pntd.0000706-Diallo1], [@pntd.0000706-Jupp1]. In Asia, CHIKV is mainly transmitted within an urban cycle in an inter-human transmission achieved essentially by the human-biting Ae. aegypti and the less anthropophilic Ae. albopictus, which prefers suburban and rural areas where it colonizes both artificial and natural containers [@pntd.0000706-Lam1], [@pntd.0000706-Laras1]. CHIKV mainly induces high fever and severe arthralgia, and had limited impacts on public health before its emergence in the Indian Ocean in 2005. This major epidemic started in the Comoro Islands in January 2005 then spread rapidly to the other islands of the region, Mayotte, Seychelles, La Reunion and Mauritius [@pntd.0000706-Schuffenecker1]. In April 2006, one third of the population in La Reunion Island had been in contact with the virus [@pntd.0000706-Renault1]. Surprisingly, the vector in this epidemic was not Ae. aegypti, only present as residual populations on the island, but Ae. albopictus [@pntd.0000706-Salvan1]--[@pntd.0000706-Delatte1]. This latter species was proved to be a very efficient vector of a mutated strain CHIKV harboring a switch from an alanine to a valine in the E1 glycoprotein, mutation that appeared in the course of the outbreak and was then selected as a major epidemic genotype [@pntd.0000706-Vazeille1], [@pntd.0000706-Schuffenecker1]. CHIK outbreaks spread rapidly and caused several million clinical cases in the Indian Ocean Islands and India, where outbreaks had been absent for 32 years [@pntd.0000706-Pialoux1], [@pntd.0000706-Santhosh1]. One consequence was an increasing overlapping in the distribution of the two arboviral diseases, DEN and CHIK leading to the detection of a higher number of co-infections in humans. Co-infection by DENV and CHIKV in patients has been known for a long time. First described in 1964 in South India [@pntd.0000706-Myers1], it has been more frequently reported since the re-emergence of CHIK: Sri-Lanka [@pntd.0000706-Hapuarachchi1], India [@pntd.0000706-Chahar1], [@pntd.0000706-Schilling1], Malaysia [@pntd.0000706-Nayar1] where the main vector involved is Ae. aegypti and in Gabon [@pntd.0000706-Leroy1], Madagascar [@pntd.0000706-Ratsitorahina1] and La Reunion Island [@pntd.0000706-World1] where viral transmission is achieved by Ae. albopictus. On La Reunion Island, the CHIK outbreak was preceded by a small outbreak due to DENV-1 with 228 cases reported between March and July 2004 [@pntd.0000706-Pierre1]. Despite the limited impact of the DEN outbreak, in January 2006, suspected cases of co-infection DENV-CHIKV in patients were reported [@pntd.0000706-World1] and the same phenomenon was also observed in Madagascar in January 2006 [@pntd.0000706-Ratsitorahina1]. Dual arboviral infections in humans can occur following the bite of two mosquitoes, each infected by one virus, or the bite of one mosquito infected by the two viruses. If isolations of either virus from mosquitoes collected in the course of an epidemic have been already noticed, to our knowledge, doubly infected mosquitoes have never been described. A recent study even concluded on the failure to prove co-infection of Ae. aegypti by these two viruses [@pntd.0000706-Rohani1]. Here, we orally infected in a single blood-meal, Ae. albopictus from La Reunion Island with autochthonous viral strains isolated during the 2004--2005 outbreak of DEN-1 and the 2005--2006 CHIK outbreak. Dissemination of both viruses within the vector was checked by immunofluorescence assay on female head squashes. Furthermore, saliva was collected from each female to check their ability to deliver simultaneously both DENV and CHIKV infectious particles. Additionally, we performed secondary infections by inoculating first DENV-1 then providing CHIKV in an infectious blood-meal. Methods {#s2} ======= Ethics statement {#s2a} ---------------- All experiments on live vertebrates were performed in compliance with French and European regulation and according to the Institut Pasteur guidelines for laboratory animal husbandry and care. Mosquitoes {#s2b} ---------- Ae. albopictus Providence (ALPROV) were collected as eggs in 2006 on La Reunion Island and provided by the DRASS (Direction Régionale des Affaires Sanitaires et Sociales). The F5 or F6 generation was used for experimental infections. Colonies were maintained at 28±1°C with a light:dark cycle of 16 h:8 h and a 80% relative humidity. Larvae were reared in pans containing 1 yeast tablet in 1 liter of tap water. Adults were provided with 10% sucrose solution ad libitum and fed three times a week on anaesthetized mice. Virus {#s2c} ----- The CHIKV 06.21 isolated in November 2005 from a new-born male from La Reunion Island presenting meningo-encephalitis symptoms was used for all experiments. This strain harbored the A-\>V mutation at the position 226 in the E1 glycoprotein (E1-226V) [@pntd.0000706-Schuffenecker1]. Viral stock used was a third passage on Ae. albopictus C6/36 [@pntd.0000706-Iragashi1] stored at −80°C in aliquots. Procedure for C6/36 cell infections and passages are described elsewhere [@pntd.0000706-Vazeille1]. The DENV-1 185/04 was isolated in May 2004 from the plasma of a patient in La Reunion Island. The strain belonged to the Brazil group of the Pacific genotype which was the main genotype isolated during the outbreak (GenBank: DQ285558.1). The virus was provided as a second passage on C6/36 cells. DENV-1 production on mosquito cell cultures being insufficient to allow mosquito oral infections, the virus stock was produced by inoculating intra-thoracically mosquitoes with the viral strain [@pntd.0000706-Rosen1]. Inoculated mosquitoes were incubated 10 days at 28°C and their infectious status checked by indirect immunofluorescent assay (IFA) on head squashes [@pntd.0000706-Kuberski1]. Bodies were then pooled and triturated in heated (56°C for 30 min) FCS (Fetal Calf Serum). The supernatant fluid recovered after low speed centrifugation was used as a source of virus in mosquito blood-meals. Both viruses were provided by the French National Reference Center for Arboviruses at the Institut Pasteur which had obtained the verbal consent from patients or parent\'s patients who provided blood sera. Intrathoracic inoculation of mosquitoes {#s2d} --------------------------------------- Adult females were inoculated using the protocol described by Rosen and Gubler [@pntd.0000706-Rosen1], each mosquito receiving 0.17 µL (i.e. 10^2.8^ PFU/mL) of the DENV-1 strain. Oral infection of mosquitoes {#s2e} ---------------------------- Infection assays were performed with 7 day-old females which were allowed to feed for 15 min through a chicken skin membrane covering the base of a glass feeder containing the blood-virus mixture maintained at 37°C. The infectious blood-meal was composed of a virus suspension diluted (1∶3) in washed rabbit erythrocytes isolated from arterial blood collected 24 h before the infectious blood-meal [@pntd.0000706-VazeilleFalcoz1]. A phagostimulant, ATP, was added at a final concentration of 5×10^−3^ M. Fully engorged females were transferred to small cardboard containers and maintained with 10% sucrose at 28±1°C for 14 days. Viral suspension provided in the blood-meal contained one or two viruses. For the secondary infection experiment, the blood-meal with CHIKV yielding 10^6^ FFU (foci forming unit)/mL was provided 7 or 13 days after inoculation with DENV-1. As control, females were inoculated with DENV-1 alone or orally infected with CHIKV alone. For the two trials where both viruses were provided by oral route, titers of the blood-meals were respectively: 10^6^ FFU/mL for CHIKV in both trials and 10^4,5^ FFU/mL for DENV-1 in the trial 1 and 10^5,9^ FFU/mL for DENV-1 in the trial 2. Saliva collection and virus detection/titration {#s2f} ----------------------------------------------- At day 14 post-infection, females were chilled, and their wings and legs removed, the stress inducing a forced salivation. Proboscis was inserted into 1 µL micropipette (microcaps®, Drummond Scientific Company, USA) filled with FCS. After 45 min, medium containing the saliva was expelled under pressure into 1.5 mL tubes containing Leibovitz L15 medium supplemented with 10% FBS. To allow a specific detection of both viruses, each sample was inoculated in two wells, one for the detection of CHIKV and one for the detection of DENV-1. 20 µL of each sample were added to monolayers of C6/36 cells in 24 wells plaque to detect infectious particles by the foci forming technique using an immunoperoxydase assay. Cells were incubated 3 days for CHIKV and 5 days for DENV-1 at 28°C under an overlay consisting of 50% of Leibovitz L-15 medium supplemented with 10% FBS and 50% of carboxyl methyl cellulose. Cells were then fixed with 3.6% formaldehyde at room temperature (RT) for 20 min and a immunoperoxydase assay staining was performed to detect foci. After a first incubation of 4 min with PBS 0.1% Triton X-100 (Sigma) at RT, cells were incubated 20 min at 37°C with a mouse ascitic fluid at a dilution of 1∶1000 for CHIKV and 1∶100 for DENV-1 (both ascetic fluids were provided by the French National Reference Center for Arbovirus at the Institut Pasteur). After a wash in PBS 1X, cells were incubated at 37°C for 45 min with peroxydase-conjugated goat anti-mouse IgG antibody (Pierce biotechnology, Rockford, USA) at a 1∶100 dilution in PBS 1X. After final wash in PBS 1X, Fast 3,3′ Diaminobenzidine (Sigma) was used to reveal the staining and foci were counted. The titer of infectious particles per saliva was expressed as FFU/mL (mean ± standard deviation). Female status analyzed by IFA {#s2g} ----------------------------- After salivation, females were tested for the presence of CHIKV and DENV-1 by IFA on their head squashes [@pntd.0000706-Kuberski1]. CHIKV and DENV-1 antigens were detected with the same mouse ascitic fluids used for saliva titration. Head squashes being performed between two slides, infection status of females fed with both viruses, could be checked for both antigens by using one slide for each IFA. Mosquitoes inoculated with CHIKV and DENV-1 were used as positive controls, negative controls were inoculated with cell culture media. Results {#s3} ======= Secondary infection with CHIKV of Ae. albopictus previously inoculated with DENV-1 {#s3a} ------------------------------------------------------------------------------------ As shown on [Table 1](#pntd-0000706-t001){ref-type="table"}, only few females inoculated with DENV-1 were willing to feed on an artificial blood-meal in the BSL-3 insectarium. When a blood-meal was proposed at day 7 post-inoculation, eight females out of 106 fed and among them, three survived until day 13 post-inoculation. By IFA on head squashes, we found that all 3 females had disseminated both viruses. Besides, when a blood-meal was offered at day 13 post-inoculation, eight females out of 54 females fed and four survived until day 20 post-inoculation. These four females had disseminated both viruses. Control females inoculated or orally infected by only one virus were all positive. 10.1371/journal.pntd.0000706.t001 ###### Superinfection with CHIKV of Aedes albopictus Providence previously inoculated with DENV-1. ![](pntd.0000706.t001){#pntd-0000706-t001-1} at day 7 post-inoculation at day 13 post-inoculation ---------------------------------- ------------------------------------ -------------------------------------- ----------------------------------- ------------------------------------ -------------------------------------- Engorged females at day 7 pi (N) Surviving females at day 13 pi (N) Co-infected females at day 13 pi (N) Engorged females at day 13 pi (N) Surviving females at day 20 pi (N) Co-infected females at day 20 pi (N) 8 (106) 3 (8) 3 (3) 8 (54) 4 (8) 4 (4) N, number of females tested. Oral co-infections {#s3b} ------------------ Females were exposed to both viruses in a same blood-meal and disseminated infection rates were estimated at day 14 post-infection (pi) ([Table 2](#pntd-0000706-t002){ref-type="table"}). In the trial 1, 71.6% of females have only disseminated CHIKV, 0% only DENV-1, 18.6% both viruses and 9.8% did not disseminate any virus. In the trial 2, 30.8% of females have only disseminated CHIKV, 7.7% only DENV-1, 50.8% both viruses and 10.7% did not disseminate any virus. When providing a higher titer of DENV-1 in the blood-meal (trial 2), a higher proportion of females were co-infected with both viruses. 10.1371/journal.pntd.0000706.t002 ###### Disseminated infection rates (%) of Aedes albopictus Providence 14 days after a blood-meal with both CHIKV and DENV-1. ![](pntd.0000706.t002){#pntd-0000706-t002-2} CHIKV DENV-1 CHIKV + DENV-1 Non-infected Total --------- ------- -------- ---------------- -------------- ------- Trial 1 71.6 0 18.6 9.8 102 Trial 2 30.8 7.7 50.8 10.7 65 Disseminated infection rates were estimated by IFA on head squashes. For CHIKV, both trials used a blood-meal at a titer of 10^6^ FFU/mL. For DENV-1, trial 1 corresponded to a blood-meal at a titer of 10^4.5^ FFU/mL and trial 2 to a titer of 10^5.9^ FFU/mL. Total corresponds to the total number of females tested. For females having disseminated both viruses, saliva was collected at day 14 pi and titrated. Relative transmission of the two viruses are shown in [Table 3](#pntd-0000706-t003){ref-type="table"}. In the trial 1, out of 19 saliva, 4 contained simultaneously CHIKV and DENV-1, 4 only CHIKV, 3 only DENV-1 and 8 no virus. In the trial 2, out of 33 saliva, 9 presented concomitantly CHIKV and DENV-1, 8 only CHIKV, 2 only DENV-1 and 14 no virus. Mean titers, expressed as FFU per saliva, and standard deviation are shown on [Table 4](#pntd-0000706-t004){ref-type="table"}. 10.1371/journal.pntd.0000706.t003 ###### Relative transmission of the two viruses by dually infected mosquitoes. ![](pntd.0000706.t003){#pntd-0000706-t003-3} DENV-1 ------- -------- --- --- --- ---- CHIKV Yes 4 4 9 8 No 3 8 2 14 Saliva from females having disseminated both viruses 14 days after an infectious blood-meal was tested for the presence of both viruses. For CHIKV, both trials used a blood-meal at a titer of 10^6^ FFU/mL. For DENV-1, trial 1 corresponded to a blood-meal at a titer of 10^4.5^ FFU/mL and trial 2 to a titer of 10^5.9^ FFU/mL. 10.1371/journal.pntd.0000706.t004 ###### Mean numbers ± standard deviations of infectious viral particles in saliva of Aedes albopictus Providence co-infected with CHIKV and DENV-1. ![](pntd.0000706.t004){#pntd-0000706-t004-4} Trial 1 (N) Trial 2 (N) ------------------------------------- -------- ------------- ------------- Saliva with both viruses detected CHIKV 16±10 (4) 44±75 (9) DENV-1 74±134 (4) 46±66 (9) Saliva with only one virus detected CHIKV 4±1 (4) 45±64 (8) DENV-1 4±1 (3) 17±0 (2) The saliva of females detected positive for both viruses by IFA on head squashes, 14 days after an infectious blood-meal were titrated. Titers were expressed in FFU (foci forming units). For CHIKV, both trials used a blood-meal at a titer of 10^6^ FFU/mL. For DENV-1, trial 1 corresponded to a blood-meal at a titer of 10^4.5^ FFU/mL and trial 2 to a titer of 10^5.9^ FFU/mL. N refers to the total number of females tested. Discussion {#s4} ========== We report here the ability of Ae. albopictus from La Reunion Island to replicate simultaneously autochthonous strains of DENV-1 and CHIKV provided in the same blood-meal and to deliver both infectious viral particles in their saliva. To our knowledge, such co-infection has never been shown neither under laboratory conditions nor in the field. Lastly, we succeeded in inducing a secondary infection with CHIKV 7 or 13 days after a first infection with DENV-1 virus. CHIKV and DENV are both transmitted by Ae. aegypti and Ae. albopictus, the former being considered the major vector and the latter, the secondary vector. However Ae. albopictus is able to sustain DEN outbreaks in the absence of Ae. aegypti [@pntd.0000706-Gratz1]. Indeed, in the Indian Ocean, Ae. albopictus was predominant in Seychelles and in La Reunion Island where the species took part of DEN outbreaks in 1976-77 [@pntd.0000706-Metselaar1], [@pntd.0000706-Kless1] and in 2005 [@pntd.0000706-Pierre1]. On La Reunion Island, Ae. aegypti populations are scarce and do not exhibit a high anthropophily [@pntd.0000706-Salvan1], [@pntd.0000706-Paupy1] while Ae. albopictus has favored the emergence of a new CHIKV strain harboring a substitution (alanine → valine) at the position 226 of the E1 glycoprotein during the 2005--2006 outbreak. This variant presents high levels of replication in Ae. albopictus [@pntd.0000706-Vazeille1] and a short extrinsic incubation period as the virus could be found in saliva as early as two days after infection [@pntd.0000706-Dubrulle1]. Subsequent outbreaks due to the new CHIKV variant were often related to transmission by Ae. albopictus corroborating an adaptative mutation in response to a requirement for transmission by this species. Moreover, co-infections with both DENV and CHIKV have been detected in some patients from La Reunion Island [@pntd.0000706-World1] and Madagascar in 2006 [@pntd.0000706-Ratsitorahina1]. Co-infections CHIKV-DENV in patients have been first described in 1967 and since the emergence of CHIKV in the Indian Ocean, reports of co-infections are increasing: in the Indian Ocean, as mentioned above, but also in Sri-Lanka [@pntd.0000706-Hapuarachchi1], Malaysia [@pntd.0000706-Nayar1], in India [@pntd.0000706-Chahar1], [@pntd.0000706-Schilling1] where the main vector involved is Ae. aegypti and in Gabon [@pntd.0000706-Leroy1]. Except for the Americas still free of CHIK infection, the geographic range of CHIK is now largely overlapping that of DEN. Furthermore, the emergence of CHIK outbreaks due to the new variant coincided with the recent establishment of Ae. albopictus in Central Africa, in Cameroon [@pntd.0000706-Fontenille1], [@pntd.0000706-Peyrefitte1] and Gabon [@pntd.0000706-Coffinet1], [@pntd.0000706-Peyrefitte2]. In 2007, patients with co-infections DENV-CHIKV were indeed detected for the first time in Africa in Gabon [@pntd.0000706-Leroy1]. Co-infection of a mosquito vector by two different viruses can occur by the way of two successive infectious blood-meals taken on two different viremic hosts or by one blood-meal taken on a co-infected host. We chose to mix both viruses in the same meal since Ae. albopictus females from La Reunion Island were reluctant to feed twice at 7 or 13 days interval on an infectious meal in the BSL-3 conditions. If Ae. albopictus females can be readily fed on an artificial meal in a BSL-3 laboratory it is very difficult to make them take a second artificial meal in the same conditions after 1 or 2 weeks of incubation and even to make then take a first meal if they have been kept more than 48 h in the depression conditions of the BSL-3. Temperature and humidity were optimal, females had the opportunity to lay their eggs if needed, were starved prior blood-feeding as usual. Therefore, we were unable to offer viruses in sequential meals and choose to test the possibility of a superinfection by first injecting one virus (DENV-1) then offering an infectious blood meal with the second virus (CHIKV) 7 or 13 days latter. Even then, we had very few individuals blood-fed but could however detect replication and dissemination of both viruses. Ae. albopictus from La Reunion Island were already known to be a very efficient vector for the new CHIKV variant and able to sustain DEN outbreaks. We demonstrated that replication of both viruses can take place simultaneously and vectors become able to transmit the two viruses in a single bite. In Ae. albopictus, CHIKV infectious particles can be found in saliva from two days after oral infection [@pntd.0000706-Dubrulle1] when this delay is much longer for DENV at least 10 days [@pntd.0000706-Gubler4]--[@pntd.0000706-Salazar1]. We performed the saliva analysis 14 days after feeding the females on a co-infected blood-meal but were not able to detect infectious viral particles in all saliva collected from females presenting a disseminated infection of both viruses. As CHIKV was not detected, when it should have been considering its rapid excretion in saliva [@pntd.0000706-Dubrulle1], females with negative saliva must have been females unable to salivate using our technique [@pntd.0000706-Dubrulle1] or excreting a very small amount of infectious particles that we could not detect due to the low sensitivity of the technique of viral detection. It would be interesting to perform a kinetic study of excretion of infectious particles from females exposed to one or both viruses while controlling the presence of saliva in the collected samples. We performed our study with Ae. albopictus from La Reunion Island with two viral strains isolated in the same area during the DENV-1 outbreak of 2005 and the CHIKV outbreak of 2005--2006. This combination vector/virus fits well with the natural context strengthening the findings of this work. Nevertheless, it seems difficult to infer to other epidemiological contexts as shown by the failure of Rohani et al. [@pntd.0000706-Rohani1] to prove co-infection of Ae. aegypti by CHIKV and DENV. The mechanism of co-infection in Ae. albopictus by two different arboviruses needs to be further investigated and tested using additional trials with different viral strains to determine if this phenomenon is an exception due to the particularly well adapted partners or a quite common mechanism. The still ongoing expansion of this species, particularly in Africa where numerous arboviruses are transmitted, in of particular concern and, if the co-infection was a quite usual phenomenon could have great implication on human health. It should be noted that superinfection is possible in mosquitoes as well as in cells infected with heterologous viruses (i.e., different genus) and not with homologous ones [@pntd.0000706-Beaty1]--[@pntd.0000706-Zou1]. Little is known about the molecular and cellular basis of co-infection which should be explored. Acknowledgments {#s5} =============== We are grateful to the DRASS in the Reunion for providing the ALPROV mosquito strain and to the French National Reference Center for Arboviruses for the E1-226V CHIKV and the DENV-1 strains. We also wish to thank Sara Moutailler for critical reading of the manuscript. We are grateful to Michèle Bouloy and Félix Rey for their support. The authors have declared that no competing interests exist. This work was funded by the ANR EntomoChik drive, the Fondation Duranton de Magny, and the Institut Pasteur (PTR 201). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MV ABF. Performed the experiments: MV LM EM. Analyzed the data: MV ABF. Wrote the paper: MV ABF.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1} =============== Acute scrotal pain includes some urgent diagnoses, such as testicular torsion and incarcerated hernia or omentum. Traditionally, an in-depth interview and color Doppler ultrasonography (US) can lead to an acute diagnosis, although some cases exist where clinicians cannot eliminate emergency etiologies because the case history and physical examination are not performed accurately or the US revealed equivocal imaging. Here, we present a case where a presurgical diagnosis compromised incarcerated omentum, and the operation revealed that intense hydrocele inflammation was the cause of the symptoms. 2. Case Presentation {#sec2} ==================== A 14-month-old boy who had been diagnosed with a hydrocele in the right spermatic cord was referred to our department from the outpatient clinic with a right swollen scrotum. He presented to our hospital for the management of intermittent right scrotum pain for 4 days. He was in grumpy mood, his right scrotum was red and swollen, and the cremasteric reflux was absent ([Figure 1](#fig1){ref-type="fig"}). The urine test showed no white blood cells, and blood tests showed a slight increase in CRP (1.27 ng/mL). The US showed normal testes and epididymis with normal blood flow by the color Doppler US ([Figure 2(a)](#fig2){ref-type="fig"}). On the cranial of the right testis, low-echoic hydrocele with multiseptum was present, and its capsulizing wall was as thick as 3--5 mm ([Figure 2(b)](#fig2){ref-type="fig"}). In the right inguinal canal, omentum was present in the patient\'s processus vaginalis with blood flow by the color Doppler US ([Figure 2(c)](#fig2){ref-type="fig"}). However, we could not completely rule out the partial omental incarceration considering the sensitivity of US, and thus, emergent operative intervention was performed. Under inguinal exploration, processus vaginalis (PV) was present with the omentum inserted, but there was no evidence of necrotic or adherent tissue. PV was not communicating with hydrocele. Surgical findings did not indicate that the omentum itself was the cause of scrotal pain, and we next went on to the testis exploration. Upon opening the right tunica albuginea, the testis was normal colored, and the spermatic cord did not experience torsion. However, the hydrocele wall was ubiquitously adherent with the surrounding tissue and was much thicker and more solid than typical pediatric testicular hydrocele. The enlarged hydrocele wall was removed, and its edges were oversewn in the surgery. Pathology of the hydrocele wall showed edema in the wall, fibrin precipitation, and lymphocyte infiltration ([Figure 3(a)](#fig3){ref-type="fig"}), indicating that severe inflammation caused scrotal pain and redness. 3. Discussion {#sec3} ============= Acute scrotum is characterized by scrotal pain that has rapid or acute onset. The most serious condition in pediatrics is testicular torsion that requires emergent surgery to avoid testicular necrosis. As the sensitivity of the US against testicular torsion is as high as 88.9--100% \[[@B1], [@B2]\] and the US in this case showed clear blood flow into both testes, we had to rule out another disease that required emergent surgery---incarcerated intestine or omentum into the inguinal canal. Even the highest sensitivity of the US for incarcerated inguinal hernia is 91% \[[@B3]\], and thus, the possibility of an incarcerated hernia could not be eliminated. Therefore, we proceeded to emergent surgical intervention. Although the accurate etiology is uncertain, there are three reasons that suggest the painful red scrotum was derived from inflammation of hydrocele wall. First, the pathology findings revealed a thick hydrocele wall, a high number of inflammatory cells, and firm fibrin precipitation, those of which are not usually seen in typical pediatric hydroceles ([Figure 3(b)](#fig3){ref-type="fig"}). Second, there were no other differential diagnoses that could explain the cause of the pain. The other possible etiology was compression of the testicular vessels by a massive hydrocele or edematous incarcerated inguinal hernia \[[@B4]\]. These conditions are rare, with only five cases being reported to date \[[@B4]--[@B8]\]. In each of those cases, the preoperative US clearly showed reduction in testicular blood flow. Based on these findings, the effect of compression could be ruled out in our case. The third observation is the type of onset. Acute scrotum pain due to ischemia is caused by vessel obstruction; therefore, the onset should be sharp and sudden. In the present case, the patient was referred to our department after 4 days of pain initiation. Retrospectively, the time course suggests that the cause is not likely to be ischemic, although we could not be certain that the testis or omentum was intact. All these observations suggest that the cause of the pain was inflammatory changes in the hydrocele. The reason for hydrocele inflammation is uncertain. It is widely known that incarcerated hernias sometimes cause hydroceles. In our case, however, there was no ischemia of the omentum; nor was the hydrocele communicating with abdominal cavity, suggesting other reasons exist for hydrocele inflammation. Clinically, it was reported that omentum often form adhesions to inflamed regions such as the patient\'s PV \[[@B3]\]. In our case, possibly, omentum caused inflammatory changes in the PV and the inflammation spread to the hydrocele wall, although this is just a speculation. Pediatric hydroceles are common and often asymptomatic. To our knowledge, this is the first case of a pediatric hydrocele that caused acute scrotum pain. In retrospect, surgery was not necessary in this case; the intention was to ensure that we were ruling out tissue ischemia, a critical condition. Physicians should not hesitate to perform emergency surgery if the cause of scrotal pain cannot be determined with confidence. The authors appreciate Naomi Kasuga, who is a laboratory technician, for supporting tissue staining. Conflicts of Interest ===================== The authors declare that they have no conflicts of interest. ![The scrotum was red and swollen.](CRIM2018-2862514.001){#fig1} ![(a) Both testes were normal sized with normal blood flow. (b) The hydrocele wall was as thick as 3--5 mm (scale bar = 1 cm). (c) The omentum is inserted into the right inguinal canal. The blood flow was detected.](CRIM2018-2862514.002){#fig2} ![(a) The stroma was thick and edematous with fibrin precipitation and lymphatic cell infiltration. (b) Pediatric noncommunicating hydrocele aged at 1 year as a control. The tissue was not as edematous as in our case, and less lymphatic cells were infiltrated (scale bar = 100 µm).](CRIM2018-2862514.003){#fig3} [^1]: Academic Editor: Bruno Megarbane	{ "pile_set_name": "PubMed Central" }
All relevant data are within the paper and its Supporting Information files. Introduction {#sec001} ============ In agriculture, insecticides are commonly used to control insect pest species. However, pest management is hindered by the increasing level of insecticide resistance in various pest species worldwide. One of the most serious insect pest species to evolve resistance against many insecticides from different chemical classes is the Old World cotton bollworm, Helicoverpa armigera (Hübner). Its pest status is further strengthened by its highly polyphagous nature and extremely wide geographical distribution, with the latter due to its ability to migrate long distances \[[@pone.0197760.ref001]--[@pone.0197760.ref004]\]. Recent work confirmed a major incursion of H. armigera into Brazil, as well as Argentina, Uruguay and Paraguay and it has been detected in Puerto Rico and Florida \[[@pone.0197760.ref005]--[@pone.0197760.ref008]\]. Pyrethroid resistance of H. armigera was first reported in Australian populations in 1983, six years after the introduction of these insecticides \[[@pone.0197760.ref009]\]. Since then, pyrethroid resistance has been reported from various populations globally but, in most cases, the molecular resistance mechanism in H. armigera is yet to be determined. The two major pyrethroid resistance mechanisms common to other species are target site insensitivity, and metabolic resistance by cytochrome P450 monooxygenases (P450s) and carboxylesterases \[[@pone.0197760.ref010], [@pone.0197760.ref011]\]. Target site resistance against pyrethroids, also known as knockdown resistance (kdr), is based on one or more point mutations in the insect sodium channel protein, which is the target of pyrethroid insecticides. In H. armigera, two mutations (D1549V and E1553G) in the sodium channel protein have been described \[[@pone.0197760.ref012], [@pone.0197760.ref013]\]. In the case of metabolic resistance, 18 P450s have been identified as capable of metabolizing one or more pyrethroid insecticides after heterologous expression \[[@pone.0197760.ref011], [@pone.0197760.ref014]--[@pone.0197760.ref019]\]. One of them is the chimeric CYP337B3, recently identified in Australian H. armigera and thought to have arisen through an unequal crossing-over between the parental genes, CYP337B1 and CYP337B2 \[[@pone.0197760.ref018]\]. Joussen et al. \[[@pone.0197760.ref018]\] demonstrated that, after heterologous expression, CYP337B3 is capable of metabolizing fenvalerate, a type II pyrethroid ester, to the nontoxic 4\'-hydroxyfenvalerate. In contrast, the parental enzymes exhibit no detectable fenvalerate metabolism. CYP337B3 has been shown to confer 42-fold resistance towards fenvalerate in Australian H. armigera lines differing solely in the presence of CYP337B3 and its parental genes \[[@pone.0197760.ref018]\]. Similarly, a 49-fold resistance factor was reported by Forrester et al. \[[@pone.0197760.ref020]\] for an Australian field-collected population, indicating that metabolism of fenvalerate by CYP337B3 is the detoxification mechanism in vivo. Recently, CYP337B3 was also identified in field-collected populations from Pakistan \[[@pone.0197760.ref021]\] and China \[[@pone.0197760.ref022]\], confirming its presence outside Australia. Sequence analysis revealed a distinct CYP337B3 allele (CYP337B3v2), in the Pakistani population and three distinct alleles in the Chinese populations (CYP337B3v2, CYP337B3v3, CYP337B3v4) that differ from the Australian allele (CYP337B3v1), by a number of synonymous and non-synonymous SNPs, in addition to variability of the intron sequence and size. This variation may result from different crossing-over positions during recombination of the CYP337B1 and CYP337B2 parental genes, with different alleles of CYP337B1 and CYP337B2 involved in the crossing-over. Such a pattern is indicative of independent origins of the various CYP337B3 alleles, however further work would be required to determine whether this is accurate across a wide geographic distribution of samples. Given the high frequency of this allele around the world, it is possible that the population of H. armigera now established in South America \[[@pone.0197760.ref005], [@pone.0197760.ref023], [@pone.0197760.ref024]\] may be carrying this allele and that the possession of CYP337B3 may confer a selective advantage over local species. Previous studies with allozymes and mitochondrial DNA indicated that very little population structure among H. armigera populations, though more recent work using many thousand genomic markers has shown that the population of H. armigera in Australia can be differentiated from the rest of the world \[[@pone.0197760.ref025], [@pone.0197760.ref026]\]. This lack of structure makes the identification of the CYP337B3 chimeric origin difficult, as well as more generally hampering the identification of source populations in the case of incursions into the New World. However, examining patterns of distribution and diversity among recently selected genes can shed light on such processes. CYP337B3 is a strong candidate for having undergone rapid selection due to its role in pyrethroid resistance, which has a short (approximately 50 year) history across much of the globe \[[@pone.0197760.ref027]\]. Here, we use field-collected populations from 17 countries distributed through Africa, Asia, Oceania, Europe, and South America to first examine the evolutionary history of the chimeric CYP337B3 gene and the global spread in H. armigera. We use global allelic frequency data to show that CYP337B3 is commonly found around the world, with at least six likely independent unequal crossing-over events giving rise to the same resistance phenotype in different regions. We also confirm recent data showing evidence of hybridisation between invasive H. armigera from Brazil and the local Helicoverpa zea population, including the exchange of the CYP337B3 gene \[[@pone.0197760.ref026], [@pone.0197760.ref028]\]. Furthermore, we then use the different CYP337B3 alleles to shed light on the nature of selection of CYP337B3, finding evidence to support recent selection operating on standing variation as opposed to novel mutations arising as a result of pyrethroid application. Materials and methods {#sec002} ===================== Population samples {#sec003} ------------------ A total of 1063 insect samples were used in the current study, including several specimens collected and used in previous work \[[@pone.0197760.ref005], [@pone.0197760.ref025], [@pone.0197760.ref026], [@pone.0197760.ref028], [@pone.0197760.ref029]\]. H. armigera was collected from 17 countries, H. zea from two ([Fig 1](#pone.0197760.g001){ref-type="fig"}; [S1 Table](#pone.0197760.s007){ref-type="supplementary-material"}). All insect samples were collected from agricultural areas with the permission of the landholder or by trapping using species-specific pheromones. No collecting permits were required because both species are common agricultural pests. No endangered or protected species were used in this work. Samples were usually collected as larvae from wild and crop host plants, as adult moths via light/pheromone traps or as larvae after bioassay. For one set of samples from China (n = 240), larvae were collected from colonies derived from eight recently collected field samples (established with 20 to 50 individuals). Collection dates range from 2002 to 2013. Samples were preserved in ethanol (\>95%), in RNAlater, or at -20°C prior to DNA extraction. DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen) or the Fast Tissue-to-PCR Kit (Fermentas, Thermo Scientific). ![Collection localities.\ H. armigera (n = 986) were sampled from 17 countries (black stars): Australia, New Zealand, China, Korea, India, Pakistan, Greece, Spain, France, Burkina Faso, Cameroon, Chad, Ghana, Madagascar, Senegal, Uganda, and Brazil. H. zea were sampled from Brazil (n = 43) and the United States of America (n = 16). Modified from public domain map from Natural Earth (<http://www.naturalearthdata.com>).](pone.0197760.g001){#pone.0197760.g001} The species status of several preserved specimens was confirmed by mitochondrial gene (COI and Cytb) sequencing, either from previous work \[[@pone.0197760.ref005], [@pone.0197760.ref029]\] or, where new samples were available, by amplifying and sequencing the same regions ([S2 Table](#pone.0197760.s008){ref-type="supplementary-material"}). PCR amplification followed the protocols of Behere et al. \[[@pone.0197760.ref029]\] and Tay et al. \[[@pone.0197760.ref005]\]. PCR products were sequenced at Macrogen (Seoul, Korea) and the Biological Resources Facility (Australian National University, Canberra, Australia). Assembly of DNA trace sequences was performed using CLC Genomics Workbench version 8.0. Screening for CYP337B3 {#sec004} ------------------------ Following species identification, samples were screened for the presence of CYP337B3 using the CYP337B2F and CYP337B1R primers described in Joussen et al. \[[@pone.0197760.ref018]\]. Heterozygote/homozygote status was determined through relevant band detection on 1.5--2% agarose gels containing 1% (w/v) of GelRed (Biotium) and visualised under UV light. Initial sequence was generated from these short fragments for a subset of samples following the PCR amplification protocol of Joussen et al. \[[@pone.0197760.ref018]\]. However, in order to generate more information, primers were designed to amplify the intron of the gene ([S2 Table](#pone.0197760.s008){ref-type="supplementary-material"}). In the chimeric CYP337B3 gene, CYP337B1 contributes the intron and several hundred base pairs of coding sequence depending on the allele, while CYP337B2 contributes much of the 5\' end of the coding sequence. Thus, the CYP337B3 intron corresponds to the CYP337B1 intronic sequence. We also screened samples from Brazil (n = 181) and Australia (n = 97) for the reciprocal recombination event using CYP337B1F and CYP337B2R primers ([S2 Table](#pone.0197760.s008){ref-type="supplementary-material"}) but no products were detected with this primer combination. Further amplification of full-length genes and transcripts was performed for selected individuals with representative alleles. PCR products and cloned fragments were sequenced at several institutes, including: Macrogen (Seoul, Korea), Biological Resources Facility (Australian National University, Canberra, Australia), UC Davis Genome Center (University of California, Davis, US), StarSEQ (Mainz, Germany), Bioneer Corporation (Daejeon, Korea) and the Department of Entomology at the Max Planck Institute for Chemical Ecology (Jena, Germany). Assembly and sequence analysis of the PCR products was done using CLC genomics workbench v8.0 and T-Coffee (<http://www.ebi.ac.uk/Tools/msa/tcoffee/>). Resistance bioassays {#sec005} -------------------- To establish the role of CYP337B3 in pythrethroid resistance, fenvalerate field resistant and susceptible individuals were identified by bioassay of field-collected material. Fenvalerate (95.3%) was provided by Sumitomo Chemical (Sydney, Australia) and dissolved in analytical grade acetone to produce a diagnostic concentration (0.125 μg/μL). A 1 μL volume of the discriminating concentration was applied to the dorsal thorax of late 3^rd^/early 4^th^ instar larvae within a weight range of 30-40mg using a 50 μL micro-syringe in a repeating dispenser (Hamilton Company, Reno, NV, USA). Bioassays were maintained for 7 d at 25°C with a 14:10 (L:D) h cycle. Relative humidity (RH) was not controlled. Dead and moribund larvae were collected and described as sensitive by using one or more of the following criteria: larvae unable to demonstrate coordinated movement when prodded; paralysis of prolegs; larvae very slow to right themselves (time exceeding 3 s). Larvae that were actively feeding and developing normally were described as resistant. Population genetic analysis {#sec006} --------------------------- The elimination of variation in regions linked to an adaptive allele on one or even a few haplotypes is referred to as a "selective sweep" \[[@pone.0197760.ref030]\]. This process will lead to low nucleotide diversity (π) across the region of the genome under selection as well as an excess of rare variants as highlighted by Tajima's D \[[@pone.0197760.ref031]\]. This statistic identifies deviation from a neutral model of evolution, whereby a negative score signifies purifying selection or population expansion that one would expect following a selection event. To provide evidence that the CYP337B3 gene is under selection in H. armigera, we searched for genomic signals of a recent selection event using high-throughput sequencing data. DNA samples from 12 Australian H. armigera homozygous for CYP337B3v1 were sequenced on an Illumina HiSeq. Briefly, Nextera libraries were produced following the manufacturer's instructions and sequence was generated as 100 bp PE reads (Illumina HiSeq 2000, Biological Resources Facility, Australian National University, Canberra, Australia). Raw reads were aligned to the bacterial artificial chromosome (BAC) 33J17 (JQ995292.1) sequence using BBMAP v33.43 (<http://sourceforge.net/projects/bbmap/>), trimming reads when quality in at least 2 bases fell below Q10. We did not align reads to the reference H. armigera genome sequence, which contains only CYP337B1 and CYP337B2 but not CYP337B3 \[[@pone.0197760.ref032]\]. Only uniquely aligning reads were included in the analysis, to prevent spuriously inferring evolutionary processes occurring independently on each BAC. Outputted BAM files were sorted before duplicate reads were removed and files were annotated with read groups using Picard v1.138 (<http://picard.sourceforge.net>). The reference sequences were indexed using Samtools \[[@pone.0197760.ref033]\]. UnifiedGenotyper in GATK v3.3--0 \[[@pone.0197760.ref034]\] was used to estimate genotypes across all individuals simultaneously, implementing a heterozygosity value of 0.01. Genotypes annotated as "LowQual" were removed prior to subsequent analysis with VCFtools v0.1.12b \[[@pone.0197760.ref035]\], whereby π and Tajima's D were calculated in sliding windows of 2500 bp that progressed by 1250 bp across biallelic sites of 33J17. Results were plotted in R v3.1.2 using ggplot2 v1.0.1 \[[@pone.0197760.ref036]\], while gene annotations were derived via tblastx \[[@pone.0197760.ref037]\] and visualised with CLC Genomics workbench v8.0. To examine potential mechanisms underlying the origin of the CYP337B3 alleles, an intron dataset containing sequences for representative CYP337B1 and CYP337B3 individuals (since the chimeric CYP337B3 intron comes from the CYP337B1 parental gene; see above) was created. MAFFT ver. 7.182 \[[@pone.0197760.ref038]\] was used to align these sequences using the -linsi option. Subsequently, IQ-TREE ver. 1.3.0 \[[@pone.0197760.ref039]\] was used, first in the -m TESTNEWONLY mode to determine the appropriate nucleotide substitution model, and then in full mode to generate a maximum likelihood tree. Subsequently 10,000 bootstrap replicates were carried out, and the bootstrap values transferred from the consensus tree to corresponding nodes of the maximum likelihood tree. BEAST ver. 2.3.0 \[[@pone.0197760.ref040]\] was used for the intron-only dataset to test whether the CYP337B3 allele is monophyletic in origin (i.e., has arisen once from a CYP337B1/CYP337B2 cross-over and subsequently diverged independently). Two analyses were performed--one, in which monophyletic taxon sets for each of the CYP337B1 and CYP337B3 introns was enforced, and the other in which no monophyly was assumed. In BEAST, all runs used the evolutionary model identified in IQ-TREE (HKY + G4), a constant population coalescent tree prior, and a chain length of 10 x 10^6^. Other coalescent tree priors were also tested, but these did not quantitatively impact the results. BEAST results were examined in Tracer to confirm convergence and check ESS values (all exceeded 1,500). The likelihood of each analysis was then compared using Bayes Factors and the ACIM model, with 1,000 replicates, in Tracer ver. 1.6 \[[@pone.0197760.ref041]\]. Results {#sec007} ======= A total of 1063 insect samples was collected from 18 countries around the world ([Fig 1](#pone.0197760.g001){ref-type="fig"}). The identification of H. armigera was initially based on morphological assessments by collectors and subsequently confirmed with the mitochondrial markers, cytochrome oxidase I and cytochrome b \[[@pone.0197760.ref029]\]. Any individuals not identified as H. armigera by mitochondrial DNA were excluded from subsequent analysis (H. armigera n = 999, H. zea n = 59, Chloridea (Heliothis) virescens n = 5). A PCR assay using one primer from each of the two parental genes CYP337B1 and CYP337B2 was used to identify the presence of the chimeric CYP337B3 gene. CYP337B3 was extremely common in field collected H. armigera throughout its native range ([Fig 2](#pone.0197760.g002){ref-type="fig"} and [S1](#pone.0197760.s001){ref-type="supplementary-material"}--[S5](#pone.0197760.s005){ref-type="supplementary-material"} Figs). From a total of 999 individuals tested, CYP337B3 was present in 969 (97%) and only 30 (3%) individuals were homozygous for CYP337B1 or CYP337B2. 878 individuals were CYP337B3 homozygotes (88%) and 91 were heterozygotes, meaning they carried one chromosome carrying CYP337B3 and the other chromosome with the two parental genes CYP337B1 and CYP337B2 (9%). In some countries, almost all individuals were at least heterozygous for the CYP337B3 gene, but homozygous CYP337B3 individuals were the most common everywhere. In India, Pakistan, Korea, China, and Australia, almost all individuals were homozygous for CYP337B3, while heterozygotes (CYP337B1, CYP337B2 and CYP337B3) at the CYP337B3 locus were more common in European and African samples ([Fig 2](#pone.0197760.g002){ref-type="fig"} and [S1](#pone.0197760.s001){ref-type="supplementary-material"}--[S5](#pone.0197760.s005){ref-type="supplementary-material"} Figs). ![The frequency of CYP337B genotypes in selected geographic regions.\ For a more detailed breakdown see [S1 Fig](#pone.0197760.s001){ref-type="supplementary-material"} through [S5 Fig](#pone.0197760.s005){ref-type="supplementary-material"}.](pone.0197760.g002){#pone.0197760.g002} The PCR product of CYP337B3 used in the screening PCR reaction \[[@pone.0197760.ref018]\] was sequenced from positive individuals (alleles n = 460) and a pattern of single nucleotide polymorphisms was identified. By comparing this short (355 bp) sequence between individuals, 6 alleles (CYP337B3v1-2, 5--8) were identified worldwide ([Fig 3](#pone.0197760.g003){ref-type="fig"}). An allele originally identified in an Australian (Toowoomba) strain \[[@pone.0197760.ref018]\] was found only in Australia in the current study, where it was the most common allele (v1: n = 91), and v2 the next most common allele (v2: n = 23). In Asian samples, v2 was the most common allele (n = 203) and two other rare alleles were identified (v7 and v8; n = 1 each) in addition to v3 and v4 identified by Han et al. \[[@pone.0197760.ref022]\]. Three alleles were found in Africa including two that were new (v2: n = 11, v5: n = 79, v6: n = 8). In the European samples, we identified an approximately 50:50 ratio of CYP337B3v2 and v5 (v2: n = 20; v5 n = 17). In addition, a proportion of samples from Australia (n = 11), Africa (n = 2), and Europe (n = 7) appeared from the sequencing traces to be heterozygous for different CYP337B3 alleles. Overall, the various CYP337B3 alleles appeared to show distinct patterns for each region, suggesting different alleles are present for pyrethroid resistance around the world. ![Frequencies of the different alleles of CYP337B3 (B3vX) in different geographic regions.\ Africa = Burkina Faso, Cameroon, Chad, Ghana, Madagascar, Senegal, Uganda; Australasia = Australia and New Zealand; Europe = Greece, France and Spain. n = the number of alleles sequenced, homozygotes and heterozygote alleles summed.](pone.0197760.g003){#pone.0197760.g003} Sequencing of the entire coding sequences of all variants revealed a number of non-synonymous SNPs resulting in different predicted proteins when compared to CYP337B1v1: 3 in v2 \[[@pone.0197760.ref021]\], 3 in v5, 8 in v6, 9 in v7, and 11 in v8. All variants of the chimeric gene appear to show a similar pattern with slightly different cross over points between CYP337B1 and CYP337B2 during the formation of CYP337B3 and different intron sequences and sizes that derived from CYP337B1 determined by sequencing genomic DNA ([Fig 4](#pone.0197760.g004){ref-type="fig"}; [S6 Fig](#pone.0197760.s006){ref-type="supplementary-material"}; [S3 Table](#pone.0197760.s009){ref-type="supplementary-material"}). This suggests that multiple origins of the same functional phenotype from different genetic backgrounds define the development of this chimeric gene. ![A schematic of the different crossing-over points and intron sizes observed in the chimeric CYP337B3 gene.](pone.0197760.g004){#pone.0197760.g004} All of the 93 H. armigera individuals collected in Brazil were positive for CYP337B3, and all except one were homozygous. The majority of CYP337B3 alleles detected in Brazilian H. armigera were identical to CYP337B3v2, which is predominantly found in Asia. However, the single heterozygous individual was carrying an African allele (CYP337B3v5). This suggests that Asia may be the most likely source for the Brazilian incursion of H. armigera, however, other source populations are also possible, as is the occurrence of multiple incursions. As well as examining H. armigera, we also examined H. zea from Brazil (n = 59) and we were able to amplify CYP337B3 from one individual from Goias while from the others (n = 23) only CYP337B1 provided consistent results. DNA from the suspect individual ([S1 Table](#pone.0197760.s007){ref-type="supplementary-material"}) was re-extracted twice in a different laboratory, using new reagents (extraction kit, primers, PCR reagents) and PCR was performed for the CYP337B3 gene, intron and the mitochondrial marker COI. Extraction blanks were performed and were negative for both the COI sequence and the CYP337B3. Sequencing of both fragments from the three different extractions confirmed the presence of the CYP337B3v2 allele (as found in the Brazilian H. armigera) in an individual with a mitochondrial sequence from H. zea. Bioassays for pyrethroid resistance on H. armigera from cotton growing areas in eastern Australia using the discriminating dose clearly showed a higher frequency of homozygotes for CYP337B3 in survivors (100% CYP337B3 homozygous; n = 26) than in dead or dying insects (9 CYP337B3 homozygotes, 2 heterozygotes and 1 homozygous CYP337B1/B2 individual; n = 12), although a subset of the dead or dying caterpillars was actually positive for CYP337B3. In addition to bioassayed individuals, samples were collected from crops in Kununurra, Ord River Irrigation Scheme, Western Australia seven days after a field application of fenvalerate on cotton. Survivors were found to possess CYP337B3, with a 100% homozygous frequency (n = 7). Insects collected from the same region on unsprayed crops (cotton, chickpea and chia) showed a lower frequency of CYP337B3 (2 homozygote CYP337B3, 1 heterozygote and 3 homozygote CYP337B1/B2 individuals; n = 6). Although these very small samples from field populations are insufficient for claims of statistical significance, they are consistent with the significant genotypic differences previously found in the Toowoomba laboratory population derived from Australia \[[@pone.0197760.ref018]\]. Using individuals from Australia that were homozygous for the CYP337B3v1 allele, high throughput sequencing data was aligned to the BAC clone 33J17 which was the original source of the CYP337B3v1 identification and originates from the Australian Toowoomba strain \[[@pone.0197760.ref018]\]. Of the 100,377 genotypes called across 33J17, 32,381 remained after "LowQual" SNPs were removed. Both π and Tajima's D were variable, however values were decisively differentiated where CYP337B3 lies; nucleotide diversity reaches its lowest point at the CYP337B3 locus (0.00057) relative to the BAC-wide mean at 0.0179 ([Fig 5](#pone.0197760.g005){ref-type="fig"}; [S4 Table](#pone.0197760.s010){ref-type="supplementary-material"}). This is the expectation following a "hard" selective sweep, whereby adaptive alleles appear as a single haplotype before increasing in frequency \[[@pone.0197760.ref030]\]. Tajima's D is also lowest across the region where CYP337B3 is located (-2.18) and points towards the process of purifying selection. It is worth noting that the BAC-wide mean value for Tajima's D is -0.43; such a low negative value may signify that other evolutionary processes have also acted upon this region \[[@pone.0197760.ref042]\]. While this variability is likely a result of small sample sizes and relatively small sliding window values, the small size (\~ 5 kbp) of the region around CYP337B3v1 may signify ongoing high recombination rates and gene flow \[[@pone.0197760.ref043], [@pone.0197760.ref044]\]. ![Selective sweep among Australian H. armigera homozygous for CYP337B3v1.\ (a) Nucleotide diversity (π) and (b) Tajima's D calculated across sliding windows of the 33J17 BAC (JQ995292.1), with the red dashed line showing the average value of π (0.0179). Locations of gene bodies are indicated between the plots, with those in red identified as potential reverse transcriptases and those in yellow as exons of CYP337B3v1.](pone.0197760.g005){#pone.0197760.g005} To establish the potential mechanisms that underlie the creation of the chimeric gene and account for its global distribution, we sequenced the intron of CYP337B3. This region should contain more polymorphisms and allow a better diagnosis of the frequency of the different genotypes. As for the coding region data (above), we found a pattern of eight alleles in the intron data. Introns sequenced from B3v3, v4 and v8 are very similar and differ only by two SNPS whereas introns from B3v1, v2, v5, v6 and v7 show more significant differences in both length and sequence ([Fig 4](#pone.0197760.g004){ref-type="fig"}). Individuals that were homozygous for the CYP337B3 gene were predominantly homozygous at the allele level, but as in the coding region, heterozygotes were again observed in the sequence traces of a small number of samples from Africa, Australia and Europe. Cloning and sequencing of the heterozygous PCR products showed that, in Australia, v1/v2 (Australia/Asia) heterozygotes were present, while in Africa, there were heterozygotes of v2/v5 (Asia/Africa) and v5/v6 (both African). European heterozygotes possessed v2/v5 (Asia/Africa). Phylogenetic relationships were inferred for the intronic region of a subset of representative CYP337B1 and CYP337B3 sequences. If the chimeric CYP337B3 were due to a single cross-over event followed by subsequent divergence of allelic variations, then the phylogenetic estimates should show a monophyletic clade with the CYP337B3 clade nested within the CYP337B1 clade. Our tests with BEAST indicated that this was not the case; specifically Bayes Factor (BF) analysis found overwhelming support for a paraphyletic origin of CYP337B3 (BF score for paraphyly vs. monophyly = 494.412, with BF \> 10 indicating very strong evidence against the null model of monophyly). This does not rule out the possibility of some subsequent recombination between derived CYP337B3 alleles and either parent gene. [Fig 6](#pone.0197760.g006){ref-type="fig"} presents the results of the maximum likelihood phylogenetic analysis, which also suggests multiple origins of the various CYP337B3 alleles. Although bootstrap support values for some nodes of the tree are quite low, several well-supported relationships are present, and these are consistent with multiple geographic origins for the CYP337B3 alleles. For example, CYP337B3v1 (found only in Australia) is most closely related at the sequence level to a CYP337B1v1 allele from Australia. Furthermore there is a cluster of closely related but rare CYP337B3 alleles found in China (CYP337B3v3, v4, and v7) which could indicate local origins for those alleles. Thus, while some CYP337B3 alleles may be recently derived from local CYP337B1 alleles, the overall signal indicates that multiple unequal crossing-over events are responsible for the geographic distribution of chimeric CYP337B3 while gene flow as well as some intergenic recombination may also have played a role producing the current geographic distribution patterns of CYP337B3 alleles. ![Maximum likelihood phylogeny of CYP337B1 and CYP337B3 alleles.\ The maximum likelihood tree was generated in IQ-TREE with the HKY+G4 model of sequence evolution, using representative CYP337B1 and CYP337B3 intron sequences. Bootstrap values from 10,000 replicates are given on tree nodes. The eight CYP337B3 alleles are indicated with labels (v1:v8), different CYP337B1 alleles are indicated by B1- followed by an allele ID and the country of origin. Sequences and related NCBI identifiers can be found in [S3 Table](#pone.0197760.s009){ref-type="supplementary-material"} and [S1 File](#pone.0197760.s011){ref-type="supplementary-material"}.](pone.0197760.g006){#pone.0197760.g006} We performed an additional calculation to determine the molecular rate that would be required to produce the measured CYP337B3 allelic diversity if it had all arisen in the last \~50 years. Even using the rapidly evolving mitochondrial divergence rate of 1.5--2.5% divergence per million years as a baseline \[[@pone.0197760.ref045]\], we found that a 1000-fold increase in this rate would be required to place all CYP337B3 divergence events inside a 50-year time window. This rate is so high, it suggests that rather than recently arising through de novo mutation in response to pyrethroid applications and then spreading around the world, cytochrome P450-mediated pyrethroid resistance is most likely the result of selection acting over the last 50 years on existing genetic variation present in the populations at low frequency. Discussion {#sec008} ========== Helicoverpa armigera is one of the world's most destructive pests and has developed resistance to numerous pesticides over the years \[[@pone.0197760.ref027]\]. Globally, pyrethroids are one of the most commonly used insecticides and resistance occurred rapidly at a global scale after the first use of pyrethroid insecticides in the late 1970s \[[@pone.0197760.ref027]\]. A number of mechanisms have been proposed for this resistance, including target site insensitivity \[[@pone.0197760.ref027], [@pone.0197760.ref046], [@pone.0197760.ref047]\], and metabolism by carboxylesterases \[[@pone.0197760.ref048], [@pone.0197760.ref049]\] and P450s \[[@pone.0197760.ref050]--[@pone.0197760.ref056]\]. Recent work suggests that a chimeric gene, CYP337B3, may play an important role in cytochrome P450-mediated pyrethroid resistance and we set out to examine the frequency of this gene worldwide \[[@pone.0197760.ref018], [@pone.0197760.ref021]\]. We detected the chimeric CYP337B3 gene in every population examined and found a total of eight allelic variants of CYP337B3 globally. Of the eight global alleles identified, one was found only in Australia (CYP337B3v1), one was dominant across Asia (India, Pakistan, China, Korea), and Brazil (CYP337B3v2), four were rare and uniquely found in China (CYP337B3v3, v4, v7 and v8) and two were predominantly African (CYP337B3v5, v6). European samples showed a mixture of alleles (CYP337B3v2, v5). CYP337B3v2 was the most widespread allele, identified on every continent examined but was present at the highest frequencies in Asia and Brazil. One issue in using CYP337B3 as a population marker is that diversity in CYP337B3 is not new diversity, but rather a re-shuffling of diversity already present in the parental CYP337B1 and CYP337B2 genes. However, the presence of multiple alleles at the CYP337B3 locus suggests that this chimeric gene has formed independently several times over the evolutionary history of the locus. In the overall patterns, we found that three alleles, v1, v2 and v5, together account for \~98% of all detected CYP337B3 alleles. The geographic distribution of these common alleles may suggest older recombination events in Australia, Asia, and Africa producing the common alleles and potentially more recent additional events in Asia and Africa producing the rare alleles. The alternative explanation that a single event has occurred, and then been distributed globally by migration and subjected to subsequent local recombination, is much less likely as it is not supported by the phylogeny or the molecular rate analysis. Indeed, our phylogenetic analysis found strong support for a paraphyletic origin of CYP337B3 rather than the alternative of a single unequal cross-over event followed by subsequent allelic divergence. In addition, our calculations suggest that a 1000-fold increase in a commonly accepted baseline divergence rate (1.5--2.5% divergence per million years) would be required to place all CYP337B3 divergence events inside a 50-year time window \[[@pone.0197760.ref045]\]. Taken together, the allele frequencies and phylogenetic analysis suggest that, rather than recently arising through de novo mutations in response to pyrethroid application, cytochrome P450-mediated pyrethroid resistance is most likely the result of recent selection acting on existing genetic variation maintained in the population at low frequency. Our selective sweep analysis supports this scenario, with patterns of estimated π and Tajima's D across the 33J17 BAC indicating that the CYP337B3 locus has undergone recent selection, at least in Australia and with regard to CYP337B3v1. Recent work confirms that this can also be shown for the other alleles \[[@pone.0197760.ref026]\]. Further work will reveal how the diversity in this one region compares to the rest of the genome. There are a number of examples in the literature where the same insecticide resistance mechanisms have developed independently in different locations. In particular pyrethroid resistance (perhaps because of the short period between the introduction and the development of resistance) associated with identical point mutations in the voltage-gated sodium channel has developed in different haplotypes with different geographical origins in mosquitoes and houseflies \[[@pone.0197760.ref057], [@pone.0197760.ref058]\]. In the sheep blowfly, the same mutations in an esterase gene conferring organophosphate resistance were identified in very different haplotypes \[[@pone.0197760.ref059]\]. Although further work incorporating whole genome scale markers will be necessary to determine the origin of H. armigera in the New World and global patterns of movement in this species \[[@pone.0197760.ref026], [@pone.0197760.ref028]\], we gain insight into some of these processes here. The vast majority of the Brazilian insects examined possessed the allele CYP337B3v2 found predominantly in Asia. However, we also detected the CYP337B3v2 allele in other regions of the world, and so it is possible that the Brazilian populations of H. armigera could have come from localities where the CYP337B3v2 allele also exists. For example, the closest native population of H. armigera to the New World is in West Africa and it is possible that the New World H. armigera are the result of a similar incursion from West Africa. However, only Asia has close to the same proportion of CYP337B3v2 alleles. A recent survey \[[@pone.0197760.ref008]\] showed that the pattern of mitochondrial markers in H. armigera populations collected from southern Brazil, Argentina, Uruguay and Paraguay is very different from H. armigera populations from central and northern regions of Brazil, suggesting multiple incursions of this global pest into the New World \[[@pone.0197760.ref060]\]. It would be interesting to examine the CYP337B3 locus in these individuals to see if they share the same pattern as found in Brazil. Recently, evidence for other pyrethroid detoxicative mechanisms in addition to CYP337B3 has been found in China \[[@pone.0197760.ref022]\], and our own data showed some mortality in CYP337B3- homozygous individuals. This might suggest that possession of the CYP337B3 is not enough for resistance alone and other mechanisms related to expression or other genes may exist. However, as the resistance based on CYP337B3 is metabolic, it is not unusual that a high dose of insecticide will kill even homozygous individuals, for example if the enzyme is not abundant enough to decrease the amount of insecticide at its target site beneath the toxic threshold. Furthermore, we clearly see evidence of conservation of the CYP337B3 alleles at this locus across the globe and we would not expect the observed level of conservation in the absence of selection, particularly in the non-coding regions. Indeed, when we examined individuals from Australia we found evidence of a selective sweep around the gene, indicating maintenance of the allele at this locus. Throughout its native range, phenotypic pyrethroid resistance is very common in H. armigera and it is of course possible that other or additive resistance mechanisms are at work in populations around the world. The arrival of H. armigera in the New World represents a significant expansion of its geographc range. It remains to be seen how large the problem H. armigera will become but, clearly, arriving with a resistance gene already in place would provide H. armigera a selective advantage over local susceptible populations of pests such as H. zea and Chloridea (Heliothis) virescens. One further complication is that H. zea has been shown to interbreed with H. armigera in the laboratory to produce viable offspring \[[@pone.0197760.ref061]--[@pone.0197760.ref063]\] and there is evidence of hybridisation in Brazil using genome wide markers \[[@pone.0197760.ref028]\]. In the current study, we applied the same PCR reaction conditions to individuals identified as H. zea from Brazil and we able to amplify CYP337B3v2 from one individual. Clearly this one result is not enough to suggest that hybridisation is widespread between the two species but the detection of a presumably highly selectable gene indicates that this is something that should be considered a risk. While it is unclear what risk hybrids may pose, it is certain that an exchange of resistance and host association genes could lead to significant problems in control. The fact that the presence of the chimeric resistance gene CYP337B3 can be ascertained by a simple screening PCR enables the rapid determination of susceptible vs. pyrethroid-resistant populations, which could lead to improved pest management strategies and the detection of hybrid individuals in the field. This work highlights a case of resistance alleles in H. armigera evolving on a local scale in a global context and provides an example of a species invading new continents, already able to overcome some of the control measures deployed for other species. We also show that resistance genes can arise locally and then mix on a global scale as a result of either natural movement or human-mediated transport. The implications for H. armigera and potentially H. zea as species are unclear but what is not in doubt is that the pesticide resistance selection pressure on H. armigera, in terms of numbers exposed, has dramatically increased with this incursion. This is true not just for conventional pesticide resistance but also for transgenic insecticidal traits. Transgenic cotton and especially corn and soybean which express Bt toxins are grown over a far larger area in the New World and the management of Bt and conventional pesticide resistance should be a priority. Ultimately, we grow the same crops around the world, control pests using the same chemistries, and transport commodities through global trade routes, so it is not a surprise that we are developing global, well-adapted pests. Our paper not only sets the scene for further work in this area but also highlights the value of considering biosecurity threats in terms of both the actual pest as well as its resistance profile (both phenotype and genotype). Supporting information {#sec009} ====================== ###### The frequency of CYP337B3 on the Asian Sub-continent by region. Northern, Central and Southern represent general geographic regions of India. (TIF) ###### Click here for additional data file. ###### The frequency of CYP337B3 in selected African countries. Homozygous and heterozygous frequencies are shown. (TIF) ###### Click here for additional data file. ###### The frequency of CYP337B3 in China and Korea. \Larvae from lab colonies derived from each field collection (20--50 individuals) and maintained at a population size of 300--500 adults per generation without exposure to any insecticide for 3--5 generations. (TIF) ###### Click here for additional data file. ###### The frequency of CYP337B3* in Australia and New Zealand. Homozygous and heterozygous frequencies are shown. (TIF) ###### Click here for additional data file. ###### The frequency of CYP337B3 in Europe. Homozygous and heterozygous frequencies are shown. (TIF) ###### Click here for additional data file. ###### Alignment of the crossover points for the various CYP337B3 alleles. CYP337B1 is in red, CYP337B2 is green and the boxed sequence represents crossover point for each CYP337B3 allele. (TIF) ###### Click here for additional data file. ###### Collection sites of H. armigera and H. zea from 18 different countries. (DOCX) ###### Click here for additional data file. ###### PCR primers used in this work. (DOCX) ###### Click here for additional data file. ###### Sequence identifiers, gene names, origin and NCBI numbers used for the maximum likelihood estimate of phylogenetic relationships. (DOCX) ###### Click here for additional data file. ###### Sequencing and alignment statistics for the 12 individuals used to demonstrate the selective sweep around CYP337B3v1. (DOCX) ###### Click here for additional data file. ###### Intron sequence data used for the maximum likelihood estimate of phylogenetic relationships. (DOCX) ###### Click here for additional data file. We thank George Mironidis (IMBB-FORTH, Greece), Athanase Badolo (University of Ouagadougou), Juil Kim (National Institute of Horticultural and Herbal Science, Korea), and Muhammad Ashfaq (National Institute for Biotechnology and Genetic Engineering, Pakistan; now: Biodiversity Institute of Ontario, Canada) for sampling and providing insects for this study from Greece, Burkina Faso, Korea, and Pakistan, respectively; Hae Kyung Kwon (National Institute of Horticultural and Herbal Science; Korea) for technical assistance; and Domenica Schnabelrauch (MPI for Chemical Ecology, Germany) for sequencing. In Brazil, we thank Miguel Soria and Cecelia Czepak for the collection of specimens. In Northern Australia, we also thank Penny Goldsmith from Ord Co. [^1]: Competing Interests:The authors have declared that no competing interests exist. [^2]: Current address: Department of Crop and Soil Science, Oregon State University, Corvallis, Oregon, United States of America [^3]: Current address: Centre for Animal Sciences and Fisheries, University of Swat, Odigram, Khyber Pakhtunkhwa, Pakistan	{ "pile_set_name": "PubMed Central" }
Due to confidentiality restrictions and personal data protection, data are available upon request. Requests for data may be sent to: <mila.vulchanova@ntnu.no>/<sobh.chahboun@ntnu.no>. Introduction {#sec001} ============ Figurative language is a pervasive phenomenon in daily communication, and is reflected in a range of expressions, such as metaphors, idioms, clichés, irony, jokes. Such expressions vary in structure and degrees of transparency and familiarity, and can fluctuate from a single word to a complex sentence. A common feature that unites the categories of figurative language is the fact that reference in such expressions is indirect. More importantly, the interpretation of figurative expressions is non-compositional (non-literal). This deviation from compositional on-line computation and the heterogeneity of figurative uses makes figurative language more demanding in terms of processing \[[@pone.0168571.ref001]\]. In language development, the ability to interpret figurative language develops gradually, and comprehension of such expressions can be especially demanding in deficit populations where language is affected. Individuals with High Functioning Autism (HFA) are distinguished by relative preservation of language and cognitive skills. Intelligence in this group is at normal or above average levels, and quite often such individuals demonstrate specific advantages in certain cognitive areas. The term high functioning captures the level of adequate cognitive functioning in that group, while often, when referring to the level of language competence (i.e., intact structural language), they are described as highly verbal. However, problems with pragmatic language have been consistently reported across the autistic spectrum, even when structural language is intact \[[@pone.0168571.ref002]--[@pone.0168571.ref006]\]. In addition, some studies have found evidence of weaknesses in the figurative and idiomatic language domains, contrasting with clear strengths in areas of grammar \[[@pone.0168571.ref004], [@pone.0168571.ref006]\]. Several behavioral and imaging studies have shown that there are many issues to take into account regarding the interpretation of figurative language. It depends on the type of expression, its degree of compositionality/transparency, its linguistic structure, its source domain of knowledge, and not least, its novelty/conventionality. Thus, the main question is how these factors impact on the comprehension of the categories of interest (idioms, metaphors, irony), and whether the dissociations between structural language skills and figurative language competence reported in highly verbal individuals with autism are still maintained with a careful choice of experimental paradigm. The current study is designed to answer these questions by focusing on figurative language processing in highly verbal individuals with autism. The main goal was to investigate whether individuals with autism and preserved language competence understand idiomatic expressions and novel metaphors at the same level as carefully matched controls, and, specifically, whether in this process they benefit from visually presented context and can integrate information from multiple sources (e.g. visual modality and oral language presented simultaneously). Idiomatic expressions and their processing {#sec002} ------------------------------------------ The mechanisms underlying the processing of figurative language have been widely debated. Idiomatic expressions are one of the most common and popular forms of figurative language used in daily interaction. Idioms are multi-word expressions whose interpretation is non-literal, meaning it cannot be derived by using regular compositional processes. As such, they require the language user to go beyond what is stated literally in order to infer the communicative intent of the speaker \[[@pone.0168571.ref007]\]. In addition, idioms have attracted attention both on theoretical and empirical grounds. The reason is that they cannot be placed immediately within a grammar-lexicon dichotomy of language. On the one hand, it can be assumed that they belong in, and are retrieved from, the mental lexicon, following the assumption that they need to be acquired and stored in the same way as lexical items. On the other, their (complex) structure suggests that they may be processed like structures generated by grammar \[[@pone.0168571.ref008]--[@pone.0168571.ref009]\]. Idiom processing has been addressed extensively in recent research, but mainly, two different types of theoretical approaches have emerged regarding the processing and comprehension of idioms \[[@pone.0168571.ref010]\]. Representative of the first type of approach is the Lexical Representation Hypothesis. Originally proposed by Swinney & Cutler \[[@pone.0168571.ref011]\] (see also Bobrow & Bell \[[@pone.0168571.ref012]\] and Gibbs \[[@pone.0168571.ref013]\]), this hypothesis claims that idioms are stored as lexical items, and understanding an idiom involves two parallel processes, a retrieval process, and a literal compositional computation process based on decomposing every element separately \[[@pone.0168571.ref008]--[@pone.0168571.ref010]\]. Further, it is assumed that the retrieval process is faster, while the compositional process demands longer time. In a further development, Hamblin & Gibbs \[[@pone.0168571.ref014]\] suggest that idioms are essentially decomposable. For these authors, idiomatic interpretation rests on identifying the individual constituents, mainly because of the decomposability of the idiomatic expressions. Along similar lines, some authors have claimed that the processing and understanding of idioms cannot be reduced to lexical access or lexical retrieval only, and an additional interpretation process might be necessary \[[@pone.0168571.ref015]--[@pone.0168571.ref017]\]. Thus, Gibbs \[[@pone.0168571.ref016]\] suggests that idioms invoke complex meanings which are motivated by independently existing conceptual metaphors which are part of everyday thought. The second type of approach, labelled the Configuration Hypothesis, aims to capture the dynamic process involved in idiom interpretation. The authors in favour of this position argue that idioms are represented in a concrete and distributed way and are processed as complex expressions \[[@pone.0168571.ref015]\]. This latter approach can be described as a "compositional" approach whereby each constituent is assumed to contribute semantically to the idiom interpretation. In a comprehensive review of approaches to idioms, Titone & Connine \[[@pone.0168571.ref007]\] suggest that either view fails to account adequately for findings in research or capture the nature of idioms. These authors thus suggest a "hybrid" model, which acknowledges both the arbitrary "word" nature and the compositional aspect of idioms. Their model takes into account an important parameter on which idioms vary: degree of decomposability. A similar view, albeit from a slightly different perspective, is defended in Sanford \[[@pone.0168571.ref018]\] who describes idioms as the intersection of conceptual and syntactic schemas. Another factor which has been shown to impact on idiom processing is the mode of presentation. Tabossi, Fanari & Wolf \[[@pone.0168571.ref019]\] found that idiom identification differs depending on whether the expression was presented auditorily or visually (as text). This clearly implies that the different modalities in which the idioms are presented, demand different forms of processing. Another related figurative and formulaic language category are proverbs. Proverbs are fixed sentential expressions which invoke well-received wisdom, social or moral norms \[[@pone.0168571.ref020]--[@pone.0168571.ref021]\]. They reflect themes from everyday life and often refer to general principles of reasonable human behaviour and action \[[@pone.0168571.ref021]--[@pone.0168571.ref022]\], and, as such, may seem to express self-evident truths. However, like idioms, proverbs display a wide range of interpretations depending on the type of mechanism involved: they can be metaphorical, as You can't judge a book by its cover, or based on personification, hyperbole or a paradox (No news is good news). Unlike idioms, however, proverbs are typically based on some kind of cause-effect reasoning (Once bitten, twice shy) or reasoning by analogy, but may also reflect a certain type of cultural practice (Never look a gift horse in the mouth). Although proverbs may have prototypical figurative meanings, they are very much vital due to the necessity to draw precise inferences in each specific context of encounter \[[@pone.0168571.ref021]\]. Thus, the correct interpretation of a proverb requires the ability to navigate between their concrete, literal meaning and their figurative meaning by establishing a metaphorical connection between these two interpretations. These properties of proverbs make them an interesting category to study, especially in comparison to idioms. In the current study we refer to this type of figurative expression as "instructive expressions" reflecting their communicative role in language use. Unlike idioms and proverbs, metaphors are a category which is often less conventionalised and more transparent. It is often assumed that metaphors are a device which maps concepts from two distinct domains, which are not otherwise conceptually linked to each other \[[@pone.0168571.ref023]\]. Like idioms, however, metaphors vary in degree of novelty/conventionality which affects the way they are processed. In the current study we use novel metaphors as a category which does not depend on stored vocabulary knowledge, but rather invites active processing based on analogy. Thus, the processing of novel metaphors can act as a base-line for the processing of novel expressions, in contrast to all other idiomatic categories included in the study. Figurative language in High Functioning Autism (HFA) {#sec003} ---------------------------------------------------- Despite the long-standing interest in idioms, their nature and processing in typical populations, research on idiom comprehension in atypical populations overall, and in autism, in particular, has been surprisingly scanty. Mashal & Kasirer \[[@pone.0168571.ref024]\] conducted a study where individuals with HFA, children with learning disorder and typically developing controls were compared. The authors used 11 subtests, ranging from interpretation of figurative expressions, including visual metaphors, idioms, conventional metaphors, novel metaphors, to comprehension of homophones and semantic tests (synonymy, similarity). Their results showed that children with learning deficit scored significantly lower than typically developing peers on comprehension of conventional metaphors and idioms, but no significant differences were observed between the two groups on comprehension of visual and novel metaphors. Furthermore, both deficit groups displayed a dissociation between the visual modality and the auditory modality, supporting the assumption that modality is a salient factor in figurative language processing. In contrast, the control group demonstrated an association between idioms and conventional metaphors. This result supports the idea that idioms and conventional metaphors share a common ground, in that both are relatively frozen conventionalised expressions and most likely rest on similar mechanisms for their processing. This is consistent with findings in a recent study of metaphor processing in highly verbal individuals with autism \[[@pone.0168571.ref025]\]. Vogindroukas & Zikopoulou \[[@pone.0168571.ref026]\] conducted a study aimed at the developmental trajectory of idiom comprehension in Asperger's syndrome (AS) and HFA. The authors compared children with HFA/AS to their typically developing peers and to control adults. They found that children with AS/ HFA performed more poorly on the idiom task compared to the other two groups, while the two typically developing groups did not display any differences in terms of performance. In addition, the authors did not find a significant correlation between IQ scores and performance on the task for the HFA group, but IQ did correlate with idiomatic comprehension and age for the two control groups. It is important to mention that this study used idioms out of context. The fact that the idioms were presented without a supportive context means that the participants had to rely exclusively on a semantic analysis or retrieval. Thus, the study design might have introduced an additional problem and a potential confound between expression familiarity/frequency and decomposability. In that study, Vogindroukas & Zikopoulou \[[@pone.0168571.ref026]\] confirm that children with AS/HFA have a tendency to make literal interpretations, but that the difficulty in idiom comprehension is definitely not due to an intellectual deficit. Whyte et al \[[@pone.0168571.ref027]\] addressed what factors predict performance on idiom comprehension in highly-verbal children with autism in comparison to an age-matched control group and a control group matched on grammar-age equivalence. This study focused specifically on the contribution of syntax competence and advanced Theory of Mind (ToM) abilities to idiom comprehension in autism. The results document that children with autism indeed perform more poorly compared to age-matched controls, but that there is no significant difference with the syntax competence matched group. Furthermore, while advanced ToM ability did not predict performance in the control groups, it correlated with idiom comprehension in the participants with autism. Lee et al \[[@pone.0168571.ref028]\] tested Korean high-functioning children with autism compared to similarly aged children with Attention Deficit Hyperactivity Disorder (ADHD) and a control group of typically developing children. The test design included matching idioms to a picture, which either described the figurative meaning of the expression or the literal interpretation. Non-figurative sentences paired with their literal interpretation were used as a base-line condition. This study found that both the children with autism and the ADHD group performed significantly worse than controls on the two idiom conditions, and especially in the condition where the idiom was paired with its literal interpretation. No differences were observed on the non-figurative expressions. Norbury \[[@pone.0168571.ref029]\] investigated idiom comprehension in context in children with communication disorders and age-matched peers. Children belonging to the communication disorders group were selected to represent a spectrum of communication difficulty, including Specific Language Impairment (SLI), Pragmatic Language Impairment (PLI) (with or without presence of autistic symptoms), autistic disorder, and Asperger syndrome. An idiom definition task was chosen to measure idiom comprehension in context. The main finding of that study was that children with language and communication impairments benefit from context in interpreting unfamiliar idioms. However, children with deficits in structural language did not benefit from context as much as typically developing peers. Furthermore, children with autism spectrum disorder (ASD) symptoms and pragmatic impairments, but without concomitant structural language deficits, were not impaired to the same degree in their idiom understanding and were able to use context as effectively as their peers. Importantly, language ability was one of the most significant predictors of idiom understanding. These findings highlight the role of language skills as a key determinant in the use of context to aid idiom comprehension, and further suggest that contextual cues can provide sufficient information to derive at least a cursory understanding of new idiomatic expressions. Factors in idioms processing {#sec004} ---------------------------- Previous research has documented that idioms are more easily processed in the presence of supportive context (\[[@pone.0168571.ref016]\]; see Vulchanova et al \[[@pone.0168571.ref010]\] for a review). Findings from several studies suggest that the main role of context is to provide semantic support for decoding the target (appropriate) meaning of a sentence or an expression \[[@pone.0168571.ref017], [@pone.0168571.ref030]--[@pone.0168571.ref033]\]. In contrast, it is widely argued that individuals with autism are impaired in processing ambiguous linguistic information in context \[[@pone.0168571.ref034]--[@pone.0168571.ref036]\]. In addition, they often fail to attach context to their memories, and are specifically impaired in processing the social aspects of contextual information \[[@pone.0168571.ref037]\], while Ozonoff & Miller \[[@pone.0168571.ref038]\] demonstrate deviant use of context. However, the evidence concerning a contextual deficit is somewhat controversial. Thus, other studies have shown that competence at making use of context depends on language status, and children with higher verbal age can perform at the level of controls \[[@pone.0168571.ref039]\]. Also, success in that group may depend on the exact position of the ambiguous word (homograph) in the sentence. López & Leekam \[[@pone.0168571.ref036]\] found that, even though children with ASD performed worse than controls, they were more successful with homographs occurring later in the sentence (middle or end), suggesting sensitivity to context also in that group. A study by Brock, Norbury, Einav & Nation \[[@pone.0168571.ref040]\] showed that participants with autism display a tendency similar to controls to use predictions based on the meaning of the lexical verb, as reflected in an increase in looks to the picture matching the object of the verb. At the same time both groups were less likely to be distracted by a phonological competitor of the object word. This tendency, however, was mediated by language status in the autism group. In that study context was restricted to the level of the sentence, with a focus on the information encoded in the lexical verb and the ability to use that information to successfully orient to the possible object. Also, Frith \[[@pone.0168571.ref041]\] argues that individuals with autism need a concrete context to be able to understand words and expressions, and Saldaña & Frith \[[@pone.0168571.ref042]\] document that children with autism have a normal reduction in reading times for expressions which are congruent with previous events. Given this evidence, we expect that context will not place any additional burden on the interpretation of the stimuli used in the current design. Idiomatic expressions vary in degree of transparency/decomposability, and it seems clear that the more transparent the expression, the easier it is to understand, and conversely, the more opaque, the more difficult. A number of studies \[[@pone.0168571.ref043]--[@pone.0168571.ref044]\] demonstrate that the lexical makeup of idioms is a matter of degree and largely depends on the idiom's compositional properties, and that idioms are accessed differentially according to their degree of syntactic frozenness. These authors thus claim that idioms do not form a unique class of linguistic items, but rather share properties with "more" literal language. Following this tradition, and with Titone & Connine \[[@pone.0168571.ref007]\], we expect that comprehension difficulty will be a direct function of the transparency/decomposability of the idiom \[[@pone.0168571.ref031], [@pone.0168571.ref033]\]. It can be assumed that the ability to understand idioms relies on the competence to process these expressions beyond the literal interpretation of individual words. This means that understanding idiomatic expressions demands making inferences, as well as integrating contextual information from both verbal and non-verbal sources. In our daily life, our knowledge about the world around us is based on the semantic and conceptual information gathered through our senses, implying that we perceive the world through different modalities \[[@pone.0168571.ref045]\]. Consequently, it seems logical to expect that some of the difficulties that individuals with autism demonstrate in figurative language may be due to a more generic deficit in multimodal integration. Indeed, problems with multi-sensory information have been documented in autism, and specifically at speech-specific multisensory integration \[[@pone.0168571.ref046]\]. Irwin, Tornatore, Brancazio & Whalen suggest that there are fundamental differences in audio-visual integration in children and adolescents with autism in comparison to typical peers \[[@pone.0168571.ref047]\]. Furthemore, individuals with autism have been shown to present with a reduced McGurk effect in the processing of audio-visual speech \[[@pone.0168571.ref048]\]. It has also been suggested that the neurodevelopmental abnormalities in autism may be due to poor information integration across brain regions \[[@pone.0168571.ref049]\]. Van Der Smagt, Engeland & Kemner \[[@pone.0168571.ref050]\] further suggest that any problems arising from integrating auditory and visual information must stem from higher processing stages, as is the case in processing complex linguistic material, and not low-level integration processes (e.g., at the level of individual sound/visual stimulus). Hypotheses {#sec005} ========== In this study, we investigate the interaction of linguistic and visual context in the interpretation of figurative expressions in highly-verbal individuals with autistic spectrum disorder compared to typically developing peers. The aim of this research is to test a more homogeneous sample of participants within the broad group of individuals with autism by focusing on the higher end of the spectrum (high functioning individuals with intact structural language) in two homogeneous chronological age samples. Thus, we are interested in establishing a potential developmental trajectory, albeit through a cross-sectional design. Consistent with the evidence from previous research, we expected to find effects of age and group. In particular, we hypothesized that the two groups with autism will perform significantly worse than the age- and verbal abilities-matched control groups. Specifically, we expected to find higher reaction times (RTs) and error means in the participants with autism compared to their typically developing peers, despite an improvement with age, as observed in the case of metaphors in a recent study (Chahboun et al., \[[@pone.0168571.ref025]\]). Furthermore, it can be expected that, since individuals with HFA have a tendency for literal interpretation, they will have difficulties in appreciating the figurative nature of non-literal language and more often select interpretations that match the literal meaning of the expression. Given that figurative language skills develop constantly through childhood, we also expect that younger participants (the two children groups) will perform at a lower level than the young adult groups. While cross-modal tasks, which require the integration of information from different modalities may not present a problem for typically developing participants, they may tax processing in individuals with autism and make the task more demanding. Concerning the presence of visually supported meaning (pictures that matched the interpretation of the idioms), it can be expected that visual context may facilitate the processing of the figurative meaning, or, at least, will not present an additional problem, as suggested by López & Leekam \[[@pone.0168571.ref036]\] and indicated by the results in Brock et al \[[@pone.0168571.ref040]\], also consistent with the findings in Norbury about the facilitatory role of context \[[@pone.0168571.ref029]\]. Alternatively, it may be the case that the presence of both auditory and visual information might rather present a problem, as a result of the demand for information integration in the face of documented problems in that domain in autism. Thus, we anticipated higher error rates and RTs in the condition when stimuli were presented in a cross-modal fashion (auditory stimulus/visual target) in the groups with autism, and the absence of such an effect in the control groups. One of the most important criteria in the processing of figurative language is expression transparency. We expected that degree of transparency would impact both on reaction latencies and accuracy in both participants with autism and controls. More specifically, the most transparent category, the one labelled biological idioms would be easiest to process, while cultural idioms, which are more highly conventionalised and less transparent would, in comparison, present greater difficulty, especially for the child participants and participants with autism. We also expected novel metaphors and instructive idioms to be more easily processed than the less transparent cultural idioms. Method {#sec006} ====== Participants {#sec007} ------------ Two age groups of HFA (n = 45) and typically developing (TD) controls (n = 39) were recruited, all native speakers of Spanish. The first age group included children in the age range ten to twelve years. The second age group included young adults from sixteen to twenty-two years old (see [Table 1](#pone.0168571.t001){ref-type="table"} for the participants' descriptive and demographic data). Participants and their legal guardians (usually the parents) provided written informed consent for entry into the study according of the declaration of Helsinki principles. The study was approved by the Andalusian Biomedical Research Ethics Committee. The diagnosis of HFA was confirmed with the Autism Diagnostic Observation Schedule (ADOS) \[[@pone.0168571.ref051]\] (Children, $\overline{x}$ = 10,8; Young adults, $\overline{x}$ = 12,3). 10.1371/journal.pone.0168571.t001 ###### Mean (and standard deviation) of background measures for each age and group. HFA (high functioning autism), FSIQ (intelligence quotient, Full Scale score), VCI (verbal comprehension index), RG (receptive grammar scores), RC (reading comprehension scores). Note: p \< .05 for scores with the same superscript. ![](pone.0168571.t001){#pone.0168571.t001g} Children Adults ---------------------- -------------- -------------- --------------- ---------------- n 25 19 20 20 Age (months, SD) 11,3 ±0,96 11,9±0,89 18,1±1,65 18,3±1,92 Gender (M) 21 17 17 14 FSIQ 110,71±14,58 105,76±11,56 108,3±13,39 118,34±8,46 VCI 118,7±17,47 107,4±15,03 122,62±16,71 117,63±12,97 RG 72,72±5,56 74,31±4,30 74,40±3,53^a^ 78,42±1,74 ^a^ RC 0,58±0,17 0,59±0,16 0,63±0,15 0,72±0,12 The individuals with HFA and their typically developing peers were matched on age, gender and the verbal comprehension based on the Wechsler scale (WISC-IV \[[@pone.0168571.ref052]\] or WAIS-IV \[[@pone.0168571.ref053]\] were used, depending on the chronological age of the participant). The Verbal Comprehension Index (VCI) was used as an overall measure of verbal reasoning and includes measures of semantic/conceptual reasoning (similarities and vocabulary), and verbal comprehension. The descriptive statistics are given in [Table 1](#pone.0168571.t001){ref-type="table"}. The matching of the groups was based on the Wilcoxon Test and the smallest p value was p = .222 suggesting that there were no significant differences observed between the groups (See [Table 1](#pone.0168571.t001){ref-type="table"}). We also considered reading comprehension scores and receptive grammar (based on CEG, which is the Spanish adaptation of TROG \[[@pone.0168571.ref054]\]) as background measures to ensure our participants display scores within the norm (see [Table 1](#pone.0168571.t001){ref-type="table"}). No differences were observed between participants for reading comprehension. However, despite all our participants being within the norm for receptive grammar, differences were observed between the two groups of young adult participants. For this reason, we included receptive grammar as a co-variate in the analyses of the data. Apparatus and stimuli {#sec008} --------------------- In a pilot study with 50 adults, all native speakers of Spanish, we determined the degree of familiarity and frequency of the figurative expressions to be used in the main study. We selected 124 idioms and instructive expressions and 20 metaphors, for which participants in the pilot study had to determine: 1. "Do you know this expression?" 2. "If yes, do you know what it means?" 3. "Do you use this expression yourself?". Participants were asked to answer truthfully, and to take into account that there were no wrong answers. We established ratings for each expression by assigning numerical values to the answers (1 for yes and 0 for no) and averaging the scores from the 3 questions. The idiomatic expressions which received a rating of over 0.80 were selected for the study and classified into three main categories: a/ biological idioms; b/ cultural idioms; and c/ instructive expressions, as used in Vulchanova et al. \[[@pone.0168571.ref008]\]. For the first two categories they followed the typology adopted in Penttilä, Nenonen & Niemi \[[@pone.0168571.ref055]\], in turn inspired by Searle's idea of deep background (the human biological nature) vs. local background (local cultural practices), to account for the grounding of language in human experience and practice. The stimuli in these two categories were selected by 2 independent expert linguist raters (91% interrater agreement). Examples of biologically-based idioms are e.g., estar con el agua hasta el cuello (to be with water up to the neck); or venir como anillo al dedo (to come as ring to the finger), and of culturally-based idioms, e.g estar como una cabra (to be like a goat). The difference between the two categories is that, while biological idioms are derived from human (bodily) experience and interaction with the environment, and as such, tend to be more transparent and available to compositional parsing, cultural idioms are more idiosyncratic, tend to vary from one culture to another, and are, in general, less transparent. Our third category were instructive expressions (i.e., proverbs), whose meaning can be computed on-line, by drawing inferences on the basis of their individual constituent phrases (e.g., Las mentiras tienen las patas cortas---Lies have short legs---). The inclusion of this category was dictated by the need to compare such expressions, whose interpretation may be more directly available in appropriate context, to the comprehension of idioms. Finally, novel metaphors were included as a category that may be assumed to be less frozen/fixed, and as such more readily interpretable. From the pilot study, we selected for inclusion in the novel metaphors category those metaphors which earned a rating below .20 on the familiarity and frequency scale (e.g.: Estas flotando en el aire---You are floating in the air). The study was designed as a sentence-picture matching task (38 visual---presented as text---and 38 auditory expressions (a total of 76 expressions---20 biological idioms/20 cultural idioms/16 instructive expressions/20 metaphors) randomized and counterbalanced between groups and participants). Each expression was linked to four possible images: One reflecting the figurative target meaning of the expression, a second image reflecting the literal meaning, a third image corresponding to a figurative meaning, but not the target one, and finally, a distractor as forth image (Cf. [Fig 1](#pone.0168571.g001){ref-type="fig"}). Each expression was presented in a short appropriate context (See [S1 Appendix](#pone.0168571.s001){ref-type="supplementary-material"} for examples). Thus, the study design included both verbal (presented as text/discourse) and visual (pictures) context. ![Images for the idiom "estar con el agua hasta el cuello!" (lit.: be with water up to the neck/fig.: drown in work)](pone.0168571.g001){#pone.0168571.g001} Matlab (R2014b) (matrix laboratory) was used as a multiparadigm numerical programming language to design, build and run the experiment. Procedure {#sec009} --------- In this experiment, we used visual stimuli (images and sentences) that reflected literal meanings and idiomatic meanings. First, a fixation point (+) appeared for 500ms. Participants then saw on the screen the idiom or metaphor. Depending on the experimental block, the participant heard the expression in context via loud-speakers or read it in context in a short text presented on the computer screen orthographically. Thus, the condition in which stimuli were presented auditorily presented a cross-modal environment whereby participants had to select a visual target which matched a figurative expression they had heard via the loud-speakers. In the auditory experimental block, first the figurative expression appeared on the computer screen followed by a blank screen, while the expression in appropriate context were being presented auditorily. In the text (orthographic) condition both stimulus (expression and context) and target picture were all in the same (visual) modality. First the expression appeared on the screen, and was then followed by the context. Then, in both experimental conditions a fixation point (+) reappeared in the middle of the screen followed by four images in each corner, each one reflecting the figurative meaning (target) meaning, the literal meaning, figurative, but not target, meaning, and a distractor image. The position of presenting the images was counterbalanced between participants and between the modality of presenting the context and the order of the expressions. The participant's task was to click on the stimulus he/ she considered most appropriate for the meaning of the expression. Accuracy and reaction times were collected to determine the ease of processing. Results {#sec010} ======= The data of both the control and experimental groups (n = 84) were analysed with R \[[@pone.0168571.ref056], [@pone.0168571.ref057]\]; and the lme4, afex and lmerTest packages \[[@pone.0168571.ref058]\]. We used two performance measures in the study; reaction time which reveals the ease with which the figurative expression is processed, and accuracy which essentially measures the ability to link the figurative expression to its target image. We ran linear mixed models on reaction times and generalized linear mixed models on accuracy, including the variables of Group (ASD vs. control), Age (Children vs. Young Adults), Modality (Visual vs Auditory), and Type of Expression (Biological, Cultural, Proverbs, and Novel Metaphors). We report relevant F and p-values below (based on the Satterthwaite approximation for degrees of freedom) for reaction times and Chi square and p-values for accuracy. We ran four analyses. The first one, the reaction time analysis, provided us with a general overview of the ease with which the participants responded to the different experimental manipulations in the study. The next analyses explore the ability to identify the correct target compared to other options. Thus, a second model provided us with participants' accuracy and compared overall target and non-target responses. The third and fourth specifically explore correct to other specific non-appropriate responses. The third analysis, called "comparison between figurative target responses and literal responses", provides a comparison between responses to target images and images corresponding to the literal meaning. Finally, our fourth model is based on a comparison between figurative target responses and figurative non-target responses. This model only includes responses to images reflecting target meanings and responses to images which reflected a figurative albeit not target meaning. Reaction time analyses {#sec011} ---------------------- Reaction times were analysed using Log(RT) (ms) as a measure rather than reaction times to correct for right skewedness of the reaction time data. Only correct responses are included in the analyses. We also excluded extremely fast or slow responses. We excluded two trials with reaction times longer than 18s and we calculated the Z value of Speed (Log(RT) (ms)) for each participant (Z~p~) and for each item per age and group (Z~i~). If the sum of the squares of the Z values was smaller or equal than 8 (Z~p~^2^+Z~i~^2^ ≤ 8), the trial was included, and values bigger than 8 were excluded. An overall linear mixed model analysis on Log (RT) (ms) performance has been done with R, which provided us with the best fitting model by step-wise backward selection based on the Akaike Information Criterion (See [S2 Appendix](#pone.0168571.s002){ref-type="supplementary-material"}). We were left with a model including a three-way interaction of group, modality and type of expression, a two-way interaction of age and type of expression, and all lower effects. Also included in the model were the by-subject random intercepts and slopes for modality and by-item random intercepts and slopes for Age and Group. The overall model reveals that there are differences in reaction latency for Age (children/young adults) (F (1, 82.64) = 20.38, p \< .001) (Satterthwaite approximation for degrees of freedom) and Group (control/HFA) (F (1, 86.85) = 10.64, p = .001), with slower responses by children and individuals with autism (see [Fig 2](#pone.0168571.g002){ref-type="fig"}). Furthermore, an interaction between age and type of expression was found (F (3, 66.11) = 2.903, p = .041) (see [Fig 3](#pone.0168571.g003){ref-type="fig"}). A multiple comparison with Tukey contrasts showed that the interaction is due to differences between different types of expressions in children, but not adults. Here, a significant difference was observed between novel metaphors and instructive expressions (p = .0014). In addition, a marginally significant difference between cultural idioms and novel metaphors was observed (p = .07) showing that instructive expressions were the most demanding and difficult to process for the children. ![Average reaction times (ms) for each age (children/young adults) and group (control/HFA).\ Error bars denote one standard error of the mean.](pone.0168571.g002){#pone.0168571.g002} ![Average reaction times (RTs) for each age (children/young adults) and group (control/HFA) for each type of expression.\ Error bars denote one standard error of the mean.](pone.0168571.g003){#pone.0168571.g003} As we have pointed out, a three-way interaction between type of expression, group and modality was observed (p \< .001) (see [Fig 4A and 4B](#pone.0168571.g004){ref-type="fig"}). In order to establish the origin of the three-way interaction, additional multiple comparisons with Tukey contrasts were run. For the HFA group in the auditory modality, no significant differences were observed between different types of expressions. However, for the visual modality, an almost significant difference was observed between instructive expressions and novel metaphors (p = .0501), with novel metaphors processed faster than instructive expressions. For the control group in the auditory modality, significant differences were observed between instructive expressions and biological idioms (p = .013), with slower responses in the instructive expressions category, and between novel metaphors and cultural idioms (p = .033), with faster responses in the novel metaphor category, and between novel metaphors and instructive expressions (p \< .001), with faster responses for novel metaphors. No significant differences were observed in the visual modality in the control group between different types of expressions. These results suggest a different interaction pattern between participants with autism and controls. While for controls there is differential processing depending on type of expression in the oral language modality, with more transparent expressions processed faster, this pattern is absent in participants with autism. ![4a, 4b. Average reaction times (ms) group (control/HFA) type of expression and modality. Error bars denote one standard error of the mean.](pone.0168571.g004){#pone.0168571.g004} Moreover, we explored the differences between groups and modalities for each type of expression. For biological idioms, while the control group was faster in the auditory modality, the HFA group was faster in the visual modality (p = .026), but always significantly slower than controls (p = .001). For cultural idioms, the HFA group was significantly slower than controls (p = .001). In addition, no modality differences were observed in the HFA group for this type of expression, while the control group were faster in the visual modality (p = .02). For instructive expressions, the control group was significantly slower in the auditory modality (p = .01), while the HFA group showed no modality preference in this category and was slower in comparison to the control group overall in this type of expression (p = .015). Finally, for the novel metaphor category, again the HFA group was much slower overall (p \< .001), and showed no modality differences, while the control group showed faster responses in the auditory modality (p = .01). To sum up those results, with increase of expression non-transparency, the advantage of modality changes only for controls. Thus, for controls transparent expressions are processed faster in the oral modality and less transparent expressions are processed faster in the orthographic modality. No such trend is observed for the participants with autism. These results suggest a different processing pattern in autism. Comparison between figurative target responses and non-target responses {#sec012} ----------------------------------------------------------------------- An overall generalized linear mixed model analysis on ACC (mean errors) performance has been done with R, which provided us with the best fitting model by step-wise backward selection based on the Akaike Information Criterion. We were left with a model including a two-way interaction of age and group and main effects. Also included in the model were the by-subject random intercepts and slopes for modality and by-item random intercepts and slopes for Age and Group. The overall model thus reveals differences in accuracy in Age (children/young adults) (χ^2^ (1, 23) = 5.73, p = .0016) and Group (control/HFA) (χ^2^ (1, 23) = 11.21, p \< .001), with more errors by children and individuals with autism (see [Fig 5](#pone.0168571.g005){ref-type="fig"}). In addition, a main effect of type of expression was observed as well (χ^2^ (1, 21) = 8.49, p = .036). Multiple Comparisons of Means with Tukey contrasts (See [S3 Appendix](#pone.0168571.s003){ref-type="supplementary-material"}) revealed that this effect was due to marginally significant differences between instructive expressions and biological idioms (p = .07), and between instructive expressions and novel metaphors (p = .06), with more errors in the instructive expressions category. ![Average response accuracy (mean errors) for each age (children/young adults) and group (control/HFA).\ Error bars denote one standard error of the mean.](pone.0168571.g005){#pone.0168571.g005} Moreover, a two-way interaction between age and group was observed (χ^2^ (1, 23) = 4.98, p = .025). Multiple Comparisons of Means with Tukey contrasts (See [S3 Appendix](#pone.0168571.s003){ref-type="supplementary-material"}) revealed that the interaction was due mainly to the differences between young adults (HFA and control) (p \< .001), and between the two control groups (children and young adults) (p = .01). No differences were observed between the two HFA groups (p = 0.99) or children (HFA and control) (p = 0.74). These results suggest that participants with autism perform at the same level regardless of age. Comparison between figurative target responses and literal responses {#sec013} -------------------------------------------------------------------- In order to establish responses to target figurative meaning contra literal interpretations, we re-ran the model, while excluding non-target figurative, and distractor responses (See [S4 Appendix](#pone.0168571.s004){ref-type="supplementary-material"}). In this model, Target responses were scored with 1 and literal responses were scored with 2. The model includes a significant three-way interaction between modality, age and group (control/HFA), and all lower interactions and main effects. Also included in the model were the by-subject random intercepts and slopes for modality and by-item random intercepts and slopes for Age and Group. This model reveals a main effect of group (control/HFA) ((χ^2^ (1, 26) = 5.22, p = .022), with more literal responses by children and young adults with autism and a marginally significant difference in accuracy between Age (children/young adults) ((χ^2^ (1, 26) = 3.51, p = .06) (see [Fig 6](#pone.0168571.g006){ref-type="fig"}). In addition, a main effect of type of expression was observed (χ^2^ (1, 24) = 10.37, p = .015). Multiple Comparisons of Means with Tukey contrasts (See [S4 Appendix](#pone.0168571.s004){ref-type="supplementary-material"}) revealed that this effect was due to a difference between instructive expressions and novel metaphors (p = .02), with more literal responses in the instructive expressions category, and to a marginally significant difference between cultural idioms and novel metaphors, with more literal responses in the cultural idioms category. These results suggest that less transparent expressions are likely to trigger more literal responses. ![Average literality for each age (children/young adults) and group (control/HFA).\ Target responses were scored with 1 and literal responses were scored with 2. Higher means indicate more literal responses. Error bars denote one standard error of the mean.](pone.0168571.g006){#pone.0168571.g006} Furthermore, a two-way interaction between age and group was observed (χ^2^ (1, 26) = 4.89, p = .02). We ran multiple comparisons of means with Tukey contrasts which revealed that this interaction was due to a significant difference between control young adults and HFA young adults (p = .015), with the young adults with autism giving more literal responses than their respective typically developing peers. Also, the difference between both ages of control participants was significant (p = .03), where the young adults detected the figurative target image more accurately. No differences were observed between the two HFA groups (p = 0.99) or children (HFA and control) (p = 0.99). These results suggest that children and participants with autism are more likely to provide literal responses. A three-way interaction between age, group and modality was observed (χ^2^ (1, 26) = 3.97, p = 0.046) (see [Fig 7](#pone.0168571.g007){ref-type="fig"}). Again, a multiple comparison of means with Tukey contrasts showed no significant effects in the visual modality in any of the age groups, but clear differences in the auditory modality between the control young adults and the young adults with autism (p = 0.001), where the older groups of participants had inverted patters of responses. The young adults with HFA were less literal in the visual modality than the auditory modality, while the control young adults were better in the auditory modality, suggesting a differential pattern in response. Moreover, a marginally significant difference was observed between the control children and the control young adults (p = .055), where the children did not show modality differences, but the older group was better in the auditory modality. ![Average literality for each age (children/young adults) and group (control/HFA) and modality (visual/auditory).\ Target responses were scored with 1 and literal responses were scored with 2. Higher means indicate more literal responses. Error bars denote one standard error of the mean.](pone.0168571.g007){#pone.0168571.g007} Comparison between figurative target responses and figurative non-target responses {#sec014} ---------------------------------------------------------------------------------- The final part of the analyses compared figurative target responses and figurative non-target ones. Therefore, we ran the model again without responses to the literal image and responses to the distractor (See [S5 Appendix](#pone.0168571.s005){ref-type="supplementary-material"}). From the overall model it is clear again, that there are differences in accuracy between groups (control/HFA) (χ^2^ (1, 11) = 6.49, p = .01), with more errors by individuals with autism. Additionally, a main effect of age (children/young adults) was found as well (χ^2^ (1, 11) = 12.92, p \< .001 and (see [Fig 8](#pone.0168571.g008){ref-type="fig"}), with more errors in the two younger groups. These results suggest that figurative language comprehension is not yet stabilized and still developing in the child participants. No other significant effects were found in this comparison. Despite this, figurative non-target responses only appeared in very few instances (4.5% of the responses in comparison with 81.8% for figurative target ones). Consequently, the comparison between figurative target and figurative non-target responses shows a very small error range. ![Average accuracy in figurative meanings for each age (children/young adults) and group (control/HFA) and modality (visual/auditory).\ Target responses were scored with 1 and literal responses were scored with 0. Higher means indicate more accurate responses. Error bars denote one standard error of the mean.](pone.0168571.g008){#pone.0168571.g008} General linear model---Receptive grammar as a covariate {#sec015} ------------------------------------------------------- We were also interested in the extent to which the difference between the two young adult groups on receptive grammar influenced these groups' performance on the tasks. Repeated measures ANOVA were run in SPSS \[[@pone.0168571.ref059]\], where we considered age and group as between subject factors, modality and type of expression as within subject variables and receptive grammar as a covariate. Bonferroni confidence interval adjustment was considered and observed power was calculated in this analysis (See [S6 Appendix](#pone.0168571.s006){ref-type="supplementary-material"}). We followed the same procedure for all the previously considered dependent variables (reaction time, overall accuracy, scores for target and literal responses, and scores for target and figurative non target responses). Receptive grammar as a covariate did not affect in any way any of the models and none of the main effects or interactions changed with the inclusion of receptive grammar scores as a covariate in the regression model. We report the power effects analyses in [S7 Appendix](#pone.0168571.s007){ref-type="supplementary-material"}. Discussion {#sec016} ========== Our main hypothesis was that we would find poorer performance on figurative language processing by children and the participants with autism compared to their peers, carefully matched on age, non-verbal IQ, and verbal comprehension. Consistent with our hypotheses, the analyses of reaction times in the current study found a main effect of age and group, with slower responses for children and participants with autism. While no significant difference was observed between the two children's groups on a test of receptive grammar, there was a difference between the two young adult groups of participants, and oral language differences could have explained the different performance between these groups. However, the fact that our results were not affected by the inclusion of receptive grammar scores as a co-variate in the analysis, ensures that the obtained results were not due to a structural language problem (e.g., in receptive grammar), but to the experimental manipulation. These analyses also revealed that two types of expressions were particularly difficult for children, cultural idioms and instructive expressions (proverbs), while novel metaphors were processed much faster. These results suggest that the impact of expression transparency tends to diminish with age and as figurative skills stabilize in young adulthood. The analyses of accuracy revealed a similar pattern: more errors were found in the two child groups and in the two groups with autism. Here a main effect of type of expression was also observed, whereby instructive expressions emerged as the most difficult category to process in comparison to both novel metaphors and biological idioms. The analyses of accuracy also revealed that participants with autism were significantly more likely than controls to select the literal interpretation of the figurative expression, as represented in a visual image. In this study, we hypothesized that the participants with autism would display a relative weakness in performance on idiomatic language, despite intact structural language skills. Although individuals with autism obtain high accuracy scores and can obviously solve the task, the current findings confirm that the adult individuals with autism are not reaching an equivalent level of performance to typically developing matched controls, both with regard to accuracy and reaction time. These results are consistent with results from the few previous studies of idiom comprehension in autism, where the participants with HFA performed more poorly and more slowly than their typically developing matched peers, suggesting a different pattern of processing figurative language in general, and idiomatic competences, in particular. These results are also consistent with the finding that the contrast in ability in autism in comparison to controls is more evident on the higher end of the spectrum \[[@pone.0168571.ref060]\]. The current results, however, contradict findings of comparable semantic and language processing skills in studies where the controls and participants with autism are matched on structural language skills \[[@pone.0168571.ref027],[@pone.0168571.ref040]\]. The participants in the current study were highly-verbal individuals with autism carefully matched to the control groups on IQ and verbal skills, including semantic, and oral and reading comprehension skills. The current results thus challenge the idea that the problems in figurative language in autism arise exclusively from verbal status and language skills, also consistent with results in Eigsti, de Marchena, Schuh & Kelley \[[@pone.0168571.ref061]\] and Olofson et al. \[[@pone.0168571.ref062]\] and the review in Vulchanova et al. \[[@pone.0168571.ref010]\]. More importantly, both the reaction latencies and accuracy analyses reveal a different processing pattern in the interaction between modality (auditory/visual) and type of expression, especially evident in the young adult groups. This pattern of responses is suggestive of different processing or perhaps alternative strategies for the comprehension of figurative language in autism, despite overall high accuracy. We consider that this different pattern of processing is due, in part, to a delayed trajectory in developing the necessary skills to process figurative language. This is clearly seen in the similar response pattern of the young adults with autism and the control children on both accuracy and reaction speed. This may suggest that rather than an inability to process figurative language or idioms altogether, in this case, there is a delay in the ability to detect such expressions, and to differentiate between the different categories of figurative language. These results support the findings from another study with the same participants on metaphorical language \[[@pone.0168571.ref025]\]. Yet, even though participants with autism seem to improve with age in terms of reaction time, suggesting faster processing, they still do not reach the same levels of performance as control matched peers. The improvement with age does not only apply to the individuals with HFA, but also to the control groups, which is consistent with Nippold & Duthie \[[@pone.0168571.ref063]\], Levorato & Cacciari \[[@pone.0168571.ref064]\] and Vulchanova et al \[[@pone.0168571.ref008]\], and our original hypothesis, but not with Vogindroukas & Zikopoulou \[[@pone.0168571.ref026]\], who, surprisingly, did not find any reliable difference in performance between the younger and older controls. Furthermore, our study explored the impact of the nature and the transparency of the type of figurative expression. In the study we included biological idioms, which have been argued to be more closely grounded in human experience and interaction with the environment (Searle's deep background). We expected such expressions to be more readily accessible and more decomposable, despite their idiomatic nature, compared to cultural idioms, also for the participants with autism. In line with Chahboun et al \[[@pone.0168571.ref025]\], we also expected that novel metaphors would be easier to process. These expectations were confirmed. Even though the participants with autism were significantly delayed in comparison to controls, they displayed similar patterns of difficulties and ease across the test categories as the control groups. Biological idioms and novel metaphors were the easiest to process. We reason that these two categories are the most transparent, and their meaning can be easily inferred from the context, while cultural idioms are less transparent and more "frozen" \[[@pone.0168571.ref065]\]. These findings are consistent with Vulchanova et al. \[[@pone.0168571.ref008]\] and Vega-Moreno \[[@pone.0168571.ref017]\] who claim that inferential skills are not only involved, but mandatory in idiom comprehension. In our study, the figurative expressions were elicited based on a pilot, where we selected the most familiar and frequent expressions for the idioms category, and the least frequent (most novel) for the metaphors category. The current results confirm that novel metaphors were easier to process compared to less transparent expressions, such as idioms. Our findings are partly consistent with Giora, Gazal, Goldstein, Fein, & Stringaris \[[@pone.0168571.ref066]\], who found that familiarity/novelty of the expression were reliable predictors of performance in both typically developing individuals and individuals with HFA. The current results are also consistent with the findings in Norbury \[[@pone.0168571.ref029]\], where participants with autism were facilitated by the presence of context in the comprehension of novel idioms and performed at levels similar to controls. The current findings, however, contradict the results obtained by Mashal & Kasirer \[[@pone.0168571.ref024]\]. These authors found that the most novel figurative expressions were the most difficult to process in both adults and children with HFA. However, this difference might be due to experimental design and choice of experimental stimuli, as is often the case in research on figurative language. Contrary to our expectations, instructive expressions (proverbs) were the most difficult category to process, along with cultural idioms. We included this category in the design because of the expectation that they may not be stored in the mental lexicon, but rather rely on the immediate activation of richer "memory organization packages" \[[@pone.0168571.ref021],[@pone.0168571.ref067]\], thus allowing more readily for interpretations on the fly when encountered in appropriate context. This expectation was not confirmed, however. Moreover, both children and participants with autism were significantly more likely to select the literal interpretations of both instructive expressions and cultural idioms. These results support the findings in Honeck, Welge & Temple \[[@pone.0168571.ref068]\] concerning the processing of proverbs, where participants were faster to make appropriateness judgements when the proverb was related literally to the context compared to when it was related figuratively. These authors interpret the findings as supporting a multi-stage model of proverb comprehension, with literal interpretations becoming available faster than figurative ones. The current results suggest that instructive expressions (proverbs) indeed require some ability to think abstractly \[[@pone.0168571.ref020]\] and cannot be interpreted so readily based on their constituent words. We also find evidence that this skill takes time to acquire, as evident in the child participants. We also speculate that the more culturally-based an expression is, the more reliant it is on learning it in social interaction, thus consistent with the limitations documented in this domain in autism. Regarding the effect of modality, we hypothesized that the control group would not display dissociations between the visual and auditory presentation modalities, while for the participants with autism we expected certain problems when the expressions were presented auditorily and had to be matched to a visual stimulus (a picture) in the cross-modal condition. This was confirmed, but only for the most transparent expressions (biological idioms and novel metaphors). The control groups performed better when biological idioms and novel metaphors were presented in the auditory modality. Thus, it appears that the control group have an advantage for oral language, and can easily integrate oral stimuli with visual contextual support (the images). This was not the case for the participants with autism, and it could suggest a a difficulty even in highly-verbal individuals with autism in multimodal information integration, even when the idiomatic expressions are decomposable (cf. also Ozonoff & Miller, \[[@pone.0168571.ref038]\]). These differential responses to modality in the autism group could be indicating the involvement of different processes in the resolution of the idioms, or different approaches when attempting to respond to this task. These findings point to the possibility of a problem of simultaneous processing in both modalities in this population. An additional possibility may be that there are residual oral language problems also in individuals with intact structural language skills. This is consistent with Eigsti \[[@pone.0168571.ref069]\], where the problems in ASD are attributed to low-level impairments in generic cognitive and processing mechanisms. Alternatively, it can be assumed that "seeing" text is an advantage, especially in the groups with autism. Indeed, there is evidence that orthographic decoding is a strength in autism and this may, in turn, lead to better comprehension outcomes, if a simple view of reading is adopted \[[@pone.0168571.ref070]\]. Another interesting result was the difference in degree of literalness observed in response accuracy. As expected, the younger participants and participants with autism interpreted the stimuli more often literally than the older and the control groups. Figurative language competence takes longer to develop and requires a number of other skills (inferencing, discourse skills; \[[@pone.0168571.ref071]\]), and the older we are, the more resources for processing we have. In the case of participants with autism, both children and young adults were very literal and did not appear to improve in this respect with age (the two age groups in the study performed similarly). This was not the case for the control groups. For children presentation modality did not affect response literalness, while for the older participants there was an advantage in the auditory modality, suggesting that idiom comprehension most probably resides in mature oral language skills. The results of the participants with autism are also consistent with other research suggesting a tendency for literal (compositional) interpretation in autism \[[@pone.0168571.ref026],[@pone.0168571.ref072]\]. The overall results of this study support our hypothesis that idiomatic expressions require more competences than just good structural language. These clearly provide a necessary condition, but may not be sufficient. This is evidenced by the findings supporting the presence of a delay in the two groups with HFA. Additionally, the current study contributes to research studying the processing of idioms specifically, and figurative language, more broadly, in providing empirical evidence on how different categories of figurative expressions are processed in both highly verbal individuals with autism and typically developing individuals. We find support that expressions, which are more firmly grounded in human bodily experience (biological idioms), and are also more transparent (novel metaphors) are processed more readily, despite their idiomatic nature, regardless of participant group. Finally, we find evidence that young adult controls have a clear advantage in processing auditory language stimuli, most likely as a result of subtle differences in oral language skills. The problems observed in the participants with autism are suggestive of either multi-sensory information integration difficulty or simply a residual problem in oral language and need to be addressed in future research potentially leading also to better target intervention for that group. Supporting Information {#sec017} ====================== ###### Examples of expressions in context. (DOCX) ###### Click here for additional data file. ###### Overall model (Reaction times). (RTF) ###### Click here for additional data file. ###### Overall model (Accuracy analysis: Target vs non-target). (RTF) ###### Click here for additional data file. ###### Overall model (Accuracy analysis: Target vs literal responses). (RTF) ###### Click here for additional data file. ###### Overall model (Accuracy analysis: Figurative target vs Figurative non-target responses). (RTF) ###### Click here for additional data file. ###### General linear model analyses with receptive grammar as covariate. (DOCX) ###### Click here for additional data file. ###### Observed powers. (DOCX) ###### Click here for additional data file. We would like to thank the participants and the families for their effort and time in participating in this study. We are also grateful to Martina Micai for help in the data collection. Last but not least, we would like to thank schools and associations for putting us in contact with both the HFA groups and the TD groups. [^1]: Competing Interests:The authors have declared that no competing interests exist. [^2]: Conceptualization: MV VV DS SC.Data curation: HE SC.Formal analysis: MV VV HE SC.Funding acquisition: MV VV DS.Investigation: SC.Methodology: MV VV DS SC.Project administration: MV VV DS.Resources: MV VV DS HE SC.Software: HE SC.Supervision: MV VV DS.Validation: MV VV HE SC.Visualization: MV VV DS SC.Writing -- original draft: SC.Writing -- review & editing: MV VV DS HE SC. [^3]: Current address: Department of Language and Literature, Norwegian University of Science and Technology, Trondheim, Norway [^4]: Current address: Departamento de Psicología Evolutiva y de la Educación. Universidad de Sevilla, Avda. Camilo José Cela s/n, Seville, Spain	{ "pile_set_name": "PubMed Central" }
INTRODUCTION {#sec1-1} ============ Pre-eclampsia and eclampsia are hypertensive disorders of pregnancy that cause significant morbidity and mortality in the fetus and mother. Globally, pre-eclampsia and eclampsia account for 10-15% of maternal deaths.\[[@ref1]\] Pre-eclampsia is a multi-organ group of related disease processes that occur in up to 5-8% of pregnancies after 20 weeks of gestation. The presentation is variable, but generally includes the combination of maternal hypertension and proteinuria.\[[@ref2]\] Severe pre-eclampsia is defined as any of the following: (i) severe hypertension (systolic blood pressure \>160 mmHg or diastolic blood pressure \>110 mmHg); (ii) proteinuria 5 g per 24 h; (iii) oliguria \<400 ml urine per 24 h; (iv) cerebral irritability or visual disturbances; (v) epigastric or right upper quadrant pain (liver capsule distension); or pulmonary edema.\[[@ref3]\] Eclampsia is the new onset of grandmal seizures occurring during and after pregnancy that do not have another identifiable cause.\[[@ref2]\] The development of eclampsia is associated with increased risk of adverse outcome for both mother and fetus.\[[@ref1][@ref4]\] In Nigeria, eclampsia is the 3^rd^ commonest cause of maternal mortality.\[[@ref5]\] Eclampsia could be prevented in majority of the cases if there was early booking for antenatal care services, permitting early identification of pre-eclampsia and institution of appropriate therapy.\[[@ref6]\] The high maternal morbidity and mortality due to eclampsia in the developing countries has been ascribed to late referral, delay in hospitalization, late transportation, unbooked status of patients, and multiple seizures prior to admission.\[[@ref5]\] Progression of pre-eclampsia to eclampsia may occur during pregnancy, labor or in the postpartum period.\[[@ref4]\] Reviewing the published literature, the authors of this report were unable to find any record of eclamptic seizures occurring during cesarean section under spinal anesthesia. We, therefore, report the case of a primigravida who presented with severe pre-eclampsia, which progressed to eclampsia during cesarean section under spinal anesthesia and was successfully managed. CASE REPORT {#sec1-2} =========== A 26-year-old primipara booked for antenatal care at 36 weeks gestation at the University of Port Harcourt Teaching Hospital. On presentation, she was found to have 2++ of protein in urine and blood pressure was 190/120 mmHg. A diagnosis of severe pre-eclampsia was made, and the patient was subsequently admitted to be managed on diazepam, hydrallazine, and magnesium sulfate. The patient decided to go home to inform her relatives and get her personal belongings. She came back at night and refused medications. She was subsequently booked for emergency cesarean section on the next day 'due to fetal bradycardia and unfavorable cervix.' She was counseled for surgery and spinal anesthesia, which she then accepted. 'She still refused the administration of magnesium sulfate.' 'The platelet count was 130 × 10^9^/liter.' Spinal anesthesia was established at L4, L5 interspace with 1.8 ml of 'preservative-free' 0.5% plain bupivacaine after preloading her with 1 liter of normal saline and oxygen was administered at a flow rate of 6 liters per minute by face mask. The level of block was T8 and surgery commenced. The blood pressure fell from a pre-induction value of 190/110 mmHg to 182/105 mmHg, and the patient had generalized tonic-clonic seizures twice while under spinal anesthesia. 'The first seizure occurred just after delivery of the baby, but before administration of 5 i.u. of bolus intravenous oxytocin.' 'This was about 9 minutes after intrathecal injection of bupivacaine and 5 minutes after commencement of surgery.' 5 mg of intravenous diazepam was administered, and it stopped after 10 seconds. A second fit occurred 5 minutes later, and 10 mg intravenous diazepam was administered. It stopped after about 50 seconds and did not recur. 'Oxygen administration by face mask with the patient breathing spontaneously, which was commenced immediately after establishing the subarachnoid anesthesia, was continued throughout the surgical procedure. A size 3 Guedel oropharyngeal airway was inserted and retained from the onset of the first seizure. Arterial oxygen saturation remained essentially normal, and there was no vomiting or obvious regurgitation.' A live female baby was delivered with Apgar scores of 5 and 6 at 1 and 5 minutes, respectively. 'The baby was resuscitated by the neonatology senior registrar in the operating room. He suctioned the oral cavity, initially ventilated with 100% oxygen by a Rendel-Baker-Soucek pediatric facemask and later placed a size 2.5 mm orotracheal tube through which he continued artificial ventilation of the baby. The baby was wrapped in warm clothing and transferred to the special care baby unit.' She regained consciousness 15 minutes after the second seizure, but was drowsy. She was managed in the first 48 hours after surgery in the intensive care unit. 'Magnesium sulfate was not administered to the patient all through,' and she did not have any other seizure. She was discharged home on the 7^th^ post-operative day with an advice to complete her course of antibiotics, analgesics, and hematinics. DISCUSSION {#sec1-3} ========== The diagnosis of eclampsia was made in this case based on the occurrence of 'generalized' seizures against a background of pre-eclampsia.\[[@ref2]\] 'Spinal myoclonus associated with subarachnoid anesthesia with bupivacaine has been reported,\[[@ref7]\] both in Caucasians\[[@ref8]\] and patients of African origin.\[[@ref9]\] Whereas, our patient had generalized tonic-clonic seizures, spinal myoclonus associated with subarachnoid anesthesia with bupivacaine affected the lower limbs only.' When this patient with severe pre-eclampsia consented to undergo an emergency cesarean section, she was counseled for spinal anesthesia, which she also accepted. Neuroaxial block techniques such as spinal anesthesia are considered a safe method of providing anesthesia for the patient with pre-eclampsia and severe pre-eclampsia. This is due to avoidance of the risks associated with general anesthesia such as exacerbated hypertension, failed intubation, and aspiration.\[[@ref10]\] Recent studies have shown that spinal and combined spinal-epidural anesthesia can be safely administered without significantly increasing the risk to the mother and fetus, even in severe pre-eclampsia.\[[@ref11][@ref12]\] Visalyaputra et al.,\[[@ref11]\] showed that while there was a brief period of increased hypotension in severely pre-eclamptic patients receiving spinal versus epidural anesthesia, there were no clinical differences in fetal or maternal outcomes. Judicious volume expansion as was done with normal saline in this case may be considered prior to regional anesthesia, but caution should be used to avoid fluid overload given the increased risk of pulmonary edema.\[[@ref2]\] Although eclampsia may occur in the absence of hypertension,\[[@ref2]\] the rarity of progression from pre-eclampsia to eclampsia during cesarean sections under spinal anesthesia could be due to the tendency for the blood pressure to be lower when spinal anesthesia is used. The intra-operative eclamptic fits were successfully controlled with intravenously administered diazepam. However, magnesium sulfate is the anti-convulsant of choice for treating eclampsia. It is more effective than diazepam, phenytoin, or lytic cocktail.\[[@ref1]\] Magnesium sulfate should have been used even before the cesarean section if not for the patient\'s refusal because it has been shown to reduce the incidence both of eclampsia complicating severe pre-eclampsia and further fits in eclamptic patients.\[[@ref13]\] Eclampsia could be prevented in majority of cases if there was early booking, early identification of pre-eclampsia, and institution of appropriate therapy.\[[@ref6]\] The patient in this case, was residing in Port Harcourt, an urban city, which had two government-owned tertiary health care facilities, and several privately-owned clinics, which offer antenatal care services. She failed to utilize the services available in any of these healthcare facilities. When she presented with severe pre-eclampsia, she initially refused medications, thereby further delaying interventions that could have prevented eclampsia. Eclampsia remains a continuing problem in developing countries despite improvements in antenatal care facilities as seen in this case. Poor utilization of available antenatal care services has been identified in a study as contributing to a high incidence of eclampsia.\[[@ref6]\] It has been suggested that more awareness and enabling factors be created for more women to access antenatal care facilities.\[[@ref5]\] Also, factors preventing pregnant mothers from utilizing available antenatal care services need to be studied so as to maximize the effectiveness of these services in reducing maternal and perinatal morbidity and mortality. CONCLUSION {#sec1-4} ========== 'Progression of pre-eclampsia to eclampsia during caesarean section under spinal anaesthesia is rare,' but it can occur. Although it could be successfully treated with anticonvulsant drugs, it is better to prevent its occurrence' Eclampsia could be prevented in majority of cases by early booking for antenatal care, early identification of pre-eclampsia and prompt institution of appropriate therapy. Factors preventing utilization of available quality antenatal care services by pregnant women should be studied. Source of Support: Nil Conflict of Interest: None declared	{ "pile_set_name": "PubMed Central" }
Introduction ============ Mycoplasma pneumoniae (MP) pneumonia (MPP) is well known for its diverse radiological findings. In a previous report, we developed a system for scoring the affected area on thoracic computed tomography (CT) in adult patents with MPP, and reported that the affected area was associated with the presence of hypoxemia (1). However, to the best of our knowledge, no one has elucidated the correlation between the areas of lung involvement on chest X-ray and the clinical findings in both children and adults with MPP. In the present study, we retrospectively analyzed the data of MPP, to investigate the association between the values obtained by a chest X-ray scoring system and the clinical findings. Materials and Methods ===================== We retrospectively examined the data of MPP patients who visited our hospital during from April 2006 to July 2014. This study was approved by the ethics committee of Kyorin University on July 2014 (approval number: H26-032). For the purpose of the present study, "child" was defined as \<15 years of age, while "adult" was defined as ≥15 years of age. The diagnostic criteria for MPP were as follows: 1) the presence of new abnormal lung shadows on a chest X-ray, and one of the following: 2) a ≥4-fold titer rise \[complement fixation (CF) or particle agglutination (PA) tests\] during the convalescent phase in comparison to the acute phase; or a single PA titer of ＞1:160; or 3) the isolation of MP in sputum culture. The chest X-ray films were divided into three levels: 1) the bronchial bifurcation, 2) the upper level of the diaphragm, and 3) halfway between 1) and 2) ([Fig. 1](#g001){ref-type="fig"}), and total visual scores were defined as follows: grade 0, no opacity; grade 1, 5% opacity; grade 2, 5-24% opacity; grade 3, 25-49% opacity; grade 4, 50-74% opacity; and grade 5, ≥75%. The scoring system that was applied in the present study consisted of a total of six areas, the scores for each area were summed to determine the total score (range, 0-30). Two pulmonologists and one radiologist with ＞10 years of experience independently reviewed each chest X-ray film; the final decisions were made after the examiners reached a consensus. The correlations between the total score and the clinical and/or laboratory data were also evaluated. ![The chest X-ray was divided into three levels: 1) the bronchial bifurcation, 2) the upper level of the diaphragm, and 3) halfway between levels 1) and 2). Each area was calculated as follows: Grade 0, no opacity; Grade 1, \<5% opacity; Grade 2, 5-24% opacity; Grade 3, 25-49% opacity; Grade 4, 50-74% opacity; and Grade 5, \>75% opacity. The total score was calculated as the sum of the scores for the six areas.](1349-7235-56-2845-g001){#g001} Statistical analysis -------------------- The distribution and variance of numeric data were evaluated using the Kolmogorov-Smirnov test and Levene\'s median test, respectively. Categorical data are presented as percentages of the total or numerically, as appropriate. Statistical comparisons of nonparametric data were performed using the Mann-Whitney test. Categorical data were compared using the chi-squared test. All tests were two-sided. p values of ＜0.05 were considered to indicate statistical significance. All statistical analyses were performed using the SPSS software program (version 19.0 for Windows). Results ======= The clinical characteristics of the children and adults with MPP ---------------------------------------------------------------- We identified a total of 71 children and 54 adults with MPP. The characteristics of the two groups were comparable with regard to the proportions of sex, underlying respiratory diseases, the incidence of hypoxemia, and incidence of antecedent treatment with macrolide therapy ([Table 1](#t001){ref-type="table"}). Furthermore, the duration from the onset of symptoms until the referral of the patient to our hospital (with/without treatments at the local hospital) was similar in the two groups (mean±SD: children, 7.1±2.4 days versus adults, 6.9±4.6 days; not significant) ([Table 1](#t001){ref-type="table"}). However, the child group showed significantly higher body temperatures (children, 38.4 ± 0.8 vs. adults, 37.7 ± 1.1°C, p = 0.03) and higher serum levels of lactate dehydrogenase (LDH; 369 ± 142 vs. 237 ± 68.2 IU/L; p ＜0.001) and aspartate aminotransferase (36.4 ± 14.8 vs. 28.0 ± 16.8 IU/L, p ＜0.001) in comparison to the adult group. In contrast, the serum C-reactive protein levels in the adult group were significantly higher in comparison to the child group (adult group, 9.2 ± 9.0 mg/dL vs. child group, 3.0 ± 3.6 mg/dL; p ＜0.001). ###### Comparisons of Clinical Characteristics between the Child and Adult MPP Groups. Child (n=71) Adult (n=54) p value ------------------------------------------------------- -------------- -------------- ---------- Age (years) 7.9 ± 3.6 37.6 ± 18.1 p\<0.001 M:F 34:37 18:36 NS Underlying diseases (%) 11.1 (7/63) 23.4 (11/47) NS Asthma 7 9 NS Emphysema 0 1 NS Lung caner 0 1 NS Initial onset to first visit to our hospital (days)\* 7.1±2.4 6.9±4.6 NS BT (°C) 38.4 ± 0.8 37.7 ± 1.1 p=0.03 Hypoxemia (%) 18.1 (4/22) 12.2 (5/41) NS Antecedent macrolide treatment (%) 46 (23/50) 38.2 (18/47) NS WBC (×10^3^/µL) 8.5 ± 5.5 8.5 ± 4.4 NS CRP (mg/dL) 3.0 ± 3.6 9.2 ± 9.0 p\<0.001 LDH (IU/L) 369 ± 142 237 ± 68.2 p\<0.001 AST (IU/L) 36.4 ± 14.8 28.0 ± 16.8 p\<0.001 ALT(IU/L) 23.9 ± 22.3 22.4 ± 18.2 NS Diagnostic method Single titer (PA≥1:320 or CF≥1:64) 66 (93.0%) 40 (74.1%) p=0.005 Pair (×4) 5 (7.0%) 14 (25.9%) p=0.005 Culture 0 (0) 1 (1.9%) NS AST: aspartate aminotransferase, ALT; alanine aminotransferase, BT: body temperature, CRP: C-reactive protein, F: female, LDH: lactase dehydrogenase, M:male, WBC: white blood cell counts \* Data are presented as mean±standard deviation The radiological findings between children and adults with MPP -------------------------------------------------------------- Regarding the chest X-ray findings, consolidation was a major radiological finding in both groups ([Table 2](#t002){ref-type="table"}). Air bronchogram, bronchial wall thickening, and atelectasis were observed significantly more frequently in the child group than in the adult group. Other findings, such as reticular shadowing, tiny nodules, and pleural effusion, did not differ to a statistically significant extent. ###### Comparisons of Radiological Findings between the Child and Adult MPP Groups. CHILD (%) ADULT (%) p value --------------------------- ----------- ----------- --------- Total number of patients 71 54 Consolidation 62 (87.3) 45 (83.3) NS Air bronchogram 43 (60.6) 18 (33.3) p=0.004 Reticular shadowing 20 (28.2) 11 (20.4) NS Tiny nodules 17 (24.0) 12 (22.2) NS Bronchial wall thickening 24 (33.8) 8 (14.8) p=0.022 Pleural effusion 1 (1.4) 4 (7.41) NS Atelectasis 7 (9.9) 0 (0) p=0.019 The clinical significance of the chest X-ray score and the zonal predominance of MPP ------------------------------------------------------------------------------------ The correlation between hypoxemia and the total affected lung areas (calculated according to the total score) was not statistically significant in either of the groups ([Fig. 2A, D](#g002){ref-type="fig"}). Importantly, no correlation was found between the time from the onset of symptoms to referral to our hospital and the total score in either group (data not shown). Regarding zonal predominance, MPP predominantly affected the middle-to-lower lung fields in both groups ([Fig. 2B, E](#g002){ref-type="fig"}). Interestingly, a moderate positive correlation was found between the total score and the maximum PA titer value (r = 0.409, p ＜0.001), but only in the child group ([Fig. 2C, F](#g002){ref-type="fig"}). The total score was not significantly correlated with the serum inflammatory markers \[white blood cell count (WBC), LDH, and C-reactive protein (CRP)\] in either group (data not shown). ![No significant correlation between hypoxemia and the total score was found in the child group or the adult group (A, C). MPP was predominantly located in the middle-to-lower lung fields in both groups (B, D). A moderate positive correlation was found between the total score and maximum PA titer value (r=0.409, p\<0.001) only in the child group (C, F). \p value \<0.05. \\\p value \<0.001](1349-7235-56-2845-g002){#g002} The correlations between the serum LDH levels and the proportion of hypoxemia in children and adults with MPP ------------------------------------------------------------------------------------------------------------- No correlation was found between the presence of hypoxemia and the serum LDH levels in either the child group or the adult group; however, a non-significant positive correlation was observed in the latter group (p = 0.074) ([Fig. 3](#g003){ref-type="fig"}). ![No correlation was found between the presence of hypoxemia and the serum LDH level in either group.](1349-7235-56-2845-g003){#g003} Discussion ========== This study provides the first evidence of the relevance of the total affected lung area (calculated as a total score) on chest X-ray films and the clinical and laboratory findings in both children and adults with MPP. Firstly, no correlation was found between the total score and the presence of hypoxemia in either children or adults with MPP. This was contrary to our previous report, in which a scoring method was applied to assess the thoracic CT scans of adults with MPP ([@B1]). This may be because chest X-rays show lower sensitivity in the detection of faint shadowing in comparison to thoracic CT or due to absence of severe MPP patients who required admission to an ICU and/or mechanical ventilation. Secondly, this study clearly demonstrated that the middle-to-lower lung fields were predominantly affected by MPP. Previous studies in which chest X-ray films were analyzed implied that the lower lung field was predominantly affected in both children and adults with MPP ([@B2], [@B3]), while peribronchial and perivascular interstitial infiltration and/or air space consolidation were the common radiological patterns. These trends were quantitatively confirmed in the present study, with the use of the chest X-ray scoring system and by the assessment of the radiological patterns. Although the precise reason why the incidence of atelectasis in children with MPP was significantly higher in comparison to adults with MPP is unknown, it was possibly caused by the obstruction of the airway by a mucus plug due to the increased expression of mucins by the bronchial epithelial cells after MP infection ([@B4]). Thirdly, a positive correlation between the maximum PA titer and the total score was only observed in children with MPP . This discrepancy suggested the possibility that an excessive immune reaction to MP antigens ([@B5]-[@B7]), which was predominantly attributed to the inflammatory process, occurred in children rather than adults. Indeed, the findings of Tanaka et al. ([@B8], [@B9]) and the data from our previous studies ([@B5]-[@B7]) demonstrated that the enhanced cellular-mediated and/or humoral immune response to MP can exaggerate lung inflammation in mouse models. Notably, the total lung score was not correlated with the presence of hypoxemia. This finding was anticipated by general physicians who treat MPP pneumonia. The serum LDH level has been considered a marker for refractory MPP ([@B10]); however, no one has elucidated the significance of LDH levels in predicting hypoxemia. In this regard, the present study demonstrated a lack of correlation between these factors. This study is associated with some limitations. Firstly, the extent of lung abnormalities might have been underestimated due to the low sensitivity of chest X-rays in the detection of tiny nodules and reticular shadowing, which might have affected the relationship between the total score and the laboratory/clinical findings. Furthermore, this study was conducted at a regional referral center; thus, the adult MPP patients seemed to have a greater incidence of underlying respiratory disease in comparison to the typical MPP patients described in previous reports ([@B11]) who were usually treated in local hospitals. This might have affected the results of our study. However, physicians who treat MPP patients have usually recognized that no correlation exists between the extent of lung involvement on chest X-ray and the presence of hypoxemia. The results of the present study are compatible with their position. Conclusion ========== This study showed the first evidence of a correlation between the quantitatively calculated area of lung abnormalities on chest X-ray films and the clinical findings, including the presence of hypoxemia, in both children and adults with MPP. The authors state that they have no Conflict of Interest (COI). Takeshi Saraya and Takayasu Watanabe contributed equally to this work. [^1]: Correspondence to Dr.　Takeshi Saraya, <sara@yd5.so-net.ne.jp>	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1} =============== Dental pulp is highly vulnerable to various physicochemical and microbiological stimuli, such as acute injury, bacterial activity and metabolites, dental bleaching, and nonpolymerized resin monomers \[[@B1]--[@B3]\]. These stimuli can cause dental pulp cell (DPC) death, eventually leading to irreversible pulp inflammation and necrosis \[[@B1]\]. One main pathogenic mechanism of these stimuli is oxidative stress- (OS-) mediated damage in DPCs \[[@B4], [@B5]\]. OS is characterized by the excessive generation of reactive oxygen species (ROS), e.g., H~2~O~2~, O~2~ ^−^, and ^·^OH \[[@B6]\]. OS stimulated by dental composites and lipopolysaccharides has been reported to induce cell cycle alteration and death of human DPCs (HDPCs) \[[@B7]\]. During dental bleaching, exogenous hydrogen peroxide (H~2~O~2~) released from the bleaching solution can result in the cell death of DPCs \[[@B5]\]. Mitochondria are the main resource and also the major attacking target of ROS \[[@B8]\]. In response to the abovementioned pathological stimuli, the excessively generated ROS will cause rapid depletion of antioxidants and then induce oxidative damage to mitochondria, which subsequently lead to mitochondrial dysfunction in HDPCs \[[@B9]\]. Such a cell death can be attenuated by the application of mitochondria-specific antioxidants \[[@B10]\]. These results indicate that mitochondrial dysfunction is highly involved in OS-induced HDPC death. However, the underlying molecular mechanisms remain largely unknown. The mitochondrial permeability transition pore (mPTP), assembled between the inner and outer mitochondrial membranes, opens with the relatively severe disturbance of intracellular redox and/or Ca^2+^ homeostasis \[[@B11], [@B12]\]. The mPTP opening can lead to a solute exchange between mitochondrial matrix contents and the surrounding cytoplasm, which is commonly linked to mitochondrial dysfunction \[[@B13]\]. Cyclophilin D (CypD) is a critical protein for mPTP opening \[[@B14]\]. CypD-dependent mPTP opening has been shown to play a key role in ROS-induced mitochondrial dysfunction \[[@B15]\] and cell death \[[@B14]\]. The pharmacological inhibition or genetic ablation of CypD can rescue mitochondrial dysfunction and cell damage induced by OS \[[@B16]\]. However, it remains unknown whether the CypD-dependent mitochondrial pathway is involved in the OS-mediated death of HDPCs. In this study, we aimed to identify the potential role of CypD in the regulation of mPTP and mitochondrial dysfunction in the OS-induced HDPC death. 2. Materials and Methods {#sec2} ======================== 2.1. Informed Consent and Ethical Approval {#sec2.1} ------------------------------------------ The human research commission of the School and Hospital of Stomatology, Wenzhou Medical University, Wenzhou, China (2018001), gave approval to the study. For the human tooth collection, a written informed consent was obtained from all the subjects. 2.2. Reagents {#sec2.2} ------------- Cell culture medium and additional supplements were bought from Life Technologies (Grand Island, NY, USA). The antibodies were obtained from Cell Signaling Technology (Beverly, MA, USA). The chamber slides and 4′,6-diamidino-2-phenylindole (DAPI) were from Invitrogen (Carlsbad, CA, USA). MitoSOX Red, TMRM, and MitoTracker Green (MTGreen, Molecular Probes) were from Life Technologies (Grand Island, NY, USA). A terminal deoxynucleotidyl transferase deoxyuridine triphosphate nick-end labeling (TUNEL) kit was from Roche (Mannheim, BW, Germany). 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), an annexin V-fluorescein isothiocyanate (FITC) detection kit, H~2~O~2~, and N-acetylcysteine (NAC) were from Sigma-Aldrich (St. Louis, MO, USA). Cyclosporine A (CsA) was from Cell Signaling Technology (Beverly, MA, USA). An adenosine triphosphate (ATP) assay kit was from Beyotime Institute of Biotechnology (Shanghai, China). 2.3. Cell Culture, Characterization, and Treatment {#sec2.3} -------------------------------------------------- HDPCs were isolated from the dental pulp tissues of noncarious third molars and grown in Dulbecco\'s modified Eagle\'s medium (DMEM) containing 10% fetal bovine serum (FBS), 100 U/mL penicillin, and 100 mg/mL streptomycin (Gibco, Grand Island, NY, USA). Cultures were maintained in a humidified atmosphere containing 5% CO~2~ at 37°C. For HDPC lineage characterization, the morphological analysis was performed, and the cells were immunocytochemically stained for vimentin and keratin. The working concentrations of the compounds were as follows: NAC (2.5 mM) and CsA (2 μM). The final concentration of dimethyl sulfoxide (DMSO) in the culture was less than 0.5% in all the experiments. The cells with or without H~2~O~2~ were treated with the indicated test compounds according to the experimental protocol. 2.4. Cell Viability {#sec2.4} ------------------- HDPCs were seeded in 96-well plates and treated with the indicated reagents. Then, HDPCs were incubated with 10 μL MTT solution (5 mg/mL). After 4 h, the formazan crystals were dissolved by DMSO. The plates were then measured in a microplate reader. The optical density (OD) of the control group in MTT was taken as 100% viable. 2.5. siRNA-Mediated CypD Knockdown in HDPCs {#sec2.5} ------------------------------------------- CypD siRNA targeting human peptidylprolyl isomerase F (PPIF) and the negative control (NC) were purchased from RiboBio (Guangdong, China) and transfected in HDPCs with Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer\'s protocol. 2.6. Measurement of Cell Death by Flow Cytometry and TUNEL Assays {#sec2.6} ----------------------------------------------------------------- The cell death was detected by annexin V labeled with FITC. Propidium iodide (PI) was used to determine cell necrosis. After exposure to various experimental conditions, the cells were trypsinized and labeled with fluorochromes, and then, cytofluorometric analysis was performed with a FACScan (Becton Dickinson, NY, USA). For the TUNEL assays, different groups of cells were grown on a coverslip, incubated with a TUNEL reaction mixture, counterstained with DAPI, and observed under a fluorescence microscope. 2.7. Western Blot Analyses {#sec2.7} -------------------------- HDPCs were collected and lysed in the cell lysis buffer. Proteins were separated by electrophoresis and transferred onto a polyvinylidene difluoride (PVDF) membrane. Proteins bound by primary antibodies were visualized with an appropriate secondary antibody. The protein bands were detected using the Bio-Rad imaging system (Bio-Rad, Hercules, CA, USA) and quantified using NIH ImageJ software. 2.8. Intercellular ATP Level Determination {#sec2.8} ------------------------------------------ The ATP levels were detected by an ATP detection kit according to the manufacturer\'s instructions. The data was measured via a luminescence plate reader. 2.9. Functional Imaging Assays {#sec2.9} ------------------------------ After treatment with the indicated reagents, the cells were incubated with MitoSOX (2.5 μM) or TMRM (100 nM). Images were captured under the fluorescence microscope. The NIH ImageJ software was used to measure and quantify the fluorescence signals. More than 100 clearly identifiable mitochondria in 10-15 randomly selected cells per experiment were measured in three independent experiments \[[@B17], [@B18]\]. 2.10. Detection of Ca^2+^ Level {#sec2.10} ------------------------------- The cells were treated with the indicated reagents. Then, the cells were incubated with 10 μM Fluo-4-AM (Beyotime, Shanghai, China), a Ca^2+^-sensitive fluorescent probe. The Ca^2+^ level was detected under a fluorescence microscope. 2.11. Data Analysis {#sec2.11} ------------------- Data were presented as the mean of three independent replicates ± standard deviation (SD) and considered significant at p \< 0.05. Statistical analysis was performed using StatView software (version 5.0.1, SAS Institute, USA). For comparisons between the multiple groups, one-way ANOVA was used followed by individual post hoc Fisher tests when applicable. 3. Results {#sec3} ========== 3.1. H~2~O~2~ Induced Cell Death and Mitochondrial Dysfunction in HDPCs {#sec3.1} ----------------------------------------------------------------------- The cells around the pulp tissue, which showed a spindle shape and fibroblast-like morphology after being cultured for 5 days (Supplementary [Figure 1a](#supplementary-material-1){ref-type="supplementary-material"}), were characterized as primary HDPCs. In addition, the cells exerted strong character of proliferation (Supplementary [Figure 1b](#supplementary-material-1){ref-type="supplementary-material"}) and showed mesenchymal origin by the positive expression of vimentin in the cytoplasm but were negative for keratin (Supplementary Figures [1c](#supplementary-material-1){ref-type="supplementary-material"} and [1d](#supplementary-material-1){ref-type="supplementary-material"}). As shown in [Figure 1(a)](#fig1){ref-type="fig"}, H~2~O~2~ reduced HDPC viability in a time- and dose-dependent manner. The annexin/PI staining assay revealed a significant enhancement in the incidence of cell death. 250 μM H~2~O~2~ markedly enhanced the rate of cell death compared with the control group (Figures [1(b)](#fig1){ref-type="fig"} and [1(d)](#fig1){ref-type="fig"}). The H~2~O~2~-induced DNA damage was further confirmed by the TUNEL staining (Figures [1(d)](#fig1){ref-type="fig"} and [1(e)](#fig1){ref-type="fig"}). Compared with the control group, H~2~O~2~ remarkably increased the protein level of proapoptotic Bax but did not influence antiapoptotic Bcl-2 (Figures [1(f)](#fig1){ref-type="fig"} and [1(h)](#fig1){ref-type="fig"}). For the parameters of mitochondrial dysfunctions, H~2~O~2~ significantly increased the mitochondrial ROS (mtROS) level (MitoSOX staining) (Figures [1(i)](#fig1){ref-type="fig"} and [1(j)](#fig1){ref-type="fig"}) and decreased the mitochondrial membrane potential (MMP) (TMRM staining) (Figures [1(k)](#fig1){ref-type="fig"} and [1(l)](#fig1){ref-type="fig"}) in HDPCs. Furthermore, H~2~O~2~ markedly increased the intracellular Ca^2+^ level (Figures [1(m)](#fig1){ref-type="fig"} and [1(n)](#fig1){ref-type="fig"}) and reduced the ATP level ([Figure 1(o)](#fig1){ref-type="fig"}). In addition, H~2~O~2~ also significantly upregulated the expression of CypD (Figures [1(p)](#fig1){ref-type="fig"} and [1(q)](#fig1){ref-type="fig"}). 3.2. NAC Attenuated H~2~O~2~-Induced Cell Death and Mitochondrial Dysfunction in HDPCs {#sec3.2} -------------------------------------------------------------------------------------- NAC, a nonspecific antioxidant, observably preserved the cell viability in HDPCs treated with H~2~O~2~ ([Figure 2(a)](#fig2){ref-type="fig"}) and significantly mitigated cell death (TUNEL staining) (Figures [2(b)](#fig2){ref-type="fig"} and [2(c)](#fig2){ref-type="fig"}) and Bax expression (Figures [2(d)](#fig2){ref-type="fig"} and [2(e)](#fig2){ref-type="fig"}). NAC attenuated the effects of H~2~O~2~ by significantly suppressing mtROS (Figures [2(f)](#fig2){ref-type="fig"} and [2(g)](#fig2){ref-type="fig"}), increasing MMP (Figures [2(h)](#fig2){ref-type="fig"} and [2(i)](#fig2){ref-type="fig"}), decreasing the intracellular Ca^2+^ level (Figures [2(j)](#fig2){ref-type="fig"} and [2(k)](#fig2){ref-type="fig"}), and enhancing the ATP level ([Figure 2(l)](#fig2){ref-type="fig"}). NAC also decreased the CypD expression (Figures [2(m)](#fig2){ref-type="fig"} and [2(n)](#fig2){ref-type="fig"}). 3.3. Inhibition of CypD by CsA Reversed H~2~O~2~-Induced Cell Death and Mitochondrial Dysfunction in HDPCs {#sec3.3} ---------------------------------------------------------------------------------------------------------- CsA, a pharmaceutical inhibitor of CypD, significantly inhibited the H~2~O~2~-induced cytotoxicity and death, as was evident by the MTT assay ([Figure 3(a)](#fig3){ref-type="fig"}) and TUNEL staining (Figures [3(b)](#fig3){ref-type="fig"} and [3(c)](#fig3){ref-type="fig"}), respectively. Furthermore, CsA significantly downregulated the expression level of Bax and CypD (Figures [3(d)](#fig3){ref-type="fig"} and [3(f)](#fig3){ref-type="fig"}) and abrogated mtROS (Figures [3(g)](#fig3){ref-type="fig"} and [3(h)](#fig3){ref-type="fig"}) and the intracellular Ca^2+^ level (Figures [3(k)](#fig3){ref-type="fig"} and [3(l)](#fig3){ref-type="fig"}). CsA also effectively increased the MMP (Figures [3(i)](#fig3){ref-type="fig"} and [3(j)](#fig3){ref-type="fig"}) and ATP level ([Figure 3(m)](#fig3){ref-type="fig"}). 3.4. Blockade of CypD by siRNA Mitigated H~2~O~2~-Induced Cell Death and Mitochondrial Dysfunction in HDPCs {#sec3.4} ----------------------------------------------------------------------------------------------------------- The genetic knockdown of CypD with siRNA-PPIF was further conducted to validate the effect of CypD on HDPC death and mitochondrial dysfunction. Western blotting proved that siRNA-PPIF significantly downregulated the CypD expression (Figures [4(a)](#fig4){ref-type="fig"} and [4(b)](#fig4){ref-type="fig"}). We found that siRNA-PPIF significantly attenuated the H~2~O~2~-induced cytotoxicity ([Figure 4(c)](#fig4){ref-type="fig"}) and death (Figures [4(d)](#fig4){ref-type="fig"} and [4(e)](#fig4){ref-type="fig"}). siRNA-PPIF significantly inhibited the H~2~O~2~-induced expression of Bax and CypD (Figures [4(f)](#fig4){ref-type="fig"} and [4(h)](#fig4){ref-type="fig"}) and alleviated the mtROS and intracellular Ca^2+^ level (Figures [4(i)](#fig4){ref-type="fig"}, [4(j)](#fig4){ref-type="fig"}, [4(n)](#fig4){ref-type="fig"}, and [4(o)](#fig4){ref-type="fig"}). siRNA-PPIF also protected the mitochondrial function from the adverse effect of H~2~O~2~ (Figures [4(k)](#fig4){ref-type="fig"} and [4(m)](#fig4){ref-type="fig"}). 4. Discussion {#sec4} ============= Substantial evidence has shown that many pathogenic stimuli can cause significant oxidative damage, leading to inflammation and necrosis in dental pulp tissues. However, hitherto, the underlying molecular mechanisms accounting for the OS-induced death of HDPCs remain elusive. In this study, we, for the first time, reported the key role of CypD in the OS-induced death of HDPCs. We found that pharmacological blockage and genetic reduction of CypD could significantly attenuate the oxidative damage to HDPCs. Our findings delineated novel insights into the crucial role of CypD-dependent mPTP opening and mitochondrial dysfunction in the OS-induced HDPC death. Excessive ROS production can be induced in dental pulp by pathogenic stimuli, such as bacterial metabolite, dental bleaching, and unpolymerized resin monomers \[[@B2], [@B3], [@B19]\]. H~2~O~2~ is the most stable and commonly existing form of ROS and is widely used to establish cell OS injury models \[[@B20], [@B21]\]. In this study, we found that H~2~O~2~ significantly decreased the viability of HDPCs. Furthermore, H~2~O~2~ dramatically enhanced the death of HDPCs in comparison with the control group, as indicated by the results of flow cytometry and TUNEL assays. Mitochondrial outer membrane permeation (MOMP) is a key event in apoptotic cell death, which is controlled by the Bcl-2 family proteins \[[@B22]\]. Bax and Bcl-2, Bcl-2 family proteins, are important in regulating apoptotic cell death. Bax is upregulated in response to DNA damage \[[@B23]\] and regulates mitochondria-dependent cell death \[[@B24]\]. In response to apoptotic stimuli, Bax oligomerizes and forms pores perforating the outer mitochondrial membrane \[[@B25], [@B26]\]. MOMP occurs during pore formation, and the size of the pore can vary according to the number of recruited Bax dimers \[[@B27]\]. Bcl-2 plays an important part in promoting cellular survival and bating the actions of proapoptotic proteins \[[@B28]\]. Bcl-2 antagonizes MOMP by blocking Bax oligomerization and pore-forming activity \[[@B29], [@B30]\]. Our immunoblot analyses showed that H~2~O~2~ significantly increased the expression of proapoptotic protein Bax in HDPCs, which was consistent with the previous findings from osteoblastic cells \[[@B18]\] and human fetal lung fibroblast cells \[[@B31]\]. On the other hand, different from these studies, we found that Bcl-2 was not significantly affected by H~2~O~2~, which might be due to the different cell types. Increased ROS triggers cell death by inducing mitochondrial dysfunction \[[@B17], [@B32], [@B33]\]. NAC primarily abrogates mtROS and therefore mitigates myocardial cell death \[[@B20]\]. In the current study, H~2~O~2~ resulted in significant mitochondrial dysfunction, which was indicated by the increased mtROS, enhanced intracellular Ca^2+^ level, and decreased ATP level and MMP in HDPCs. NAC, in this study, not only blunted OS but also attenuated mitochondrial dysfunction in HDPCs. These results further confirmed that ROS-induced mitochondrial dysfunction played a key role in HDPC death. A mitochondria-specific antioxidant, mitoquinone, has been proved to be effective in ameliorating OS-related diseases \[[@B18], [@B34]\]. It would be highly interesting to apply these agents in this HDPC injury model to confirm the role of mitochondrial OS and dysfunction in OS-induced HDPC death in the future study. The mPTP, initiated by severe OS and cytosolic Ca^2+^ overload, is a key regulator in cell death \[[@B35], [@B36]\]. The opening of mPTP leads to rapid loss of MMP (which abolishes all MMP-dependent mitochondrial activities, including ATP synthesis), osmotic breakdown of mitochondrial membranes, and structural breakdown of the organelle, eventually causing the regulated cell death \[[@B13], [@B35], [@B36]\]. In this study, OS was shown to lead to mPTP opening in HDPCs, as reflected by MMP dissipation and Ca^2+^ disorder, which would result in the increase in mtROS and energy failure (decrease in ATP). A similar pattern was also found in H~2~O~2~-treated RAW264.7 macrophages \[[@B37]\]. NAC successfully inhibited mPTP opening and rescued mitochondrial dysfunction, thereby attenuating HDPC death. Taken together, these results showed that (i) mPTP opening was implicated in HDPC death and (ii) OS was the major inducer of mPTP opening-induced death in the H~2~O~2~-treated HDPCs. In our study, we mainly focused on the mPTP opening-induced cell death. Other than CypD, several proteins including the voltage-dependent anion channel (VDAC), adenine nucleotide translocator (ANT), and inorganic phosphate carrier (PHC) are supposed to be the putative modulators of mPTP \[[@B35]\]. However, previous studies using robust genetic tools confirm that CypD is the only vital protein in vivo required for mPTP opening \[[@B14], [@B38]\]. The knockout of VDAC, ANT, or PHC fails to protect cells from the mPTP opening \[[@B39]--[@B42]\]. In addition, mitochondrial F~1~F~O~ ATPase is also demonstrated to be involved in mPTP formation \[[@B43]\]. However, this conclusion is challenged by further reports by calculating their ion conductance and selectivity and especially the persistence of the mitochondrial permeability transition in the absence of the c-subunit of human F~1~F~O~ ATPase \[[@B44]--[@B46]\]. A very recent study largely gives structural explanation for the abovementioned inconsistent phenomena \[[@B47]\]. This study shows that there is no conventional mPTP formation but a CsA-sensitive channel when c-subunit is knocked out, still contributing to the depolarization of the inner mitochondrial membrane \[[@B47]\]. Further evidence will help to confirm the role of F~1~F~O~ ATPase in mPTP. In this study, we mainly focused on CypD and hypothesized that CypD was a key molecule in the H~2~O~2~-induced HDPC death. As expected, CsA and CypD siRNA inhibited the H~2~O~2~-induced mPTP opening, attenuated mitochondrial dysfunction, and prevented the death of HDPCs. Moreover, our results found that the CypD expression level significantly increased in the H~2~O~2~-treated HDPCs. This was consistent with a previous study showing that the increased expression of CypD contributed to the oxidative injuries in endothelial cells induced by a thyroid-stimulating hormone \[[@B16]\]. An in vivo study also confirmed that OS-induced cell death was mediated by CypD overexpression in the fibroblasts of patients with X-linked adrenoleukodystrophy \[[@B48]\]. Nonetheless, the overexpression of CypD was not always indispensable in its mediated cell death \[[@B49]\]. CypD may mediate the mPTP opening through three main mechanisms: (1) physical relationships with pore components \[[@B50]\], (2) ANT-independent pore formation \[[@B39]\], and (3) mitochondrial crista remodeling on the complete discharge of cytochrome C \[[@B51]\]. Consequently, CypD-mediated mPTP opening, especially in response to OS and Ca^2+^ overload, can be attenuated by either CypD genetic deletion \[[@B14]\], CypD peptidyl-prolyl cis- or trans-isomerase activity inhibition, or CsA molecular conformation \[[@B39]\]. All findings underscore a vital role of CypD-driven mPTP opening in OS-induced mitochondrial dysfunction and cell death of HDPCs. However, we only analyzed the role of CypD in H~2~O~2~-induced toxicity in dental pulp in vitro, and in vivo evidence should be achieved to corroborate the role of CypD. The upstream molecules regulating CypD remain largely unknown in the H~2~O~2~-induced death of HDPCs. A previous study indicates that gallic acid potentiates the extracellular signal-regulated kinase (ERK) phosphorylation, leading to a decrease in CypD expression, which contributes to the neuroprotective effect on cerebral ischemia/reperfusion injury \[[@B52]\]. Some studies also show that an increased thyroid-stimulating hormone can activate CypD by inhibiting sirtuin-3 (SIRT3) in endothelial cells \[[@B16]\]. Therefore, the search for vital upstream regulators of CypD in HDPCs, such as ERK, SIRT3, or p53, needs to be further explored in the future study. 5. Conclusions {#sec5} ============== Our data offered new insights into the role of CypD-mediated mitochondrial dysfunction in the H~2~O~2~-induced HDPC death (Supplementary [Figure 2](#supplementary-material-1){ref-type="supplementary-material"}) and the possible usage of the CypD inhibitor in the clinical treatment of dental pulp injury in the future. This work was supported by the Natural Science Foundation of China (81870757, 81500817, and 81870777), Zhejiang Provincial Natural Science Foundation of China (LY16H140005), Health Science and Technology Project of Zhejiang Province (2016KYB184), and Wenzhou Public Technical Research Medical Program (2015Y0507). Data Availability ================= The data used to support the findings of this study are available from the corresponding authors upon request. Conflicts of Interest ===================== The authors declare that there is no conflict of interest regarding the publication of this paper. Authors\' Contributions ======================= Shengbin Huang, Bingbing Zheng, and Xing Jin contributed equally to this work. Supplementary Materials {#supplementary-material-1} ======================= ###### Figure S1: characterization of HDPCs. Figure S2: working hypothesis. ###### Click here for additional data file. ![H~2~O~2~-induced cell death and mitochondrial dysfunction in the HDPCs. (a) Cell viability was determined by MTT reduction in the HDPCs with or without the presence of H~2~O~2~. (b, c) Flow cytometric quantification of cell death. HDPCs were treated for 24 h with culture medium or H~2~O~2~ (150 μM, 250 μM). (d, e) TUNEL staining and assay. (f) Representative immunoreactive bands with different densities for Bcl-2 and Bax in the HDPCs in the presence of H~2~O~2~. Quantification of immunoreactive bands for Bax (g) and Bcl-2 (h) relative to β-actin. Representative images showing MitoSOX staining (i) and quantification (j) in the indicated groups. Representative images with TMRM staining (k) and quantification (l) in the indicated groups. Representative images showing Fluo-4-AM staining (m) and quantification (n) in the indicated groups. ATP (o) in the indicated groups. (p) Densitometry of immunoreactive bands for CypD in the HDPCs in the presence of H~2~O~2~. (q) Quantification of immunoreactive bands for CypD relative to β-actin. HDPCs were treated for 24 h with H~2~O~2~ (250 μM) (+) or culture medium (-). Data represent the mean values ± SD of three independent experiments.](OMCL2019-1729013.001){#fig1} ![NAC attenuated H~2~O~2~-induced cell death and mitochondrial dysfunction in the HDPCs. (a) Cell viability determined by MTT reduction in the HDPCs in the presence of H~2~O~2~ with or without NAC. (b) TUNEL staining and (c) assay after NAC treatment. (d) Representative immunoreactive bands for Bax in the HDPCs with (+) or without (-) NAC treatment in the presence of H~2~O~2~ (+) or culture medium (-). Quantification of immunoreactive bands for Bax (e) relative to β-actin. Representative images showing MitoSOX staining (f) and quantification (g) in the indicated groups. Representative images showing TMRM staining (h) and quantification (i) in the indicated groups. Representative images showing Fluo-4-AM staining (j) and quantification (k) in the indicated groups. ATP (l) in the indicated groups. (m) Densitometry immunoreactive bands for CypD in the HDPCs with (+) or without (-) NAC treatment in the presence of H~2~O~2~ (+) or culture medium (-). (n) Quantification of immunoreactive bands for CypD relative to β-actin. HDPCs were treated for 24 h with (+) or without (−) NAC (2.5 mM) in the presence of H~2~O~2~ (250 μM) (+) or culture medium (−). Data represent the mean values ± SD of three independent experiments.](OMCL2019-1729013.002){#fig2} ![CsA attenuated H~2~O~2~-induced cell death and mitochondrial dysfunction in the HDPCs. (a) Cell viability determined by MTT reduction in the HDPCs in the presence of H~2~O~2~ with or without CsA. (b) TUNEL staining and (c) assay after CsA treatment. (d) Representative immunoreactive bands for Bax and CypD in the HDPCs with (+) or without (-) CsA treatment in the presence of H~2~O~2~ (+) or culture medium (-). Quantification of immunoreactive bands for Bax (e) and CypD (f) relative to β-actin. Representative images showing MitoSOX staining (g) and quantification (h) in the indicated groups. Representative images showing TMRM staining (i) and quantification (j) in the indicated groups. ATP (m) in the indicated groups. Representative images showing Fluo-4-AM staining (k) and quantification (l) in the indicated groups. HDPCs were treated for 24 h with (+) or without (−) CsA (2 μM) in the presence of H~2~O~2~ (250 μM) (+) or culture medium (−). Data represent the mean values ± SD of three independent experiments.](OMCL2019-1729013.003){#fig3} ![CypD siRNA attenuated H~2~O~2~-induced cell death and mitochondrial dysfunction in the HDPCs. (a) Representative immunoreactive bands for CypD in the HDPCs with CypD siRNA (+) or NC (-) treatment at a final concentration of 50 nM for 48 h. Quantification of immunoreactive bands for CypD (b) relative to β-actin. (c) Cell viability determined by MTT reduction in the HDPCs in the presence of H~2~O~2~ with or without siRNA-PPIF. (d) TUNEL staining and (e) assay after siRNA-PPIF treatment. (f) Representative immunoreactive bands for Bax, Bcl-2, and CypD in the HDPCs with (+) or without (-) siRNA-PPIF treatment in the presence of H~2~O~2~ (+) or culture medium (-). Quantification of immunoreactive bands for Bax (g), Bcl-2 (h), and CypD (i) relative to β-actin. Representative images showing MitoSOX staining (j) and quantification (k) in the indicated groups. Representative images showing TMRM staining (l) and quantification (m) in the indicated groups. ATP (n) in the indicated groups. Representative images showing Fluo-4-AM staining (o) and quantification (p) in the indicated groups. HDPCs were treated for 24 h with H~2~O~2~ (250 μM) (+) or culture medium (−) in the CypD siRNA group (+) or the NC group (−). Data represent the mean values ± SD of three independent experiments.](OMCL2019-1729013.004){#fig4} [^1]: Academic Editor: Manuela Corte-Real	{ "pile_set_name": "PubMed Central" }
![](glasgowmedj75363-0001){#sp1 .161} ![](glasgowmedj75363-0002){#sp2 .162} ![](glasgowmedj75363-0003){#sp3 .163} ![](glasgowmedj75363-0004){#sp4 .164} ![](glasgowmedj75363-0005){#sp5 .165} ![](glasgowmedj75363-0006){#sp6 .166} ![](glasgowmedj75363-0007){#sp7 .167} ![](glasgowmedj75363-0008){#sp8 .168} ![](glasgowmedj75363-0009){#sp9 .169} ![](glasgowmedj75363-0010){#sp10 .170} ![](glasgowmedj75363-0011){#sp11 .171} ![](glasgowmedj75363-0012){#sp12 .172} ![](glasgowmedj75363-0013){#sp13 .173} ![](glasgowmedj75363-0014){#sp14 .174} ![](glasgowmedj75363-0015){#sp15 .175} ![](glasgowmedj75363-0016){#sp16 .176} ![](glasgowmedj75363-0017){#sp17 .177} [^1]: Read before a meeting of the Monklands Medical Society, on 24th November, 1914.	{ "pile_set_name": "PubMed Central" }
Introduction {#s1} ============ Glutaredoxins (Grxs) are oxidoreductases sharing a conserved 3D structure with members of the thioredoxin (Trx) superfamily. They are present in most living organisms except in some bacterial and archaeal phyla (Rouhier et al., [@B35]; Alves et al., [@B1]). The primary function of Grxs was long thought to be the reduction of disulfide bonds and more particularly those formed between reduced glutathione (GSH) and a protein cysteinyl residue, a process known as glutathionylation, but as explained below, specific Grx members could rather serve as iron-sulfur (Fe-S) cluster transfer proteins (Rouhier, [@B31]). Glutathionylation is a post-translational modification which potentially fulfills several functions. It can constitute an intermediate of some reaction mechanisms, but it can also modulate protein function and serve as a signaling mechanism or protect cysteine residues from irreversible oxidation in plants (Zaffagnini et al., [@B47]) and in animals (Dalle-Donne et al., [@B14]). Although a few additional proteins, as some specific thioredoxins, can catalyze deglutathionylation reactions, Grxs are likely the major deglutathionylation system, at least in plants (Bedhomme et al., [@B5]; Chibani et al., [@B7]). They are usually regenerated via a NADPH/glutathione reductase (GR)/GSH system, but a few Grxs can be reduced by ferredoxin- or NADPH-dependent thioredoxin reductases (Johansson et al., [@B23]; Zaffagnini et al., [@B48]). Depending on the Grx isoforms and target proteins or substrates considered, Grxs can use different catalytic mechanisms, called monothiol and dithiol mechanisms (Rouhier et al., [@B35]). The monothiol mechanism is used for the reduction of glutathionylated proteins and requires a priori only the catalytic cysteine (the first or more N-terminal of the two active site cysteines, referred to as Cys~A~). It performs a nucleophilic attack on the protein-glutathione adduct, the Grx becoming glutathionylated. This oxidized Grx is regenerated by reduction via a GSH molecule. The dithiol mechanism requires the catalytic cysteine but also a recycling cysteine which could be either the second active site cysteine (Cys~B~) or an additional extra active site cysteine (Cys~C~). If the target protein is glutathionylated, the first step is similar to the monothiol mechanism, but the glutathione-adduct formed on Grx is solved by one of these recycling cysteines instead of GSH. The resulting disulfide is then either reduced by two molecules of GSH or by a thioredoxin reductase. If the target protein has an intra or inter-molecular disulfide, the mechanism is comparable to the one used by Trxs. The catalytic cysteine forms a transient mixed disulfide with one of the cysteines of the target protein, which is then resolved by the recycling cysteine. As above, final regeneration occurs via two molecules of GSH or via a thioredoxin reductase. Based on the active site sequence, Grxs were initially categorized into two classes, a dithiol (CP\[Y/F\]C motif) and a monothiol (CGFS motif) class (Rodriguez-Manzaneque et al., [@B29]) that were subsequently renamed classes I and II, respectively (Couturier et al., [@B10]). Analyses of Grx distribution and evolution in archaea, bacteria, and eukaryotes suggested that Grx domains of classes I and II have evolved through duplication and divergence from one initial gene present in the last common ancestor of all organisms (Alves et al., [@B1]). Two additional classes are found in land plants (Couturier et al., [@B10]). The Grx isoforms belonging to class III are specific to land plants and those from class IV are also present in few algae and animals. There is no biochemical or functional information on the latter Grxs yet. From genetic analyses, it appears that class III Grx isoforms also named CC-type Grxs or ROXY play a role in plant development and in pathogen defense mechanisms, reviewed in (Meyer et al., [@B25]). These functions are related to their ability to interact with TGA transcription factors, likely regulating their redox states. However, the biochemical properties of these Grxs have not been studied in detail yet, because of the difficulty to produce soluble recombinant proteins (Couturier et al., [@B9]). Only one study demonstrated that Arabidopsis thaliana GrxS10 is able to regenerate in vitro the mitochondrial type IIF peroxiredoxin (Finkemeier et al., [@B16]). The Grxs from class II are particular considering their involvement in the maturation of Fe-S clusters and in the regulation of iron homeostasis (Rouhier et al., [@B32]), which is likely related to their capacity to bind labile Fe-S clusters and to transfer them to target proteins (Bandyopadhyay et al., [@B4]). On the other hand, though many studies performed with class II Grxs from various sources indicated that they have no or very poor efficiency for the reduction of the traditionally tested glutathionylated substrates using a GR/GSH regeneration system, at least one study indicated that they have a reductase activity. Indeed, Chlamydomonas reinhardtii Grx3 is able to reduce in vitro a glutathionylated A~4~-glyceraldehyde-3-phosphate dehydrogenase (Zaffagnini et al., [@B48]). Interestingly, it can be regenerated by the ferredoxin-thioredoxin reductase but not by the GSH/GR system. The existence of alternative regeneration systems may explain why the characterization of such activities for class II Grxs has been retarded. Overall, the demonstrated capacity of several class I Grxs to provide electrons to a battery of enzymes, as peroxiredoxins and methionine sulfoxide reductases, enzymes that go through a catalytic cycle which involves glutathionylation of their catalytic cysteine (Rouhier et al., [@B34]; Vieira Dos Santos et al., [@B43]; Tarrago et al., [@B42]), suggests that they are the central and favorite deglutathionylating agents in cells. In plants, class I regroups six different glutaredoxin members called GrxC1, C2, C3, C4, S12, and C5, the latter being restricted to Brassicaceae (Couturier et al., [@B13]). The cytoplasmic GrxC1 and the plastidial GrxC5 are able to bind an \[Fe~2~S~2~\] cluster, at least in vitro into recombinant proteins (Rouhier et al., [@B36]; Couturier et al., [@B13]). The biochemical and structural characterization of the two plastidial isoforms, GrxS12 and GrxC5, confirmed that the apoforms of these Grxs reduce glutathionylated proteins with catalytic properties quite similar to other plant and non-plant class I Grxs (Couturier et al., [@B11], [@B13]). Moreover, poplar GrxS12 was shown to be sensitive and temporarily inactivated after treatments with oxidizing compounds such as hydrogen peroxide (H~2~O~2~), nitrosoglutathione (GSNO), and oxidized glutathione (GSSG) (Zaffagnini et al., [@B46]). Considering the very electronegative redox potential of the glutathione adduct formed on the catalytic Cys~A~ (−315 mV at pH 7.0), the reduction of GrxS12 necessitates a quite low redox potential of the GSH/GSSG couple in the chloroplast. Thus, GrxS12 may act as a stress-related redox regulator allowing glutathione to play a signaling role in some oxidizing conditions by maintaining the glutathionylation of its target proteins over the duration of the stress period. Several class I Grxs, C. reinhardtii Grx1 and Grx2 (Gao et al., [@B19]), A. thaliana GrxC1, C2 (Riondet et al., [@B28]) and GrxC5 (Couturier et al., [@B13]) and poplar GrxC1, GrxC4, and GrxS12 (Rouhier et al., [@B33], [@B36]; Couturier et al., [@B11]; Zaffagnini et al., [@B46]) have been partially or extensively characterized at the biochemical level, but some members have not yet been studied and there is no thorough and comparative study allowing to understand if the existence of several members is only related to their sub-cellular localization or to their expression pattern or whether some biochemical properties can partially explain their diversity. From the biochemical analyses performed here with the four dithiol class I Grxs existing in poplar, GrxC1 to GrxC4 (the fifth member, GrxS12, has a monothiol active site and was characterized previously), two sub-groups with distinct properties emerged. In addition, this work provides new insights on the existence of different oxidation forms involving the active site cysteines (Cys~A~ and Cys~B~) or an additional C-terminal cysteine (Cys~C~) and on their potential role for the reaction mechanism. Materials and methods {#s2} ===================== Site-directed mutagenesis ------------------------- The construction of pET3d expression plasmids containing poplar GrxC1, C2, C3, and C4 was described previously (Rouhier et al., [@B33], [@B36]). All cysteine residues of GrxC2, C3 were individually substituted into serines using two complementary mutagenic primers (listed in Table [S1](#SM7){ref-type="supplementary-material"}). The corresponding mutated proteins are called GrxC2 C23S, C26S, or C80S, GrxC3 C37S or C40S. The recombinant proteins GrxC1 C31S, C34S, C88S, and GrxC4 C27S and C30S were prepared previously (Rouhier et al., [@B33], [@B36]). Two additional mutants, in which only the catalytic cysteine is remaining, GrxC1 C34/88S and GrxC2 C26/80S, have also been cloned. Heterologous expression and purification of recombinant proteins in E. coli ----------------------------------------------------------------------------- For protein expression, cultures of the E. coli BL21(DE3) strain, containing the pSBET plasmid and co-transformed with the different recombinant pET3d plasmids, were successively amplified up to 2.4 L in LB medium at 37°C supplemented with 50 μg/mL of ampicillin and kanamycin. Induction of protein expression was performed at exponential phase by adding 100 μM isopropyl β-D-thiogalactopyranoside for 4 h at 37°C. After centrifugation (20 min at 4400 g), the cell pellets were resuspended in about 20 mL of TE NaCl buffer (30 mM Tris-HCl pH 8.0, 1 mM EDTA, 200 mM NaCl) and eventually conserved at −20°C. Cell lysis was performed by sonication and the soluble and insoluble fractions were separated by centrifugation for 45 min at 27,000 g. Purification of proteins was carried out in three steps. The soluble fraction was first precipitated by ammonium sulfate from 0 to 40% and then to 80% of the saturation. The 40--80% ammonium sulfate-precipitated fraction was subjected to gel filtration chromatography (ACA44 gel) equilibrated with TE NaCl buffer. After dialysis against TE (30 mM Tris-HCl pH 8.0, 1 mM EDTA) buffer and concentration, the interesting fractions were loaded to a DEAE (diethylaminoethyl) sepharose column equilibrated in TE buffer. GrxC1 WT, GrxC2 WT and their corresponding cysteinic mutants passed through the column, whereas GrxC3 WT, GrxC4 WT and their respective cysteinic mutants were retained and eluted using a linear 0--0.4 M NaCl gradient. The purest fractions, as judged by SDS-PAGE analysis, were pooled and dialyzed against TE buffer by ultrafiltration in Amicon cells equipped with a YM10 membrane. Finally, the fractions were concentrated and stored at −20°C until further use. Protein purity was checked by SDS-PAGE, and protein concentrations were determined spectrophotometrically using the corresponding molar extinction coefficients at 280 nm of 10,095 M^−1^ cm^−1^ for GrxC1 WT and its monocysteinic mutants, 9970 M^−1^ cm^−1^ for GrxC1 C34/88S, 8605 M^−1^ cm^−1^ for GrxC2 WT and its monocysteinic mutants, 8480 M^−1^ cm^−1^ for GrxC2 C26/80S, 7575 M^−1^ cm^−1^ for GrxC3 WT, 7450 M^−1^ cm^−1^ for GrxC3 C37S and C40S, 4595 M^−1^ cm^−1^ for GrxC4 WT and 4470 M^−1^ cm^−1^ for GrxC4 C27S and C30S. Enzymatic activities -------------------- The thioltransferase activities \[2-hydroxyethyl-disulfide (HED) and dehydroascorbate (DHA) assays\] were measured as described previously (Couturier et al., [@B13]) using as purified non-reduced proteins. Briefly, measurements were performed at 25°C in steady-state conditions by following NADPH oxidation at 340 nm in the presence of a Grx reducing system which is composed of NADPH, GR, and GSH. The reaction was started by adding Grx after a 3-min pre-incubation time and Grx activity was corrected by subtracting the spontaneous reduction rate observed in the absence of Grx. Because of the spontaneous reaction of GSH with the substrates, we can generally not work at saturating concentrations of all substrates using these assays. Enzyme and substrate concentrations used are indicated in the legend of figures and tables. The activity was expressed as nmol of NADPH oxidized/nmol of Grx/s using a molar extinction coefficient of 6220 M^−1^ cm^−1^ at 340 nm for NADPH. Three independent experiments were performed at each substrate concentration, and k~cat~ and apparent K~m~ values were calculated by non-linear regression using the Michaelis--Menten equation using the program GraphPad Prism 4. Preparation and analysis of oxidized Grxs ----------------------------------------- Around 10 mg of protein was reduced using 40 mM DTT in 500 μl of 30 mM Tris-HCl pH 8.0 buffer for 2 h at 25°C followed by desalting on G25 column pre-equilibrated with 30 mM Tris-HCl pH 8.0 buffer. Oxidized proteins were prepared by incubating 500 μM reduced poplar Grxs in ca 200 μL with a 10 fold excess GSSG or H~2~O~2~ or a 5 fold excess GSNO for 2 h at 25°C before desalting on G25 columns. The protein oxidation state was then analyzed by electrospray ionization mass spectrometry analysis as described previously (Couturier et al., [@B13]). For alkylation assays, reduced Grxs treated or not with oxidants were diluted to 10 μM into 50 μ l of 30 mM Tris-HCl pH 8.0 buffer and precipitated on ice for 30 min with one volume of 20% trichloroacetic acid (TCA). After centrifugation (10 min at 13,000 g) and washing with 100% acetone, the pellet was resuspended into 10 μl of 100 mM Tris-HCl pH 8.0, 1% SDS containing 2 mM of methoxyl-PEG maleimide of 2 kDa (mPEG maleimide) which alkylates free thiol groups. The protein mixture was then separated on non-reducing 15% SDS-PAGE. For midpoint redox potential titrations, Cys~A~ glutathionylated forms were obtained by the above described GSNO treatment of the double cysteinic mutants (GrxC1 C34/88S, GrxC2 C26/80S) and of the monocysteinic mutants (GrxC3 C40S and GrxC4 C30S). The intramolecular disulfide bond (Cys~A~-Cys~B~) containing proteins were obtained by the above described H~2~O~2~ treatment but using GrxC1 C88S, GrxC2 C80S and GrxC3 and GrxC4 WT isoforms. PK~a~ determination of the catalytic cysteine ------------------------------------------------- The pK~a~ measurements of N-terminal active site cysteines (Cys~A~) have been performed with WT Grx isoforms following a procedure described in (Gallogly et al., [@B17]). Briefly, 3 μM reduced Grx was incubated with or without 300 μM iodoacetamide (IAM) in 100 mM sodium citrate or phosphate buffers ranging from pH 2.0 to 7.0. Following this pre-incubation step, Grx activity was determined by adding an aliquot of the pre-incubation mixture to the HED assay described above. The percentages of remaining activity at each pH were determined by comparing the activity of the enzyme incubated with and without IAM and an adaptation of the Henderson--Hasselbach equation (Gallogly et al., [@B17]) was used for pK~a~ value determination using the program GraphPad Prism 4. Midpoint redox potential (E~m~) determination ------------------------------------------------- Oxidation-reduction titrations using the fluorescence of the adduct formed between protein free thiols and monobromobimane (mBBr) were carried out at ambient temperature in 200 μl of 100 mM HEPES pH 7.0 buffer containing 10 μM of oxidized proteins, either glutathionylated or with an intramolecular disulfide-bond, and defined mixtures of oxidized and reduced DTT or glutathione for more positive values to set the ambient potential (E~h~). Total concentration of DTT or glutathione was 2 mM. After 2 h incubation, mBBr was added at a final concentration of 2 mM and the reaction was carried out in the dark for 20 min. The reaction mixture was then precipitated on ice for 30 min with one volume of 20% TCA. After centrifugation (10 min at 13,000 g) and washing with 100% acetone, the pellet was resuspended into 400 μl of 100 mM Tris-HCl pH 8.0, 1% SDS. Fluorescence emission of the resulting solution was measured at 472 nm after excitation at 380 nm using a variant Cary Eclipse (Agilent). Values were transformed into percentages of reduced protein and fitted to the Nernst equation using non-linear regression for E~m~ value determination. Results {#s3} ======= Kinetic comparison of poplar class I Grxs ----------------------------------------- An amino acid sequence alignment of the predicted mature forms of all poplar class I Grxs reveals several conserved residues, particularly those around the active sites and those involved in glutathione binding (Figure [1](#F1){ref-type="fig"}). However, GrxC1 and GrxC2 present CGYC and CPFC active site motifs, respectively, different from the CPYC active site found in GrxC3 and GrxC4. Interestingly, contrary to GrxC3 and GrxC4, GrxC1 and GrxC2 isoforms possess an additional cysteine (Cys~C~) located in their C-terminal part in a IGGCD motif, as in human Grx1, E. coli Grx3 and the two other class I Grxs found in plants, GrxC5 and GrxS12 (Figure [1](#F1){ref-type="fig"}). With the aim of comparing the redox and kinetic properties of all dithiol Grxs (GrxC5 is not found in poplar) and understanding why these duplicated genes have been retained during evolution, we have performed a thorough biochemical analysis of the corresponding recombinant proteins. ![Amino acid alignment of class I glutaredoxins. The alignment was done using ClustalW from sequences of the recombinant proteins that are deprived of putative targeting sequences. The strictly conserved amino acids are depicted in white on black. Other conservative amino acid changes are indicated in black on gray. Residues involved in GSH binding based on the 3D structures of poplar GrxC4 and GrxS12 and Arabidospis GrxC5 are marked by an asterisk.](fpls-04-00518-g0001){#F1} The catalytic properties of each isoform were first determined by measuring their reductase activity using the classical HED and DHA assays, which measure the reduction of glutathionylated-β-mercaptoethanol or of dehydroascorbate, respectively. All proteins display a better catalytic efficiency (k~cat~/K~m~) in the HED test with a difference of two orders of magnitude (Table [1](#T1){ref-type="table"}). In both assays, the kinetic parameters determined for each protein follow the same tendency. GrxC3 and C4 have quite comparable turnover numbers, with k~cat~ values about 10 fold higher than those of GrxC1 and C2. In terms of catalytic efficiencies, the dichotomy between GrxC1/C2 and GrxC3/C4 is clear but it is often below this factor of 10 because GrxC3 and C4 have higher apparent K~m~ values for both substrates. Overall, GrxC1 always exhibited the lowest catalytic efficiency and GrxC4 the highest. Concerning the apparent affinity for GSH, although there are differences in the K~m~ values for GSH between the four isoforms, ranging from \~2.0 mM for GrxC4 to \~3.9 mM for GrxC2, the GrxC1/C2 and GrxC3/C4 dichotomy is not visible using this parameter. ###### Kinetic parameters of poplar dithiol class I glutaredoxins. PtGrxC1 PtGrxC2 PtGrxC3 PtGrxC4 PtGrxS12[^a^](#TN1){ref-type="table-fn"} AtGrxC5[^a^](#TN1){ref-type="table-fn"} --------- ------------------------------------------ ------------- ------------- ------------- ---------------------------------------------- --------------------------------------------- ------------- β-ME-SG K~m~ (mM) 0.10 ± 0.01 0.06 ± 0.01 0.50 ± 0.06 0.32 ± 0.04 0.31 ± 0.04 0.20 ± 0.02 k~cat~(s^−1^) 7.9 ± 0.2 11.7 ± 0.1 77.0 ± 4.4 90.3 ± 4.9 23.1 ± 0.9 1.21 ± 0.03 k~cat~/K~m~(x 10^5^ M^−1^.s^−1^) 0.79 1.95 1.54 2.82 0.745 0.06 DHA K~m~ (mM) 0.55 ± 0.06 0.37 ± 0.03 1.14 ± 0.13 0.60 ± 0.09 0.38 ± 0.19 0.21 ± 0.03 k~cat~(s^−1^) 0.6 ± 0.1 1.3 ± 0.1 6.5 ± 0.4 8.2 ± 0.6 1.7 ± 0.2 0.23 ± 0.01 k~cat~/K~m~(x 10^3^ M^−1^.s^−1^) 1.09 3.51 5.70 13.67 4.6 1.1 GSH K~m~ (mM) 2.76 ± 0.26 3.87 ± 0.53 3.60 ± 0.42 1.96 ± 0.28 4.0 ± 0.5 nd k~cat~(s^−1^) 13.8 ± 0.6 30.1 ± 2.3 131.3 ± 8.3 102.9 ± 6.5 92.8 ± 5.3 nd k~cat~/K~m~(x 10^3^ M^−1^.s^−1^) 5.00 7.78 36.47 52.50 23.20 nd The apparent K~m~ values for β-ME-SG (glutathionylated β-mercaptoethanol) were determined using a HED concentration range of 0.1--1 mM in presence of 2 mM GSH and 100 nM enzyme for GrxC1, 50 nM for GrxC2, 10 nM for GrxC3 and GrxC4. The apparent K~m~ values for DHA were determined using a DHA concentration range of 0.1--1 mM in presence of 2 mM GSH and 2 μM enzyme for GrxC1, 1 μM for GrxC2, 200 nM for GrxC3, and 150 nM for GrxC4. The apparent K~m~ values for GSH were determined using a GSH concentration range of 0.25--5 mM in the presence of 0.7 mM HED and 100 nM enzyme for GrxC1, 50 nM for GrxC2, 10 nM for GrxC3 and GrxC4. The k~cat~ and K~m~ values were calculated by non-linear regression using the Michaelis--Menten equation. Values are the means ± SD of three replicates. Data extracted from the following references Couturier et al. ([@B11], [@B13]), Zaffagnini et al. ([@B46]). nd, not determined. Then, the involvement for the catalytic mechanism of the different cysteine residues present in each isoform was investigated by comparing the activity of WT proteins to their corresponding cysteinic mutants using similar assays at fixed substrate concentrations (Figure [2](#F2){ref-type="fig"}). As expected, all monocysteinic mutants for the catalytic cysteine Cys~A~ (GrxC1 C31S, GrxC2 C23S, GrxC3 C37S, and GrxC4 C27S) were inactive, confirming that it is indispensable for the reductase activity. On the contrary, mutating Cys~C~, in positions 88 and 80 in GrxC1 and GrxC2, respectively, did not affect catalytic activity. Finally, the mutation of Cys~B~ has a different impact depending on the isoforms. The protein variants for this cysteine in GrxC1 and C2 (GrxC1 C34S and GrxC2 C26S) were 5 fold more efficient and 2--3 fold more efficient than the corresponding WT proteins in the HED and DHA assays, respectively. On the other hand, the mutation of this cysteine in GrxC3 and GrxC4 has almost no impact, since, only a slight decrease was observed in the HED assay for the GrxC3 C40S variant. ![Reductase activity of GrxC1, C2, C3, and C4 and of cysteine-to-serine variants. (A) HED reduction was measured in the presence of 2 mM GSH and 0.7 mM HED using either 100 nM GrxC1, 20 nM for GrxC1 C34S, 50 nM GrxC2, 10 nM for GrxC2 C26S, 10 nM GrxC3, or 10 nM GrxC4. (B) DHA reductase activity was measured in the presence of 2 mM GSH and 1 mM DHA using either 2 μM GrxC1, 1 μM for GrxC1 C34S, 1 μM GrxC2, 250 nM GrxC2 C26S, 200 nM GrxC3, or 150 nM GrxC4. The data are expressed as nmol NADPH oxidized per nmol enzymes per s and represent means ± SD of three separate experiments.](fpls-04-00518-g0002){#F2} Oxidative modifications of cysteine residues -------------------------------------------- Several observations prompted us to analyze the sensitivity of all dithiol class I Grxs to oxidizing treatments and the resulting changes in their redox state. First, the mutation of Cys~B~ increased the catalytic efficiency of GrxC1 and C2 but not of GrxC3 and C4 suggesting possible difference in the reaction mechanisms. Second, although Cys~C~ does not influence their catalytic activity, the migration of the purified proteins in non-reducing SDS PAGE gels indicated the formation of covalent linkages for GrxC1 and C2 that are not visible for GrxC3 and C4 (data not shown). For this purpose, reduced proteins have been treated for 2 h with a 10 fold excess of GSSG or H~2~O~2~, or a 5 fold excess of GSNO. The resulting redox state was analyzed by mass spectrometry and by protein migration on non-reducing SDS PAGE with or without thiol alkylation by 2 kDa mPEG maleimide, which allows distinguishing reduced and oxidized forms of Grxs. Since reduced mPEG-alkylated proteins and covalent dimers of GrxC1 and GrxC2 migrate roughly at the same position, controls without alkylation have been included (Figure [3](#F3){ref-type="fig"}). With non-alkylated proteins, we confirmed that an H~2~O~2~ treatment, but not the GSSG and GSNO treatments, promoted the formation of covalent dimers of GrxC1 and GrxC2 (Figures [3A,B](#F3){ref-type="fig"}) but not of GrxC3 and GrxC4 (Figures [3C,D](#F3){ref-type="fig"}). Only a small part of GrxC2 dimer is visible after these 2 h GSSG and GSNO treatments. From the procedure that included an alkylation step, we can further conclude that all treatments led essentially to a complete oxidation of all thiol groups present on all Grxs since almost no shift of migration was observed. A minor form with one free thiol group is visible for GrxC2 after a GSSG treatment. It is even more discrete upon GSNO treatment. For GrxC3 and C4, these oxidized monomeric forms are interpreted as containing intramolecular disulfides between active site cysteines (Cys~A~-Cys~B~) (Figures [3C,D](#F3){ref-type="fig"}). For GrxC1 and C2, these two cysteines should also form this intramolecular disulfide, and Cys~C~ is either modified by a glutathione or a nitroso (in the case of GSNO) adduct as confirmed by mass spectrometry analyses (see below) or it is involved in an intermolecular disulfide upon H~2~O~2~ treatment. From similar experiments performed with the GrxC1 C88S and GrxC2 C80S mutants (Figures [3E,F](#F3){ref-type="fig"}), the absence of covalent dimers indicated that Cys~C~ is indeed involved in dimerization. Moreover, in the context of these variants, after GSSG and GSNO treatments, the proteins mainly contain an active site disulfide bond whereas a small part of the proteins has one remaining free thiol, likely corresponding to glutathionylated intermediates. ![Sensitivity of poplar Grxs to oxidative treatments. Pre-reduced Grxs were subjected to 2 h oxidative treatments with H~2~O~2~, GSSG, and GSNO as described in the "materials and methods" section. Prior to separation on SDS-PAGE gels in non-reducing conditions, GrxC1 WT (A), GrxC2 WT (B), GrxC3 WT (C), GrxC4 WT (D), GrxC1 C88S (E), and GrxC2 C80S (F) were alkylated or not with 2 kDa mPEG maleimide. The shift observed following the alkylation of the free thiol groups is larger than expected (ca. 4 kDa instead of 2 kDa), but this has been already observed previously (Chibani et al., [@B8]). The numbers on the right correspond to the number of alkylated thiols that remained free after reducing or oxidizing treatment of Grxs. The stars in panels (A) and (B) indicate the position of covalent dimer.](fpls-04-00518-g0003){#F3} In order to identify the type of protein modification formed by these oxidative treatments, all proteins have been analyzed by mass spectrometry (Table [2](#T2){ref-type="table"} and Figures [S1](#SM1){ref-type="supplementary-material"}--[S6](#SM6){ref-type="supplementary-material"}). First, all reduced proteins gave a single species with a mass decrease of ca 131 Da compared to the calculated theoretical molecular masses. It does undoubtedly correspond to the cleavage of the first methionine as expected from the presence of an alanine in second position. The results obtained with GrxC3 and GrxC4 are more easily interpreted since these proteins contain only two cysteines. In the presence of H~2~O~2~, and in accordance with the absence of detected free thiols in the previous experiments (Figures [3C,D](#F3){ref-type="fig"}), GrxC3 and GrxC4 are exclusively detected as a monomeric oxidized form with a decrease of \~2 Da confirming the formation of a Cys~A~-Cys~B~ intramolecular disulfide bond. The same forms were primarily obtained upon GSSG and GSNO treatments but a minor monomeric form with a mass increment of 305 Da corresponding to one glutathione adduct was also detected (Table [2](#T2){ref-type="table"}). It was not detected on gels presumably because of the presence of minute amounts. Nevertheless, this suggests that the glutathionylated form represents an intermediate for the formation of the intramolecular disulfide. ###### Electrospray ionization mass spectrometry analysis of the redox states of poplar class I Grxs. Protein Theoretical size (Da) Theoretical size without Met (Da) DTT~red~ treatment GSSG treatment Relative percentage values (%) GSNO treatment Relative percentage values (%) H~2~O~2~ treatment Relative percentage values (%) Deduced redox state -------------- --------------------------- --------------------------------------- ------------------------ ----------------------- ------------------------------------ ----------------------- ------------------------------------ ------------------------ ------------------------------------ ------------------------- ---- ------------------ PtGrxC1 WT 12,514.3 12,383.1 12,382.7 12,686.3 (+303.6) 76.4 12,380.0 (−2.7) 5.7 12,412.2 (+29.5) 72.1 SH SH SH 12,993.8 (+611.1) 23.6 12,686.3 (+303.6) 81.8 24,756.5 (×2) 27.9 12,993.4 (+610.7) 12.5 S-S SH S-S SO~2~H S-S SSG SSG SH SSG S-S S-S (interchain) PtGrxC2 WT 12,178.9 12,047.7 12,047.3 12,045.0 (−2.3) 11.1 12,044.9 (−2.4) 23.7 12,077.2 (+32.2) 50 SH SH SH 12,351.5 (+304.2) 68.4 12,350.7 (+303.4) 76.3 24,088.9 (×2) 50 12,657.9 (+610.6) 20.5 S-S SH S-S SO~2~H S-S SSG SSG SH SSG S-S S-S (interchain) PtGrxC3 WT 12,516.2 12,385.0 12,384.7 12,382.7 (−2.0) 66.3 12,382.9 (−1.8) 75.9 12,382.7 (−2.0) 100 SH SH 12,690.0 (+305.3) 33.7 12,690.0 (+305.3) 24.1 S-S SSG SH PtGrxC4 WT 12,526.1 12,394.9 12,394.4 12,392.0 (−2.4) 62.5 12,392.1 (−2.5) 90.8 12,392.3 (−2.7) 100 SH SH 12,699.6 (+305.2) 37.5 12,699.2 (+304.8) 9.2 S-S SSG SH PtGrxC1 C88S 12,498.2 12,367.0 12,366.6 12,364.7 (−1.9) 64.0 12,364.5 (−2.1) 79.6 12,364.6 (−2.0) 100 SH SH 12,671.8 (+305.2) 36.0 12,672.0 (+305.4) 20.4 S-S SSG SH PtGrxC2 C80S 12,162.8 12,031.7 12,030.6 12,029.3 (−1.3) 81.5 12,028.8 (−1.8) 100 12,028.7 (−1.9) 100 SH SH 12,336.5 (+305.9) 18.5 S-S SSG SH Reduced proteins and proteins treated with GSNO, GSSG, or H~2~O~2~ were analyzed by mass spectrometry. The mass differences are indicated between parentheses. The mass accuracy is generally ±0.5 Da. Comparing mass differences and migration profiles of corresponding alkylated proteins under non-reducing conditions enabled to deduce the redox state of each cysteine residue. When several oxidized forms are detected and with clear different abundance, the major form is indicated in bold. The relative percentage of each form has been deduced from the spectra assuming equal ionization of all GRX forms. The results obtained for GrxC1 and GrxC2 are slightly more complex. In response to an H~2~O~2~ treatment, two different species are formed for both proteins. In accordance with the results from alkylation experiments, a major form corresponds to a disulfide-bridged dimer having also an intramolecular Cys~A~-Cys~B~ disulfide bond. The detected monomeric forms, which have no free thiols or a very small fraction as observed for GrxC2, exhibited a mass increment of 29.5 Da for GrxC1 and 32.2 Da for GrxC2, suggesting the presence of two oxygen atoms. Considering that this overoxidation is not observed for GrxC1 C88S and GrxC2 C80S variants which only form monomeric proteins with an intramolecular disulfide (as indicated by the mass decrease of 2 Da), the results obtained for the WT proteins are best interpreted as the formation of an intramolecular disulfide between the active site cysteines (Cys~A~-Cys~B~) and of a sulfinic acid on Cys~C~. In the presence of GSSG or GSNO, GrxC1 and GrxC2 exist only under monomeric forms but with variable oxidation states. Except for the GSSG-treated GrxC1, there is a minor form with a mass decrease of about 2 Da consistent with the formation of an intramolecular disulfide. Another minor form, found in all samples except the GSNO-treated GrxC2, exhibited a mass increment of 610.5--611 Da, which corresponded to the presence of two glutathione adducts. However, the major species formed by these two treatments is a monomer with a mass increment of about 303.5 Da which is consistent with the presence of one glutathione adduct on Cys~C~ in addition to the Cys~A~-Cys~B~ disulfide bond. Accordingly, the major species detected in protein variants mutated for Cys~C~ (Grx C1 C88S and GrxC2 C80S) is the one with an intramolecular disulfide (decrease of 2 Da upon GSSG or GSNO treatments). The glutathionylated forms detected by mass spectrometry were visible on gels (see Figures [3E,F](#F3){ref-type="fig"}). The in vitro observation of intermolecular disulfide bonds in GrxC1 and C2, involving Cys~C~, raised the question of their possible reduction. Thus, we tested the ability of DTT but also of a physiological GSH reducing system composed of NADPH, GR, and GSH to reduce this intermolecular disulfide, the formation of which was initiated by a 2h H~2~O~2~ treatment. As shown in Figure [4](#F4){ref-type="fig"}, both DTT and GSH are able to reduce GrxC1 and GrxC2 dimers to a similar extent. Whether minute amounts of reduced monomeric Grx initially present or formed can catalytically increase the reduction of dimeric Grxs in the presence of GSH cannot be excluded. Overall, it indicates that the formation of these disulfide-bridged forms of GrxC1 and GrxC2 isoforms is reversible, notably by GSH, a relevant cellular physiological reductant. ![GSH-dependent reduction of the intermolecular disulfide in poplar GrxC1 and GrxC2. The ability of GSH to reduce poplar disulfide-bridged dimer GrxC1 and disulfide-bridged dimer GrxC2 generated by a 2 h H~2~O~2~ treatment was evaluated in the presence of 1 mM DTT and 1 mM GSH coupled to NADPH and GR. The lanes Ox represent the initial oxidized proteins. Proteins were separated on SDS-PAGE gels under non-reducing conditions.](fpls-04-00518-g0004){#F4} Redox properties of class I glutaredoxins ----------------------------------------- As in other thiol-dependent oxidoreductases, the catalytic efficiency of class I Grxs should be at least partially governed by thermodynamic parameters as the redox midpoint potentials of dithiol-disulfide couples and by the reactivity of the catalytic cysteine which is dependent on its pK~a~. Hence, we determined the Cys~A~ pK~a~ of the four Grx isoforms using a method relying on iodoacetamide which is an alkylating reagent reacting with thiolates but not thiols. The pH-dependent inactivation of WT Grx by iodoacetamide was followed by comparing the activity of WT proteins in the HED assay, after pre-incubation of the protein in different buffers ranging from pH 2.0 to 7.0 in the presence or not of iodoacetamide. From these titration curves, we obtained pK~a~ values of 5.3 ± 0.1 for GrxC1, 5.0 ± 0.1 for GrxC2, 4.6 ± 0.1 for GrxC3, and 4.6 ± 0.1 for GrxC4 (Figure [5](#F5){ref-type="fig"}). These results indicate first that at physiological pH, the thiolate ion will be the dominant species for all Grxs. Although the differences are not huge, it also confirmed the existence of two subgroups, GrxC1 and C2 having higher pK~a~compared to GrxC3 and C4. ![*pK~a~ determination of the catalytic cysteine of poplar GrxC1, C2, C3, and C4. Reduced GrxC1 (A), GrxC2 (B), GrxC3 (C), and GrxC4 (D)* WT proteins were incubated in different buffers ranging from pH 2.0 to 7.0 in presence or not of iodoacetamide (IAM) prior to measurement of their activity by using the HED assay. The percentages of remaining activity at each pH were determined by comparing the activity of the enzyme incubated with and without IAM and an adaptation of the Henderson--Hasselbach equation was used for pK~a~ value determination. The data are represented as mean ± SD of three separate experiments.](fpls-04-00518-g0005){#F5} Two disulfide bond forms may be catalytically relevant for all these Grxs, the one with a glutathionylated Cys~A~ and the one with a Cys~A~-Cys~B~ intramolecular disulfide bond. For measuring the midpoint redox potentials (E~m~) of each of these disulfide/dithiol couples, we used different protein variants and two distinct oxidative treatments in order to discriminate between these two oxidized forms. To obtain proteins containing an intramolecular disulfide bond, reduced Grxs possessing only the two active site cysteines (GrxC1 C88S, GrxC2 C80S, GrxC3 WT, and GrxC4 WT) have been treated with H~2~O~2~ before desalting on G25 columns. On the other hand, glutathionylated proteins have been obtained by treating reduced proteins in which only Cys~A~ was remaining (GrxC1 C34/88S, GrxC2 C26/80S, GrxC3 C40S, and GrxC4 C30S) with GSNO before desalting. The E~m~ values determined at pH 7.0 for the intramolecular disulfide range from −233 mV for GrxC3 to −263 mV to GrxC1 (Figure [6](#F6){ref-type="fig"}). On the other hand, only little variations were visible when measuring E~m~ values for the glutathione adduct formed on the catalytic cysteine. They varied from −242 mV for GrxC3 to −254 mV for GrxC1. Overall, the intramolecular active site disulfide bond and glutathionylated forms of GrxC2 and GrxC4 present similar E~m~ values. A 10 mV difference was measured between both forms for GrxC1 and C3, the more electronegative redox potential being the active site disulfide in GrxC1 and the glutathione adduct in GrxC3. However, the differences are small and should not explain the observed differences in catalytic efficiencies. In reducing cellular environments, all these oxidation forms should be compatible with glutathione reduction. Considering their possible targeting to secretory pathways, this may not be the case for Grx3 and C4 if they are in the endoplasmic reticulum or in the apoplasm. ![Determination of redox potential of oxidized poplar GrxC1, C2, C3, and C4. Representative curve of oxidation-reduction titration of Cys~A~-Cys~B~ intramolecular disulfide forms (black circles) and glutathionylated Cys~A~ (white circles) forms of poplar GrxC1 (A), GrxC2 (B), GrxC3 (C), and GrxC4 (D) at pH 7.0. The titration was carried out using a total DTT or GSH concentration of 2 mM for 2 h and resulting free thiol groups were labeled by mBBr. Values in the text are the means ± SD of three replicates.](fpls-04-00518-g0006){#F6} Discussion {#s4} ========== Duplication and neo-functionalization into class I glutaredoxins ---------------------------------------------------------------- Among all living organisms, higher plants have the highest number of genes coding for Grxs. In addition to specific isoforms such as those belonging to the class III, land plants also possess an expanded class I, usually with 5--6 isoforms (Couturier et al., [@B10]). From previous phylogenetic analyses, it was quite clear that, there are three Grx couples, GrxC1/C2, GrxC3/C4, and GrxC5/S12 that probably evolved by more or less recent duplication (Couturier et al., [@B10]). Interestingly, GrxC1 and C2 proteins are cytoplasmic proteins (Riondet et al., [@B28]), GrxC3 and C4 have N-terminal extensions predicted to target the proteins to secretory pathways but the final destination is not known, and GrxC5 and S12 are chloroplastic (Couturier et al., [@B11], [@B13]). The case of GrxC1 and C2 is interesting. Whereas GrxC2 prototypes are found in all land plants (mosses, lycophytes, monocots, dicots), GrxC1 is specifically found in dicots. A redundancy between both isoforms is suggested in A. thaliana by the observation that single knock-out mutants in grxc1 or grxc2 are aphenotypic, whereas a double mutant is lethal at an early stage after pollinization (Riondet et al., [@B28]). Considering that GrxC1, but not GrxC2, is able to bind an Fe-S cluster (at least in vitro), this would rather suggest that the essential nature of these two paralogs is linked to their reductase activity. These results point to the need of investigating at the genetic level a possible redundancy among the GrxC3/C4 and GrxC5/12 couples. This study primarily focused on the biochemical properties of dithiol class I Grxs since a detailed structure-function analysis was previously achieved on the two other prototypes of class I Grxs, the poplar monothiol GrxS12 and the Arabidopsis dithiol GrxC5 (Couturier et al., [@B11], [@B13]). GrxC5 isoforms are only found in species of the Brassicaceae family. We have demonstrated in particular that, although being very similar in sequence due to their recent duplication, a single substitution in the active site sequence (from WCSYS in GrxS12 to WCSYC in GrxC5) explains the capacity of GrxC5, but not GrxS12, to bind an \[Fe~2~-S~2~\] cluster into a dimer. Among other class I Grxs, only GrxC1 can also exist under a holoform binding an Fe-S cluster (Rouhier et al., [@B36]). In this case, the mutational analysis showed that the presence of the glycine in the CGYC active site instead of a proline in other Grxs (CPFC in GrxC2 or CPYC in GrxC3 and C4) was sufficient to explain this property. Overall, this indicates that subtle sequence differences affecting key residues may provide different properties to otherwise quite close proteins. This prompted us to further analyze the various glutaredoxins of this class as GrxC2 and GrxC3 have been only poorly studied so far. Two subgroups in dithiol Grxs exhibit distinct kinetic properties ----------------------------------------------------------------- Considering the divergence observed in the amino acid sequences, the first aim of this work was to perform a thorough comparative analysis of the biochemical properties of GrxC1, GrxC2, GrxC3, and GrxC4 from poplar. All these proteins exhibit deglutathionylation or reductase activities using classical HED and DHA assays with catalytic efficiencies (10^5^ M^−1^ s^−1^ for HED and 10^3^--10^4^ M^−1^ s^−1^ for DHA) in the same range as those previously reported for plant and non-plant class I Grxs (Table [3](#T3){ref-type="table"}). Two subgroups can however be distinguished, GrxC1 and GrxC2 appeared to be less efficient catalysts than GrxC3 and GrxC4. The difference is particularly clear when considering the turnover numbers (k~cat~) and the k~cat~/K~DHA~ and k~cat~/K~GSH~. This is not visible with k~cat~/K~β\ −ME−SG~ because the differences in the turnover number are compensated in this case by differences in apparent affinity in favor of GrxC1 and C2. An observation that may be interesting to understand these differences is that GrxC1/C2 and GrxC3/C4 should have quite different charge distribution at the protein surface that could account for substrate recognition and thus for the observed differences in catalytic efficiency. Indeed, their behavior on ion exchange chromatography indicates that GrxC1 and GrxC2 should be more basic having a pI close or higher than 8, whereas GrxC3 and GrxC4 should be more acidic having a pI value below 8 (theoretical pI estimations are indeed comprised between 5 and 6). ###### Kinetic and biochemical parameters of selected plant and non-plant class I Grxs. Name Active site sequence HED DHA GSH *pK~a~* **E~m~ (pH7.0) (mV) References** ---------- -------------------------- ------------- -------------- ------------- --------------- --------------------------- ---------------- ----------- -------------------- -------------- ------------------ ------ ----------------------------------------------------------------------- AtGrxC1 YCGYC 0.74 ± 0.01 38.91 ± 0.43 5.3 × 10^4^ 0.08 ± 0.01 3.03 ± 0.82 3.8 × 10^4^ nd nd nd nd nd Riondet et al., [@B28] AtGrxC2 YCPYC 0.34 ± 0.02 18.44 ± 0.57 5.4 × 10^4^ 0.08 ± 0.01 3.26 ± 0.03 4.1 × 10^4^ nd nd nd nd nd Riondet et al., [@B28] ScGrx1 YCPYC 0.14 3.5 2.5 × 10^4^ nd Nd nd 6.2 17.1 (HED) 2.75 × 10^3^ 3.2 to 4.0 ± 0.2 nd Discola et al., [@B15] ScGrx2 YCPYC 0.6 85 1.4 × 10^5^ nd nd nd 0.9 129 (HED) 1.43 × 10^5^ 3.1 to 3.5 ± 0.2 nd Discola et al., [@B15] CrGrx1 HCPYC 0.34 ± 0.04 30.40 ± 4.31 8.9 × 10^4^ 0.39 ± 0.03 1.67 ± 0.09 4.3 × 10^3^ 2.6 ± 0.5 161.6 ± 15.0 (HED) 6.1 × 10^4^ 3.9 ± 0.1 nd Zaffagnini et al., [@B48] CrGrx2 YCPYC 0.20 ± 0.04 7.10 ± 0.27 3.5 × 10^4^ 0.17 ± 0.05 0.84 ± 0.06 4.9 × 10^3^ 3.7 ± 0.7 26.5 ± 4.1 (HED) 7.1 × 10^3^ 4.8 ± 0.1 nd Gao et al., [@B19] HsGrx1 TCPYC 1.07 8.16 7.6 × 10^3^ nd Nd nd 2.2 293 (HED) 1.4 × 10^5^ 3.6 −232 Johansson et al., [@B23]; Jao et al., [@B22]; Sagemark et al., [@B38] HsGrx2 SCSYC 0.11 1.3 1.2 × 10^4^ nd Nd nd 5.9 71.3 (HED) 1.2 × 10^4^ 4.6 −221 Jao et al., [@B22]; Sagemark et al., [@B38]; Gallogly et al., [@B17] nd, not determined. In order to further explain the differences between poplar class I Grxs, we determined the pK~a~ value of their catalytic cysteine, as it was proposed that reactivity of human and Chlamydomonas class I Grxs is partially related to the pK~a~ of N-terminal active site cysteine (Gallogly et al., [@B17]; Gao et al., [@B19]). The values obtained for poplar Grxs are globally close to those reported for other plant and non-plant class I Grxs (Table [3](#T3){ref-type="table"}), including the value of 3.9 determined for poplar GrxS12 using the same procedure (Zaffagnini et al., [@B46]; Roos et al., [@B30]). Interestingly, a difference of ca 0.5 unit is visible between GrxC1/C2 (pK~a~ around 5) and GrxC3/C4 (pK~a~ around 4.6). Such differences between paralogs have been described already since C. reinhardtii Grx1, S. cerevisiae Grx2, and human Grx1 exhibit higher turnover numbers often translated into higher catalytic efficiencies compared to C. reinhardtii Grx2, S. cerevisiae Grx1, and human Grx2, respectively (Table [3](#T3){ref-type="table"}). For the human and C. reinhardtii enzymes, it is also associated to a difference in pK~a~ values of ca 1 unit, the enzymes with the lowest pK~a~ being the most efficient. According to the Brønsted theory, in a thiol-disulfide exchange reaction, each unit decrease in the pK~a~ value of the leaving group is responsible for a 4 fold increase in the second-order rate constant (Szajewski and Whitesides, [@B40]). This could fit with the conclusion that differences in pK~a~ values between class I poplar Grx isoforms are at least partially responsible for the differences in catalytic efficiencies observed. On the other hand, only small differences were observed for the oxidation-reduction midpoint potentials of both the glutathione adduct or the intramolecular disulfide and they are not discriminating GrxC1/C2 vs. GrxC3/C4. The measured values (from −230 to −260 mV) are in the same range than those reported for E. coli Grx1 (−233 mV) and human Grx1 and 2 (−232 and −221 mV) (Table [3](#T3){ref-type="table"}) (Aslund et al., [@B2]; Sagemark et al., [@B38]). The role of the recycling Cys~B~: monothiol or dithiol reaction mechanism? -------------------------------------------------------------------------- Interestingly, the differences in protein activities between GrxC1 or C2 and GrxC3 or C4 are attenuated when Cys~B~ of GrxC1 or C2 is removed (see Figure [2](#F2){ref-type="fig"}). This has been reported also for Arabidopsis GrxC5, yeast Grx1, human Grx1 and Grx2, and pig Grx (Yang and Wells, [@B45]; Yang et al., [@B44]; Johansson et al., [@B23]; Gallogly et al., [@B17]; Discola et al., [@B15]; Couturier et al., [@B13]). In clear contrast, mutating this cysteine in yeast Grx2, E. coli Grx1 and Grx3 led to a decrease of Grx activity (Bushweller et al., [@B6]; Nordstrand et al., [@B26]; Peltoniemi et al., [@B27]; Discola et al., [@B15]; Saaranen et al., [@B37]). For E. coli Grx1 and by extension for Grxs for which mutation of this cysteine decreased their activity, this decrease was attributed to a change in glutathione recognition, binding, or affinity, because Cys~B~ would determine the glutathione specificity of the glutathionylated Grx reduction step (Saaranen et al., [@B37]). For GrxC1 and GrxC2 and other eukaryote Grxs mentioned above in which the mutation of this cysteine increased their activity, the data indicates that the presence of Cys~B~ slows down the reaction for some reason. It may eventually modify the pK~a~ of Cys~A~. This has been experimentally confirmed for a pig Grx, for which the pK~a~ of the catalytic Cys22 is 3.8 in a WT protein but 4.9 and 5.9 in C25S and C25A variants (Yang and Wells, [@B45]). Another possibility is that a Cys~A~-Cys~B~ intramolecular disulfide is formed during activity measurements either because it constitutes a regular step of the reaction cycle in some situations, especially for dithiol substrates or because it is a side reaction reflecting the competition between Cys~B~ and GSH for the reduction of the glutathionylated Cys~A~ (Gallogly et al., [@B18]). Then, the reduction of this Cys~A~-Cys~B~ intramolecular disulfide intermediate would require two molecules of GSH vs. one molecule of GSH for the glutathionylated form and add two steps in the reaction mechanism. Thus, it seems unfavorable in terms of efficiency and energetic cost. We have previously analyzed possible variations in the redox state of the dithiol AtGrxC5 (WCSYC active site) in response to 15 min treatments with GSSG or oxidized DTT by measuring the tryptophan intrinsic fluorescence of the protein, since the oxidation of the catalytic cysteine quenched the fluorescence of the adjacent tryptophan (Couturier et al., [@B13]). It was concluded that the dithiol AtGrxC5 cannot form a Cys~A~-Cys~B~ intramolecular disulfide first because only the GSSG treatment led to the formation of a glutathionylated protein and to fluorescence quenching and second because the WT protein was crystallized with a glutathionylated adduct. However, at that time, we did use neither H~2~O~2~ and GSNO treatments, nor prolonged incubation, nor alkylation assays. In fact, reconsidering the results of mass spectrometry showing, in addition to the peaks with one or two glutathione adducts, a peak with a decrease of 2 Da between the reduced protein and the GSSG-treated protein, it is very likely that a small fraction of the protein contained an intramolecular disulfide formed between active site cysteines (Couturier et al., [@B13]). Here, we have clearly observed that a prolonged oxidative treatment of reduced class I Grxs with H~2~O~2~, GSSG, or GSNO led preferentially to the formation of an intramolecular disulfide bond between Cys~A~ and Cys~B~. For the H~2~O~2~ treatment, it likely goes first through the formation of a sulfenic acid on Cys~A~ before Cys~B~ performs a nucleophilic attack on it. For the GSSG or GSNO treatments, it is very likely that a nucleophilic attack of Cys~A~ onto these oxidized glutathione species leads to the formation of a glutathione adduct on Cys~A~ which is then slowly reduced by a Cys~B~ nucleophilic attack leading to the formation of the intramolecular disulfide and to the concomitant release of a GSH molecule. It is worth mentioning here that these in vitro experiments have been done in particular with the aim of exploring all possible post-translational modifications but the conditions are not exactly physiological first because the concentrations of the oxidants may not be realistic and more importantly because this is done in the absence of GSH, which is normally found at high amounts and is highly reduced in most cell compartments. The presence of GSH would certainly prevent to a large extent the formation of these intramolecular disulfides because it may preferentially reduce either the sulfenic acids or the glutathione adducts over the Cys~B~ recycling cysteine. This is certainly the case for GrxC3 and GrxC4 since mutating CysB has no effect on the activity. The formation of an intramolecular disulfide in these proteins could thus constitute uniquely a side reaction, protecting the cysteine or inactivating the proteins only under specific conditions. On the contrary, if we consider that the increase of protein activity observed for CysB mutated GrxC1 and GrxC2 in HED and DHA assays in the presence of a fully reduced GSH pool reflects the formation of this intramolecular disulfide, it may be physiologically relevant for GrxC1 and GrxC2. This is interesting in several respects. First, while the current dogma is that the preferential substrates of Grxs are glutathionylated proteins, recent studies performed with mammalian Grx2 indicated that they can reduce dithiol substrates, e.g., proteins with intra- or intermolecular disulfide (Hanschmann et al., [@B20]; Schutte et al., [@B39]). In this case, the Grxs should employ a reaction mechanism similar to the one used by Trxs involving both Cys~A~ and Cys~B~. Second, based on the observation that some Grxs can be reduced by thioredoxin reductases (Johansson et al., [@B23]; Zaffagnini et al., [@B48]; Couturier et al., [@B12]), another possibility could be that in specific physiological situations or sub-cellular compartments where glutathione is depleted, absent or oxidized, the formation of an intramolecular disulfide may be determinant either to favor the use of an alternative reducing system as the one constituted by NADPH and thioredoxin reductase or to simply constitute a protective mechanism, preventing the irreversible overoxidation of this residue and thus protein inactivation. In conclusion, the role of Cys~B~ in dithiol Grxs is yet uncertain and may depend on the Grx isoform as this residue influences differentially the protein activity. The fact that the formation of an intramolecular disulfide bond between Cys~A~ and Cys~B~ is possible in all cases lets open the possibility to use both monothiol and dithiol catalytic mechanisms, contrary to Grxs possessing only one cysteine residue. The role of the semi-conserved Cys~C~ in GrxC1 and GrxC2 isoforms ------------------------------------------------------------------- The Cys~C~ cysteine residue, present in the IGGCD motif (at position 88 and 80 in GrxC1 and C2, respectively), is found in several other class I Grxs, including plant GrxC5 and S12 as well as in human Grx1 and E. coli Grx3. Moreover, it is also present in many class II Grxs. It was shown for instance that a Cys~A~-Cys~C~ intramolecular disulfide bond can be formed in S. cerevisiae Grx5 and C. reinhardtii Grx3 (Tamarit et al., [@B41]; Zaffagnini et al., [@B48]), suggesting that Cys~C~ can serve as a recycling cysteine in the absence of Cys~B~. Reduced glutathione does not seem to be an efficient reductant for this intramolecular disulfide bond and in the case of CrGrx3, it is reduced by a ferredoxin-thioredoxin reductase system (Zaffagnini et al., [@B48]). Although GrxS12 has a monothiol WCSYS active site sequence, a Cys~A~-Cys~C~ intramolecular disulfide has not been observed whatever the oxidizing conditions tested. Moreover, in the crystal structure of a glutathionylated GrxS12, the sulfur atom of Cys87 is 9.6Å away from the Cys27-SG adduct (Couturier et al., [@B11]). Hence, this cysteine may not serve as a recycling cysteine in class I Grxs and does not a priori participate to the reaction mechanism. Consistently, its mutation in GrxC1 and C2 did not influence protein activity, as previously observed for poplar GrxS12, Arabidopsis GrxC5, or E. coli Grx3 (Nordstrand et al., [@B26]; Couturier et al., [@B11], [@B13]). However, considering its position very close to the active site residues and to residues involved in GSH binding an effect on activity could have been observed. Indeed, it is positioned between the so-called GG kink which is important in determining the backbone geometry of the following amino acids, and a Thr or an Asp which forms hydrogen bonds with GSH (see Figure [1](#F1){ref-type="fig"}) (Couturier et al., [@B11]; Li et al., [@B24]). In the GrxS12 structure, the backbone amino group of Cys~C~ also directly forms a hydrogen bond with the glutamyl group of GSH (Couturier et al., [@B11]), but the substitution of the cysteine into serine would not disrupt this interaction, preventing to definitely rule out a GSH stabilizing effect of this cysteine. On the other hand, it is probable that modifying the redox state of this cysteine might have more pronounced effects influencing for example GSH binding. It has been shown for example that, under non-reducing conditions, E. coli Grx3 exists under an expected monomeric form but also forms disulfide-bridged dimer or multimers via Cys~C~ (Aslund et al., [@B3], [@B2]). In addition, previously reported oxidative treatments of human Grx1, which contains three extra active site cysteines including Cys~C~, led to the identification of several possible post-translational modifications, i.e., intramolecular disulfide, disulfide-bonded dimers and oligomers, glutathione adducts, nitrosylation, some of them inhibiting its activity (Hashemy et al., [@B21]). Depending on the conditions, Cys~C~ was found to be either nitrosylated or glutathionylated or involved into a disulfide. Similarly, we have observed redox changes of GrxC1 and C2 upon treatment with oxidants that could further provide clues about the possible function of this cysteine, if any. In the presence of H~2~O~2~, this cysteine was found to be either overoxidized or involved in disulfide-bridged homodimers. Since this dimerization only occurs in presence of H~2~O~2~, the formation of an intermediate sulfenic acid is likely required before intermolecular disulfide bond formation which is consistent with the presence of a small part of overoxidized proteins. Interestingly, GSH can efficiently reduce these disulfide-bridged dimers, making this reaction physiologically reversible. In the presence of GSSG or GSNO, Cys~C~ is also prone to oxidative modification, but in the form of a glutathione adduct. Although we cannot test the impact of these modifications on Grx activity (we cannot selectively modify Cys~C~ without also oxidizing Cys~A~and Cys~B~), we can speculate that the different reversible oxidation forms of Cys~C~ observed here might represent regulatory or signaling intermediates transiently modifying the functions of these Grxs by affecting for example the nature or the time-course reduction of the target proteins. Conflict of interest statement ------------------------------ The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. This work was supported by the French National Research Agency through the Laboratory of Excellence ARBRE (ANR-12-LABXARBRE-01). The authors would like to thank Bruno Dastillung and Arnaud Hecker for their help with some experiments. Supplementary material {#s5} ====================== The Supplementary Material for this article can be found online at: <http://www.frontiersin.org/journal/10.3389/fpls.2013.00518/abstract> ###### Electrospray ionization mass spectrometry analysis of PtGrxC1 WT. Spectra of whole protein were determined for reduced protein before (A) and after treatment with GSSG (B), GSNO (C), or H~2~O~2~ (D) as described in the Methods section. ###### Click here for additional data file. ###### Electrospray ionization mass spectrometry analysis of PtGrxC2 WT. Spectra of whole protein were determined for reduced protein before (A) and after treatment with GSSG (B), GSNO (C), or H~2~O~2~ (D) as described in the Methods section. ###### Click here for additional data file. ###### Electrospray ionization mass spectrometry analysis of PtGrxC3 WT. Spectra of whole protein were determined for reduced protein before (A) and after treatment with GSSG (B), GSNO (C), or H~2~O~2~ (D) as described in the Methods section. ###### Click here for additional data file. ###### Electrospray ionization mass spectrometry analysis of PtGrxC4 WT. Spectra of whole protein were determined for reduced protein before (A) and after treatment with GSSG (B), GSNO (C), or H~2~O~2~ (D) as described in the Methods section. ###### Click here for additional data file. ###### Electrospray ionization mass spectrometry analysis of PtGrxC1 C88S. Spectra of whole protein were determined for reduced protein before (A) and after treatment with GSSG (B), GSNO (C), or H~2~O~2~ (D) as described in the Methods section. ###### Click here for additional data file. ###### Electrospray ionization mass spectrometry analysis of PtGrxC2 C80S. Spectra of whole protein were determined for reduced protein before (A) and after treatment with GSSG (B), GSNO (C), or H~2~O~2~ (D) as described in the Methods section. ###### Click here for additional data file. ###### Primers used for site-directed mutagenesis. ###### Click here for additional data file. [^1]: Edited by: Andreas Meyer, University of Bonn, Germany [^2]: Reviewed by: Mirko Zaffagnini, University of Bologna, Italy; Christophe Riondet, UMR 5096, France [^3]: This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science.	{ "pile_set_name": "PubMed Central" }
Introduction {#sec1-1} ============ Diffusion tensor imaging (DTI) is a MRI technique that enables the measurement of the diffusion of water in tissue in order to produce neural tract images ([Fig. 1](#F1){ref-type="fig"}). The idea of using diffusion data to produce images of neural tracts was first proposed by Aaron Filler & colleagues in March of 1991. Several months later (1992) the first DTI image showing neural tracts curving through the brain was produced. Recent development of magnetic resonance imaging (MRI) has made it possible to delineate detailed anatomical and functional information of both normal and pathological brain tissues in a noninvasive manner. Among them, diffusion tensor MR imaging (DTI), has attracted great interest in that it can provide neural ﬁber information by use of the ﬁber--tracking technique and can also detect the presence of tumor cell inﬁltration into the normal white matter around the tumor boarder. The target in the treatment of malignant brain tumors is to maximize tumor resection volume, information on neural ﬁber preservation and tumor cell inﬁltration is critical for preoperative neurosurgical planning \[[@R1]\]. DTI data can be used to visualize the major white matter (WM) tracts of the brain \[[@R2]--[@R4]\]. DTI is an MR technique that can indirectly evaluate the integrity of WM by measuring water diffusion and its directionality in three dimensions \[[@R5]\]. DTI has been applied to differentiate edema from tumor, in patients with brain tumor for tumor characterization and to assess structural properties of the adjacent tracts \[[@R6]--[@R11]\]. Knowledge of the anatomical relationship between tumor and WM tracts could improve preoperative risk analysis and decrease the risk of WM tract injury during surgery. Multiple studies investigated the relationships between fractional anisotropy (FA) values, mean diffusivity (MD) values and the histological tumor cell inﬁltration within the region of interest where the FA and MD were measured \[[@R12]--[@R14]\]. These studies have demonstrated that FA values tend to decrease in locations where the neural ﬁbers have been inﬁltrated by tumor cells, suggesting that relatively low FA values indicate tumor cell invasion into white matter, where neural ﬁbers reside. The relationships between tumor cell density and FA or MD are still controversial, as both positive and negative correlation between these parameters has been reported \[[@R13],[@R15],[@R16]\]. Thus, the development of preoperative histological evaluation of tumors by noninvasive imaging techniques would greatly assist the preoperative diagnosis of the lesion and also preoperative surgical planning of the operation. For example, when a tumor biopsy of a glioma is required, it is extremely important that the target for biopsy is the location with the highest cell density, to allow accurate histological diagnosis. ![1 The first images of tractography from our research project. Fibers in red color are left--right. Fibers in green color are anterior--posterior. Fibers in blue color are superior--inferior](JMedLife-04-372-g001){#F1} Methods {#sec1-2} ======= The principal application of DTI is in the imaging of white matter where the location, orientation, and anisotropy of the tracts can be measured. DTI allows clinicians to look at anisotropic diffusion in white-matter tracts, but it is limited in demonstrating spatial and directional anisotropy. Advanced methods such as color coding and tractography (fiber tracking) have been used to investigate the directionality. If diffusion gradients (i.e. magnetic field variations in the MRI magnet) are applied in at least 3 directions, 6 directions improve the accuracy that describes the 3-dimensional shape of diffusion. The fiber direction is indicated by the tensor's main eigenvector which can be color--coded, yielding a cartography of the tracts\' position and direction: red for left--right, blue for superior--inferior, green for anterior--posterior. Tractography potentially solves a problem for a neurosurgeon in terms of minimizing functional damage and determining the extent of diffuse infiltration of pathologic tissue to minimize residual tumor volume. In this way, tractography facilitates preoperative planning. Tractographic images may help to clarify whether a tumor is compressing, abutting, or infiltrating the contiguous white--matter tracts. DTI identify different tumor components, and to differentiate tumor invasion from normal brain tissue or edema. DTI has demonstrated a potential in distinguishing gliomas and solitary metastasis in the brain parenchyma. Significantly higher MD, compared with levels in normal--appearing white matter, has been demonstrated in the peritumoral regions of both gliomas and metastases. Peritumoral MD of metastases and meningioma is significantly higher than that of gliomas. The FA and MD measurement in region of interest can be selected anatomically by referring to the MR images of preoperative surgical planning and postoperative results in order to match the location from which the histological specimen can be obtained. FA and MD values were can be calculated using a computer with the image analyzing software like MedINRIA. After surgical resection, specimens can be ﬁxed in formalin and embedded with parafﬁn; hematoxylin and eosin-stained specimens were checked to determine the histological tumor type. Cell counting can be performed under a light microscope at standard magnification like ×200. A standard area for the tumor cell density can be established and can be used a program like ImageJ for counting. Also, Ki--67 labeled cells can be counted and the percentage of Ki--67 labeled cells can be calculated within the observed ﬁeld. \[[@R17]\] The localization of [tumors](http://en.wikipedia.org/wiki/Tumor) in relation to the white matter tracts (infiltration, deflection), has been one the most important initial applications. In surgical planning for some types of brain tumors, surgery is aided by knowing the proximity and relative position of tumor. Discussions and Results {#sec1-3} ======================= Tractography combined with functional MRI may potentially help in preoperative planning of brain tumors by mapping areas of active infiltration. The recent development of DTI allows for direct examination of the brain microstructure. In the surgery of patients with brain tumors, preservation of vital cerebral function is as important as maximizing tumor resection. The associated morbidity of aggressive resections can be significantly reduced by carefully preservation of vital cerebral function, and the quality of life of these patients will be largely improved. In general, cerebral tumor may alter the adjacent WM in three different ways: by (1) displacing the WM tracts but with relative preservation of the fibers, (2) infiltrating the WM tracts, and (3) disrupting of the WM tracts. Intracranial tumors may involve both functional cortical gray and white matter tracts. Preoperative characterization of tumor tissue using noninvasive imaging modalities is one of many necessary steps for establishing a successful treatment strategy against malignant brain tumors. It is widely accepted that tumor cell density correlates with tumor malignancy. In cases where a tumor biopsy is required, the target is preferentially chosen at the lesion with the highest tumor cell density, to achieve accurate diagnosis and grading of the disease. In cases of contrast agent--enhancing tumors, usually the target is set at the location with distinct enhancement on MRI, based on the known fact that the blood--brain barrier is disrupted within the lesion with higher malignancy, enabling contrast agents to enhance the area \[[@R18]\]. Perfusion weighted MR imaging (PWI), chemical shift imaging (CSI) and positron emission tomography (PET) have been proposed for retrieving information on biological characteristics of the tumor in a noninvasive fashion \[[@R19],[@R20]\]. The FA and MD can be calculated and the results indicated an FA signiﬁcantly lower in the low-grade gliomas than the malignant lymphomas. \[[@R17]\] This observation is similar to the conclusion of Toh et al. (2008) \[[@R21]\]. The average indicated MD signiﬁcantly higher in low--grade gliomas and also in high-grade gliomas compared to malignant lymphomas. These ﬁndings suggest that MD is more sensitive than FA in distinguishing gliomas from malignant lymphomas. Tumor density of the tumor core shows positive correlation with FA but negative correlation with MD. These ﬁndings are compatible with those previously reported by Beppu et al. (2003, 2005) \[[@R15],[@R16]\] but contradict those of Stadlbauer et al. (2006) \[[@R13]\]. Although a direct comparison of FA and MD was not performed in the study by Stadlbauer et al., they showed that tumor cell density has negative correlation with FA and positive correlation with MD \[[@R13]\]. Thus it seems fair to speculate that the data in the study of Stadlbauer et al. also shows an inverse relationship between FA and MD values \[[@R13]\]. In addition, an inverse relationship between FA and MD has also been reported for multiple sclerosis \[[@R22]\]. One possible explanation for FA and MD show inverse correlation is the presence of tumor cells within the tissue prevents water diffusing in a single direction, which would show both high FA and MD values. Previous studies have focused the use of diffusion tensor MR imaging (DTI) for ﬁber--tracking \[[@R23]--[@R30]\] and evaluating tumor inﬁltration near the tumor boarder \[[@R12]--[@R14]\]. These assessments are crucial for achieving maximum tumor resection of malignant brain tumors. Ki--67 labeling index is positively correlated with FA and negatively correlated with MD. Tumor cell density at the core of the tumor and Ki--67 labeling index show good positive correlation, suggesting that tumor cell density and Ki--67 labeling index have diagnostic value in malignant brain tumors. \[[@R17]\]. As FA and MD (in combination) seem to represent the histology of these tumors, in particular of tumor cell density. It is expected that locations with higher FA and lower MD will have higher tumor cell density, as well as have a relatively high Ki--67 labeling index, indicating higher malignant potential. It should be noted that higher FA values could also implicate the contamination of high FA resulting from intact neural ﬁbers. \[[@R17]\] Conclusions {#sec1-4} =========== As a conclusion, data from multiple studies of DTI suggest to represent tumor tissue characteristics in a noninvasive fashion. When the enhancement by contrast agent is vague, other indicators are necessary to assess tumor cell density within the tumor. The use of diffusion tensor MR imaging (DTI), however, for this purpose still requires investigation. Acknowledgments This investigation was supported by a Grant for Scientiﬁc Research from the Ministry of Education and Science of Romania (subject numbers: 42--149/2008).	{ "pile_set_name": "PubMed Central" }
Background {#Sec1} ========== Enterohaemorrhagic Escherichia coli O157:H7 (EHEC O157) is an important intestinal pathogenic bacterium that can causes diarrhea, hemorrhagic colitis, and in 10% of cases of systemic hemolytic uremic syndrome. EHEC O157 is the most extensively studied EHEC and is responsible for regular outbreaks of foodborne illness worldwide \[[@CR1]\]. EHEC O157 colonization involves the formation of attaching and effacing (A/E) lesions on the intestinal epithelium, which are characterized by loss of microvilli and allow intimate attachment of the bacterium to the host cell membrane \[[@CR2]\]. A/E lesion formation genes are localized on a pathogenicity island, known as the locus for enterocyte effacement (LEE), which encodes a bacterial type III secretion system (T3SS) and is capable of injecting bacterial effector proteins into the host cell cytoplasm \[[@CR3]\]. The LEE contains 41 genes that are organized in five major operons (LEE1, LEE2, LEE3, LEE5, and LEE4) \[[@CR3]\]. LEE1, LEE2, and LEE3 encode the major structural components of the T3SS, LEE4 encodes several secreted proteins \[[@CR4]--[@CR6]\], and LEE5 encodes Tir and Intimin \[[@CR7]\]. ORF1 on the LEE1 operon encodes the master regulator Ler (LEE encoded regulator). Ler is capable of activating LEE2 to LEE5 \[[@CR8], [@CR9]\]. Transcriptional regulation of the LEE is extremely complex. The regulatory system of the LEE involves at least three kinds of regulators: LEE-encoded regulators (including Ler \[[@CR9]\], GrlA and GrlR \[[@CR10]\]), global regulators (such as H-NS, IHF and Fis \[[@CR11]\]), and horizontally transferred regulators (such as EivF, EtrA, and GrvA). Although the complexity of LEE regulation in EHEC O157 has been acknowledged, the mechanism by which LEE regulation occurs in not fully understood. In EHEC O157, biofilm formation is regulated by a complex network of regulatory cascades. The biofilm master regulator CsgD (curli specific gene D) is a key transcriptional response regulator controlling the formation of curli fimbriae and cellulose production \[[@CR12]\]. Biofilm formation can also be regulated by cyclic diguanylate (c-di-GMP) concentration \[[@CR13]\]. c-di-GMP is a common second messenger in bacteria, which is synthesized by diguanylate cyclases (DGCs) and degraded by c-di-GMP-specific phosphodiesterases (PDEs) \[[@CR14]\]. Among these genes, yeaI, which is a DGC-encoding gene, increases the c-di-GMP concentration in E. coli BW25113 \[[@CR15]\] and promotes biofilm formation in uropathogenic E.coli CFT073 \[[@CR16]\]. Colonization of the mammalian gastrointestinal tract brings bacteria into contact with a strong acid barrier in the stomach and organic acids in the intestine \[[@CR17]\]. To reach their site of colonization, EHEC must traverse the acidic environment of the stomach. Although the environment within the large intestine is less acidic, EHEC must survive volatile organic acids produced via anaerobic fermentation by the local microbiota \[[@CR18]\]. Several distinct acid resistance (AR) pathways have been identified in E. coli, and are present in EHEC \[[@CR19]\]. Acid resistance and/or induction of acid tolerance may better enable pathogens to survive gastrointestinal acidity and ultimately cause disease, and may thus enhance virulence \[[@CR20]--[@CR22]\]. The gene asr is important for adaptation to the acidic stomach, as asr mutants are unable to establish colonies in the stomach \[[@CR23]\]. The periplasmic chaperones HdeA and HdeB are also important for cell survival at low pH \[[@CR24]\] by protecting periplasmic proteins from aggregation at low pH, which is crucial considering the high permeability of the outer membrane \[[@CR25]\]. Transcription of the hdeAB operon is activated by RpoS and GadE, and repressed by H-NS and MarA \[[@CR26]\]. GadE, GadW, and GadX also play a critical role in the transcriptional regulation of the glutamate-dependent acid resistance (GDAR) system in E. coli K-12 MG1655 \[[@CR27]\]. Two-component signal transduction systems (TCSs) enable bacteria to sense environmental stimuli and transfer this information across the cytoplasmic membrane to the cytoplasm \[[@CR28]\]. A typical TCS consists of a sensor histidine kinase (HK) and its cognate DNA-binding response regulator (RR). The membrane HK typically has extracellular and cytoplasmic domains linked via a transmembrane domain. Upon ligand binding to the extracellular domain and subsequent conformational change, auto-phosphorylation of the conserved histidine residue in the cytoplasmic domain takes place. The phosphate is then transferred to the aspartic residue on the RR. Phosphorylation of the RR activates an output domain that can modulate gene expression \[[@CR29]\]. Most RRs are transcriptional factors, and once phosphorylated they bind to target promoters, activating or repressing transcription \[[@CR28]\]. Recently, a functional TCS, RstAB, has been implicated in the regulation of bacterial virulence in Vibrio alginolyticus, Salmonella typhimurium, Photobacterium damselae, Clostridioides difficile, and avian pathogenic E. coli \[[@CR30]--[@CR35]\]. The regulatory function of the TCS protein RstA on bacterial virulence in EHEC O157 remains unclear. Therefore, in the present study, we investigated the global effects of RstA on gene expression in EHEC O157. Genes whose expression was affected by RstA were identified and grouped into different clusters of orthologous group (COG) categories. We aimed to contribute to the understanding of the regulatory function of RstA in EHEC O157, especially with regard to virulence, which may impact future disease control and treatment effort against this important pathogen. Results {#Sec2} ======= Transcriptional data analysis {#Sec3} ----------------------------- To gain an understanding of RstA regulation at the global level, we systematically catalogued the transcriptomes of the EHEC O157:H7 strain EDL933 wild type strain (WT) and the ΔrstA mutant using high-throughput Illumina RNA-seq analysis. After filtering low quality reads, a total of 17,129,356 to 23,535,490 reads were obtained for the EHEC O157 WT and the ΔrstA mutant, respectively. Approximately 99.5% of the total reads for the EHEC O157 WT and 99.3% of those for the ΔrstA mutant were uniquely mapped to the reference genome (Additional file [1](#MOESM1){ref-type="media"}: Table S1). A total of 1237 genes were differentially expressed in the ΔrstA mutant compared to that in the EHEC O157 WT of these, 892 and 345 genes were categorized as up- and down-regulated, respectively (Additional file [2](#MOESM2){ref-type="media"}: Excel files S1 and S2). These results indicate that RstA acts as both an activator and repressor in EHEC O157. We selected 10 of the differentially regulated genes at random for validation by qRT-PCR using the same culture conditions. The qRT-PCR results correlated well with the RNA-seq data, indicating that the RNA-seq data were robust and valid (Fig. [1](#Fig1){ref-type="fig"}).Fig. 1Confirmation of RNA-seq results by qRT-PCR. RNA-seq results were validated by comparing EHEC O157 WT and ΔrstA mutant strains using qRT-PCR to measure the relative expression of 10 randomly selected genes that were differentially expressed according to RNA-seq. The trend of expression was similar according to both RNA-seq and qRT-PCR for these 10 genes, validating the RNA-seq data. Data are presented as mean ± SD; n = 3 Genes with differential regulation in the wild type and mutant strains were classified using the NCBI COG functional categories annotation system. The COG categories that were significantly enriched in the group of up-regulated genes were primarily involved in the cell wall, membrane, envelope biogenesis, translation, ribosomal structure and biogenesis, carbohydrate transport and metabolism, nucleotide transport and metabolism, energy production and conversion, lipid transport and metabolism, and amino acid transport and metabolism. The COG categories that were significantly enriched in the list of down-regulated genes included posttranslational modification, protein turnover, and chaperones (Fig. [2](#Fig2){ref-type="fig"}, Additional file [1](#MOESM1){ref-type="media"}: Fig. S1).Fig. 2Clusters of orthologous group (COG) analysis of RstA-regulated genes in EHEC O157. Bars represent the number of up-regulated (blue) or down-regulated (orange) genes in the ΔrstA mutant compared to those in the EHEC O157 WT strain. The significant enrichment of a given COG in the sets of up- or downregulated genes was determined using one-tailed Fisher's exact test with Benjamini--Hochberg false discovery rate correction RstA regulates the LEE pathogenicity island {#Sec4} ------------------------------------------- Based on RNA-seq results, the expression of majority of LEE genes (from the LEE1 to LEE5 operon) were significantly downregulated in the ΔrstA mutant relative to the EHEC O157 WT strain (Fig. [3](#Fig3){ref-type="fig"}a, Additional file [2](#MOESM2){ref-type="media"}: Excel file S3). To determine whether RstA is involved in the virulence of EHEC O157, qRT-PCR was performed to measure the expression of seven representative LEE genes, including ler (the master regulator of LEE genes), escT (LEE1), escC (LEE2), escN (LEE3), eae (intimin, LEE5), tir (intimin receptor, LEE 5), and espB (LEE 4) in EHEC O157 WT and the ΔrstA mutant strains. The transcript levels of these representative LEE genes were down-regulated in the ΔrstA mutant compared to those in the EHEC O157 WT (Fig. [3](#Fig3){ref-type="fig"}b). We then evaluated the adherence of the ΔrstA mutant to HeLa cells, and found that deletion of rstA significantly reduced bacterial adherence to HeLa cells compared with that of the EHEC O157 WT (Fig. [3](#Fig3){ref-type="fig"}c). Both EHEC O157 WT and the ΔrstA mutant exhibited a similar growth rate, indicating the difference in adherence capacity between these two strains was not due to different growth rates (Fig. [3](#Fig3){ref-type="fig"}d). We found consistent results using fluorescent actin staining (FAS), which suggested that the ΔrstA mutant formed fewer pedestals on HeLa cells than the EHEC O157 WT (Fig. [3](#Fig3){ref-type="fig"}f, g). Mouse colonization experiments were used to determine the adherence capacity of these bacterial strains in vivo. The amount of the ΔrstA mutant recovered from the colon of infected mice was significantly lower than that of the EHEC O157 WT strain at 6 h post-infection (Fig. [3](#Fig3){ref-type="fig"}e). These differences could be restored to wild-type levels when a complementary plasmid pTRC99a-RstA was introduced into the ΔrstA mutant. Collectively, these results suggest that RstA is a positive regulator of bacterial virulence in EHEC O157.Fig. 3Effect of rstA on EHEC O157 adherence and LEE genes expression. a A heat map representing differential regulation of virulence genes in the EHEC O157 WT and mutant strains. The z-score indicates whether the genes were upregulated (red) or downregulated (green). b qRT-PCR analysis of changes in LEE genes expression in EHEC O157 WT, ΔrstA mutant, and rstA complementary strain. c Adherence of EHEC O157 WT, the ΔrstA mutant, rstA complementary strain to HeLa cells. d Growth of EHEC O157 WT, the ΔrstA mutant, rstA complementary strain in LB medium. e Adherence capacity of EHEC O157 WT, ΔrstA mutant and rstA complementary strain in the distal colon of mice at 6 h. Statistical significance was assessed using the Mann--Whitney rank-sum test. f Detection of AE lesion formation by EHEC O157 WT, ΔrstA mutant, and rstA complementary strain by FAS in HeLa cells at 3 h. The HeLa cell actin cytoskeleton (green) and nuclei of bacterial and HeLa cells (red) are shown. AE lesions are indicated by arrowheads. g The number of pedestals/infected HeLa cell of EHEC O157 WT, ΔrstA mutant, and rstA complementary strain (n = 50 cells). In b--d, data represent mean ± SD (n = 3). \P ≤ 0.05, \\P ≤ 0.01, \\\P ≤ 0.001 (Student's t-test) RstA is involved in the regulation of EHEC O157 acid tolerance {#Sec5} -------------------------------------------------------------- The immediate challenge facing EHEC O157 in an infected human host is survival in the extreme acidic environment of the stomach. In the present study, we used RNA-seq to determine that genes encoding acid resistance proteins (HdeAB, Asr and GadEWX) and we found that these genes were down-regulated in the ΔrstA mutant compared to those in the EHEC O157 WT strain (Fig. [4](#Fig4){ref-type="fig"}a, Additional file [2](#MOESM2){ref-type="media"}: Excel file S4). We therefore compared the survival of EHEC O157 WT and the ΔrstA mutant when exposed to acidified LB broth (pH 3.0).The viable cells were recovered on LB agar plates after incubation in the acidified broth at 37 °C for 0--6 h, and the number of CFUs were determined as a ratio to the initial inoculum (CFU/ml at 0 h). The survival assay showed that the survival rate of EHEC O157 WT were much higher than the ΔrstA mutant after 2 h in acidified LB broth (Fig. [4](#Fig4){ref-type="fig"}b), indicating that RstA plays a significant role for EHEC O157 survival in low pH environment.Fig. 4Effect of rstA on EHEC O157 acid tolerance. a A heat map representing differential regulation of acid tolerance related genes in the WT and mutant strains. The z-score indicates whether the genes were upregulated (red) or downregulated (green). b Survival assay of EHEC O157 WT and the ΔrstA mutant strain in acid challenge after incubation in acidified broth (pH 3.0) for 0--6 h. Data represent mean ± SD (n = 3). \P ≤ 0.05, \\P ≤ 0.01 (Student's t-test) RstA impacts biofilm formation {#Sec6} ------------------------------ In the present study, we used RNA-seq to compare the transcriptomes of the EHEC O157 WT and the ΔrstA* mutant strains, and identified several upregulated diguanylate cyclase genes in the ΔrstA mutant strain. These genes participate in the formation of the ubiquitous second messenger, cyclic-di-GMP (c-di-GMP) (Fig. [5](#Fig5){ref-type="fig"}a, Additional file [2](#MOESM2){ref-type="media"}: Excel file S5), which promotes biofilm formation in many bacteria \[[@CR36]\].Fig. 5Effect of rstA on EHEC O157 biofilm formation. a A heat map representing differential regulation of diguanylate cyclase genes in the EHEC O157 WT and the ΔrstA mutant strain. The z-score indicates whether the genes were upregulated (red) or downregulated (green). b Biofilm formation in EHEC O157 WT, ΔrstA mutant, and rstA complementary by crystal violet staining. c Intracellular c-di-GMP concentration of EHEC O157 WT, the ΔrstA mutant and rstA complementary strain. In b, c, Data represent mean ± SD (n = 3). \P ≤ 0.05, \\P ≤ 0.01 (Student's t-test) To evaluate the effect of RstA in biofilm formation in EHEC O157, a crystal violet staining assay was performed to quantify biofilm formation by EHEC O157 WT, the ΔrstA* mutant, and the complementary strain. Biofilm formation was significantly increased in the ΔrstA mutant compared to that in the EHEC O157 WT and the complementary strain (Fig. [5](#Fig5){ref-type="fig"}b), indicating that RstA is a negative regulator of biofilm formation in EHEC O157. We then measured the concentration of intracellular c-di-GMP using HPLC, and found significantly increased concentrations of intracellular c-di-GMP in the ΔrstA mutant (Fig. [5](#Fig5){ref-type="fig"}c). Taken together, these findings support the conclusion that RstA inhibits biofilm formation in EHEC O157 by controlling the biosynthesis of c-di-GMP. RstA box analysis in EHEC O157 {#Sec7} ------------------------------ RstA has been shown to bind to the conserved motif TACATNTNGTTACA, which is termed the RstA box and is present in the promoter region of many RstA-activated or repressed genes in E. coli \[[@CR37]\]. Furthermore, the consensus TACA repeat sequence is necessary for RstA binding \[[@CR38]\]. Then, we searched for this RstA box-like sequence (TACANNNNNNTACA, N = 5--6) along the entire EHEC O157 EDL933 genome, and found 19 possible targets in the intergenic region. Among these targets, 14 are located in the promoter region and 8 were identified here for the first time (Additional file [1](#MOESM1){ref-type="media"}: Table S4). No RstA box was found in the promoter region of any of the LEE operons (P~LEE1~, P~LEE2/3~, P~LEE4~, P~LEE5~). Electrophoretic mobility shift assay (EMSA) results confirmed that RstA cannot directly bind to the promoter region of LEE operons (Fig. [6](#Fig6){ref-type="fig"}a--d). This suggests that RstA may activates LEE genes expression indirectly via an unknown regulator(s). With increasing concentrations of RstA protein, we observed slowly migrating bands for the promoter region of asr, hdeA and yeaI, but not rpoS (negative control) under the same conditions. This indicates that RstA enhances acid tolerance by directly regulating the expression of hdeA and asr, and represses biofilm formation by regulating the concentration of c-di-GMP via yeaI (Fig. [6](#Fig6){ref-type="fig"}e--h).Fig. 6EMSA of the binding of RstA. EMSA of the binding of RstA to P~LEE1~ (a), P~LEE2/3~ (b), P~LEE4~ (c), and P~LEE5~ (d), rpoS (e, negative control), P~asr~ (f), P~hdeA~ (g) and P~yeaI~ (h). PCR products were added to the reaction mixtures at 40 ng each. RstA protein was added to the reaction buffer in each assay in lanes 2--5 at 0.25, 0.5, 1, and 2 μM, respectively. No protein was added in lane 1. No binding was observed in a--e, while binding was observed in f--h Discussion {#Sec8} ========== RstA is a well-known TCS regulator that decreases bacterial adhesion and virulence in different bacterial species, including avian pathogenic E. coli, Edwardsiella ictaluri, Photobacterium damselae and Clostridioides difficile. In the present study, we investigated the effects of RstA on the global gene expression of EHEC O157 using RNA-seq. We validated the RNA-seq results using qRT-PCR to evaluate the changes in expression of 10 randomly selected genes in the WT and mutant strains. For all examined genes, the fold change detected by RNA-seq had the same trend as that observed by real-time PCR. These results are consistent with the results of previous experiments, validating our use of RNA-seq and verifying the results obtained here. In EHEC O157, 33 response regulators and 30 sensor kinases have been assumed to exist on the basis of genome sequence analysis results. Several response regulators have been reported to regulate the expression of virulence genes in EHEC O157. In the present study, we observed that inactivation of RstA results in significant downregulation of LEE genes expression in EHEC O157. Several other genes related to virulence were also regulated by RstA, including stx1, stx2, nleA, nleB, nleB2, nleC, and nleL, which implies that RstA is a global virulence regulator in EHEC O157 (Additional file [2](#MOESM2){ref-type="media"}: Excel file S6). According to the results of the adherence, FAS, and colonization assays, RstA contributes to EHEC O157 adherence in vitro and colonization in vivo. These results taken together indicate that RstA is a transcriptional activator of virulence in EHEC O157. Therefore, it was somewhat surprising that no RstA box was found in the promoter region of these genes. The EMSA results also confirm that RstA does not directly bind to the promoters of LEE1, LEE2/3, LEE4, and LEE5. This suggests that the observed positive regulation of LEE genes by RstA occurs via an indirect mechanism, possibly with an unknown intermediate regulator. To be able to establish colonization in a host, EHEC O157 must survive the acidic conditions in the stomach before it reaches the intestine \[[@CR39]\]. In the present study, RNA-seq comparison of mutant and WT strains revealed that several important acid tolerance genes (hdeAB operon, asr and gadEWX) were down-regulated in the ΔrstA mutant. We performed a survival assay and found that the ΔrstA mutant was more acid sensitive than the EHEC O157 WT. According to our EMSA results, rstA can directly bind to the promoter region of both asr and hdeA. These results suggest that rstA plays important roles in acid tolerance during host colonization. Previous studies have shown that changes in temperature, pH, and starvation, dramatically affect rstA expression in V. alginolyticus \[[@CR30]\]. This suggests that RstAB may sense environmental pH changes to regulate acid tolerance genes. However, whether rstA responds directly to low pH as a signal to activate acid tolerance pathways is unknown, and requires further investigation. Biofilm formation is mediated by bacterial surface structures that are regulated by environmental conditions. RstAB inhibits biofilm formation in Salmonella enterica and promotes biofilm formation in Vibrio alginolyticus \[[@CR30], [@CR40]\]. In the present study, we demonstrated that disruption of rstA in EHEC O157 also results in significant up-regulation of several c-di-GMP synthesis genes, suggesting that biofilm formation is influenced by c-di-GMP concentration. As expected, the ΔrstA mutant exhibited increased biofilm formation ability and intracellular c-di-GMP concentration. Among these up-regulated c-di-GMP synthesis genes, yeaI was directly regulated by rstA, and thus rstA can increase biofilm formation by increasing the concentration of c-di-GMP. Our RNA-seq results show that 66 regulator genes were differentially expressed (40 up-regulated and 26 down-regulated) in the ΔrstA mutant compared with WT strain (Additional file [2](#MOESM2){ref-type="media"}: Excel file S7). COG analysis of these genes indicated that these regulators participate in multiple biological processes (including intracellular trafficking, secretion, vesicular transport, amino acid transport metabolism signal transduction mechanisms, and transcription). This suggests that rstA can be both activator and repressor, and is capable of regulating more complex pathways than expected. Among the up-regulated regulators, an RstA box was found in the promoter region of narP (z3450), which regulates nitrate/nitrite respiration \[[@CR41]\]. The regulatory effect of rstA on other regulators needs to be confirmed experimentally. Conclusions {#Sec9} =========== The present study has contributed to our understanding of the EHEC O157 RstAB regulon, and identified a number of novel genes and functions that are affected by rstA. We found that RstA positively regulates virulence and acid tolerance, but negatively regulates biofilm formation in EHEC O157. In summary, the RstAB TCS in EHEC O157 plays a major role in the regulation of virulence, acid tolerance, and biofilm formation. Further research is required to reveal the mechanisms by which RstA regulates LEE genes. This may identify novel gene targets to control infections caused by this pathogen, which is particularly important given the emergence of drug resistance. Methods {#Sec10} ======= Bacterial strains and media {#Sec11} --------------------------- Bacterial strains, plasmids and primers used in this study are listed in Additional file [1](#MOESM1){ref-type="media"}: Tables S2 and S3. The ΔrstA mutant was constructed using the λ-Red recombination system and confirmed by PCR amplification and sequencing. A complementary strain was constructed by cloning RstA into the plasmid pTRC99a, and the resulting constructs were electroporated into EHEC O157 ΔrstA mutant. Antibiotics were added at the following final concentrations as required: 100 μg/ml ampicillin, 25 μg/ml chloramphenicol, 50 μg/ml nalidixic acid. RNA isolation, purification and sequencing {#Sec12} ------------------------------------------ Overnight cultures of EHEC O157 wild type (EHEC O157 WT) and the ΔrstA mutant were 1:100 subcultured in 20 ml of fresh Dulbecco's modified Eagle medium (DMEM, virulence-inducing medium for EHEC O157, Hyclone; \#SH30022.01) without antibiotics at 37 °C with shaking at 180 rpm, until the exponential growth phase was reached (OD~600~ = 0.6--0.8). Total RNA was extracted using TRIzol Reagent (Invitrogen; \# 15596026) and purified using the RNeasy Mini Kit (Qiagen; \#74104). The RNA was quantified and qualified using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA), a NanoDrop (Thermo Fisher Scientific Inc.), and 1% agarose gel electrophoresis. One microgram total RNA with a RIN value \> 6.5 was used for library preparation. rRNA (including 16S and 23S rRNA) was depleted from total RNA using The Ribo-off rRNA Depletion Kit (Bacteria) (Vazyme; \#N407). Libraries were constructed by VAHTSTM Total RNA-seq (H/M/R) Library Prep Kit for Illumina^®^ (Vazyme; \#NR603) according to manufacturer's instructions. Libraries with different indices were multiplexed and loaded on an Illumina HiSeq instrument according to manufacturer's instructions (Illumina, San Diego, CA, USA). The sequences were processed and data were analyzed by GENEWIZ, Inc (Suzhou, China). All sequence data have been deposited in the NCBI SRA database under the accession codes SRR9678084, SRR9678085, SRR9678086, and SRR9678087. Quantitative RT-PCR (qRT-PCR) {#Sec13} ----------------------------- Total RNA was extracted as previously described. First-strand cDNA was synthesized using the PrimeScript 1st Strand cDNA Synthesis Kit (Takara; \#D6110 A), according to the manufacturer's instructions. Primers for qPCR are listed in Additional file [1](#MOESM1){ref-type="media"}: Table S3. The 16S rRNA gene (rrsH) was used as a reference to standardize expression across the samples \[[@CR42]\]. Samples were amplified by PCR and amplicons were detected using SYBR green dye and an Applied Biosystems ABI 7500 sequence detection system (Applied Biosystems, CA, USA). The relative difference in gene expression was calculated using the cycle threshold method (2^−ΔΔct^) \[[@CR43]\]. Data were collected from at least three biological replicates. Bacterial adherence assay {#Sec14} ------------------------- Overnight cultures were subcultured in DMEM at 37 °C until they reached an OD~600~ of 0.6--0.8 for adaptation. Before infection, HeLa cells were washed three times with phosphate-buffered saline (PBS). The cell culture medium was replaced with fresh DMEM without antibiotics or fetal bovine serum. Cells were then infected with bacteria in DMEM at a multiplicity of infection (MOI) of 100:1. After incubation with HeLa cells for 3 h, unattached bacteria were removed by washing with PBS six times. The HeLa cells were then lysed with 0.1% SDS in H~2~O. Lysates were plated onto LB agar plates to count the number of viable adhered bacteria. Each experiment was carried out at least three times. Gut colonization assay {#Sec15} ---------------------- Six-week-old female BALB/c mice were provided with food and water ad libitum before infection. In each group, female BALB/c mice (n = 10) were orally infected with 10^9^ CFU of bacteria in 100 μl PBS. The infected mice were anaesthetized and euthanized via cervical dislocation at 6 h after infection. The distal colons were excised and the luminal contents were removed. Each distal colon of the intestine was washed with PBS three times to remove unattached bacteria, and then weighed and homogenized in 0.5 ml of PBS. The homogenates were diluted, and samples of the O157 WT strain, the ΔrstA mutant, and complementary strain were plated on LB agar containing nalidixic acid (50 μg/ml), chloramphenicol (25 μg/ml), or ampicillin (100 μg/ml), respectively, to determine the number of CFU per gram of organ tissue. Fluorescent actin staining {#Sec16} -------------------------- Fluorescent actin staining (FAS) assays were performed as described previously \[[@CR44]\]. Overnight cultures were subcultured in DMEM at 37 °C until they reached an OD~600~ of 0.6--0.8 for adaptation. HeLa cells were grown on coverslips for 24 h at 37 °C with 5% CO~2~. HeLa cells on coverslips were then infected with bacteria in DMEM at a multiplicity of infection (MOI) of 100:1. After incubation for 3 h at 37 °C and 5% CO~2~, the coverslips were washed with PBS and the bacteria were fixed with formaldehyde, and the cells were permeabilized with 0.2% Triton-X and stained with fluorescein isothiocyanate-labeled phalloidin to visualize actin filaments. Bacteria and HeLa cell nuclei were stained with propidium iodide. AE lesions formed by each strain were calculated for at least 50 HeLa cells. Quantitative biofilm assay {#Sec17} -------------------------- Biofilm formation was quantified by crystal violet staining, as previously described \[[@CR22]\]. Overnight cultures were diluted in fresh medium (1:100) and incubated in 96-well polystyrene microtiter plates at 37 °C for 24 h. The loosely associated bacteria were removed by washing with PBS three-times, and the remaining bacteria were stained with 0.5% crystal violet for 5 min. The biofilm was then destained by adding 200 μl of 95% ethanol to each well, and quantified using an enzyme-linked immunosorbent assay plate reader at 590 nm. Each experiment was carried out at least three times. High performance liquid chromatography {#Sec18} -------------------------------------- c-di-GMP was quantified using HPLC as described previously \[[@CR45]\]. Overnight cultures were subcultured in LB medium at 37 °C until they reached an OD~600~ of 0.6 for adaptation. Approximately 100 mg of cells were harvested in a pellet by centrifugation. The pellet was washed with PBS and resuspended in H~2~O. The suspension was heated at 95 °C for 15 min, followed by sonication. Ethanol was added to the sample to a final concentration of 70%. After centrifugation, the supernatant was pooled, frozen, and subsequently lyophilized overnight. The lyophilized flakes were resuspended in 1 ml of H~2~O and filtered through a 0.2 µm pore size filter. HPLC was performed using a 5 μm, 4.6 × 250 mm reverse phase column (Agela Venusil XBP-C18, VX952505-0) at room temperature with detection at 253 nm, on a Surveyor Plus HPLC System (Thermo Finnigan). Each experiment was carried out at least three times. Acid tolerance assay {#Sec19} -------------------- Overnight cultures were washed with PBS three times then diluted to a concentration of 10^6^ CFU/ml in LB acidified to pH 3.0 with HCl. Then cultures were incubated at 37 °C for 0 to 6 h with shaking at 180 rpm. A 100 μl aliquot was removed from the flask and suitable dilutions were plated on LB agar once every hour. Experiments were performed independently three times. EMSA {#Sec20} ---- The 6 × His-tagged RstA protein was expressed and purified in E. coli BL21 (DE3). DNA target fragments were amplified by PCR and purified using a SPARKeasy Gel DNA Extraction Kit (Sparkjade; \#AE0101-C). Purified PCR fragments (40 ng) were incubated at 25 °C for 30 min with 6× His-tagged RstA protein at concentrations ranging from 0 to 2 µM in 20 μl reactions containing binding buffer (1 mM Tris--HCl \[pH 7.5\], 0.2 mM dithiothreitol, 5 mM MgCl~2~, 10 mM KCl, and 10% glycerol, 30 mM acetyl phosphate). The protein-DNA fragments were electrophoretically separated on a native polyacrylamide gel at 4 °C and 80 V/cm. The gel was stained for 10 min in a solution of 0.1% GelRed (Biotium; \#41000), and protein bands were visualized by ultraviolet transillumination. Statistical analysis {#Sec21} -------------------- Statistical analysis was conducted using MedCalc (v12.3.0.0). The mean ± SD from three independent experiments was calculated. Differences between two mean values were evaluated by two-tailed Student's t test. Statistical significance was assessed with the Mann--Whitney rank-sum test in mouse colonization experiments. The significant enrichment of a given COG in the sets of up- or downregulated genes was determined using one-tailed Fisher's exact test with Benjamini--Hochberg false discovery rate correction \[[@CR46]\]. A P value \< 0.05 was considered to indicate statistical significance. Supplementary information ========================= {#Sec22} Additional file 1: Figure S1. Distribution of differentially expressed genes in the EHEC O157 EDL933 genome, Table S1. Transcriptional reads analysis of EHEC O157 WT and the ΔrstA mutant, Table S2. Strains and plasmids used in this study, Table S3. Primers used in this study (5′--3′), Table S4. RstA box analysis in EHEC O157. Additional file 2: Excel file S1. Down-regulated genes in the ΔrstA mutant compared with EHEC O157:H7 wild type strain, Excel file S2. Up-regulated genes in the ΔrstA mutant compared with EHEC O157:H7 wild type strain. Excel file S3. Differently expressed LEE genes in the ΔrstA mutant compared with EHEC O157:H7 wild type strain, Excel file S4. Differently expressed acid tolerance genes in the ΔrstA mutant compared with EHEC O157:H7 wild type strain, Excel file S5. Differently expressed diguanylate cyclases and phosphodiesterases genes in the ΔrstA mutant compared with EHEC O157:H7 wild type strain, Excel file S6. Differently expressed Stxs and Nle genes in the ΔrstA mutant compared with EHEC O157:H7 wild type strain, Excel file S7. Differently expressed regulatory genes in the ΔrstA mutant compared with EHEC O157:H7 wild type strain. EHEC O157 : enterohemorrhagic Escherichia coli O157:H7 TCS : two-component system HUS : hemolytic uremic syndrome A/E lesions : attaching and effacing lesions LEE : locus for enterocyte effacement T3SS : type III secretion system c-di-GMP : cyclic diguanylate DGCs : diguanylate cyclases PDEs : phosphodiesterases GDAR : glutamate-dependent acid resistance COG : clusters of orthologous group HK : histidine kinase RR : response regulator Publisher\'s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Yutao Liu and Shujie Li contributed equally to this work Supplementary information ========================= Supplementary information accompanies this paper at 10.1186/s13099-019-0335-4. We wish to thank Wenchao Lin from Tianjin Biochip Corporation for his technical assistance in RNA-seq data analysis. We thank Xi Hou for her help with statistical analysis. BY designed the research; BY, YL, SL, WL, PW, PD, LL, JW, PY, QW, TX and YX performed the experiments; YL, WL, PW, PD, LL, JW, and PY collected the data; SL, QW, TX and YX analyzed the data; BY, YL and SL wrote the manuscript. All authors read and approved the final manuscript. This research was funded by the National Science Foundation of China (NSFC) Program (Grant Nos. 31530083 and 31800125) and the Natural Science Foundation of Tianjin (Grant No. 17JCQNJC09300). The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Assays performed in vivo were carried out according to the standards set forth in the Guide for the Care and Use of Laboratory Animals. The experimental protocols were approved by the Institutional Animal Care Committee at Nankai University. The research protocol was revised and approved by that committee. Not applicable. The authors declare that they have no competing interests.	{ "pile_set_name": "PubMed Central" }
INTRODUCTION {#sec1-1} ============ Oral lichen planus (OLP) is a T-cell mediated inflammatory disease of the oral mucosa. The etiology of OLP is unknown and it occurs in 0.5% -1.9% of the population.\[[@ref1][@ref2]\] Interleukin-8(IL-8) is an important mediator of host response to injury, trauma and inflammation\[[@ref3]\] and plays a role in activation of neutrophils, neutrophil chemotactic factor, T-cells and basophils.\[[@ref4]\] Interlukin-8 is produced by different cells such as monocytes/macrophages, T-cells, neutrophils, endothelial cells, fibroblasts and keratinocytes in inflammatory and pathological processes.\[[@ref3][@ref5][@ref6]\] Healthy tissues have insignificant amounts of IL-8 but its concentration quickly reaches 10--100 times its baseline level in response to pro-inflammatory cytokines such as tumor necrosis factor-alpha (TNF-α) or IL-1β, bacterial or viral products, and cellular stress.\[[@ref7]\] Keratinocytes in OLP patients can synthesize IL-1 and TNF-α.\[[@ref8][@ref9]\] Moreover, mononuclear cells infiltrated in the tissues in mucosal OLP patients, as well as mononuclear cells in the peripheral blood of these patients, can produce TNF-α.\[[@ref8][@ref9][@ref10]\] Furthermore, following an increase in releasing both locally and systemically of IL-1 and TNF-α, keratinocytes, macrophages, T-cells, endothelial cells and fibroblasts in OLP lesions release significant amounts of IL-8. Interlukin-8 results in greater infiltration of T-cells such as cytotoxic T-cells at the site of OLP. Thus, IL-8 may be involved in the pathogenesis of OLP.\[[@ref11]\] Studies have shown that serum levels of IL-6 and IL-8 are higher in OLP patients compared to healthy individuals.\[[@ref11][@ref12]\] Moreover, serum level of IL-8 is a more specific indicator of OLP than IL-6.\[[@ref11]\] Diabetes mellitus (DM) is a metabolic disease caused by impaired insulin production or resistance to it, and is characterized by the abnormal metabolism of glucose, proteins and lipids.\[[@ref13]\] In fact, type II DM is an inflammatory disease and inflammatory cytokines are involved in its pathogenesis.\[[@ref14]\] Serum level of IL-8 in type II DM is significantly higher than that in healthy individuals.\[[@ref15][@ref16]\] At present, a correlation between DM and OLP seems probable.\[[@ref17]\] The Prevalence of DM in OLP patients has been reported as1.6% -- 85%.\[[@ref18][@ref19][@ref20]\] Studies in this regard have been variable in the methodology and results. Jolly showed that 85% of patients with OLP in his study had abnormal glucose tolerance test results.\[[@ref21]\] Seyhan et al., in their study stated that half of the patients with OLP had glucose metabolism impairment and one-fourth of these patients had DM.\[[@ref22]\] However, other studies failed to show a higher incidence of abnormal glucose tolerance test in OLP patients compared to the healthy population.\[[@ref23][@ref24]\] Since IL8 is an important inflammatory mediator involved in both conditions, the first objective of this study was to measure and compare the serum level of IL8 in four groups of OLP, DM, OLP+DM and healthy control. Considering the existing controversies regarding the possible association of DM and OLP, the second objective of this study was to assess the fasting blood sugar (FBS) and 2-h postprandial blood sugar in OLP patients. MATERIALS AND METHODS {#sec1-2} ===================== This analytical cross sectional study was conducted on three groups of OLP, type II DM and control. Considering the 80% power ofthe study, d = 50 and α = 0.05, thesample size was calculated to be 35 in OLP, 20 in type II DM and 20 in control groups. Samples were selected through convenience sampling among those presenting to Isfahan School of Dentistry and clinics in the city of Isfahan. The OLP group was divided into two subgroups of OLP alone and OLP + type II DM. Thus, 30 OLP,5OLP and type II DM,20 type II DM and 20healthy control individuals were evaluated. The healthy controls were selected among those presenting for check up and had no history of systemic disease or drug intake. An internal medicine specialist confirmed the health of control subjects. The DM patients were selected among subjects suspected for DM, who referredto a diabetes treatment center for a definite diagnosis. These subjects had FBS ≥110 mg/dL or 2-h postprandial blood sugar ≥200 mg/dL and reported polyphagia, polydipsia and polyuria.\[[@ref25]\] These subjects had no other systemic disease and reported no drug intake. The diagnosis of type II DM was made by an internal medicine specialist. The OLP patients were selected among those presenting to Isfahan School of Dentistry and clinics in the city of Isfahan. Following observation of white striations and red streaks and inclusion of OLP in the list of differential diagnoses (according to the clinical criteria) by an oral medicine specialist, a biopsy sample was taken from the lesion. Upon confirmation of OLP by pathologist according to the World Health Organization criteria,\[[@ref26]\] the patient was included in the study. Patients with lichenoid reactions due to medications or contact and those with graft versus host disease (frequently Bone marrow graft) or cutaneous lesions were excluded. Fasting and 2-hpostprandial blood sugar tests were requested for subjects with definite diagnosis of OLP. Based on the test results, they were assigned to OLP or OLP+DM group until 35 samples were recruited. All participants underwent a thorough physical examination to ensurethe absence of inflammation in other parts of the body. Tobacco users, alcohol consumers and patients with periodontitis were excluded. In general, patients with a history of systemic conditions other than DM and OLP and those on any type of medication were excluded from the study. After 8 h of fasting, 3cc of blood was drawn from each subject between 7 and 9 a.m.(to prevent errors due to the circadian rhythm). Within 5--10 minutes, subjects had a usual breakfast (carbohydrate-rich) and underwent 2-h postprandial blood sugar test two hours after the breakfast. Within this 2-h period, subjects were requested to refrain from severe physical exercise. Immediately after clotting, blood samples were centrifuged at 3000 rpm for 15 minutes to isolate the serum. Immediately afterwards, the FBS and 2-h postprandial blood sugar were measured. The remaining serum isolated from the fasting blood sample was stored at 20°C for measurement of IL-8 using the ELISA Kit (Quantikine ELISA, R and D Systems, Inc., USA). The obtained data were analyzed using SPSS (IBM Stasistics 20), one-way ANOVA and post -- hocleast significant difference (LSD) test. Normality of the data was checked using Kolmogorov- Smirnov test. RESULTS {#sec1-3} ======= A total of 20 DM, 20 healthy control and 35 OLP individuals were included; 14.28% of OLP patients had type II DM. Furthermore, 20% of OLP patients had impaired fasting glucose (FBS of 100- 125 mg/dL). As seen in [Table 1](#T1){ref-type="table"}, the highest mean concentration of serum IL-8 was seen in OLP+DM group followed by OLP, DM (6.1 ± 1.91 pg/mL) and control (5.3 ± 1.23 pg/mL) groups. ###### The mean (±standard deviation) and range of changes in serum interleukin-8 concentration in the four groups and result of post hoc least significant difference test ![](DRJ-13-413-g001) One-way ANOVA showed significant differences among groups in terms of serum level of IL8 (P = 0.011). Post-hoc LSD test was then applied and showed a significant difference in serum level of IL-8 between the control and OLP groups (P = 0.002). The difference between the OLP+DM and control subjects was also significant in terms if serum level of IL-8 (P = 0.022). However, although the mean serum level of IL-8 in the DM group was higher than that in the control group, this difference was not significant (P = 0.087). The serum level of IL8 in OLP+DM patients was higher than that in DM group but this difference was not significant either (P = 0.219). The serum level of IL8 was higher in OLP+DM group compared to the OLP group but not significantly (P = 0.615). The mean levels of FBS and 2-h postprandial blood sugar in OLP+DM patients were the highest followed by DM, OLP and control groups, respectively \[[Table 2](#T2){ref-type="table"}\]. ###### The mean levels of fasting blood sugar and 2-h postprandial blood sugar in the four groups ![](DRJ-13-413-g002) DISCUSSION {#sec1-4} ========== To the best of the authors\' knowledge, no previous study has compared FBS, 2-h postprandial blood sugar and IL-8 levels in the serum of OLP, type II DM, OLP+DM and control subjects simultaneously; thus, the current study is probably the first one. The first objective of this study was comparing the serum level of IL-8 in OLP subjects and healthy individuals. The results showed that the mean serum level of IL-8 in OLP patients was significantly higher than that in controls (P \< 0.05). Sun et al.,\[[@ref11]\] in 2005 concluded that the serum level of IL-8 was significantly higher in OLP patients than controls and thus, IL-8 is a more sensitive marker than IL-6 for diagnosis of OLP. Rhodus et al. in 2005\[[@ref27]\] and Zhang et al. in 2008\[[@ref28]\] stated that the serum and salivary levels of IL8 were significantly higher than those in control group. Our findings are in agreement with the results of the above-mentioned studies. The following explanation can better elucidate the reason behind higher serum level of IL-8 in OLP patients compared to healthy controls. In general, there is a strong theory that T-cell-mediated autoimmune reactions are involved in the pathogenesis of OLP. The localand systemic release of different cytokines from the oral mucosa and blood is responsible for initiation and progression of this disease.\[[@ref29]\] Histological analysis has shown band-like infiltrates of T lymphocytes in the upper lamina propriaof OLP lesions.\[[@ref30]\] Inflammatory cytokines such as TNF-αand IL-1 stimulate T -- cells, monocytes, macrophages, keratinocytes, neutrophils, and endothelial cells and result in secretion of IL-8 in the lesion and in the peripheral blood.\[[@ref31][@ref32]\] Cell culture studies have shown that OLP keratinocytes produce higher levels ofTNF-α than gingivitis keratinocytes.\[[@ref9][@ref10]\] Thus, greater infiltration of inflammatory cells especially T -- cells as well as increased secretion of TNF-α and IL-1 in these patients can lead to higher secretion of IL-8; which explains the high serum levels of IL-8 in OLP patients compared to healthy individuals. Based on previous studies, OLP plus DM may occur concomitantly.\[[@ref18][@ref19][@ref20]\] Thus, the second objective of this study was to assess the serum level of IL-8 as a possible relevant factor in DM patients compared to healthy individuals. In the current study, the mean serum level of IL-8 in DM patients (6.1 ± 1.91 pg/mL) was higher than that in controls (5.3 ± 1.23pg/mL); however, this difference was not significant (P \> 0.05). According to a study by Abou-Shousha et al.,\[[@ref14]\] serum level of IL-8 in DM patients was significantly higher than that in healthy controls (P = 0.032). They suggested that serum level of IL-8 could be used as a predictor of DM-related micro- and macro-vascular diseases especially in high-risk individuals. Other studies also showed significantly higher serum level of IL-8 in type II DM compared to healthy individuals.\[[@ref15][@ref16]\] Although in the above-mentioned studies the serum level of IL-8 in diabetic patients was significantly higher than that in controls, this difference in our study was not significant. Such difference in results might be due to the variable sample sizes, differences in the nutritional regimens of healthy individuals, genetic variations and ethnic characteristics of subjects in different geographical locations. In general, it seems that high glucose level in diabetic patients results in attachment of monocytes to endothelial cells and subsequent production of IL-8.\[[@ref15][@ref33]\] The third objective of the current study was to measure the serum level of IL-8 in patients with OLP+DM in comparison with OLP plu DM patients as well as healthy controls. In our study, a significantly higher level of serum IL-8 was found in OLP+DM patients compared to healthy controls (despite the small sample size of the former group)-(P \< 0.05). No previous study has been conducted in this regard to compare with. Since the occurrence of DM and glucose impairments has been reported in OLP patients,\[[@ref18][@ref19][@ref20]\] the fourth objective of this study was to measure the levels of FBS and 2-h postprandial blood sugar in OLP patients in comparison with other groups. The mean level of FBS and 2-h postprandial blood sugar was the highest in OLP+DM patients followed by DM, OLP and control groups, respectively. In our study, 14.28% of OLP patients had DM and 20% had impaired fasting glucose (FBS of 100--125 mg/dL). Despite the higher mean level of FBS in OLP patients compare tocontrols, this difference did not reach statistical significance. Seyhan et al.\[[@ref22]\] found a significant difference in FBS between OLP and control subjects; whereas, some other studies failed to show a higher frequency of abnormal glucose tolerance test in OLP patients compared to the general population.\[[@ref23][@ref24]\] Ansar et al.\[[@ref34]\] reported that the mean FBS in OLP patients was 102.5 ± 32 mg/dL. This value was close to the value reported in our study (99.85 ± 14.4 mg/dL). Such controversy in the results of the afore-mentioned studies may be due to different methodology and study designs. Higher FBS levels in OLP patients compared to healthy controls points to the possible role of glucose intolerance in the pathogenesis of OLP.\[[@ref35]\] In addition, the ascending trend of the serum level of IL8 in control, DM, OLP, and OLP+DM patients may indicate the role of this factor in the pathogenesis of OLP and DB; moreover, it may have a synergistic effect on both conditions. In this study, it was not possible to detect internal inflammation (if any); thus, the IL-8 value was not exclusively indicative of OLP or DM. Small sample size in the OLP+DM group was another limitation of our study which was due to the small number of OLP patients and difficulty in encouraging them to participate in the study. Future studies with larger sample sizes are required on OLP patients to compare the serum level of IL-8 in different types of OLP and also to assess its role in the clinical course of the disease. CONCLUSION {#sec1-5} ========== The results showed higher serum levels of IL-8 in OLP+DM and OLP patients compared to healthy controls. Thus, IL-8 should be considered as an effective inflammatory factor in the pathogenesis of this disease. Administration of medications balancing the level of IL8 and decreasing its serum level may be a step forward to alleviate the symptoms of OLP and OLP+DM patients. Financial support and sponsorship {#sec2-1} --------------------------------- Grant Number: 394661. Conflicts of interest {#sec2-2} --------------------- The authors of this manuscript declare that they have no conflicts of interest, real or perceived, financial or non-financial in this article. This study financially supportedby the dental Research commitee of Isfahan University of MedicalSciences.	{ "pile_set_name": "PubMed Central" }
GLAUCOMIDES BROMELICOLA Foissner, [@b12] has been recently described in the context of a comprehensive project on the diversity and ecology of ciliates from tank bromeliads (Foissner [@b12]). Bromeliads are species-rich epiphytic and terrestrial plants that are wide-spread in Central and South America. Many are able to store water in a reservoir formed by the tightly overlapping bases of their rosette leaves. Such phytotelmata (Varga [@b34]), also known as tanks or cisterns, may contain up to 30 litres of water; they represent a highly specialized aquatic habitat (Armbruster et al. [@b1]; Foissner et al. [@b16]; Kitching [@b21]). In contrast to larger aquatic habitats, tank bromeliads are disconnected from each other. Thus, phytotelmata represent natural model systems for studying major functional characteristics of food webs and for testing community theories (Armbruster et al. [@b1]; Kitching [@b22]). As yet, ca. 40 new ciliate taxa have been discovered in tank bromeliads and several have already been described (Dunthorn et al. [@b4]; Foissner [@b6],b[@b7], [@b10], [@b12]; Foissner and Stoeck [@b13]; Foissner and Wolf [@b14]; Foissner et al. [@b16], [@b15]; Omar and Foissner [@b28], [@b29]). Glaucomides bromelicola belongs to the Bromeliophryidae and is distantly related to Glaucoma (Foissner [@b12]; Foissner and Stoeck [@b13]). While the genus Glaucoma radiated in more common freshwater habitats, G. bromelicola is widespread in tank bromeliads, commonly occurring in several bromeliad species from southern Chile to Mexico (Foissner [@b12]). The original description focused on morphological and molecular characteristics while the ecology received little attention. Details of the ecology of ciliates from tank bromeliads are hitherto known only from one species, Bromeliothrix metopoides (Weisse et al. [@b42]). Accordingly, the goal of this study was to investigate the functional ecology of G. bromelicola in comparison to that of B. metopoides. In contrast to B. metopoides and many other colpodean ciliates, G. bromelicola does not form cysts (Foissner [@b12]), which renders this species vulnerable to extinction following desiccation in the tanks. Water volume (typically ca. 0.01--3 liters), light, pH (usually ranging from 4.0 to 7.0), nutrients, and allochtonous organic input from leaf litter are the most important abiotic factors characterizing phytotelmata of tank bromeliads (Brouard et al. [@b2]; Janetzky [@b20]; Laessle [@b23]; Lopez et al. [@b24]; Marino et al. [@b25]). We hypothesized that G. bromelicola should have evolved adaptations different from those of B. metopoides to compensate for the lack of cyst formation and promote its survival in astatic aquatic environments. We focused on food (quantity and quality), competition, and pH as major environmental variables that are amenable to experimental manipulation in the laboratory. As G. bromelicola often co-occurs with B. metopoides, the ciliate species that is second in common in tank bromeliads, the traits that enable coexistence are of special interest. To this end, we performed competition experiments in the laboratory. We did not consider temperature in this study because typical temperature variation in the tanks of bromeliads is in the range of (only) 10 °C (summarized by Weisse et al. [@b42]), i.e. presumably of minor importance relative to the other physico-chemical variables. Materials and Methods ===================== Origin and maintenance of stock cultures ---------------------------------------- Glaucomides bromelicola Foissner, [@b12] represents a new tetrahymenid genus and species that is frequent in tank bromeliads of Central and South America (Foissner [@b12]). Similar to B. metopoides Foissner, [@b10], G. bromelicola forms macrostomes when bacterial food is depleted. The common in vivo cell size of cultivated specimen is ca. 57 × 30 μm (microstomes), respectively, 80 × 50 μm (macrostomes). The organisms used in this study were collected from tanks of small tree bromeliads either in a subtropical forest of Brazil (B. metopoides, Polytomella sp., heterotrophic bacteria) or in the botanical garden of Puerto Plata, Dominican Republic (G. bromelicola). Details of the origin of our study organisms have been reported by Foissner ([@b10], [@b12]). Cultures were established in Eau de Volvic (French table water) enriched with some sterilized, crushed wheat grains to promote growth of indigenous bacteria and bacterivorous flagellates. Monospecific protist cultures were obtained by repeated dilution with Volvic and pipetting of individual target cells, respectively, by removing unwanted predators or competitors. Ciliate stock cultures were maintained in "filter caps" culture flasks (Biomedica) with 50 ml of Eau de Volvic enriched with one to two wheat grains. The stock cultures were kept in an incubator at 22.5 °C under a 14:10 h light--dark cycle; pH was ca. 7.5. New cultures were inoculated once per week by transferring 25 ml of the aged culture to a new flask containing 25 ml Eau de Volvic and one new wheat grain. We kept G. bromelicola in nonaxenic stock cultures with and without the as-yet-undescribed flagellate Polytomella sp.; this flagellate, which was isolated together with B. metopoides, is variable in cell size (10--40 μm), has four flagella at the basis of a distinct papilla, four contractile vacuoles, lacks an eyespot, and has the nucleus in the anterior body half (Foissner [@b10]). The genus Polytomella comprises several colorless nutritionally versatile species that thrive on acetate and other organic acids, peptone, and yeast extract (de la Cruz and Gittleson [@b3]; Pringsheim [@b30]). We added two wheat grains to 50 ml of Eau de Volvic to provide an organic-rich medium for our Polytomella sp. cultures. Ciliate cell numbers in stock cultures usually ranged from 0.5 to 2.0 × 10^4^ cells/ml. Growth experiments with different food organisms ------------------------------------------------ We measured specific growth rates (μ) of G. bromelicola in response to different food quality (heterotrophic bacteria and the flagellates Polytomella sp. and Cryptomonas sp.) and quantity. Polytomella sp. coexists with the ciliate in tank bromeliads, Cryptomonas sp. is a cryptophyte common in many freshwater bodies and the preferred food of many planktonic ciliates (Skogstad et al. [@b31]; Weisse and Müller [@b38]). If not specified, cell numbers of G. bromelicola include microstomes and macrostomes in the following. In asexually reproducing ciliates, μ is a direct proxy of their fitness (Weisse [@b35]). The ciliate was inoculated together with the respective food organism(s) into 50-ml culture flasks. Treatments with food organisms but without ciliates served as controls. Target food levels lower than satiating (\< 2 mg C/L; experiments reported in Fig. [3](#fig03){ref-type="fig"}) were obtained by diluting strongly growing cultures with Eau de Volvic. Flagellate prey concentration ranged from a few hundred cells/ml to \> 80,000 cells/ml (equivalent to \< 0.05--\> 2 mg C/L), depending on the experimental conditions. Bacterial levels ranged from 10^6^ to 10^8^ cells/ml. Except for the experiments with Cryptomonas sp., the bacterial flora used in this study was enriched from the original habitat. The bacteria present in the Cryptomonas cultures remained unidentified but were, most likely, qualitatively different from those used in the other experiments. The initial experimental ciliate abundance was 20--110 cells/ml. All growth experiments were performed in the dark to prevent photoautotrophic food (the flagellate Cryptomonas sp.) from growing. Similarly, we removed remnants from wheat grains from the experimental containers to limit bacterial growth. The experiments lasted from several days to several weeks. ![Numerical response of the ciliate Glaucomides bromelicola. (A) With the auxotrophic flagellate Polytomella sp. as food. (B) With the phototrophic flagellate Cryptomonas sp. as food. The solid lines represent the fit to Eq. 4 (see text).](jeu0060-0578-f1){#fig01} ![Population dynamics of the ciliate Glaucomides bromelicola fed bacteria and the flagellate Cryptomonas sp. at four different food levels (A--D). Symbols represent means of triplicates; error bars denote 1 SD.](jeu0060-0578-f2){#fig02} ![Population dynamics of the ciliates Bromeliothrix metopoides and Glaucomides bromelicola fed bacteria and the flagellate Polytomella sp. (A) B. metopoides in single growth experiments. (B) G. bromelicola in single growth experiments. (C) Both ciliates in pairwise growth (= competition) experiments. Symbols represent means of triplicates; error bars denote 1 SD.](jeu0060-0578-f3){#fig03} Samples (5 ml) were taken from the experimental containers at 24 h intervals and fixed with acid Lugol\'s iodine (final concentration 2% v/v). Ciliate cell numbers were determined microscopically either in counting chambers of 3 ml volume or in Sedgewick Rafter chambers of 1 ml volume. Flagellate cell numbers were also counted microscopically, either together with the ciliates or separately in a 1-ml Sedgewick chamber. At higher abundance (\> 10^4^ cells/ml), flagellates were also counted and sized electronically by means of an automatic particle counter (CASY 1-model TTC; Schärfe System, Reutlingen, Germany; Weisse and Kirchhoff [@b37]). Bacterial levels were measured in Formalin-fixed (2% v/v) samples (2 ml) taken together with the ciliate samples. Bacterial cell numbers were assessed by flow cytometry after staining with the green fluorescent nucleic acid stain SYTO-13 (Molecular Probes, Invitrogen^™^, Carlsbad, CA, USA), using a FACSCalibur flow cytometer (Becton Dickinson BD Biosciences, San Jose, CA) equipped with an argon ion laser emitting light at 488 nm. Cell size of ciliates and flagellates was measured with Lugol\'s fixed material using an inverted microscope and a semi-automatic image analysis system (LUCIA version 4.51, Laboratory Imaging, Prague, Czech Republic). Flagellates were also measured in unfixed material. The automatic particle counter yielded an independent estimate of flagellate cell size, which was used mainly to determine the cell volume of living cells. Both methods yielded similar results. The cell volume of the prey flagellates used in this study (Polytomella sp. and Cryptomonas sp.) was converted to carbon units assuming the allometric equation provided by Menden-Deuer and Lessard ([@b26]), i.e. pg C/cell = 0.216 × cell volume^0.939^. To calculate carbon biomass of bacteria, we assumed a conservative estimate of 26 fg C/cell for our cultivated bacteria (Troussellier et al. [@b33]). Cell volume (V) of the ciliates was determined from length (l) and width (w) measurements, assuming a prolate spheroid shape: where b is cell breadth (in μm, as l and w). Measurements were made on 50 ciliates each at the end of several experiments. As we could not measure the third dimension, we assumed that b is equal to 0.6 × w (Foissner [@b12]). Competition experiments ----------------------- The experiments described above assumed that G. bromelicola is the only ciliate feeding on bacterial and flagellate food. However, in the tank bromeliads, this ciliate is often encountered together with B. metopoides, a small colpodean ciliate (Dunthorn et al. [@b4]; Foissner [@b10]). Accordingly, we compared the growth response of both ciliates when kept alone to that of pairwise experiments, i.e. with the potential competitor present. In each of four different competition experiments, we offered a combination of bacteria and Polytomella sp. as food; initial food levels were saturating for both ciliate species (\> 2 mg C/L, Weisse et al. [@b42]; this study) but declined in the course of the experiments (Fig. [3](#fig03){ref-type="fig"}, 5). To investigate the effect of competition on growth and survival of B. metopoides, we added G. bromelicola as competitor to the experimental containers (50-ml culture flasks) with B. metopoides and vice versa. To ensure that ciliate growth rates with and without competitor were directly comparable, we inoculated, in the first competition experiment, from a vigorously growing B. metopoides culture, one treatment without competitor (Fig. [3](#fig03){ref-type="fig"}A) and another one with its competitor, G. bromelicola (Fig. [3](#fig03){ref-type="fig"}C); similarly, we inoculated simultaneously from a G. bromelicola culture two different treatments without (Fig. [3](#fig03){ref-type="fig"}B) and with its competitor, B. metopoides (Fig. [3](#fig03){ref-type="fig"}C). Each treatment was then split into three replicates. The flagellate Polytomella sp. (initial abundance ca. 50,000 cells/ml, equivalent to ca. 2.7 mg C/L) served as food for the macrostomes; bacteria (initial abundance ca. 10^8^ cells/ml, equivalent to ca. 2.6 mg C/L) were present and served as food for the microstomes, but were unimportant as food for the macrostomes. This first competition experiment, therefore, consisted of three different predator--prey treatments: (A) B. metopoides plus Polytomella sp. and bacteria; (B) G. bromelicola plus Polytomella sp. and bacteria; (C) B. metopoides and G. bromelicola plus Polytomella sp. and bacteria. The duration of this experiment (1 wk) turned out to be too short to reveal whether the ciliates may outcompete each other. Further, we did not monitor bacterial levels. To this end, we repeated this competition experiment with several modifications and extended the duration up to 4 wk; we also measured bacterial concentration together with flagellate and ciliate abundance. In the second competition experiment (Fig. [5](#fig05){ref-type="fig"}A, B), we provided Polytomella sp. and bacteria at similar initial satiating amounts as in the previous experiment (ca. 50,000 Polytomella/ml, ca. 10^8^ bacteria/ml) but did not add additional food in the course of the experiment. Due to ciliate grazing, food limitation was likely after several days. To reduce food limitation, we added flagellate food 5 d after the beginning of the third competition experiment ([Fig.5](#fig05){ref-type="fig"}C, D). Finally, in a fourth competition experiment ([Fig.5](#fig05){ref-type="fig"}E, F), we added food upon each sampling occasion, beginning on experimental day 5. Initial ciliate abundance ranged from 30 to 40 cells/ml in all competition experiments. The general experimental conditions, the analyses of protist and bacterial cell numbers, and the calculation of protist growth rates were identical to the growth experiments reported above. Response to pH -------------- We investigated the growth and survival of G. bromelicola over pH ranging from 4.0 to 9.0; pH was measured using a microprocessor pH-mV meter (model pH 526; WTW, Weilheim, Germany) to the nearest 0.01 unit. The pH sensor was 2-point calibrated with standard buffer solutions of pH = 6.87 and pH = 9.18 before each series of measurement. Ciliates and their prey (provided at satiating amounts) were acclimated to the experimental conditions in steps of 0.5 pH unit change per day for 2--5 d. We measured and adjusted the pH in each experimental container twice per day; when pH differed by more than 0.2 from the target pH, it was adjusted by addition of small amounts (15--35 μl) of 0.1 mol/l NaOH or HCl (Weisse and Stadler [@b39]; Weisse et al. [@b41], [@b44]). The general experimental design followed that of the growth experiments. We used an incubator with a constant temperature of 22.5 °C without illumination for all experiments. Except for the competition experiments 2--4 (results reported in [Fig.5](#fig05){ref-type="fig"}), which were performed in duplicate, all experiments were run in triplicate. Results reported are mean values ± 1 standard deviation (SD). Data analysis and statistics ---------------------------- Ciliate growth rate (μ) was calculated from changes in cell numbers, assuming exponential growth over the experimental period according to where N~0~ and N~t~ are ciliate numbers at the beginning and at the end of the experimental period, respectively. Please note that 'experimental period' does not denote the total duration of an experiment, but specific periods of 1--4 d each during which the ciliate population increased exponentially. Details are reported in the Results section. Ciliate growth rates were related to the geometric mean prey concentration (P) during the experimental period (Frost [@b17]; Heinbokel [@b18]) according to Eq. 3: where P~0~ and P~t~ are the initial and final prey concentrations (cells/ml) during incubations. Ciliate growth rates were fit to Eq. 4, which includes a positive x-axis intercept, using the Marquardt-Levenberg algorithm (SigmaPlot; SPSS Inc., Chicago, IL). where μ is the growth rate (/d), μ~max~ is the maximum growth rate (/d), P is the geometric mean prey concentration (Eq. 3), k is a constant (cells/ml), and x′ is the x-axis intercept (i.e. threshold concentration, cells/ml, where μ = 0). This equation is similar to the Michaelis--Menten model and Holling\'s type II functional response (Holling [@b19]), but assumes a positive x-axis intercept where population growth equals mortality (Weisse et al. [@b43]). Accordingly, the constant k of Eq. 4 is similar, but not identical to the half saturation constant known from Michaelis--Menten kinetics. The curve resulting from Eq. 4 that describes the change in predator density as a function of change in prey density is known as numerical response (Solomon [@b32]). One-way ANOVA and Student\'s t-test were used to test for significant differences between several pairwise treatments. Results were considered statistically significant if p \< 0.05. Results ======= Response to food supply ----------------------- With Polytomella sp. and bacteria as food, G. bromelicola showed a typical numerical response curve (Fig. [1](#fig01){ref-type="fig"}A). The nonlinear curve fit (Eq. 4) yielded a maximum growth rate (μ~max~) of 3.55/d, a constant k = 7,035 Polytomella cells/ml, and an x-axis intercept (x′ in Eq. 4, where μ = 0) of 868 Polytomella cells/ml. All parameter estimates and the curve fit were statistically significant (p \< 0.001, R^2^ = 0.882). When converted to carbon units, food levels at near-to-maximum growth rates (μ = 3.0) were 1.92 mg C/L, k was reached at 0.38 mg C/L, and x′ amounted to 0.05 mg C/L. The percentage of flagellate-feeding macrostomes ranged from 1.0% to 8.5% of the total G. bromelicola cell number. The calculations above ignore the presence of bacteria that were, most likely, primarily used as food by the microstomes. Bacterial levels ranged from 1.1 to 10.5 × 10^6^ cells/ml in the numerical response experiment, equivalent to 0.03--0.28 mg C/L. As bacterial abundance was positively correlated with Polytomella abundance, the threshold food concentration of G. bromelicola was likely close to 0.1 mg C/L, and μ~max~ was recorded when total food concentration (i.e. Polytomella sp. plus bacteria) reached ca. 2.2 mg C/L. We did not assess the numerical response of G. bromelicola to bacteria as sole food, because we had observed in our routine cultures that ciliate growth rates are lower in the absence of the flagellate (Weisse, unpubl. data, but see below, Response to pH). When we replaced Polytomella sp. by Cryptomonas sp., G. bromelicola needed approximately 3 d to adapt to the new food (Fig. [2](#fig02){ref-type="fig"}). Irrespective of the food level, ciliate growth rates were low to moderate during the first 2--3 d of the experiment and then increased. Differences between Cryptomonas abundance in the experimental containers with ciliates and controls without ciliates were small and mostly statistically not significant during this period, suggesting that G. bromelicola did not at all or only sparsely feed upon the flagellates. However, at day 7, Cryptomonas levels were significantly lower (pairwise t-tests, p \< 0.001 in each case) in each experimental container than in the respective controls, and ciliate numbers had increased, relative to the first 4 d of the experiment (Fig. [2](#fig02){ref-type="fig"}A--D). Only at the lowest flagellate abundance (ca. 15,000 cells/ml), the ciliates did not grow during the period of observation, although bacteria were present in moderate abundance at the beginning of the experiment and continuously increased up to 9 × 10^6^ cells/ml to the end of the observation period (Fig. [2](#fig02){ref-type="fig"}A). The decline in bacterial levels recorded in Fig. [2](#fig02){ref-type="fig"}C, D demonstrates that G. bromelicola microstomes fed on bacteria, because bacterial levels did not decline in the controls without ciliates. Similarly, the initial bacterial growth in treatments A and B was lower than in the respective controls (data not shown). The difference between Cryptomonas sp. abundance in the experimental containers and controls indicates that the flagellate was ingested by G. bromelicola at the lowest flagellate abundance tested. However, the cryptophyte biomass of ca. 0. 4 mg C/L did not support positive ciliate growth. Including bacteria, the total concentration of potential food was ca. 0. 6 mg C/L at the end of this experiment. When we plotted the ciliate growth rate measured between experimental days 4--7 vs. the mean Cryptomonas abundance during this period, we obtained a numerical response curve with a statistically significant (R^2^ = 0.896) curve fit (Fig. [1](#fig01){ref-type="fig"}B). However, comparing Fig. [1](#fig01){ref-type="fig"}B to A demonstrates that μ~max~ of G. bromelicola was significantly lower (0.60/d) with Cryptomonas sp. as food and the cellular threshold prey concentration was approximately tenfold higher for Cryptomonas sp. than for Polytomella sp. In terms of biomass, the difference was lower, ca. 0. 30 mg C/L (Cryptomonas sp.), respectively, 0.05 mg C/L (Polytomella sp.). Including bacteria, we estimated a total threshold concentration of ca. 0.5 mg C/L for G. bromelicola fed Cryptomonas sp. and bacteria. Tables [1](#tbl1){ref-type="table"}, [2](#tbl2){ref-type="table"} summarize the results obtained in both numerical response experiments. ###### Parameter estimates of the numerical response curves (see text, Eq. 4) of Glaucomides bromelicola fed the flagellates Polytomella sp., respectively, Cryptomonas sp. Variable Coefficient SE Coefficient SE ---------- ------------- -------------------------------------------- ------------- ------------------------------------------- μ~max~ 3.55 0.23[\\\](#tf1-3){ref-type="table-fn"} 0.60 0.14[\\](#tf1-2){ref-type="table-fn"} x′ 868 170[\\\](#tf1-3){ref-type="table-fn"} 10,678 670[\\\](#tf1-3){ref-type="table-fn"} k* 7,035 1,470[\\\](#tf1-3){ref-type="table-fn"} 12,364 6,782[\](#tf1-1){ref-type="table-fn"} R^2^ 0.88 0.44[\\\](#tf1-3){ref-type="table-fn"} 0.90 0.07[\\\](#tf1-3){ref-type="table-fn"} \*p \< 0.102. \\p \< 0.01. \\\*p \< 0.0001. μ~max~, maximum growth rate (per d), x′, threshold food concentration (cells/ml), k, a constant (cells/ml), R, coefficient of determination, SE, standard error. ###### Key variables characterizing the life strategies of the ciliates Bromeliothrix metopoides and Glaucomides bromelicola from tank bromeliads Variable B. metopoides G. bromelicola Source -------------------------------------- -------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------- Taxonomic affiliation (class, order) Colpodea, Colpodida Oligohymenophorea, Tetrahymenida Foissner ([@b10], [@b12]) Cell length 20--55 μm 25--80 μm Foissner ([@b10], [@b12]), Weisse et al. ([@b42]) Average cell volume in cultures 8,570 μm^3^ 8,250 μm^3^ Weisse et al. ([@b42]), This study Macrostomes Yes Yes Foissner ([@b10], [@b12]) Cysts Yes No Foissner ([@b10], [@b12]) Division chains Yes No Foissner ([@b10], [@b12]) Food organisms Bacteria; Polytomella sp. Bacteria; Polytomella sp.; other flagellates Weisse et al. ([@b42]), This study Sensitivity to competition Moderate Low This study Growth rate threshold 1.4[a](#tf2-1){ref-type="table-fn"}--2.3[b](#tf2-2){ref-type="table-fn"}mg C/L 0.05[](#tf2-3){ref-type="table-fn"}--0.1[a](#tf2-1){ref-type="table-fn"}mg C/L ∼0.5[d](#tf2-4){ref-type="table-fn"}mg C/L Weisse et al. ([@b42]), This study μ~max~ (/d) 3.55 4.71 Weisse et al. ([@b42]), This study pH tolerance \< 4--\> 9 \< 4--\> 9 Weisse et al. ([@b42]), This study pH optimum ∼8 7--8 Weisse et al. ([@b42]), This study With bacteria and Polytomella sp. as food, i.e. microstome and macrostome feeding. With bacteria as sole food, i.e. only microstome feeding. With Polytomella sp. as food, ignoring the bacterial background. With bacteria and Cryptomonas sp. as food, i.e. microstome and macrostome feeding. Competition for food -------------------- When kept alone with their prey, both ciliates started growing exponentially after an initial lag phase (Fig. [3](#fig03){ref-type="fig"}A, B). In contrast to B. metopoides (Fig. [3](#fig03){ref-type="fig"}A), exponential growth of G. bromelicola continued to the final day of observation (Fig. [3](#fig03){ref-type="fig"}B); μ calculated from linear regression of ln cell numbers vs. time was 1.19 ± 0.12/d (days 1--6, n = 6, R^2^ = 0.960) for B. metopoides, respectively, 1.37 ± 0.07/d (days 2--7, n = 6, R^2^ = 0.989) for G. bromelicola. Highest growth rates from day to day were reached for both ciliates during days 3--4, with μ~max~ = 2.45 ± 0.16/d (B. metopoides) and μ~max~ = 2.01 ± 0.20/d (G. bromelicola). In the latter case, average ciliate cell number increased from 84.3 to 632.0 cells/ml during days 3--4; due to the large scale of the y-axis, this is difficult to deduce from Fig. [3](#fig03){ref-type="fig"}B. In the pairwise experiment (Fig. [3](#fig03){ref-type="fig"}C), B. metopoides reached the same continued growth rate, 1.20 ± 0.09/d, as in the single growth experiment, but for a shorter period (days 1--4, n = 4, R^2^ = 0.989); during days 4--6, μ of B. metopoides was ≤ 1.0/d. Similarly, highest day to day growth rate was reached during days 3--4; however, μ~max~ = 1.51 ± 0.02/d in the competition experiment was significantly reduced (p \< 0.001) compared to the single growth experiment. Maximum cell number in the former was 4,700 cells/ml, compared to 8,060 cells/ml in the single growth experiment. Final abundance of B. metopoides was 6,041 ± 374 cells/ml (single growth), respectively, 777 ± 166 cells/ml (pairwise growth). These differences in the duration of exponential growth and final cell numbers were affected by a switch to mass encystment observed in the competition experiment during days 6--7 (Fig. [4](#fig04){ref-type="fig"}). On day 7, cyst abundance in the competition experiment was significantly higher (p \< 0.001) than in the single growth experiment. The opposite was true for days 2 and 4; we recorded no difference in cyst abundance between the two treatments during days 0, 1, 5, and 6. ![Cyst formation of the ciliate Bromeliothrix metopoides in the experiments shown in Fig. [3](#fig03){ref-type="fig"}A and C.](jeu0060-0578-f4){#fig04} Continued exponential growth rate of G. bromelicola in the pairwise experiment (Fig. [3](#fig03){ref-type="fig"}C) was identical to that measured in the single growth experiment (1.36 ± 0.08/d; days 3--7, n = 5, R^2^ = 0.988); similarly, μ~max~ = 1.88 ± 0.10/d reached during days 3--4 and was not different (p = 0.369) from the single growth experiment. Considering that G. bromelicola grew exponentially, differences in its final abundance reported in Fig. [3](#fig03){ref-type="fig"}B, C were minor. Both ciliate species formed macrostomes in the competition experiments, with typical percentages ranging from 2% to 7% of the total cell number. In summary, the first competition experiment demonstrated some effect of G. bromelicola on B. metopoides, but virtually no effect of the latter on the former. The experimental duration was too short to reveal if B. metopoides may be outcompeted by G. bromelicola. In the following three competition experiments, we extended the duration up to 4 wk; we also measured bacterial concentration together with flagellate and ciliate abundance. Results shown in Fig. [5](#fig05){ref-type="fig"} demonstrate that both ciliates may coexist over several weeks. In the second competition experiment, without feeding (Fig. [5](#fig05){ref-type="fig"}A, B), both ciliates coexisted with small, but relatively stable populations (B. metopoides: 242 ± 150 cells/ml, G. bromelicola 313 ± 285 cells/ml) after 2 wk, when the abundance of the food flagellate Polytomella sp. remained \< 1,000 cells/ml, and bacterial levels were close to 0.2 × 10^8^ cells/ml (Fig. [5](#fig05){ref-type="fig"}A). Starting 1 wk after the beginning of the experiment, the abundance of encysted B. metopoides exceeded that of motile cells and continuously increased until the end of the experiment (Fig. [5](#fig05){ref-type="fig"}B). Maximum abundance of G. bromelicola (25,715 ± 3,692 cells/ml) was more than 10-fold higher than that of B. metopoides (1,753 ± 85 cells/ml). Both ciliates were, most likely, food limited in this experiment after the initial week. ![Competition experiment with the ciliates Bromeliothrix metopoides and Glaucomides bromelicola (bottom panels) fed bacteria and the flagellate Polytomella sp. (top panels). (A, B) No additional food was added in the course of the experiment. (C, D) Additional food was added on day 5 of the experiment (indicated by dashed lines and arrows in Fig. [5](#fig05){ref-type="fig"}C). (E, F) Additional food was added on each sampling occasion, beginning on day 5 of the experiment. Bacterial abundance is related to the left y-axis in Fig. [5](#fig05){ref-type="fig"}A, C, E; flagellate abundance is related to the right y-axis in Fig. [5](#fig05){ref-type="fig"}A, C, E. Symbols represent means of duplicates, error bars denote 1 SD.](jeu0060-0578-f5){#fig05} In the third competition experiment, we added flagellate food 5 d after the beginning of the experiment (Fig. [5](#fig05){ref-type="fig"}C). The ciliates coexisted in this experiment for the following 2 wk with higher cell numbers than in the previous experiment (Fig. [5](#fig05){ref-type="fig"}D). The average abundance of G. bromelicola (12,608 ± 4,553 cells/ml) was ca. 14× higher than that of B. metopoides (884 ± 337 cells/ml) during the second and third week of the experiment. Different from the second competition experiment without additional feeding, motile cells of B. metopoides prevailed over encysted cells throughout this experiment. In the fourth competition experiment, with food added upon each sampling occasion from experimental day 5 onwards, food levels of bacteria and Polytomella sp. were higher than in the previous experiments (Fig. [5](#fig05){ref-type="fig"}E). Ciliate growth rates cannot be calculated directly from the results shown in Fig. [5](#fig05){ref-type="fig"}F because the experimental containers were diluted 1:3 upon each addition of food. However, the nearly constant ciliate cell numbers recorded for both species during experimental days 5--19 suggest that the ciliate populations increased with a relatively constant rate close to 0.55/d. Similar to the previous experiment, (i) the average abundance of G. bromelicola (4,534 ± 1,539 cells/ml) was ca. 10-fold higher than that of B. metopoides (456 ± 58 cells/ml), and (ii) motile cells of B. metopoides prevailed over encysted cells. Individual and average ciliate cell size was variable in the different growth and competition experiments, mainly depending on food and the percentage of macrostomes in the population (data not shown). Assuming typical average dimensions of l = 42 μm and w = 25 μm and accounting for 10% shrinkage due to fixation, we calculated an average cell volume of G. bromelicola of 8,250 μm^3^ under food replete conditions. The average cell size of G. bromelicola measured in the present investigation was thus smaller than the in vivo cell size reported by Foissner ([@b12]) for well-fed laboratory specimens. Response to pH -------------- We investigated the pH response of G. bromelicola over pH ranging from 4 to 9 with combined bacterial and flagellate (Polytomella sp.) food, respectively bacteria without any other food organism (Fig. [6](#fig06){ref-type="fig"}). To avoid food limitation, prey organisms were provided in satiating amounts (\> 2 mg C/L). With the combined food, G. bromelicola was relatively insensitive to pH, reaching high growth rates (\> 2.4/d) over the entire pH range tested; growth rates peaked at pH 7--8, where μ was not different (t-test, p = 0.073). Highest growth rates, μ~max~ = 3.05/d, obtained in the pH experiment with Polytomella sp. plus bacterial food at pH 7--8 were close to μ~max~ measured in the numerical response experiment with the same food organisms (cf. Fig. [1](#fig01){ref-type="fig"}A). If bacteria were the sole food, the ciliates (i) reached lower maximum growth rates, (ii) the pH optimum was shifted to acidic conditions, and (iii) the highest pH barely supported positive growth rates (Fig. [6](#fig06){ref-type="fig"}). ![pH response of the ciliate Glaucomides bromelicola fed bacteria only, respectively bacteria plus the flagellate Polytomella sp. Bars represent means of triplicates, error bars denote 1 SD.](jeu0060-0578-f6){#fig06} Discussion ========== Different life strategies of Glaucomides bromelicola and Bromeliothrix metopoides enable their stable coexistence in tank bromeliads ---------------------------------------------------------------------------------------------------------------------------------------- Results of this study and a related recent study (Foissner [@b10], [@b12]; Weisse et al. [@b42]) demonstrate contrasting life strategies of the two common ciliates from tank bromeliads (summarized in Table [2](#tbl2){ref-type="table"}). As G. bromelicola cannot form cysts, this species is threatened by desiccation that may be common in many small bromeliads (Kitching [@b22]; Marino et al. [@b25]). Accordingly, a G. bromelicola population can escape extinction only via dispersal into new suitable habitats. The preferred flagellate prey, Polytomella sp., is common in tank bromeliads but it was not found in every reservoir (Foissner, unpubl. data). The ability to adapt to different prey such as Cryptomonas sp. may enable survival of the ciliate in less suitable habitats. The adaptation period of ca. 3 d that we observed (Fig. [2](#fig02){ref-type="fig"}) appears to be short enough to prevent extinction of the population. Although μ~max~ was significantly reduced with Cryptomonas sp. as food (μ~max~ = 0.60/d), it is still in a range typical for many aquatic ciliates known to feed on Cryptomonas sp. (Skogstad et al. [@b31]; Weisse [@b35]; Weisse and Müller [@b38]). Secondly, the food threshold of G. bromelicola, i.e. the prey level that is needed to sustain the population, is relatively low, compared to most planktonic ciliates (Weisse [@b35]), and orders of magnitudes lower than that of B. metopoides (Weisse et al. [@b42]). This nutritional versatility may compensate for the lack of cyst formation in G. bromelicola. Bromeliothrix metopoides is unable to use Cryptomonas sp. and other flagellates and small ciliates as food (Weisse et al. [@b42]). With its preferred food, Polytomella sp., the threshold concentration is close to 1.4 mg C/L (Table [2](#tbl2){ref-type="table"}), i.e. higher than that of any other aquatic ciliate investigated thus far under comparable experimental conditions (Weisse et al. [@b42]). The microstomes of B. metopoides, which feed on bacteria, require an even higher food threshold (2.3 mg C/L) to proliferate. In contrast, microstomes of G. bromelicola can ingest bacteria at concentrations \< 10^7^/ml, equivalent to ca. 0.2 mg C/L (Fig. [2](#fig02){ref-type="fig"}C, D). The lower affinity to its preferred bacterial and flagellate food is likely the cause why B. metopoides is the inferior competitor when paired with G. bromelicola. However, encystment and excystment enable B. metopoides to coexist with G. bromelicola over weeks (Fig. [5](#fig05){ref-type="fig"}). During short periods of time, the former may even reach higher growth rates than the latter (Table [2](#tbl2){ref-type="table"} and Fig. [3](#fig03){ref-type="fig"}). We repeatedly observed greater than five cell divisions per day of B. metopoides in our laboratory cultures (Foissner [@b10]; Weisse et al. [@b42]), similar to the shortest generation times known from small terrestrial colpodeans (reviewed by Weisse et al. [@b42]). If the ciliates reach such high growth rates in situ remains at present unknown. Why are both ciliates restricted to tank bromeliads? ---------------------------------------------------- A central issue of this research was to explain why the most common ciliates from tank bromeliads appear to be restricted to this peculiar habitat. Of the four environmental variables studied (pH, food quantity and quality, competition), pH was seemingly of minor importance, as both ciliate species were widely tolerant to changing pH (Weisse et al. [@b42]; this study). However, recent experimental evidence with freshwater ciliates, flagellates, and microeukaryotes revealed that the realized pH niche may be confined to a small range, relative to the fundamental pH assessed in the laboratory under optimized conditions. This is because there is a significant interaction of pH with food and temperature (Moser and Weisse [@b27]; Weisse [@b35]; Weisse et al. [@b43], [@b40], b[@b42]). Interaction of pH with food quality was obvious for both ciliate species of this study; the pH niche was narrower and, in the case of G. bromelicola, restricted to the pH range typical of bromeliad reservoirs if bacteria were the only food (Fig. [6](#fig06){ref-type="fig"}; Weisse et al. [@b42]). It appears that (metabolism of) the flagellate Polytomella sp. conditioned the medium in a favorable way for G. bromelicola, promoting high growth rates even under alkaline conditions. The unusually high food threshold and its demand for specific food items (Polytomella sp. and some bacteria) are the main reasons why B. metopoides cannot thrive in other environments (Weisse et al. [@b42]), in spite of its ability to form cysts and endure desiccation of its natural habitat. Summarizing the sparse literature on the abundance of bacteria and flagellates in tank bromeliads, Weisse et al. ([@b42]) concluded that, while it is likely that the high food levels needed by B. metopoides are occasionally met in situ, it remains at present unknown if such favorable conditions are persistent in tank bromeliads. The results of this study further suggest that motile cells of this species are inferior competitors, relative to other ciliates with higher affinity to prey common in bromeliads and elsewhere. Weisse et al. ([@b42]) concluded that B. metopoides has reached a narrow peak along its fitness landscape and any deviation from the optimum conditions will reduce its fitness. This conclusion does not apply to G. bromelicola; based upon its feeding and growth characteristics reported in this study, this species should be able to live in a wider array of aquatic habitats than B. metopoides. Indeed, G. bromelicola was found in samples from 11 different bromeliad species, while B. metopoides occurred in only three bromeliad species (Dunthorn et al. [@b4]). The lack of cyst formation severely limits dispersal of G. bromelicola. Long range dispersal of ciliates and other microbes is only possible in the form of resting stages resistant to desiccation (Foissner [@b8], [@b9], [@b11]; Weisse [@b36]). We hypothesize that dispersal of G. bromelicola over narrow ranges is primarily mediated via insects specifically attracted to bromeliads, explaining why this species is endemic in tank bromeliads. A corollary of this conclusion is that it supports the earlier notion that the high degree of endemicity observed in the bromeliad tanks points to speciation within this habitat (Dunthorn et al. [@b4]; Foissner [@b12]; Foissner et al. [@b16]). Both species used in the present investigation possess a micronucleus (Foissner [@b10], [@b12]). However, conjugation was observed only in G. bromelicola (Foissner, unpubl. data), while most colpodids very likely lack sex at all (Foissner [@b5]). Thus far, no attempts have been made at mating in the laboratory. The genetic population structure of tank bromeliads is an area that awaits future research. We appreciate the constructive comments by two anonymous reviewers on the manuscript. This study was financially supported by the Austrian Science Fund (FWF, project P20360-B17).	{ "pile_set_name": "PubMed Central" }
![](brjcancer00051-0226.tif "scanned-page"){.1122}	{ "pile_set_name": "PubMed Central" }
Introduction {#Sec1} ============ The All-Pairs Shortest Paths (APSP) problem asks to find distances between all pairs of vertices in a graph. For a directed graphs with weights in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {R}$\end{document}$, there is a classical O(n^3^) time algorithm \[[@CR11], [@CR29]\]. Currently best upper bound for this problem is due to \[[@CR30]\] who showed an $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$O\left (\frac {n^{3}}{2^{{\Omega }(\log n)^{0.5}}}\right )$\end{document}$ algorithm. It is asymptotically faster than O(n^3^/logcn) for any c \> 0 (see survey \[[@CR6]\] for earlier algorithms). Showing any algorithm that would work in O(n^3−𝜖^) time for some 𝜖 \> 0 is a major open problem \[[@CR30]\]. If we consider unweighted, directed graphs there are subcubic algorithms that exploit fast matrix multiplication. For the undirected graph \[[@CR24]\] presented the optimal $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time algorithm, where ω \< 2.373 is the matrix multiplication exponent \[[@CR16]\]. For the directed case \[[@CR35]\] presented an O(n^2.575^) time algorithm that is based on the fast rectangular matrix multiplication. Moreover, if we are interested in small integer weights from the set {−M,...,M} we have O(M^0.68^n^2.575^) algorithm \[[@CR35]\]. Because APSP in undirected graphs can be solved in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$, diameter, radius, shortest cycle, etc. can be determined in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time as well. It is surprising that for a directed case, where merely O(n^2.575^) APSP is known there are also $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ algorithms for determining these properties. After a long line of improvements \[[@CR9]\] showed an $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(M n^{\omega })$\end{document}$ time algorithms for finding minimum weight perfect matching, shortest cycle, diameter and radius (some of these results were already known \[[@CR21]\]). Also, \[[@CR9]\] showed an application of their techniques that improves upon \[[@CR31]\] $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(M n^{\omega } t)$\end{document}$ time algorithm for the following problem: determine the set of vertices that lie on some cycle of length at mostt. Cygan et al. \[[@CR9]\] managed to solve this problem in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(M n^{\omega })$\end{document}$ time using Baur-Strassen's theorem. All of these algorithm are effective merely in the case of a dense graphs. For graphs with the small number of edges there are more efficient algorithms (e.g., APSP in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(\|V\|\|E\|)$\end{document}$ time \[[@CR27]\]). But these algorithms are Θ(n^3^) when \\|E\\| = Θ(n^2^). Related Work {#Sec2} ------------ ### Distance Queries {#Sec3} Yuster and Zwick \[[@CR33]\] considered the weighted, directed graphs with weights in {−M,...,M}. They showed an algorithm that needs $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(M n^{\omega })$\end{document}$ preprocessing time. After preprocessing each distance δ(u, v) in the graph can be computed exactly in O(n) query time. In the special case M = 1 they showed $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ algorithm that solves Single Source Shortest Paths (SSSP). This is the best known algorithm for a dense, weighted graph. We will match their bounds (up to the polylogarithmic factors) using Frobenius normal form. Next we will extend that approach so it will return more information about a graph in the same query/preprocessing time. ### Counting Cycles {#Sec4} For a given graph G and k determining whether G contains a simple cycle of length exactly k is NP-hard (in particular determining whether a graph contains a Hamiltonian cycle is NP-complete). However, if we fix k to be a constant this problem can be solved in polynomial time. Alon et al. \[[@CR4]\] introduced a color coding technique. For a fixed k if a graph G(V, E) contains a simple cycle of size exactly k then such a cycle can be found in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(\|V\|^{\omega })$\end{document}$ time. Unfortunately, their algorithm depends exponentially 2^O(k)^ on the length of the cycle and in consequence is inapplicable for large k. In the next years, \[[@CR5]\] showed (using a different technique) that for k ≤ 7 one can count the number of cycles of length exactly k in a graph in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(\|V\|^{\omega })$\end{document}$ time. In \[[@CR32]\] it is shown that for any even k, cycles of length k can be found in O(\\|V \\|^2^) time in undirected graphs (if they contain such a cycle). Alon et al. \[[@CR5]\] showed more methods that depend solely on a number of edges in a graph. For example for odd k they showed $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$O\left (E^{2-\frac {2}{k + 1}}\right )$\end{document}$ algorithm for finding cycles of length k. However, for dense graphs these results are worse than \[[@CR4]\]. From the other hand, to detect whether a non-simple cycle of length exactly k exists one can use the folklore algorithm. It starts by taking the adjacency matrix A of a graph G. Subsequently, in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time compute A^k^ by repeated squaring. If Tr \[A^k^\] \> 0 then there exists a non-simple cycle of length k.[1](#Fn1){ref-type="fn"} Yuster \[[@CR31]\] considered the following problem: for every vertex in a graph find a shortest cycle that contains it. He called this problem All-Nodes Shortest Cycle (ANSC). He showed a randomized algorithm that solves ANSC for undirected graphs with weights {1,...,M} in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(\sqrt {M} n^{(\omega + 3)/2})$\end{document}$ time. He noted that for simple digraphs (directed graphs with no anti-parallel edges) it reduces to All-Pairs Shortest Path problem. The fastest known APSP algorithm for unweighted, directed graphs runs in O(n^2.575^) due to \[[@CR35]\]. Here, we will show how to solve ANSC in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ for general, unweighted, directed graphs. Unfortunately, our techniques will allow us merely to find the length of such a cycle. But we can return the set of points, that lie on some cycle of a given length. Independently to our work \[[@CR3]\] proved that ANSC can be solved in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ for unweighted, undirected graphs using a completely different technique. Yuster \[[@CR31]\] also considered following problem: given a graph and an integert. LetS(k) denote the set of all vertices lying in a cycle of length≤ k. DetermineS(t). He considered directed graphs with weights in {−M,...,M} and showed $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(M n^{\omega } t)$\end{document}$ algorithm. Recently, \[[@CR9]\] improved his algorithm. They showed that for a fixed t ∈ \[0,nM\] the set S(t) can be computed in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(M n^{\omega })$\end{document}$ randomized time. We show, that for unweighted (M = 1) directed graphs we can compute sets S(1),S(2),...,S(D) in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time with high probability. Preliminaries {#Sec5} ============= Let T(n) be the minimal number of algebraic operations needed to compute the product of n × n matrix by an n × n matrix. We say that ω is the exponent of square matrix multiplication. For now the best known upper bound on ω is due to \[[@CR16]\]: Theorem 1 {#FPar1} ------------- \[[@CR16]\] For every𝜖 \> 0,T(n) \< O(n^ω\ +\ 𝜖^) whereω \< 2.37287. In this paper, we will omit 𝜖 in definition and will assume that O(n^ω^) operations are needed to multiply two matrices. The best lower bound for the exponent of matrix multiplication is ω ≥ 2. For convenience in this paper we will assume that ω \> 2. The $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}$\end{document}$ notation hides polylogarithmic factors in complexity. We will use it to emphasize that all our algorithms need polylogarithmic number of calls to the fast matrix multiplication algorithm. In this paper we will consider the Word RAM model of computation with a word size O(log n). The $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {F}$\end{document}$ denotes a small finite field and often we will assume this field to be $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {Z}_{p}$\end{document}$ for a prime number p with O(log n) bits. Note, that in Word RAM model arithmetic operations in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {Z}_{p}$\end{document}$ can be done in constant time. In some situations we will need to stress that we want to work on larger integers. We will use W to denote the upper bound on the largest number in the field. In this case the arithmetic operations in Word RAM model take O(log W) time. For matrices $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A \in \mathbb {F}^{n\times k}$\end{document}$ and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$B \in \mathbb {F}^{n \times l}$\end{document}$ the $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A \oplus B \in \mathbb {F}^{n \times (k+l)}$\end{document}$ is the concatenation of their columns. $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$C_{a,b} \in \mathbb {F}^{n\times (b-a)}$\end{document}$ denotes a matrix constructed by concatenating columns c~a~,c~a+ 1~,...,c~b~ of matrix $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$C \in \mathbb {F}^{n \times m}$\end{document}$. In this paper we will consider the unweighted, directed graph G unless stated otherwise. We will denote by V (G) and E(G) the sets of vertices and edges of graph G. The adjacency matrix A(G) ∈{0,1}^n×n^ of a directed graph G with n vertices is defined as: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A(G)_{i,j} = \left\{\begin{array}{lll} 1 & \text{if} i \rightarrow j \in E(G) \\ 0 & \text{if} i \rightarrow j \notin E(G). \end{array}\right. $$\end{document}$$ Introduction to Cyclic Subspaces and Connection to Frobenius Matrices {#Sec6} ===================================================================== We will start with a motivational example of application of cyclic subspaces. We have a constant-recursive sequence $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$a_n = c_0 a_{n-1} + c_1 a_{n-2} + {\ldots} + c_{r-1} a_{n-r} $$\end{document}$$ of order r, where all c~i~ are constants and initial conditions a~0~,...,a~r− 1~ are given. Perhaps, the most familiar example is the Fibonacci sequence F~n~ = F~n− 1~ + F~n− 2~. We can define the companion matrix of our general sequence as: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C = \left[\begin{array}{llll} 0&\ldots&0 & -c_0\\ 1&\ddots& & -c_1\\ &\ddots& 0&{\vdots} \\ 0& & 1&-c_{r-1} \end{array}\right]. $$\end{document}$$ The crucial property of the matrix C is that we can generate the next element of sequence with multiplication by C: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C^T \left[\begin{array}{lll} a_{n-r} \\ {\vdots} \\ a_{n-2} \\ a_{n-1} \end{array}\right] = \left[\begin{array}{lll} a_{n-r + 1} \\ {\vdots} \\ a_{n-1} \\ a_{n} \end{array}\right]. $$\end{document}$$ For example, for Fibonacci sequence we have: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left[\begin{array}{lll} 0 & 1 \\ 1 & 1 \end{array}\right] \left[\begin{array}{lll} F_k \\ F_{k + 1} \end{array}\right] = \left[\begin{array}{lll} F_{k + 1} \\ F_{k + 2} \end{array}\right]. $$\end{document}$$ Now, if we want the a~n+ 1~ element of the series we can square the companion matrix and get: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C^T C^T \left[\begin{array}{lll} a_{n-r} \\ {\vdots} \\ a_{n-2} \\ a_{n-1} \end{array}\right] = C^T \left[\begin{array}{lll} a_{n-r + 1} \\ {\vdots} \\ a_{n-1} \\ a_{n} \end{array}\right] = \left[\begin{array}{lll} a_{n-r + 2} \\ {\vdots} \\ a_{n} \\ a_{n + 1} \end{array}\right]. $$\end{document}$$ And analogously to get the a~n\ +\ k− 1~ element we need to compute the k-th power of the matrix C. In linear algebra such transformations are known as the cyclic subspaces generated by the vector a. For our purposes, we will restrict ourself to a conclusion, that some columns of a companion matrix occur in its powers. These properties are well known in the linear algebra theory (see \[[@CR10], [@CR13]\] for more cyclic properties). Consequences of Frobenius Normal Form {#Sec7} ------------------------------------- Let $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {F}$\end{document}$ be a commutative field. For any matrix $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A \in \mathbb {F}^{n\times n}$\end{document}$ there exists an invertible U over $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {F}$\end{document}$ such that: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$U^{-1} A U = F = \left[\begin{array}{lllll} C_1 & & & & 0 \\ & C_2 & & \\ & & C_3 & & \\ & & & {\ddots} & \\ 0 & & & & C_k \end{array}\right]. $$\end{document}$$and F is the Frobenius-canonical-form[2](#Fn2){ref-type="fn"} of A. The diagonal block C~i~ is called the companion matrix: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C_i = \left[\begin{array}{lllllll} 0& &{\ldots} & &0 & -c_0\\ 1&0& & &0 & -c_1\\ &1&{\ddots} & &{\vdots} & -c_2\\ & &{\ddots} &0& & {\vdots} \\ & & &1&0 & -c_{r-2}\\ 0& & & &1 & -c_{r-1} \end{array}\right] \in \mathbb{F}^{r\times r} . $$\end{document}$$ Each companion matrix corresponds to the monic polynomial $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$C_{i}(x) = x^{r} + c_{r-1}x^{r-1} + {\ldots } + c_{0} \in \mathbb {F}\left [x\right ]$\end{document}$ (similarly to the sequence example) and this polynomial is called the minimal polynomial of A. Each minimal polynomial has a property that C~i~(A) = 0. To guarantee that matrix has only one decomposition into Frobenius normal form we require that every polynomial must divide the next one, i.e., C~i~(x)\\|C~i+ 1~(x). The final list of polynomials is called the invariant factors of matrix A \[[@CR25]\]. Storjohann \[[@CR25]\] proposed the deterministic algorithm to compute the Frobenius canonical-form efficiently. ### Theorem 2 {#FPar2} \[[@CR25]\] The Frobenius canonical-form of a matrix can be computeddeterministically using$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$fieldoperations. Moreover, there are also probabilistic algorithms that compute this form in expected $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time over small fields \[[@CR10]\]. In this paper, all algorithms are deterministic if we the upper bound on the number of distinct walks is W. Then, due to the time of a single field operation we need additional O(log W) factor in the complexity. However, since we are mainly interested in determining if a cycle/walk of a given length exists in a graph, we can set a sufficiently small field $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {Z}_{p}$\end{document}$ (p has O(log n) bits). This way when algorithm returns nonzero we are sure that there exists some walk. If algorithm returns zero, then with high probability there will be no such walk. Cyclic Subspaces {#Sec8} ---------------- Frobenius decomposition can be used to get the desired power of a matrix (analogously to the diagonal decomposition): $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A^k = (U F U^{-1})^k = U F (U^{-1} U) F {\cdots} F (U^{-1} U) F U^{-1} = U F^k U^{-1}. $$\end{document}$$ Moreover, we will use the property that the power of block diagonal matrix F is block diagonal: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F^k = \left[\begin{array}{lllll} C_1^k & & & & 0 \\ & C_2^k & & \\ & & C_3^k & & \\ & & & {\ddots} & \\ 0 & & & & C_l^k \end{array}\right] .$$\end{document}$$ Now, we need a property of companion matrices that will enable us to power them efficiently (Fig. [1](#Fig1){ref-type="fig"}). Fig. 1Visualisation of the cyclic property (Definition 1) ### Definition 1 (Cyclic Property) {#FPar3} Let v~1~,...,v~n~ be the columns of a matrix $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$C \in \mathbb {F}^{n\times n}$\end{document}$. Let v~n+ 1~,...,v~2n~ be the columns of matrix C^n+ 1^. If, for every 1 ≤ k ≤ n the columns of matrix C^k^ are v~k~,v~k+ 1~,...,v~k\ +\ n~ then the C has a cyclic property. It turns out, that companion matrices have a cyclic property. ### Theorem 3 (Folklore [@CR13], see \[[@CR17]\] for generalization) {#FPar4} Every companion matrix has a cyclic property. We will illustrate this property with an example: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C = \left[\begin{array}{lllll} 0 & 0 & 0 & 0 & 1 \\ 1 & 0 & 0 & 0 & 2 \\ 0 & 1 & 0 & 0 & 3 \\ 0 & 0 & 1 & 0 & 4 \\ 0 & 0 & 0 & 1 & 5 \end{array}\right], C^{2} = \left[\begin{array}{lllll} 0 & 0 & 0 & 1 & 5 \\ 0 & 0 & 0 & 2 & 11 \\ 1 & 0 & 0 & 3 & 17 \\ 0 & 1 & 0 & 4 & 23 \\ 0 & 0 & 1 & 5 & 29 \end{array}\right], C^{3} = \left[\begin{array}{lllll} 0 & 0 & 1 & 5 & 29 \\ 0 & 0 & 2 & 11 & 63 \\ 0 & 0 & 3 & 17 & 98 \\ 1 & 0 & 4 & 23 & 133 \\ 0 & 1 & 5 & 29 & 168 \end{array}\right], $$\end{document}$$$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C^{4} = \left[\begin{array}{lllll} 0 & 1 & 5 & 29 & 168 \\ 0 & 2 & 11& 63& 365 \\ 0 & 3 & 17& 98& 567 \\ 0 & 4 & 23& 13& 770 \\ 1 & 5 & 29& 16& 973 \end{array}\right], C^{5} = \left[\begin{array}{lllll} 1 & 5 & 29 & 168 & 973 \\ 2 & 11 & 63 & 365 & 2114 \\ 3 & 17 & 98 & 567 & 3284 \\ 4 & 23 & 133 & 770 & 4459 \\ 5 & 29 & 168 & 973 & 5635 \end{array}\right] C^{6} = \left[\begin{array}{lllll} 5 & 29 & 168 & 973 & 5636 \\ 11 & 63 & 365 & 2114 & 12243 \\ 17 & 98 & 567 & 3284 & 19019 \\ 23 & 133 & 770 & 4459 & 25824 \\ 29 & 168 & 973 & 5635 & 32634 \end{array}\right] $$\end{document}$$ The matrix C^i^ has 4 columns identical to matrix C^i+ 1^. C has coefficients of order equal to dimension (dimension is 5 and maximum coefficient is 5). After powering to the 5th power, the coefficients can be of order 5^5^. Over a finite field $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {Z}_{p}$\end{document}$, all those coefficients will have O(log p) bits. Matching Distance Queries on Directed Unweighted Graphs {#Sec9} ======================================================= In this section, we will present an algorithm that matches the best known upper bounds of \[[@CR33]\] for distance queries in directed unweighted graphs and uses Frobenius matrices. We take the adjacency matrix A of a graph G (i.e., n × n matrix with a~u,\ v~ = 1 when (u, v) ∈ G and 0 otherwise). The k-th power of the adjacency matrix of the graph G holds the number of walks, i.e., an a~u,\ v~ element of A^k^ is the count of distinct walks fromutovof lengthkin the graph. Observation 1 (Folklore, \[[@CR8]\]) {#FPar5} ---------------------------------------- Let A ∈{0,1}^n×n^ be the adjacency matrix of a directed graph G. The (A^k^)~u,\ v~ is the number of distinct walks from u to v of length exactly k in the graph G. Hence, the shortest path between vertices u, v is the smallest k such that A^k^ has nonzero element a~u,\ v~. This will allow us to forget about graph theory interpretation for a brief moment and focus only on finding such k with algebraic tools. In this section we will proof the following Lemma. Lemma 1 {#FPar6} ----------- Given a matrix$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A \in \mathbb {F}^{n \times n}$\end{document}$.There exists an algorithm that after some preprocessing, can answer queries for any givenpair of indicesi, j ∈{1,...,n} and integerk ∈{1,...,n}, such that:query returns an element (A^k^)~i,\ j~,preprocessing takes$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$fieldoperations and query takesO(n) field operations. The algorithm is deterministic. To proof this Lemma we decompose matrix A into the Frobenius normal form. Storjohann \[[@CR25]\] showed an algorithm that returns U and F deterministically in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ field operations (note that matrix inverse can also be computed in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ field operations). To better explain the idea, in the next section we will consider a simple case when a number of invariant factors of A is exactly 1. Then in Section [4.2](#Sec11){ref-type="sec"} we will show how to generalize it to multiple invariant factors. Single Invariant Factor {#Sec10} ----------------------- In that situation, the matrix F is a companion matrix $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$C \in \mathbb {F}^{n\times n}$\end{document}$. First, we compute the (n + 1)-th power of the companion matrix F^n+ 1^. This can be done by using $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ field operations by repeated squaring (compute F, F^2^,F^4^,...,F^n+ 1^ with O(log n) matrix multiplications). Let v~1~,...,v~n~ be the columns of matrix UF and v~n+ 1~,...,v~2n~ be the columns of matrix UF^n+ 1^. Note, that because matrix F has a cyclic property, the columns v~k~,...,v~k\ +\ n− 1~ construct UF^k^ (see Fig. [2](#Fig2){ref-type="fig"}). Fig. 2Construction of UF^k^ from matrices UF and UF^n+ 1^ This step took just two matrix multiplications, because we need to multiply U times F and F^n+ 1^. The preprocessing phase takes only $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ field operations. Now, if a query asks for the number of distinct walks from vertices u to v of length exactly k we: select u-th row of matrix UF^k^ (n numbers),select v-th column of matrix U^− 1^,multiply them in by using O(n) multiplications (dot product of two n-dimensional vectors). This will give us the u, v element of matrix UF^k^U^− 1^ = A^k^. To get the length of the shortest path (i.e., the minimal k such that (A^k^)~u,\ v~ \> 0), we will modify our matrices slightly to get the number of walks of length ≤ k. At the end, we will use in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n)$\end{document}$ query tim (by using binary search) and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ preprocessing time. Basically, for a given k we need to get the u, v element of matrix A + A^2^ + ⋯ + A^k^. It suffices to add consecutive columns of matrix UF ⊕ UF^n+ 1^ = v~1~ ⊕ v~2~ ⊕... ⊕ v~2n~ in the following manner:[3](#Fn3){ref-type="fn"}$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M^{\prime} = \left[\begin{array}{ccccccc} v_1 & v_1 + v_2 & v_1 + v_2 + v_3 & {\ldots} & {\sum}^k_{i = 1} v_i & {\ldots} & {\sum}^{2n}_{i = 1} v_i \end{array}\right] \in \mathbb{F}^{n \times 2n} .$$\end{document}$$ Now, to get A + A^2^ + ⋯ + A^k^ one would need to multiply $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$M^{\prime }_{k,k+n-1} U^{-1}$\end{document}$ and subtract $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$M^{\prime }_{1,n} U^{-1}$\end{document}$ for a balance.[4](#Fn4){ref-type="fn"} The naive algorithm can transform matrices U and F to matrix M^′^ in O(n^2^) field operations during preprocessing. During query, we will need to compute two dot products (u-th row of $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$M^{\prime }_{k,k+n-1}$\end{document}$ times v-th column of U^− 1^ and u-th row of $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$M^{\prime }_{1,n}$\end{document}$ times v-th column of U^− 1^) and subtract them. We have an algorithm that for a given vertices u, v ∈ G and integer k ∈{1,...,n} can answer: how many walks fromutovof length less or equalkare in the graphG in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n)$\end{document}$ query time and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ preprocessing time. Because the result of the query is increasing in k we can use binary search. We can determine the first k for which the query will answer nonzero value in O(log n) tries. Hence, in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n)$\end{document}$ we can find the length of the shortest path. This generalized query can also return the number of walks of length exactly k, i.e., q(u, v, k) − q(u, v, k − 1). We matched the result of \[[@CR33]\] for unweighted graphs with a single invariant factor. In the next section, we will show how to generalize this technique for graphs with any number of invariant factors. Multiple Invariant Factors {#Sec11} -------------------------- Now, we will consider a case when k ≥ 1, i.e., matrix F has multiple invariant factors. First of all, we need to note that this generalization is not perfect and will allow to compute the number of walks of length up to D (the longest distance in a graph, i.e., diameter). In a real world applications of our framework (detecting cycles, determining distance between nodes, etc.) one does not need to consider walks longer than the longest possible distance in a graph. It is natural that the diameter is considered to be a bound of an output in graph problems \[[@CR1], [@CR2], [@CR7], [@CR9]\]. ### Relation of the Graph Diameter and Frobenius Normal Form {#Sec12} We begin with relating the graph diameter to the Frobenius normal form. It turns out that the graph diameter is bounded by the degree of a smallest invariant factor. #### Lemma 2 {#FPar7} \[[@CR8]\] Given a directed, unweighted graph G with a diameter D. Letμdenote the degree of the smallest invariant factor (i.e.,the dimension of the smallest block in the Frobeniusmatrix F) of an adjacency matrix of the graph G. ThenD ≤ μ. This theorem is well known in literature \[[@CR8]\]. We include the proof of this theorem for completeness. #### Proof {#FPar8} For a contradiction assume that D \> μ and let u, v ∈ G be vertices such that δ(u, v) = D. Let A be the adjacency matrix of G. We know, that there is the minimal polynomial of degree μ: Term $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$a^{k}_{i,j}$\end{document}$ denotes the i, j element of the matrix A^k^. Now, consider the elements u, v of each matrix. The diameter D \> μ and δ(u, v) = D, so for every k ≤ μ the elements $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$a^{k}_{u,v} = 0$\end{document}$ (because there is no walk of length less than D from u to v). Now, if we multiply the minimal polynomial by the matrix A we get: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A^{\mu+ 1} = a_0 A + a_1 A^2 + a_2 A^3 + {\ldots} + a_{\mu-1} A^{\mu} . $$\end{document}$$ Hence $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$a^{\mu + 1}_{u,v} = 0$\end{document}$, because every element in the sequence $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$a^{k}_{u,v} = 0$\end{document}$ for k ≤ μ. By repeating this reasoning, we get that for every k \> 0 the element $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$a^{k}_{u,v} = 0$\end{document}$. So, for every achievable pair of vertices, there must be some k ≤ μ, such that $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$a^{k}_{u,v} \ne 0$\end{document}$ and diameter is bounded by μ. □ The bounds of this inequality are tight. There are graphs with diameter D = μ and graphs with μ = n and arbitrary small diameter \[[@CR8]\]. Our algorithms are able to return walks up to the length μ. We use the bound on D solely because it is easier to interpret diameter than the smallest degree of the invariant factor. ### Generalization to Multiple Invariant Factors {#Sec13} Let k denote the number of blocks in the Frobenius matrix F and μ be the number of columns of the smallest block. To multiply the matrix U by F we can start by multiplying strips of matrix U by appropriate blocks of F and concatenate them later (see Fig. [3](#Fig3){ref-type="fig"}). Fig. 3Multiplication of the UFU^− 1^. Example for 3 invariant factors. We know that matrix F is block and consists of F~1~,F~2~,F~3~. We divide matrix U into strips U~1~,U~2~,U~3~ that corespond to blocks of F. The observation is that we can compute U~1~ ⋅ F~1~, U~2~ ⋅ F~2~ and U~3~ ⋅ F~3~ independently and concatenate them into matrix UF We start by splitting the matrix U into k strips with rows corresponding to the appropriate blocks of F (strip U~i~ has as many columns as block F~i~). Then we multiply UF and have k strips: U~1~F~1~,U~2~F~2~,...U~k~F~k~ (each with at least μ columns). Next, we multiply UF^μ^ and also keep k strips: $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$U_{1} F_{1}^{\mu }, U_{2} F_{2}^{\mu },\ldots ,U_{k} F_{k}^{\mu }$\end{document}$. Our goal is to get a data structure such that if we need UF^k^, we can quickly choose appropriate columns and append them. The matrix U~i~F~i~ has l~i~ columns: $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$v_{1},\ldots , v_{l_{i}}$\end{document}$. Because F~i~ is a companion matrix, the $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$U_{i} F_{i}^{\mu }$\end{document}$ has the cyclic property (Definition 1). And the matrix $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$U_{i} F_{i}^{\mu }$\end{document}$ has columns: $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$v_{\mu },\ldots , v_{\mu +l_{i}}$\end{document}$. Note, that there are some duplicate columns in U~i~F~i~ and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$U_{i} F_{i}^{\mu }$\end{document}$, when μ \< l~i~. Hence, we only need to keep columns $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$v_{1},\ldots ,v_{\mu +l_{i}}$\end{document}$ for this strip. We do this for all strips U~1~F~1~,...U~k~F~k~ (see Fig. [4](#Fig4){ref-type="fig"}). Fig. 4Combining strips into a single matrix. The height of the matrix in the schema is scalled down. We computed U~1~F~1~ and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$U_{1} F_{1}^{\mu }$\end{document}$. Now we noted that companion matrices have a cyclic property so some of the rows in the strips are repeated. So in the single strip we can store only subsequent columns We are left with a matrix that has at most 2n columns (because $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$l_{1} + \mu + l_{2} + \mu + {\ldots } l_{k} + \mu = k \mu + {\sum }^{n}_{i = 1} l_{i} = n + k\mu \le 2n$\end{document}$). To generate it we need to power F to μ and do multiplications U ⋅ F and U ⋅ F^μ^. This can be computed in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ field operations via fast matrix multiplication and repeated squaring. ### Queries with Multiple Invariant Factors {#Sec14} When a query for the number of walks of length k from node u to v comes, we do: For each strip i take the u-th row of $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$U_{i} F_{i} \oplus U_{i} F_{i}^{\mu }$\end{document}$ and concatenate them (see Fig. [5](#Fig5){ref-type="fig"}) into vector $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\bar {u}$\end{document}$,Take v-th column of U^− 1^ matrix and denote it $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\bar {v}$\end{document}$,Return the dot product $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\bar {u} \cdot \bar {v}$\end{document}$.Fig. 5Schema of obtaining vector $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\bar {u}$\end{document}$ (marked red) from 3 strips. We are given the row number and the power k and the lenghts l~i~ of each strip. At the end we concatenate them Because l~1~ + l~2~ + ... + l~k~ = n the vector $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\bar {u} \in \mathbb {F}^{n}$\end{document}$. Query needs O(n) field operations. Finally, this dot product is a~u,\ v~ element of the matrix UF^k^U^− 1^, for a fixed k ≤ μ because $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\bar {u}$\end{document}$ is the concatenation of original vector u. Analogously to Section [4.1](#Sec10){ref-type="sec"} one can extend this result to return the number of walks of length less or equal k. This matches (up to the polylogarithmic factor) the result of \[[@CR33]\]. We will omit the details of this observation because in the next section we will extend this framework even further. Almost Optimal Query {#Sec15} ==================== In the previous section, we showed how to preprocess a matrix A with $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ field operations in such a way that in query that uses O(n) field operations we can return a number (A^k^)~i,\ j~. However, in linear time O(n) we return only a single number. The goal of this section is to get far richer information in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n)$\end{document}$ query time (with extra factors from field operations). Theorem 4 {#FPar9} ------------- LetA ∈{0,1}^n×n^be a matrix such that the degree of smallest invariant factor isμ. Thereexists a deterministic algorithm that after some preprocessing can answer queries for anygiveni, j ∈{1,...,n}:Query returns {a~k~\\|1 ≤ k ≤ μ},wherea~k~ = (A^k^)~i,\ j~ ,Preprocessing takes $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega }\log {W})$\end{document}$ and query takes $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n\log {W})$\end{document}$ time, where W is an upper bound ona~k~for allk ∈{1,...,μ}. Note, that this theorem has some immediate application in graph algorithms (see Section [6](#Sec18){ref-type="sec"}). Hankel Matrix {#Sec16} ------------- Now, we will focus on the proof of Theorem 4. First, we need to introduce the Hankel matrix and its properties. $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$H = \left( \begin{array}{cccccc} c_1 & c_2 & {\ldots} & c_n \\ c_2 & c_3 & {\ldots} & c_{n + 1} \\ {\vdots} & {\vdots} & & {\vdots} \\ c_n & c_{n + 1} & {\ldots} & c_{2n-1} \end{array}\right) $$\end{document}$$ Hankel matrix is defined by its first row and last column (2n − 1 numbers define n × n Hankel matrix). The numbers from the previous row are left-shifted by one and the new number is added at the end. Hankel matrices have some similarities to Topelitz and Circulant matrices. The basic property we need is that the product of Hankel matrix and vector can be computed in O(n log n) time (see \[[@CR15], [@CR26]\]) even though explicitly writing the Hankel matrix as n × n matrix takes O(n^2^) time. The algorithm takes 2n − 1 parameters that define the Hankel matrix and n parameters that define the vector. The technique is based on the Fast Fourier Transformation \[[@CR15], [@CR26]\]. Using Hankel Matrices to Get Richer Query {#Sec17} ----------------------------------------- To proof the Theorem 4 we will modify only the last step in Section [4.2.3](#Sec14){ref-type="sec"}. The algorithm from Section [4.2.3](#Sec14){ref-type="sec"} concatenates the strips U~i~F~i~ and builds a single vector. Subsequently, that vector is multiplied by a column of matrix U^− 1^. But we can also do it in a different order: first we multiply the strip by a section of matrix U^− 1^ and sum the results at the end. Thus, we perform k (number of strips) dot products of smaller vectors (see Fig. [6](#Fig6){ref-type="fig"}). Fig. 6Multiplication of strips by U^− 1^ matrix. As you can see, matrix U^− 1^ can be splited into sections, that multiply only U~i~F~i~ strips Consider the query for a number of walks of length exactly k. The strips in the matrix U^− 1^ do not depend on k (vector (u~0~,...,u~l~)). However, the vector taken from U~i~F~i~ (vectors (x~i~,...,x~i\ +\ l~)) will be left shifted if we want to compute the next one. $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{array}{ccccccccr} (x_0 & x_1 & x_2 & {\ldots} & x_l )\\ (x_1 & x_2 & {\ldots} & x_l & x_{l + 1} )\\ (x_2 & {\ldots} & x_l & x_{l + 1} & x_{l + 2})\\ {\vdots} & & & & {\vdots} \\ (x_{\mu} & &{\ldots} & & x_{\mu+l}) \end{array} \times \left( \begin{array}{cc}{u_0}\\{\vdots}\\{u_l} \end{array}\right) $$\end{document}$$ As you can see, the subsequent rows can be written as the Hankel matrix (we need to add zeros to get a square matrix, but it will not influence asymptotic complexity since there will be at most O(n) of them). By using the fast Hankel matrix-vector multiplication we can compute μ values for every strip i in time O(l~i~ log l~i~) (l~i~ was defined as the length of i-th strip). At the end, we need to sum all results into a single array. Therefore, the number of operations is $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$O\left ({\sum }_{i = 1}^{k} l_{i} \log l_{i} \right )$\end{document}$. Because $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${\sum }^{k}_{i = 1} l_{i} = n$\end{document}$ the algorithm needs O(n log n) field operations. This proves Theorem 4. Here, we silently assumed that the number of walks is bounded by W. Note, that for large W, the algorithm needs to output O(n^2^ log W) bits and the complexity of every arithmetic operation needs to be multiplied by log W. If one is only interested in the deciding if an entry of some power of adjacency matrix is nonzero, we can use a standard randomization technique to eliminate log W factors from the running time. ### Corollary 1 {#FPar10} LetA ∈{0,1}^n×n^be a matrix such that the degree of smallest invariant factor isμ. Thereexists an algorithm that after some preprocessing can answer queries for any giveni, j ∈{1,...,n}:Query returns {a~k~\\|1 ≤ k ≤ μ},wherea~k~ = 1 if (A^k^)~i,\ j~is nonzero,Preprocessing takes $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ and query takes $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n)$\end{document}$ time. The algorithm is randomized with one-sided bounded error. ### Proof {#FPar11} At the beginning we will randomly select a prime number with O(log n) bits. We can write the matrix A as a polynomial: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tilde{A}_{i,j} = \left\{\begin{array}{lll} x_{i,j} & \text{if} A_{i,j} = 1 \\ 0 & \text{otherwise}, \end{array}\right. $$\end{document}$$where x~i,\ j~ are unique variables. Now we can apply Schwartz-Zippel Lemma \[[@CR23], [@CR34]\]. The $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\tilde {A}^{k}_{i,j}$\end{document}$ is the polynomial of degree at most n (because k ≤ μ ≤ n) and if we compute it over $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {Z}_{p}$\end{document}$ for p = O(n^2^) we can correctly determine if $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\tilde {A}^{k}_{i,j}$\end{document}$ is a nonzero polynomial with a constant probability. We can repeat the above procedure O(log n) times to get a correct result in all entries with high probability. □ Applications {#Sec18} ============ In this section we will show how to use Theorem 4 to improve known algorithms on graphs. First we will develop a data structure that returns a number of distinct walks efficiently. Lemma 3 {#FPar12} ----------- LetG = (V, E) be a directed, unweighted graph with n vertices and a diameter D. There exists andeterministic algorithm that after some preprocessing can answer queries for any givenu, v ∈ V:Query returns {w~i~\\|1 ≤ i ≤ D},wherew~i~is the number of distinct walks from u to v of length exactly i,Preprocessing takes $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega }\log {W})$\end{document}$ and query takes $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n\log {W})$\end{document}$ field operations.whereWis the upper bound onw~i~for alli ∈{1,...,D}. Proof {#FPar13} ----- We encode the graph G as an adjacency matrix A(G). We use the Theorem 4 to construct the data structure that given a query (u, v) outputs (A^k^)~u,\ v~ for all 1 ≤ k ≤ μ. Finally, we use Observation 1 to note, that (A^k^)~u,\ v~ is equal to the number of distinct walks from u to v of length exactly k. Moreover we use Lemma 2 to bound the number D ≤ μ, so we will always output more numbers (but we can truncate them in O(n) time). Finally we note, that the preprocessing and query of Theorem 4 matches the statement and construction of adjacency matrix is O(n^2^). □ This algorithm is a significant improvement over \[[@CR33]\]: One can use Lemma 3 to find the distance between u, v by linearly scanning the array and returning the first k such that w~k~ \> 0,Lemma 3 can count cycles. In contrast the \[[@CR33]\] cannot, because the distance from u to u is always 0 (see Section [6.1](#Sec19){ref-type="sec"}),Lemma 3 is almost optimal, i.e., when D = O(n) then query will need to output O(n log W) bits. From the other hand, Lemma 3 is merely a functional improvement and it does not break the $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ of the Single Source Shortest Paths (SSSP) for dense, directed graphs. Now we will show the application of Lemma 3. We begin with almost optimal algorithm to compute the number of all walks between all pairs of vertices. We are not aware of any other research concerning this problem. Definition 2 (All-Pairs All Walk problem) {#FPar14} --------------------------------------------- Given a directed, unweighted graph G with a diameter D. The task is to return an array A, such that for every pair of vertices u, v ∈ G and every k ∈{1,...,D} an element A\[u, v, k\] is the number of distinct walks from u to v of length k. The folklore solution to this needs O(Dn^ω^ log W) time (where W is an upper bound on number of walks) and works as follows: take the adjacency matrixAof graphGand save it inA\[u, v, 1\]. Then, square it to getA^2^and save it inA\[u, v, 2\]. Continue until you fill out complete table. In the worst case this algorithm needs D = O(n) matrix multiplications, thus it needs O(Dn^ω^) field operations. At the first glance it is surprising that we can improve it to $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{3})$\end{document}$ field operations. Theorem 5 {#FPar15} ------------- All-Pairs All Walk problem admits an$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{3} \log {W})$\end{document}$algorithm(where W is upper bound on number of walks between every pair of vertices). Proof {#FPar16} ----- We will apply the Lemma 3 algorithm. The preprocessing takes $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time. Then, for every pair of vertices u, v ask a query. A single query takes $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n \log {W})$\end{document}$ time. Next we will save it in the table A\[u, v\] (query gives D numbers w~1~,...,w~D~, such that w~i~ is the number of walks of length i and save it A\[u, v, i\] := w~i~). Because there are O(n^2^) pairs and for each pair we need $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n\log {W})$\end{document}$ time, the complexity of our solution is $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{3}\log {W})$\end{document}$. The algorithm is almost optimal because the output in the worst case may be O(n^3^ log W) (we may need O(log W) bits to encode a single entry in the table). □ Counting and Determining the Lengths of Cycles {#Sec19} ---------------------------------------------- We will use Theorem 4 to solve All-Nodes Shortest Cycle (ANSC) problem efficiently. ### Definition 3 (All-Nodes Shortest Cycles [@CR31]) {#FPar17} Given a directed, unweighted graph G. The problem All-Nodes ShortestCycle asks to output for every vertex v the length of the shortest cycle that contains v. ### Lemma 4 {#FPar18} There exists a deterministic algorithm that for a given unweighted, directed G with a diameter D: For every vertexureturnsDnumbers:$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${c_{u}^{1}}, {c_{u}^{2}}, \ldots {c_{u}^{D}}$\end{document}$,where$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${c_{u}^{k}}$\end{document}$isthe number of non-simple cycles of length exactlyk, that contain vertex u,Algorithm works in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega } \log {W})$\end{document}$ time (where W is an upper bound on $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${c^{k}_{u}}$\end{document}$ ). ### Proof {#FPar19} We will use Theorem 4. We start by preprocessing the graph G in time $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega }\log {W})$\end{document}$. Theorem 4 allows us to ask for a number of walks from u to v and receive D numbers: $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$w_{u,v}^{k}$\end{document}$. So, we ask for the number of walks from vertex u to the same vertex u. This is exactly the number of non-simple cycles of given length that contain vertex u. Because we need to ask only n queries (it is the number of vertices in a graph) and each query takes $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n\log {W})$\end{document}$ time we have $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega }\log {W} + n^{2}\log {W})$\end{document}$ = $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega }\log {W})$\end{document}$ algorithm. □ If we are only interested in deciding if the numbers $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${c_{u}^{i}}$\end{document}$ are nonzero, instead of Theorem 4 we can use Corollary 1. It introduces the one-sided randomization but allows us to shave log W factors in the running time. ### Corollary 2 {#FPar20} There exists a randomized algorithm that for a given unweighted, directed G with a diameter D : For every vertexureturnsDnumbers:$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${c_{u}^{1}}, {c_{u}^{2}}, \ldots {c_{u}^{D}}$\end{document}$, where$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${c_{u}^{k}}$\end{document}$is1 if there exists a non-simple cycle of length exactlyk, that contain vertex uor 0 otherwise,Algorithm works in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time with one sided bounded error. Now we will show how to improve upon \[[@CR31]\] $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{(\omega + 3)/2})$\end{document}$ algorithm with Corollary 2. ### Theorem 6 {#FPar21} All-Nodes Shortest Cycles admits an $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ randomized time algorithm. ### Proof {#FPar22} We use Lemma 2 to compute the table S\[v\]. For every vertex we search for the first nonzero element linearly. This with high probability is the length of the shortest cycle that contains it. Because the output contains O(n^2^) numbers the complexity is equal to the preprocessing time $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$. □ Also the Corollary 2 improves upon \[[@CR9], Theorem 45\] for unweighted graphs. ### Corollary 3 {#FPar23} Given a directed, unweighted graph G with a diameter D. LetS(c) denote the set of vertices that lie in the cycle of length exactly c.In$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$timewe can return the setsS(1),...,S(D) with constant probability of success. ### Proof {#FPar24} Similarly to the proof of Theorem 6, we can scan the output to compute the set S(c) that contains all vertices that lie on some cycle of length ≤ c. Then, by linear scan we can return the sets S(1),...,S(D). □ Conclusion and Open Problems {#Sec20} ============================ We introduced the framework based on Frobenius normal form and used it to solve some problems on directed, unweighted graphs in matrix multiplication time. The main open question is to use this framework to prove that APSP on such graphs can be solved in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ or at least O(n^2.5^). The promising way is to use the algorithms that determine operators of matrices of polynomials (e.g., determinant, solving linear system \[[@CR14], [@CR18]\]). Additionally, algorithms for a black-box polynomial degree determination seem to be a promising path. Another interesting problem is to use this framework to obtain additive approximation for APSP. Currently, the best additive approximation of APSP is due to \[[@CR20]\]. However, no additive approximation of APSP is known that would work in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time. Application in dynamic algorithm also seems to be a promising approach. Frandsen and Sankowski \[[@CR12]\] showed an algorithm, that dynamically preserves Frobenius normal form in O(kn^2^) time. Our algorithms use fast Hankel matrix-vector multiplication that is based on FFT. Reif and Tate \[[@CR19]\] presented an $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$O(\sqrt {n})$\end{document}$ time per request algorithm for FFT. Can we use these approaches to obtain a faster dynamic algorithm? Finally, it remains open how to apply the Frobenius normal form in the weighted directed graphs with small, integer weights {−M,...,M}. Cygan et al. \[[@CR9]\] took degree M polynomials and used \[[@CR18]\] algorithms to get $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(M n^{\omega })$\end{document}$ time radius and diameter detection. We suspect that similar technique can be applied to Frobenius normal form framework. Tr \[A\] denotes the trace of a matrix A, i.e., the sum of elements on the main diagonal. Sometimes this form is called rational-canonical form. Operation ⊕ denotes concatenation. X~a,\ b~ denotes a matrix constructed by concatenating columns x~a~,x~a+ 1~,...,x~b~ of a matrix X. Extended Abstract of this paper was presented at STACS 2017 \[[@CR22]\] This article is part of the Topical Collection on Special Issue on Theoretical Aspects of Computer Science (STACS 2017) Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This work is part of a project TOTAL that has received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (grant agreement No 677651). Piotr Sankowski is supported by the Polish National Science Center, grant no 2014/13/B/ST6/00770. Karol Węgrzycki is supported by the grants 2016/21/N/ST6/01468 and 2018/28/T/ST6/00084 of the Polish National Science Center. Part of this work was done while authors were visiting the Simons Institute for the Theory of Computing.	{ "pile_set_name": "PubMed Central" }
![](hosplond72205-0024){#sp1 .52} ![](hosplond72205-0025){#sp2 .53} [^1]: The two previous articles appeared on March 8 and 29.	{ "pile_set_name": "PubMed Central" }
Cancer of the thyroid is the most common endocrine malignancy. [@b1-asm-5-428] Most patients with well-differentiated thyroid carcinoma present with an asymptomatic thyroid nodule with or without regional lymphadenopathy or rarely with distant metastasis at diagnosis. Western series report a prevalence rate of 4% for metastasis outside the neck,[@b1-asm-5-428] but in an endemic iodine deficiency area like ours the overall prevalence of distant metastasis was 20%.[@b2-asm-5-428] In 17% of cases, metastases were detected synchronously and at multiple sites. Metastatic disease as the presenting manifestation is exceedingly rare.[@b3-asm-5-428] Patients with metastatic disease are usually asymptomatic and rarely present with organ-specific symptoms. We describe 4 patients with peculiar presentations. In 3 patients the metastatic disease presented with unusual manifestations for the first time and one had a follicular thyroid carcinoma in a struma ovarii after total thyroidectomy. Patients and Methods ==================== Fifty-five consecutive patients, 15 male (27%) and 40 female (73%) with differentiated thyroid carcinoma, who had more than two follow-up visits in the Thyroid Clinic at Nehru Hospital, Postgraduate Institute of Medical Education and Research, Chandigarh, over the last 10 years were included in this study. Their mean (±SD) duration of follow-up was 4.0±3.6 years ranging from 0.25 to 10 years. Their age ranged between 13--70 years with mean (±D) age of 39.5±13.3 years. Twenty-nine (53%) patients had multinodular goitre, 24 (44%) had a solitary thyroid nodule and one had a diffuse goitre; another had no obvious thyroid swelling clinically. All were euthyroid at diagnosis except for 5 patients (9%), 4 of whom were thyrotoxic and one hypothyroid. Total thyroidectomy was done in 46 patients (84%), hemithyroidectomy in 8 (15%) and in 1 (2%) surgery was not done due to refusal. Analysis of histopathological data revealed that 35 (64%) patients had papillary, 5 (9%) had follicular variant of papillary thyroid carcinoma and 15 (27%) had follicular carcinoma. All patients except one, after total thyroidectomy, received an ablative dose of ^131^I and were on suppressive doses of L-thyroxine. Four patients (7%) had post-operative permanent hypoparathyroidism. We describe 4 peculiar cases out of 55 who had unusual presenting manifestations of the thyroid cancer ([Table 1](#t1-asm-5-428){ref-type="table"}). Case 1 ====== A 30-year-old-man from an iodine sufficient area was admitted with sudden onset of flaccid quadriparesis. He had no previous history or family history of similar episodes. However, he had a history of loss of 10 kg weight in 4 months with increased appetite, hyperdefecation, palpitation and tremulousness. On examination, his body mass index (BMI) was 20.3 kg/m^2^ and had pulse rate of 112/min, and BP of 110/60 mm Hg. He had fine tremors, Plummer's nail and a soft, diffuse grade II goiterwith bruit and generalized, firm lymphadenopathy. His muscle bulk was decreased with a power of ⅖ in the lower limbs and ⅗ in the upper limbs, and deep tendon reflexes were diminished. Sensory examination was normal and there was no cranial nerve deficit. He had no thyroid associated ophthalmopathy or dermopathy. Other systemic examinations were normal. Laboratory studies at admission showed serum Na^+^ 140 mEq/L, K+ 2.4 mEq/L, Cl- 109 mEq/L, HCO3- 26 mEq/L, Mg^++^ 2.5 mEq/L, phosphate 3.8 mg/dL, T3 3.19 ng/mL (normal, 0.8--2.1), T4 253 ng/mL (normal, 40--120) and TSH was 0.05 μU/mL (normal, 0.3--6.0). Single random urine electrolytes were Na^+^ 78 mEq/L, K^+^ 28 mEq/L, Cl 95 mEq/L. Radioactive iodine uptake at 24 hours was 35% (normal, 15--35%) and whole body ^131^I scan revealed increased tracer uptake in the thyroid, lymph nodes and L5 vertebrae. Thyroid microsomal antibody (TMA) titres were insignificant. He was managed with parenteral potassium supplementation following which the patient's muscle power improved to normal within 8 hours. He was put on neomercazole and propranolol and after achieving euthyroidism. Total thyroidectomy with lymph node resection was performed. Histopathology confirmed follicular thyroid carcinoma with lymph node metastasis. He received an ablative dose of ^131^I and is on L-thyroxine suppressive therapy. There was no recurrence of such episode on follow-up in the last 2 years. Case 2 ====== A 55-year-old woman, resident of an iodine sufficient area, had diffuse goiter for the past 25 years. She presented with rapidly increasing swelling in the right frontal region along with an increase in size of the goiter for the last 4 years. On examination, her BMI was 20 kg/m^2^, she had a pulse rate of 80/min, and no tremors. She had diffuse, firm and globular goiter (grade IV) with multiple tattoo marks and did not have cervical lymphadenopathy. The swelling in the right frontal region (6×8×6 cm) was globular, firm and pulsatile ([Figure 1a](#f1a-asm-5-428){ref-type="fig"}) and she had another firm, pulsatile swelling in the left shoulder region. On investigation, thyroid function tests and serum biochemistry were unremarkable. An x-ray skull lateral view showed destruction of the frontal bone with adjacent soft tissue mass. CT scan head showed a solitary calvarial metastasis destroying the frontal bone with intracranial as well as extracranial extension into the adjacent soft tissue. ^131^I whole body scan showed significant increased tracer uptake in the thyroid region and skull ([Figure 1b](#f1b-asm-5-428){ref-type="fig"}). 99mTcMDP bone scan showed photopenic areas in the skull, left shoulder, multiple vertebrae and the seventh rib. She underwent total thyroidectomy uneventfully and histopathology confirmed follicular carcinoma and FNAC from osseous lesions showed metastases. She was subjected to ^131^I ablation for skeletal metastases followed by L-thyroxine suppressive therapy. Case 3 ====== A 70-year-old man from an iodine sufficient area presented with progressive loss of vision in the left eye of one-year duration and pain on movement of the globe. He was not a known diabetic, nor had any history of thyroid disease or ocular trauma. On examination, he was sick looking with a BMI of 27 kg/m^2^. His vitals were stable and he was clinically euthyroid. Ocular examination revealed loss of light reflex, dilated pupil and choroidal mass in the left eye and a normal right eye. There was no visible or palpable neck swelling or lymph nodes anywhere else. For the painful blind eye due to intractable glaucoma he underwent enucleation. The histopathology of the excised tissue revealed metastatic papillary thyroid carcinoma involving the posterior choroid and superficial retina ([Figure 2a](#f2a-asm-5-428){ref-type="fig"}, [2b](#f2b-asm-5-428){ref-type="fig"}). Anterior segment and optic nerves were uninvolved, while the retina was detached posteriorly. Immunostaining for thyroglobulin was positive in the tumor tissue. Color Doppler ultrasound of the neck done subsequently revealed neovascularization and 15×15 mm nodule in the right lower pole of the thyroid and a 2×2 cm lymph node near the common carotid artery. Fine needle aspiration cytology from the thyroid swelling and lymph node revealed papillary thyroid carcinoma ([Figure 2c](#f2c-asm-5-428){ref-type="fig"}). He underwent total thyroidectomy and radical node resection followed by an ablative dose of ^131^I. Whole body ^131^I scan done subsequently found no residual disease. He was on a suppressive dose of L-thyroxine and doing well for the last 3 years. Case 4 ====== A 47-year-old woman from an iodine sufficient area presented with an euthyroid solitary nodule. It had poor ^131^I-uptake on scan and had features of follicular neoplasm on FNAC. Her thyroid profile was normal. She was planned for hemithyroidectomy but underwent total thyroidectomy for hemorrhage into the nodule on exploration. The detailed histology of the nodule excluded malignancy. She received 150 μg L-thyroxine daily for 3 years. However, even without thyroid hormone replacement she later remained euthyroid and had no palpable goiter and had normal serum T3 and T4 levels. Six years later, she complained of irregular cycles and experienced hot flashes and after screening was put on hormone replacement therapy (HRT) with conjugated equine estrogen and medroxyprogesterone continuously. With this therapy, hot flashes decreased but after 6 months she had inter-menstrual bleeding. A tender right fornix and 5×6 cm firm mobile mass was found on per vaginal examination and it was confirmed by pelvic ultrasound. Bilateral salpingo-oophorectomy and hysterectomy were performed. Histopathology revealed follicular carcinoma in a struma ovarii ([Figure 3](#f3-asm-5-428){ref-type="fig"}). No other teratomatous elements were seen. The left ovary was normal. Estrogen and progesterone receptor status by peroxidase anti-peroxidase immunochemistry on ovarian tissue after microwave retrieval were found to be negative. An ^131^I whole body scan after 6 weeks of laparotomy revealed uniform low uptake in residual thyroid tissue (2.4% at 24 h) and no tracer uptake elsewhere. Her unsuppressed serum thyroglobulin level was 4.5 ng/mL (normal \<10 ng/mL) and she had low serum T3 and T4 (0.55 and 40 ng/mL, respectively) and an elevated thyroid stimulating hormone (TSH) of 40.8 μU/mL, indicating hypothyroidism. She received 4500 rads external beam radiotherapy to the pelvic region in view of the capsular invasion and ascites, together with L-thyroxine therapy. She had no evidence of metastasis during her 10-year follow-up. Discussion ========== Most patients with thyroid cancer are euthyroid and thyroid cancer manifesting with thyrotoxicosis is rare.[@b4-asm-5-428],[@b5-asm-5-428] Leiter et al described the first patient with adenocarcinoma of the thyroid with functioning metastasis and postoperative thyrotoxicosis in 1946.[@b6-asm-5-428] Since then a handful of cases (less than 50) are described. Four out of 55 patients in our study were toxic at presentation, three had pre-existing MNG and one had a functioning metastasis and presented with periodic paralysis. The different mechanisms by which thyroid malignancy causes thyrotoxicosis include excessive production of thyroid hormones by functioning thyroid and/or metastatic tissue, excessive release of thyroid hormones by infiltration of the thyroid gland by the tumor tissue, underlying Graves' disease, and iodine-induced toxicosis as a part of diagnostic or therapeutic interventions.[@b6-asm-5-428]--[@b9-asm-5-428] The association between thyrotoxicosis and periodic paralysis was initially described by Rosenfeld in 1902.[@b10-asm-5-428] The prevalence being 8% to 34% in oriental males and 0.2% in females, while in non-Asian population the reported prevalence is 0.1% to 0.2%.[@b11-asm-5-428]--[@b13-asm-5-428] Thyrotoxicosis usually precedes or occurs concurrently with periodic paralysis in 80% of cases, like ours.[@b14-asm-5-428] The disorder is clinically indistinguishable from familial hypokalemic periodic paralysis. The pathogenesis of periodic paralysis in thyrotoxicosis includes an alteration in either Na+/K+ ATP-ase pump activity due to thyrotoxicosis or increased (2 adrenergic receptor sensitivity. Increased insulin sensitivity and altered voltage gated calcium channels are also postulated.[@b15-asm-5-428] Follicular thyroid carcinoma with thyrotoxic periodic paralysis as a presenting manifestation has not been reported previously. Differentiated thyroid cancer commonly metastasizes to lymph nodes, bones, lung, brain and other organs. Osseous metastasis occurs in 3% to 12% of patients with differentiated follicular and papillary thyroid cancer.[@b14-asm-5-428] However, ^131^I scan or 99mTcMDP scan can reveal osseous metastasis in the majority (74%) of these patients.[@b14-asm-5-428] Of the two, a relatively higher prevalence of bone metastasis occurs in follicular (15.2%) than in papillary carcinoma (0.6%).[@b16-asm-5-428] Frequent sites of osseous metastases are sternum, vertebrae, pelvis and ribs. Case 2 presented with a pulsating metastatic deposit in the frontal bone, which though classical, is very rare as the presenting manifestation. The overlying skin was also involved as follicular carcinoma has a greater preponderance than papillary carcinoma for cutaneous metastases and the majority of skin metastases from thyroid cancer are localized to the head and neck area with underlying bone involvement.[@b17-asm-5-428] However, one such case has been described in which the sacrum was the site of pulsating metastases.[@b18-asm-5-428],[@b19-asm-5-428] Although secondaries are the commonest malignancy of the choroid, their occurrence is quite rare. In a study of 8712 patients with carcinoma, only 6 had metastasis to the choroid.[@b20-asm-5-428] The reasons for rarity of choroid metastasis are early diagnosis of cancer and right angle branching of the ophthalmic artery, and therefore malignant emboli lodge into the brain and meninges by means of a straight route.[@b21-asm-5-428] The other explanation in patients with disseminated malignancy could be the clinical picture dominated by a poor general condition, slow growth of choroid metastasis and death of the patient before attention is attracted to the ocular condition. Routine ophthalmic or pathological examinations of the eye are usually omitted in patients with terminal carcinomatosis.[@b20-asm-5-428] The majority of emboli travel to the 20-odd short posterior ciliary arteries rather than the long anterior branches. Therefore the posterior region of the choroid near the macula is the site of predilection. The commonest malignancies metastasizing to the choroid are malignancy of the breast, lungs, colorectal region and urinary bladder.[@b20-asm-5-428] Choroid metastasis of thyroid carcinoma is rare. Of 227 cases of choroid metastasis only one was from thyroid.[@b22-asm-5-428] A review of the literature shows 14 reported thyroid cancers metastasizing to the eye; papillary thyroid carcinoma has been reported thrice before.[@b22-asm-5-428]--[@b24-asm-5-428] Patients with choroid metastasis may present with progressive loss of vision, limitation of ocular movement and retro-orbital/orbital pain, though the majority are asymptomatic. Ocular examination usually reveals neovascularization of the iris, raised intraocular pressure, and exudative retinal detachment with an orange-colored nodule. B-mode ultrasound, MRI and biopsy in a appropriate setting clinch the diagnosis. The treatment modalities are ^131^I ablation, photocoagulation, cryopexy and rarely enucleation.[@b24-asm-5-428] Our case is unique because the choroid metastasis antedated the diagnosis of thyroid carcinoma and he had no clinically obvious thyroid swelling or regional lymphadenopathy. Struma ovarii is a slow growing ovarian neoplasm within thyroid tissue because it is the only or the predominant (\>50%) constituent. The ovarian thyroid is histologically and functionally identical to the cervical thyroid. The tumour is a highly specialized subclass of benign cystic teratoma; 95% remain benign while the remainder undergo malignant transformation, with peak frequency during the fifth decade of life.[@b25-asm-5-428],[@b26-asm-5-428] The left ovary is more frequently involved than the right and in 6% of instances struma ovarii are bilateral. Struma ovarii often present as an abdominal mass, with lower abdominal pain, ascites and, uncommonly, hyperthyroidism (5%). Past or concomitant thyroid enlargement (18%) has been described with struma ovarii, which may cause difficulty in its diagnosis. Retrospectively, the fact that this patient remained euthyroid without L-thyroxine replacement for 4 years, could have given a clue to the presence of a functioning struma ovarii. This was further substantiated when the patient became hypothyroid (TSH 40.8 μU/mL) promptly after the removal of the ovarian mass. The majority of malignant struma present as follicular carcinoma, while papillary, anasplastic, and Hurthle cell carcinoma has also been described.[@b27-asm-5-428] It is noteworthy that the lack of demonstration of estrogen and progesterone receptors in struma in our case can be explained by the complete transformation of ovarian tissue into struma, as normal thyroid follicular cells do not express estrogen receptors.[@b27-asm-5-428] In conclusion, differentiated thyroid cancer can present with atypical manifestations. A routine preoperative physical examination of the thyroid gland should be done in all cases of metastatic lesions of unknown origin. Clinical suspicion and aggressive work-up and treatment are usually rewarding. The high incidence of metastatic disease in an iodine deficiency area like ours is postulated due to the advanced stage at presentation or inherent aggressive biological behavior. ![Pulsating metastasis on the right side of the frontal area with prominent blood vessels.](asm-5-428f1a){#f1a-asm-5-428} ![^131^I whole body scan showing significantly increased tracer uptake in the thyroid region and the skull.](asm-5-428f1b){#f1b-asm-5-428} ![Low-power photograph showing papillary configuration of the tumor in relation to the choroid (H&E×160).](asm-5-428f2a){#f2a-asm-5-428} ![Photomicrograph showing retinal detachment (H&E×220)](asm-5-428f2b){#f2b-asm-5-428} ![Optically clear nuclei of tumor cells (H&E×440).](asm-5-428f2c){#f2c-asm-5-428} ![High power phtotomicrograph showing follicular thyroid carcinoma in struma ovarii and infiltration by malignant cells in capsular lymphatics (H&E×440).](asm-5-428f3){#f3-asm-5-428} ###### The clinical profile of 4 patients who had unusual features as a presenting manifestation. Patient Age Sex Cause of referral Functional status Histopathology --------- ----- ----- --------------------------------------- ------------------------------------- ---------------------------------- 1 30 M Hypokalemic paralysis Toxic Follicular 2 55 F Pulsatile scalp swelling Euthyroid Follicular 3 70 M Metastasis to globe Euthyroid Papillary 4 47 F Solitary thyroid nodule, ovarian mass Euthyroid after total thyroidectomy Struma ovarii with follicular Ca	{ "pile_set_name": "PubMed Central" }
INTRODUCTION ============ As a clean and carbon-neutral energy source, solar energy has drawn increasing attention in recent years. To date, solar energy utilization has largely revolved around photovoltaic power generation ([@R1]--[@R7]). Although significant progress has been made in photovoltaics, electricity alone only accounts for less than 20% of the global delivered energy consumptions, and the remaining energy is provided by solid, liquid, or gaseous fuels ([@R8]). Therefore, novel approaches that convert solar energy into fuels offer significant opportunities to reduce fossil fuel consumption in industrial and transportation sectors, thereby providing potentially effective means to mitigate anthropogenic CO~2~ emissions. Solar fuels can be produced from water and CO~2~ by photocatalytic splitting under mild conditions ([@R9]--[@R14]). However, the efficiency and yields of existing photocatalytic processes are far from satisfactory due to the high thermodynamic and activation barriers of these reactions and challenges in band structure optimization for photocatalysts, electron-hole recombination, and photocatalyst stability ([@R13]). In addition, product separation for these processes can be problematic ([@R14]). As an alternative to photocatalytic processes, solar-thermal water/CO~2~ splitting has shown excellent potential ([@R15]--[@R21]). For instance, solar-to-hydrogen efficiency can be as high as 18% for solar-thermal water splitting, whereas state-of-the-art photocatalytic processes have an energy conversion efficiency of less than 0.2% ([@R22]). Typical solar-thermal water/CO~2~-splitting processes convert concentrated solar energy in a two-step, cyclic redox scheme: A redox material (usually a metal oxide) is first decomposed (or reduced) under an inert environment to (partially) release its lattice oxygen (MO~x~ → MO~x−δ~ + δ/2O~2~) at elevated temperatures (\>1200°C), and then the reduced metal oxide is reoxidized (or regenerated) by putting it in contact with water or CO~2~ to replenish its lattice oxygen, producing hydrogen or carbon monoxide (MO~x−δ~ + δH~2~O/CO~2~ → MO~x~ + δH~2~/CO). The resulting hydrogen (H~2~) and/or carbon monoxide (CO) can be used as gaseous fuels or converted into liquid fuels or chemicals ([@R23], [@R24]). Such a two-step redox scheme eliminates the need for gas separation and has demonstrated notably higher CO/H~2~ production rates (1.35 mmol~CO~ kg~oxide~^−1^ s^−1^ and 0.71 mmol~H2~ kg~oxide~^−1^ s^−1^) when compared to state-of-the-art photocatalytic approaches \[\~0.07 mmol~CO~ kg~cat~^−1^ s^−1^ ([@R21], [@R25])\]. The challenge for solar-thermal processes resides in the high reaction temperatures for metal oxide decomposition. Reasonable decomposition rates are only achieved at temperatures above 1500°C for nonvolatile monometallic metal oxides ([@R15], [@R21], [@R26], [@R27]). Although recent studies on mixed metal oxides have led to notable decreases in the decomposition temperatures, most solar-thermal redox processes require an operating temperature above 1200°C ([@R17], [@R26]--[@R30]). In addition to the high reaction temperatures, water or CO~2~ conversions in these schemes are far from satisfactory (typical below 10%), posing significant challenges for process heat integration and downstream product separations ([@R31]). Introducing reducing agents can lower the decomposition (or reduction) temperature ([@R32]). In such an open-loop thermochemical approach, reducing agents such as carbonaceous fuels are used as oxygen "sinks" to facilitate lattice oxygen extraction from redox materials ([@R15]). Although significantly lower reduction temperatures (\<800°C) have been demonstrated, steam or CO~2~ conversion in the subsequent water- or CO~2~-splitting step is still very low ([@R33]--[@R35]). To address this challenge, we proposed a hybrid solar-redox scheme for water splitting, wherein concentrated solar energy is used to drive the reduction reaction in the presence of methane (CH~4~) that undergoes partial oxidation (POx) to produce Fischer-Tropsch (F-T) ready syngas. Using perovskite-supported Fe~3~O~4~ as the redox material, up to 77% steam-to-hydrogen conversion was achieved ([@R16], [@R36]). Although such a methane-assisted solar-thermal water-splitting scheme is applicable for CO~2~ splitting, further increases in water/CO~2~-splitting conversion is desirable from process efficiency and product separation standpoints. Another challenge is that iron oxide--based materials tend to overoxidize syngas products in the methane POx step, limiting the syngas selectivity to less than 70%. Herein, we report high-performance redox materials for CO~2~ splitting using the aforementioned hybrid solar-redox concept. Perovskite nanocomposites (NCs) exhibit exceptional efficacy for both CO~2~ splitting and methane POx: Over 98% CO~2~-to-CO conversion was achieved in CO~2~ splitting, and syngas selectivity reached 96% in methane POx. As a stand-alone process, methane-assisted CO~2~ splitting offers an effective approach for both CO~2~ utilization and F-T/methanol ready syngas production. As an example, the resulted syngas and CO products can be readily used for green acetic acid production ([Fig. 1](#F1){ref-type="fig"} ). Compared to conventional acetic acid synthesis, the proposed scheme has the potential to reduce fossil fuel consumption by 67% and produce 84% less CO~2~ when compared to coal-based process. ![Hybrid solar-redox process and its application for green acetic acid production from methane and CO~2~.\ This process has three sections: hybrid solar-redox for syngas and CO coproduction, methanol synthesis from the syngas, and acetic acid production.](1701184-F1){#F1} RESULTS ======= Rationale for redox material selection -------------------------------------- Performance of the redox material is critical to the hybrid solar-redox scheme. Desirable properties for the redox materials include high syngas yield and productivity, excellent CO~2~ conversion and CO productivity, and good redox stability. Ideal redox materials should also be affordable, environmentally benign, and attrition-resistant. Among the various metal oxides investigated to date, iron oxide--containing materials are the most promising ones because of their low cost, abundance, and low toxicity ([@R16], [@R17], [@R30]--[@R36]). However, pure iron oxides are not suitable for the proposed redox reactions from a thermodynamics viewpoint. A methane-assisted thermochemical CO~2~-splitting scheme can be described by the reactions listed in [Table 1](#T1){ref-type="table"}. From a thermodynamic standpoint, the redox material can be considered as an oxygen "source" in the methane POx step and an oxygen sink in the CO~2~-splitting step. Hence, a redox pair (MeO~x~/MeO~x−1~) with a high equilibrium oxygen pressure (P~O2~) can lead to overoxidation of syngas (Eqs. 3 and 4) and low CO~2~ conversion (Eq. 7). In contrast, redox pairs with a low P~O2~ will result in low methane conversion (Eq. 2). CO~2~ splitting, on the other hand, is monotonically favored at lower P~O2~. The relationship among equilibrium P~O2~, syngas yield, and CO~2~ conversion, derived from minimizing Gibbs free energy of the reaction set (Eqs. 2 to 6), is illustrated in [Fig. 2](#F2){ref-type="fig"}. As can be seen, equilibrium P~O2~ of Fe/FeO redox pair, which is a commonly used redox material, is away from the ideal (high performance) region for the hybrid solar-redox scheme. This is confirmed by a number of studies that reported low syngas selectivity and limited CO~2~ conversions ([@R16], [@R34]--[@R38]). Thermodynamic analysis suggests that no first-row transition metal oxide has redox properties located in this region ([@R39]). Unlike monometallic oxides whose redox properties are fixed, mixed metal oxides offer potentially tunable thermodynamic properties. Using iron oxide (FeO) as an example, it can form mixed oxides with strontium oxide (SrO) at various Sr/Fe atomic ratios ([@R40]). Although SrO is not reducible, formation of Sr-Fe mixed oxides changes the coordination environments and electron density of oxygen anions (O^2−^). Thus, equilibrium P~O2~ of Sr-Fe mixed oxides can be tailored by adjusting Sr/Fe ratios ([Fig. 2](#F2){ref-type="fig"}). For example, spinel-type SrFe~2~O~4~ exhibits lower P~O2~ than FeO. Further increases in Sr/Fe ratios leads to the formation of perovskite oxides such as Brownmillerite-type Sr~2~Fe~2~O~5~ and Ruddlesden-Popper (R-P)--phase Sr~3~Fe~2~O~6~ (Sr~n+1~Fe~n~O~3n+1−δ~ with n = 2), both having notably lower P~O2~ than FeO ([Fig. 2](#F2){ref-type="fig"}). Among the three Sr-Fe mixed oxides investigated, SrFe~2~O~4~, with a marginally improved redox performance over FeO, is not an excellent redox material because of low equilibrium syngas yields (\<90%) and CO~2~ conversions (\<95%) at the temperature range of interest. To compare, both Sr~2~Fe~2~O~5~ and Sr~3~Fe~2~O~6~ can be effective for methane POx and CO~2~ splitting, with Sr~3~Fe~2~O~6~ offering improved redox properties for the hybrid solar-redox scheme. ###### Main reactions that occur in methane POx and CO~2~ splitting in the redox scheme. Methane POx CO~2~ splitting ------------------------------------------ ----------------------------------------- MeO~x~ ⇆ MeO~x−1~ + ^1^/~2~O~2~ (1) CO~2~ ⇆ CO + ^1^/~2~O~2~ (7) CH~4~ + ^1^/~2~O~2~ ⇆ CO + 2H~2~ (2) MeO~x−1~ + ^1^/~2~O~2~ ⇆ MeO~x~ (8) CO + ^1^/~2~O~2~ ⇆ CO~2~ (3) 2H~2~ + O~2~ ⇆ H~2~O (4) 2CO ⇆ C + CO~2~ (5) CO + H~2~O ⇆ H~2~ + CO~2~ (6) ![Thermodynamic analysis of methane POx/CO~2~-splitting reactions and selected redox materials.\ Standard Gibbs free energy change (Δ~r~G°) of the oxidation reactions (①2Fe + O~2~ = 2FeO; ②^2^/~3~SrO + ^4^/~3~Fe + O~2~ = ^2^/~3~SrFe~2~O~4~; ③^4^/~3~SrO + ^4^/~3~Fe + O~2~ = ^2^/~3~Sr~2~Fe~2~O~5~; and ④2SrO + ^4^/~3~Fe + O~2~ = ^2^/~3~Sr~3~Fe~2~O~6~) and corresponding equilibrium oxygen partial pressures (P~O2~), for ≥95% CO~2~ conversions (blue region) and ≥90% methane-to-syngas yield (green region) at 1 atm. The overlapping (cyan) zone is the ideal (high performance) region for both methane POx and CO~2~-splitting reactions.](1701184-F2){#F2} Redox performance of Sr~3~Fe~2~O~7−δ~ ----------------------------------------- As-prepared Sr~3~Fe~2~O~7−δ~ (SF~7~) exhibits excellent redox activity (figs. S1 and S2), but it deactivates continuously (fig. S3) at 900°C under a weight hourly space velocity (WHSV) of 60,000 cm^3^ g~SF7~^−1^ hour^−1^. Specifically, syngas productivity (eq. S1) dropped by 82% in the methane POx step, and CO productivity (eq. S2) decreased by 72% in the CO~2~-splitting step over 15 redox cycles. It is noted that even with significant deactivation, CO productivity in the 16th cycle (2.6 mol kg~SF7~^−1^) is still an order of magnitude higher than that in state-of-the-art solar-thermal CO~2~-splitting schemes \[\<0.3 mol kg~oxide~^−1^ for ceria and perovskites ([@R21], [@R29], [@R41], [@R42])\]. Sr~3~Fe~2~O~7−δ~ is also superior to CeO~2~-modified Fe~2~O~3~ in a H~2~-assisted thermochemical CO~2~-splitting scheme, where CO productivity is about 1.1 mol kg~oxide~^−1^ ([@R35]). To shed light on the cause of deactivation, we acquired energy-dispersive x-ray spectroscopy (EDS) mapping and x-ray diffraction (XRD) patterns of as-prepared and spent samples (fig. S4). XRD measurements and thermogravimetric analysis suggest that the as-prepared sample has an R-P--phase Sr~3~Fe~2~O~6.75~ (tetragonal I4/mmm), whose oxygen stoichiometry decreases to about 6 in an inert atmosphere at 900°C. Metallic iron, iron carbides, Fe~3~O~4~, and SrO were detected after the first reduction. It was found that CO~2~ can fully reoxidize metallic iron back to the R-P phase, but a significant portion of Sr~3~Fe~2~O~6~ becomes unreducible in the 16th reduction (fig. S4). Elemental mappings indicate that significant phase segregation and sintering occurred in redox reactions (fig. S4); this may have caused the redox activity to decrease. To retard sintering and deactivation, the redox active-phase (SF~7~) was dispersed in an inert and earth-abundant calcium manganese oxide (Ca~0.5~Mn~0.5~O) phase. Sr~3~Fe~2~O~7−δ~-Ca~0.5~Mn~0.5~O NCs ------------------------------------ Formation of Sr~3~Fe~2~O~7−δ~-Ca~0.5~Mn~0.5~O NCs (SF~7~-CM NCs) is revealed by electron microscopy (figs. S5 and S6), and reducibility of the sample is characterized by hydrogen temperature-programmed reduction (H~2~-TPR) and in situ XRD (figs. S7 and S8). The redox stability of SF~7~-CM NCs is shown in [Fig. 3](#F3){ref-type="fig"}. Both syngas productivity in the methane POx step and CO productivity in the CO~2~-splitting step decreased slightly in the first 21 cycles and then stabilized in the subsequent 9 cycles at 900°C at a WHSV of 24,000 cm^3^ g~SF7-CM~^−1^ hour^−1^. Specifically, syngas productivity decreased from 4.42 to 4.28 mol kg~SF7-CM~^−1^, whereas CO productivity varied from 1.74 to 1.53 mol kg~SF7-CM~^−1^ over the first 21 cycles. In the CO~2~-splitting step, CO productivity was significantly higher for SF~7~-CM NCs than for CeO~2~-Fe~2~O~3~ \[1.1 mol kg~oxide~^−1^ ([@R35])\]. In the last 10 cycles, syngas selectivity was higher than 95% in the methane POx step. Moreover, the molar ratio of H~2~/CO remained around 1.96, and coke formation was negligible between 20 and 30 cycles. The excellent stability of the NCs is attributed to the confinement effect that involves dispersing an active phase in an inert medium. XRD measurements demonstrated that Ca~0.5~Mn~0.5~O acted as a dispersing medium ([Fig. 3](#F3){ref-type="fig"}), which itself had no activity toward the redox reactions. In addition, elemental mappings demonstrate that all elements are uniformly distributed in as-prepared and spent samples (figs. S9 and S10). ![Redox stability of SF~7~-CM NCs.\ (A) Productivity of H~2~, CO, and CO~2~ in methane POx and CO in CO~2~ splitting. (B) XRD patterns of spent samples. Reaction conditions: m~SF7-CM~ = 0.25 g; T = 900°C, F = 74.4 μmol s^−1^, P = 1 atm, y~CH4~ = 0.1, and y~CO2~ = 0.1. a.u., arbitrary units.](1701184-F3){#F3} [Figure 4A](#F4){ref-type="fig"} shows typical gaseous product elution profiles in the methane POx step of the 30th cycle. After feeding CH~4~ to the NC bed, CO~2~ coeluted with CO and H~2~, but its elution profile was different from those of CO and H~2~. The peak production rates for CO~2~, CO, and H~2~ were 0.51, 4.14, and 8.03 mmol kg~SF7-CM~^−1^ s^−1^ at 16, 250, and 250 s, respectively. Such a pattern is typical for the redox-based methane POx ([@R43]). The gaseous compound elution profiles in the CO~2~-splitting step of the 30th cycle are presented in [Fig. 4B](#F4){ref-type="fig"}. The breakthrough time of CO~2~, which is the time when CO~2~ flow rate at the outlet of a reactor reaches 2% of that at the inlet, was 35 s, and it coincided with the time at which the CO production rate maximized (22 mmol~CO~ kg~SF7-CM~^−1^ s^−1^). The peak CO production rate in this open-loop thermochemical scheme was one order of magnitude higher than that in conventional solar-thermal CO~2~ splitting, which was carried out at temperatures 200°C higher than that in the former ([@R21]). It should be noted that the breakthrough curve of CO~2~ depends on factors like the flow rate of feed, amount of redox materials, and inlet concentration of CO~2~. As shown in [Fig. 4B](#F4){ref-type="fig"}, CO~2~ flow rate leveled off after 180 s. In the first 3 min, the average CO production rate was 8.23 mmol kg~SF7-CM~^−1^ s^−1^, significantly higher than that (\<3 mmol kg~oxide~^−1^ s^−1^) in a recently reported methane-assisted thermochemical scheme ([@R33]). ![Instantaneous production rates of gaseous products from the redox reactions using SF~7~-CM NCs.\ (A) Elution profiles of H~2~, CO, and CO~2~ in methane POx at the 30th cycle. (B) Elution profiles of CO and CO~2~ in CO~2~ splitting at the 30th cycle. Inset in (B) is the cumulative CO molar fraction in the product and moles of CO produced in CO~2~ splitting for a short feeding period (the blue dot denotes the total moles of CO produced for complete regeneration). Reaction conditions: m~SF7-CM~ = 0.25 g; T = 900°C, F = 74.4 μmol s^−1^, P = 1 atm, y~CH4~ = 0.1 (A), and y~CO2~ = 0.1 (main figure) and 0.05 (inset) (B).](1701184-F4){#F4} One key issue with thermochemical CO~2~-splitting studies conducted to date is the incomplete CO~2~ conversion (\<90%). For SF~7~-CM NCs, near-complete CO~2~ conversion was achieved during the early stage of the reaction (fig. S11). However, the average CO~2~ conversion in the first 3 min was only 32.3% in the 30th cycle ([Fig. 4B](#F4){ref-type="fig"}). To improve CO~2~ conversion, we attempted to shorten the duration of the CO~2~-splitting step. As shown in [Fig. 4B](#F4){ref-type="fig"}, CO~2~ conversion was above 98%, and the average CO production rate was 13.8 mmol kg~SF7-CM~^−1^ s^−1^ for a shorter feeding period. Such a short regeneration period was adequate to replenish over 80% of the active lattice oxygen of the reduced NCs. To the best of our knowledge, such a high CO~2~ conversion has yet to be reported for thermochemical CO~2~ splitting. These findings clearly demonstrate that this NC outperforms the other redox materials for thermochemical CO~2~ splitting. SrFeO~3−δ~-CaO NCs ------------------ To further verify the role of the dispersing medium, we prepared another NC with SrFeO~3−δ~ as the active phase and CaO dispersing medium. SrFeO~3~ has the highest oxygen capacity among the R-P--type perovskite family (Sr~n+1~Fe~n~O~3n+1,~ n = ∞) while still having desirable redox properties ([Fig. 2](#F2){ref-type="fig"}). XRD and Rietveld refinement do not indicate significant Ca substitution of Sr in SrFeO~3−δ~ phase in the NCs, and excess Sr atoms were incorporated into the crystal structure of CaO (figs. S12 and S13). Redox testing indicated that the CaO phase significantly enhanced the redox kinetics, leading to more than 10-fold faster methane POx and CO~2~ splitting (figs. S14 and S15). Apart from the faster redox kinetics, the redox stability of SrFeO~3−δ~-CaO NCs (SF~3~-C NCs) was also excellent at 900°C. After the first two cycles, syngas (H~2~/CO ≅ 2) productivity and selectivity (eq. S3) maintained at around 5.8 mol kg~SF3-C~^−1^ and 96%, respectively, in the methane POx step (fig. S16). Meanwhile, CO productivity in the CO~2~-splitting step decreased slightly over the first 10 cycles (less than 9%) and then remained unchanged between 10 and 16 cycles (fig. S16). The stability was further confirmed by XRD analysis, which demonstrated that the 1st and 16th reduced samples exhibited identical XRD patterns with iron reduced to its metallic form (fig. S17). As demonstrated earlier, it is possible to achieve high CO~2~ conversion by partially replenishing the active lattice oxygen. Meanwhile, high methane conversion can be reached with partially reoxidized NCs (for example, 90% reoxidation). For these POx/splitting cycles, both methane and CO~2~ conversions were high compared to complete reduction/oxidation cycles but with slightly lowered syngas and CO productivity. The redox stability of SF~3~-C NCs that underwent partial reduction and oxidation at 980°C with a WHSV of 12,000 cm^3^ g~SF3-C~^−1^ hour^−1^ is shown in [Fig. 5](#F5){ref-type="fig"}. In the methane POx step, syngas (H~2~/CO ≅ 2.13) productivity maintained at around 4.63 mol kg~SF3-C~^−1^ after 13 cycles with insignificant CO~2~ production (\<0.08 mol kg~SF3-C~^−1^). After the redox material stabilized, methane conversion reached 59%, and syngas selectivity was around 90% (fig. S18 and table S1). In the CO~2~-splitting step, CO~2~ conversion was 98%, the average CO production rate was 9.13 mmol kg~SF3-C~^−1^ s^−1^, and CO productivity was 2.17 mol kg~SF3-C~^−1^ in the last 17 cycles. A further increase in syngas yield can be realized by tuning the redox window of SF~3~-C NCs. Over partially reoxidized SF~3~-C NCs, methane conversion of 90% was achieved with 89% syngas selectivity (table S2) at 980°C with a WHSV of 2045 cm^3^ g~SF3-C~^−1^ hour^−1^. Moreover, CO~2~ conversion was nearly 100% in the CO~2~-splitting step before methane POx (fig. S19). ![Redox stability of SF~3~-C NCs.\ (A) Productivity of syngas and CO~2~ and H~2~/CO molar ratio in the methane POx step. (B) Productivity of CO and CO molar fraction in the product during the CO~2~-splitting step; the dashed line denotes the CO mole fraction of 0.95. In all 30 redox cycles, methane conversion is greater than 58%, and CO~2~ conversion is above 94%. Reaction conditions: m~SF3-C~ = 0.25 g; T = 980°C, F = 37.2 μmol s^−1^, P = 1 atm, y~CH4~ = 0.1, and y~CO2~ = 0.1.](1701184-F5){#F5} DISCUSSION ========== The hybrid solar-redox scheme offers various attractive options for CO~2~ utilization and liquid fuel/chemical production. Here, acetic acid production was investigated as an example. Direct conversion of methane and CO~2~ to acetic acid is thermodynamically unfavorable, albeit being atom-economic ([@R44]). State-of-the-art approach is the syngas route: Coal is first gasified to produce syngas, and then the syngas is cleaned and conditioned to remove excess CO and CO~2~. This is followed by methanol synthesis (CO + 2H~2~ ⇆ CH~3~OH) and subsequent carbonylation of methanol (CH~3~CHO + CO ⇆ CH~3~COOH). Methane is not an ideal feedstock for acetic acid synthesis via the conventional syngas route because the CO/H~2~ ratio is too low in methane-derived syngas. The solar-driven redox process, as shown in [Fig. 1](#F1){ref-type="fig"} , produces a pure CO stream and a syngas stream with ideal compositions for methanol synthesis. Thus, the need for costly separation processes is eliminated. The solar-driven redox process has three main sections ([Fig. 1](#F1){ref-type="fig"} and fig. S20): (i) the hybrid solar-redox section where syngas and CO are produced, (ii) the methanol synthesis section that converts syngas into methanol ([@R45]), and (iii) the acetic acid synthesis section in which methanol and CO are converted to acetic acid by the Cativa process ([@R46]). Using ASPEN Plus, the overall energy required for the process was calculated. This process was compared with a coal-based scheme (fig. S21) ([@R47]). For each metric ton of acetic acid produced, 20.4 GJ~thermal~ is required in the solar-driven process, whereas 37.8 GJ~thermal~ is needed in the coal-based process ([Fig. 6](#F6){ref-type="fig"}). Regarding the productivity, the former produces 2.7 times as much acetic acid as the latter per gigajoule of feed fuel ([Fig. 6](#F6){ref-type="fig"}). Simulation results suggest that production of 1 metric ton of acetic acid results in 0.37 metric ton of CO~2~ in the solar-driven process but 2.2 metric tons CO~2~ in the coal-based approach. ![Comparison of the conventional and hybrid solar-redox processes for acetic acid production.\ The hybrid solar-redox process (see [Fig. 1](#F1){ref-type="fig"} ) is compared with the traditional coal-based process to obtain syngas. For the coal-based process, all energy demand is provided by coal. For the hybrid solar-redox process, the total energy demand is 20.4 GJ~thermal~ per metric ton of acetic acid (AcOH), and 39.2% (8 GJ~thermal~ metric ton~AcOH~^−1^) is provided by solar energy.](1701184-F6){#F6} We report iron-containing mixed metal oxides as effective CO~2~-splitting and methane POx agents in a cyclic redox scheme. Unsupported Sr~3~Fe~2~O~7−δ~ demonstrated extraordinary methane POx and CO~2~-splitting activity at 900°C. However, it deactivated over redox cycles. Dispersing Sr~3~Fe~2~O~7−δ~ in a Ca~0.5~Mn~0.5~O matrix significantly enhanced the redox stability of the R-P--structured Sr~3~Fe~2~O~7−δ~. Further investigation of the perovskite-structured SrFeO~3−δ~ compositing with CaO indicated that the inert CaO phase not only enhanced the redox kinetics but also improved its redox stability at 900° to 980°C. At 980°C, the redox activity decreased slightly in the first 13 cycles and then remained unchanged in the last 17 cycles. In the last 17 cycles, the average CO production rate was 9.13 mmol kg~SF-C~^−1^ s^−1^, and CO productivity was 2.17 mol kg~SF3-C~^−1^. The former is 6.8 times higher than state-of-the-art solar-thermal CO~2~-splitting schemes (1.35 mmol kg~oxide~^−1^ s^−1^) carried out at significantly higher temperatures (above 1200°C), and the latter is sevenfold higher than that in solar-thermal schemes (\<0.3 mol kg~oxide~^−1^). For partially reoxidized SF~3~-C NCs, methane conversion reached 90% with 89% syngas selectivity at 980°C with a WHSV of 2045 cm^3^ g~SF3-C~^−1^ hour^−1^. Moreover, CO~2~-to-CO conversion maintained at near 100% in the CO~2~-splitting step. The exceptionally high CO~2~ utilization efficiency and high syngas yield make the redox materials and the hybrid solar-redox scheme potentially attractive. Process simulations indicate that the fossil energy consumption for acetic acid production can be reduced by 67% by the hybrid solar-redox process when compared to the state-of-the-art one while reducing CO~2~ emission by as much as 84%. MATERIALS AND METHODS ===================== Synthesis of the redox materials -------------------------------- For the synthesis of SF~7~-CM NCs \[18 weight % (wt %) SF~7~\], 0.83 g of strontium nitrate \[Sr(NO~3~)~2~\], 3.71 g of calcium nitrate \[Ca(NO~3~)~2~·4H~2~O\], 4.94 g of manganese nitrate \[Mn(NO~3~)~2~·4H~2~O\], and 18.90 g of citric acid (C~6~H~8~O~7~) were dissolved in 80 ml of deionized (DI) water. The solution was then heated up to 40°C under agitation (600 rpm) on a hot plate, followed by maintaining at this temperature for 30 min. After that, 9.18 g of ethylene glycol (C~2~H~6~O~2~) was added to the above solution. The resulted solution was subsequently heated up to 80°C and then kept at this temperature until a polymeric gel (ca. 25 cm^3^) was formed. Iron oxide dispersion was prepared by ultrasonicating 0.50 g of nanoparticles in 51.5 g of 22.3 to 77.7 wt % ethylene glycol/water solution. The dispersion was then transferred to a 50-ml centrifuge tube, followed by centrifuging at 3500 rpm for 5 min. Around 30 ml of the upper part of the nanoparticle suspension was mixed with 20 ml of gel. The mixture was ultrasonicated for 15 min and then heated up to 80°C, followed by holding at this temperature for 2 days. Afterward, the gel was dried at 120°C overnight to evaporate the residual water. The dry precursor was put in a high-purity alumina crucible, which was then placed in a high-purity alumina ceramic tube that was mounted on an MTI furnace (GSL-1500X-50-UL), into which air continuously flew. The tube was heated up from room temperature to 1200°C for 300 min and then held at this temperature for 12 hours, followed by cooling down to 300°C for 300 min. For the synthesis of SF~3~-C NCs (26 wt % SrFeO~3−δ~), 0.70 g of Sr(NO~3~)~2~, 0.90 g of iron nitrate \[Fe(NO~3~)~2~·9H~2~O\], 4.57 g of Ca(NO~3~)~2~·4H~2~O, and 11.97 g of C~6~H~8~O~7~ were dissolved into 80 ml of DI water. The solution was then heated up to 40°C under agitation (600 rpm), followed by maintaining at this temperature for 30 min. After that, 5.80 g of C~2~H~6~O~2~ was added. The resulted solution was further heated up to 80°C and then kept at this temperature until a polymer gel formed. Afterward, the gel was dried at 120°C overnight to evaporate the residual water. Finally, the dry precursor was calcined at 1200°C in air for 12 hours, as described before. For the synthesis of Sr~3~Fe~2~O~7−δ~ and SrFeO~3−δ~, these two redox materials were prepared by the Pechini method as described above. The molar ratio of Sr/Fe in the precursor solutions was 3:2 and 1:1 for Sr~3~Fe~2~O~7−δ~ and SrFeO~3−δ~, respectively. The molar ratio of C~2~H~6~O~2~/C~6~H~8~O~7~/total cations was 3.75:2.5:1 in the gel. The calcination temperature was 1200°C. Material characterization ------------------------- The oxygen stoichiometry of as-prepared redox materials was determined by H~2~-TPR, which was carried out on a TA Instruments thermogravimetric analyzer (SDT Q600). About 50 mg of the sample was loaded into an alumina crucible. Before measurements, the as-prepared sample was treated at 900°C for 1 hour in 20 volume percent (volume %) O~2~/80 volume % Ar (74.4 μmol s^−1^). After treatment, the sample was cooled down and subsequently heated up to 100°C and then kept at this temperature for 30 min in 10 volume % H~2~/90 volume % Ar (74.4 μmol s^−1^). Then, it was heated up to 1150°C at a ramping rate of 10°C min^−1^. To obtain O~2~- or CO~2~-regenerated sample, the sample was first reduced in 10% CH~4~/90% Ar (74.4 μmol s^−1^) and then oxidized in 10% O~2~ (or CO~2~)/90% Ar (74.4 μmol s^−1^) at 900°C, followed by cooling down to room temperature. The oxygen stoichiometry of redox materials under inert conditions was estimated from the amount of lattice oxygen released from the as-prepared sample (after 20% O~2~ pretreatment) at a temperature range of \~25° to 900°C in Ar. Around 580 mg of the sample was loaded to a U-type quartz tube (Ø~in~ = 4 mm), and quartz wool was packed on both ends of the oxide bed to keep the particles in place. The reactor was heated up to 900°C in Ar (22.32 μmol s^−1^). The effluent stream was monitored by an MKS Cirrus 2 quadrupole mass spectrometer (QMS). Rietveld analysis of the sample's XRD patterns at room temperature before and after treating at 900°C in nitrogen was also performed. The XRD patterns were acquired on an Empyrean PANalytical XRD using a Cu Kα radiation (λ = 0.1542 nm) operating at 45 kV and 40 mA, which was equipped with an XRK 900 reactor chamber. A sample pellet was placed on a glass ceramics holder (Ø = 15 mm) in the high-temperature chamber, and the sample was scanned at a 2θ range of 20° to 110° with a step size of 0.012°. The sample was heated up from 25° to 900°C in 30 min and then held at this temperature for 1.5 hours in nitrogen. Finally, the sample was cooled down to room temperature in nitrogen. Powder XRD measurements of as-prepared and spent samples under ambient conditions were performed on a Bragg-Brentano x-ray diffractometer (Rigaku SmartLab) at a 2θ range of 20° to 80° with a step size of 0.05°, using graphite monochromatic Cu Kα radiation (λ = 0.1542 nm) with a nickel filter and operating at 40 kV and 44 mA. To examine the phase properties during the reduction of SF~7~-CM NC, we performed in situ crystallographic analyses of the sample on an Empyrean PANalytical XRD using a Cu Kα radiation operating at 45 kV and 40 mA, which is equipped with a disk-shaped high-temperature chamber. A sample pellet was placed on an alumina sample holder (Ø = 15 mm) in the high-temperature chamber, and the sample was scanned at a 2θ range of 20° to 80° with a step size of 0.1° and a scanning time of 0.1 s at each step. The sample was heated up from 25° to 675°C at an interval of 50°C in 5% H~2~/95% He with a ramping rate of 5°C min^−1^. At each step, the temperature was held for 5 min. High-resolution transmission electron microscope (TEM) images were obtained with a field-emission scanning transmission electron microscope (STEM) (JEOL 2010 F) operating at 200 kV. STEM images and mapping were obtained on an aberration-corrected (S)TEM (FEI Titan 80-300) operating at 200 kV. The sample was prepared by drop-casting a suspension of sample in ethanol (200 proof, anhydrous) on a carbon-coated copper grid and then drying it under ambient conditions. Scanning electron microscope (SEM) images were obtained on a Verios 400 (FEI) field-emission microscope and a JSM-6010LA. To determine the bulk composition of NC, samples were digested with hot concentrated nitric acid. The aliquots were filtered with a 0.45-μm filter (Whatman), and the metal concentrations were analyzed by using an inductively coupled plasma (ICP) optical emission spectrometry (PerkinElmer ICP Optical Emission Spectrometer model 8000). N~2~ physisorption was performed at −196°C on a volumetric adsorption analyzer (Micromeritics ASAP 2020). Before the analysis, the sample was outgassed under vacuum at 300°C for 2 hours. The specific surface area was calculated by using the Brunauer-Emmett-Teller method in the relative pressure range from 3 × 10^−6^ to 0.2. Reactor setup and redox experiments ----------------------------------- The redox performance was evaluated in a microreactor (U-type quartz tube; Ø~in~ = 4 mm) that was vertically placed inside an electric furnace. The furnace was equipped with a K-type thermocouple whose tip touched the outside of the quartz tube at the location of the redox material. The gas flows were controlled by MCQ mass flow controllers (Alicat). The composition of the effluent stream was monitored by QMS. In a typical experiment, 0.25 g of the redox material (250 to 425 μm) was loaded into the reactor, and quartz wool was packed on both ends of the oxide bed to keep the particles in place. The reactor was heated up to the targeted temperature in Ar (66.96 μmol s^−1^), and then these conditions were maintained for around 30 min to remove the residual air in the reaction system. After that, a CH~4~ stream (7.44 μmol s^−1^) was introduced to the Ar stream, and methane POx lasted for 15 min, followed by purging the reactor with Ar (66.96 mmol s^−1^) for 10 min. Next, a CO~2~ stream (7.44 μmol s^−1^) was introduced to the Ar stream, and CO~2~ splitting lasted for 10 min, followed by purging the reactor with Ar (66.96 μmol s^−1^) for 10 min. The above POx-splitting cycle was repeated 30 times. In the redox test with SF~3~-C NCs at 980°C, the amount of coke formed in the POx step was determined by in situ oxidizing the NCs in 10% O~2~/90% Ar (37.2 μmol s^−1^) at 900°C while measuring the CO~x~ production with QMS. The first reduced sample was recovered after the first methane POx step, and the last oxidized sample was collected after the redox experiment was completed. The last reduced sample was prepared by reducing part of the last oxidized sample in 10 volume % CH~4~/90 volume % Ar (74.4 μmol s^−1^) at 900°C for 15 min. We thank A. Shafiefarhood and Y. Liu for collecting the TEM images, Y. Gao for acquiring the SEM micrographs, N. Galinsky for collecting the in situ XRD patterns, and T. Y. Li for the Rietveld analysis of XRD patterns at the North Carolina (NC) State University. Funding: This work was supported by the NSF (CBET-1254351 and CBET-1510900) and the Kenan Institute at NC State University. Author contributions: F.L. conceived and supervised the study. J.Z. and F.L. designed the study and wrote the manuscript. J.Z. carried out the experiments and analyzed the data. V.H. carried out the ASPEN Plus simulation and collected the XRD data. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors. Supplementary material for this article is available at <http://advances.sciencemag.org/cgi/content/full/3/8/e1701184/DC1> Supplementary Text fig. S1. Characterization of Sr~3~Fe~2~O~7−δ~. fig. S2. Redox kinetics on Sr~3~Fe~2~O~7−δ~. fig. S3. Redox stability of Sr~3~Fe~2~O~7−δ~. fig. S4. Characterization of as-prepared and spent Sr~3~Fe~2~O~7−δ~ under ambient conditions. fig. S5. Characterization of as-prepared and spent (oxidized) SF~7~-CM NCs. fig. S6. STEM image and EDS elemental mapping of as-prepared SF~7~-CM NCs. fig. S7. H~2~-TPR of SF~7~-CM NCs. fig. S8. In situ XRD patterns of as-prepared SF~7~-CM NCs during reduction in hydrogen atmosphere at a temperature range of 25° to 675°C. fig. S9. STEM image and EDS elemental mapping of spent (reduced) SF~7~-CM NCs after 30 cycles. fig. S10. STEM image and EDS elemental mapping of spent (oxidized) SF~7~-CM NCs after 30 cycles. fig. S11. CO~2~ splitting on SF~7~-CM NC after methane POx. fig. S12. Characterization of SrFeO~3−δ~. fig. S13. Characterization of SF~3~-C NCs. fig. S14. Redox kinetics on SrFeO~3−δ~. fig. S15. Redox kinetics on SF~3~-C NCs. fig. S16. Redox stability of SF~3~-C NCs. fig. S17. Characterization of as-prepared and spent SF~3~-C NCs under ambient conditions. fig. S18. Redox kinetics over SF~3~-C NCs in a short redox period. fig. S19. Redox kinetics over SF~3~-C NCs. fig. S20. Block diagram for the hybrid solar-redox (HSR)--AcOH. fig. S21. Block diagram for AcOH synthesis using coal-slurry gasifier (CG). fig. S22. Comparison in the overall energy demand and production of AcOH for CG and HSR processes. fig. S23. CO~2~ footprint of the two processes. table S1. Product analysis of the last redox cycle over SF~3~-C NCs. table S2. Product analysis of a short POx over SF~3~-C NCs at 980°C. table S3. Description of the ASPEN Plus model. table S4. ASPEN Plus modules, property methods, and databanks. table S5. Comparison of CG and HSR-AcOH. References ([@R48]--[@R52]) [^1]: Present address: School of Chemical Engineering and Technology, Xi'an Jiaotong University, Xi'an, Shaanxi 710049, China.	{ "pile_set_name": "PubMed Central" }
![](hosplond70322-0011){#sp1 .191} ![](hosplond70322-0012){#sp2 .192}	{ "pile_set_name": "PubMed Central" }
Introduction ============ Retinitis pigmentosa (RP) is the most common form of hereditary retinal degeneration, occurring in around 1 in 3,500 people \[[@r1]\]. It is characterized by progressive degeneration of the peripheral retina (leading to night blindness), loss of the peripheral visual fields, and an abnormal electroretinogram. RP is clinically and genetically heterogeneous, with all three modes of Mendelian inheritance: autosomal dominant RP (adRP), autosomal recessive RP (arRP), X-linked RP (xlRP), as well as mitochondrial inheritance running in families. To date some 47 loci have been identified for nonsyndromic RP and for these 36 of the causative genes have been identified (see [RetNet](http://www.sph.uth.tmc.edu/RetNet/disease.htm) for full list). These include genes encoding components of the phototransduction cascade and the visual cycle, by which the chromophore component of rhodopsin is recycled, as well as retinal transcription factors, structural proteins, and proteins thought to be involved in intracellular transport within photoreceptors \[[@r2]\]. In addition, mutations in four ubiquitously expressed splicing factors, [p]{.ul}re-m[R]{.ul}NA [p]{.ul}rocessing factor 8 (PRPF8 \[[@r3]\]), PRPF31 \[[@r4]\], PRPF3 \[[@r5]\], and PIM1 associated protein (PAP-1) \[[@r6],[@r7]\] have also been described in dominant RP patients \[[@r8]\]. Patients with splicing factor mutations do not appear to exhibit any disease phenotype other than retinal degeneration, and the basis of this specificity remains to be determined. Splicing is a complex process that results in the precise excision of introns from pre-mRNA by a macromolecular structure called the spliceosome \[[@r9],[@r10]\]. The majority of introns are spliced by the major (U2-dependent) spliceosome, which consists of auxiliary protein factors and four small nuclear ribonucleoprotein particles (snRNPs): U1, U2, U5, and U4/U6. However, a small percentage of introns (about 1 in 700 for primates) \[[@r11]\] are spliced by the minor (U12-dependent) spliceosome, which differs from the major spliceosome in four of the snRNPs it contains (U11, U12, U4atac, and U6atac with canonical U5). Signals in intron and exon sequences further define intron and exon recognition and the execution of splicing. In both U2 and U12 introns, canonical (GT/AG) and various noncanonical consensus splice site sequences are used. U12-dependent introns are characterized by a more highly conserved branch site and lack of a polypyrimidine tract \[[@r12],[@r13]\]. It has been suggested that the splicing of U12-dependent introns may be a rate-limiting step in gene expression \[[@r14]\]. PRPF8 is a 220 kDa protein that is highly conserved both in sequence and size, varying between 220 and 280 kDa in different organisms. It is a component of the U5snRNP and U5·U4/U6 tri-snRNP \[[@r15]\]. PRPF8 can also be photochemically cross-linked to the 5′ splice site, the branchpoint and the 3′ splice site in the pre-mRNA \[[@r16]--[@r18]\] and to the U5 and U6 snRNAs \[[@r19],[@r20]\]. PRPF8 is therefore thought to be at the catalytic center of the spliceosome \[[@r21],[@r22]\]. This close association with crucial RNA sequences and proteins in the spliceosome suggests that PRPF8 could directly affect the function of the catalytic core, perhaps acting as a splicing cofactor. Unlike most other known splicing factors, as a component of the U5 snRNPs, PRPF8 participates in both U2 and U12 splicing \[[@r15]\]. Mutations in PRPF8 cause a severe form of dominant RP \[[@r3]\]. To date, 16 different mutations, including missense, premature stop and deletions, have been identified \[[@r3],[@r23]--[@r26]\]. They cluster in a highly conserved region within the last exon, encoding the C-terminus of the protein, making it unlikely that these transcripts will be removed by nonsense-mediated decay \[[@r27]\]. This is supported by real-time PCR analysis that has shown similar expression of both wild-type and mutant alleles in cell lines from RP patients carrying a nonsense mutation \[[@r28]\]. This remarkable clustering of mutations and the relatively high number of missense changes could imply that the mutations alter a specific domain of particular importance to retinal function. Crystal structural analysis of the last 273 amino acids of Caenorhabditis elegans PRPF8 shows that the RP mutations are contained in a C-terminal extension that is not thought to affect the overall structure of this domain \[[@r29]\]. However, it is hypothesized that the C-terminal peptide (2310--2335) constitutes a direct binding surface between PRPF8 and other partners, and mutations on this surface could therefore lead to disruption of critical protein--protein interactions necessary for function \[[@r29]\], a finding recently confirmed by Boon et al. \[[@r30]\]. PRPF31 is a 61 kDa splicing factor first identified in yeast \[[@r31]\]. It is a U4/U6 snRNP-associated protein that promotes tri-snRNP association between U4/U6 snRNP and U5 snRNP by direct interactions with a 102 kDa U5-specific protein \[[@r32]\]. At present, there are 19 known RP-causing mutations in PRPF31. Only three of these are missense changes while the remainder encode severely truncated or grossly abnormal transcripts (six deletions ranging from one base pair to the whole gene, five splice-site mutations, two insertion/deletion events, one duplication, and one insertion) \[[@r24],[@r33]--[@r41]\]; these changes may be targeted by nonsense-mediated decay \[[@r27]\]. In support of this, allele-specific measurement of PRPF31 transcripts have revealed a strong reduction in the expression of mutant alleles, with no truncated proteins being detected. Blocking of nonsense-mediated mRNA decay restored the amount of mutant PRPF31 mRNA but did not restore the synthesis of mutant proteins \[[@r42]\]. This suggests that mutations in PRPF31 cause RP due to haploinsufficiency, an observation further supported by the finding that high-expressing alleles of PRPF31 from the normal parent compensate for a potentially RP-causing mutation on the opposite chromosome \[[@r43]\]. This phenomenon explains the unusually high level of nonpenetrance associated with RP due to mutations in PRPF31 \[[@r44]\]. Three hypotheses can be put forward to explain how mutations in splicing factors might cause RP in the absence of any other defect. The first is that mutations specifically affect the splicing of one or several retina-specific genes. An obvious target would be rhodopsin (RHO), which makes up 80% of the protein content of the rod outer segment \[[@r45]\]. The constant process of disc shedding and renewal of the rod outer segments means that demand for RHO mRNA might be high, which could be met through new synthesis as well as a function of mRNA stability. Consistent with this first hypothesis reduced splicing of an exon 3--4 minigene construct of rhodopsin was associated with cells transfected with a mutated version of PRPF31 \[[@r46]\]. A second hypothesis would be that mutations cause a generalized splicing deficiency but this is only pathological in the fast-metabolizing, end-differentiated retina \[[@r47]\]. This could be consistent with a defect in the translocation of these splicing factor proteins to the nucleus, as proposed by Deery et al. \[[@r48]\], or with a deficit in U12-dependent spliceosome function, as proposed by McKie et al. \[[@r3]\]. A third hypothesis is that PRPF8 and PRPF31 have another function unrelated to mRNA splicing, as proposed for another disease-implicated splicing factor, survival of motor neuron 1 (SMN1) \[[@r49]\]. The finding of mutations in four spliceosomal components leading to adRP would however tend to suggest involvement of the whole tri-snRNP, rather than separate components of it, in this alternative function. Hypotheses one and two, that the RP-causing mutations cause splicing deficiency, may be more consistent with the haploinsufficiency mechanism proposed for PRPF31 mutations, rather than the dominant negative effect implied by the clustered missense mutations found in PRPF8. In this study we compared the efficiency with which U2 and U12 introns (both canonical and noncanonical) were spliced in cell lines derived from RP patients carrying PRPF8 or PRPF31 mutations and controls. The study also aimed to determine whether RP-causing mutations resulted in measurable splicing defects in lymphoblastoid cell lines derived from these patients. Methods ======= Cell culture ------------ Samples of blood from individuals carrying mutations in PRPF8 were collected by D. Mackey (Melbourne, Australia) and R. Ramesar (Cape Town, South Africa; see [Table 1](#t1){ref-type="table"} for details) and deposited with European Collection of Cell Cultures (ECACC; Salisbury, UK), where cell lines were then prepared by EBV transformation. All individuals with PRPF8 mutations had RP. Individuals recruited were otherwise healthy. Peripheral venous blood was obtained from patients and other family members by venepuncture using lithium heparin or sodium citrate vacutainers. Samples were stored and transported at room temperature for a maximum of 5 days before immortalisation with EBV. All blood samples were collected from patients after obtaining informed consent from each participant and receiving local ethics committee approval. The remaining cell lines were obtained directly from ECACC. These included cell lines from patients and asymptomatic carriers with PRPF31 mutations and also unaffected relatives of PRPF31 mutation carriers, who were used as controls. The individuals from whom these lines were prepared were again otherwise healthy. Cells were grown in RPMI1640 medium with L-glutamine supplemented with 10% heat-inactivated FBS (Gibco, Carlsbad, CA) at 37 °C in 5% CO~2~ and maintained according to ECACC protocols. ###### Cell lines used in this study with their corresponding mutations and patient characteristics Group Cell line Mutation Sex Age ---------------- ---------------------- ----------------------------------- --------- --------- Control AG0318 Control M 22 AG0300 Control F 32 AG0309 Control F 34 AG0267 Control M 46 AG0326 Control F 58 AG0261 Control F 77 AGO296 Control F 56 PRPF31 Carrier AG0353 Exon 11, 11 bp Del, nt 1115--1125 M 22 AG0311 Exon 11, 11 bp Del, nt 1115--1125 F 23 AG0298 Exon 11, 11 bp Del, nt 1115--1125 M 58 PRPF31 Severe AG0306 Exon 11, 11 bp Del, nt 1115--1125 F 47 AG0325 Exon 11, 11 bp Del, nt 1115--1125 M 56 HG0003 IVS 6 +3 A\>G F 62 HG0005 Whole gene deletion F 65 AG0316 Exon 11, 11 bp Del, nt 1115--1125 F 72 PRPF8 HG0007 H2309P F 33 HG0008 H2309P F 67 HG0009 HG0011 HG0012 H2309P F 72 H2309R F 60 H2309R F 30 Samples were collected by D.Mackey (Melbourne, Australia), or R.Ramesar (Cape Town, South Africa), or obtained from the ECACC as stated in Materials and Methods. RNA extraction and cDNA synthesis --------------------------------- Total RNA was extracted from pelleted cells using the Qiagen RNeasy extraction kit according to the manufacturer's protocol. The process included a 15-min treatment with DNaseI (Qiagen) to digest any contaminating genomic DNA. Next, 1 µl of RNA was mixed with 1 µg random primer (Invitrogen, Glasgow, UK) and 10 µl nuclease-free water. This mixture was incubated at 70 °C for 10 min and chilled on ice. To each reaction was added 8 µl of master mix: 4 µl 5X Moloney Murine Leukemia Virus Reverse Transcriptase (M-MLV RT) buffer, 2 µl 0.1 M dithiothreitol (DTT), 2 µl 10 mM deoxynucleosides (dNTPs; all Invitrogen), and 0.25 µl RNAsin (Promega, Hampshire, UK). This then was equilibrated at 37 °C for 2 min. Next, 0.5µl Moloney Murine Leukemia Virus (M-MLV) reverse transcriptase (Invitrogen) was added to each reaction, followed by incubation at 37 °C for 1 h and at 95 °C for 5 min. For each cell line a control cDNA synthesis without reverse transcriptase was performed and the presence of genomic DNA was excluded by failure to PCR with GAPDH primers shown in [Table 2](#t2){ref-type="table"}. ###### Primers used in real time PCR analysis Primer name Sequence (5′-3′) Primer name Sequence (5′-3′) ---------------------- -------------------------- ---------------------- ------------------------ GAPDH ref F AACAGCGAGACCCACTCCTC APRT ex 2--3 F GGGCCGCATCGACTACAT GAPDH ref R CATACCAGGAAATGAGCTTGACAA APRT ex 2--3 R AGCCCAGTCCAAGCTCCTG 18S F CGGCTTTGGTGACTCTAGATAACC APRT in 2--ex 3 F AACCAGGTACCCCTTGCCAC 18S R AAGTTGATAGGGCAGACGTTCG APRT in 2--ex 3 R CAAAGAGGAAGCCTCGGGAG β-actin ex 2--3 F CCTGGCACCCAGCACAA PLCδ ex 3--4 F AGATCATCCACCACTCAGGCTC β-actin ex 2--3 FR CCGATCCACACGGAGTACTT PLCδ ex 3--4 R TGTCAGCTTTTCGCAAGCAG β-actin in 2-ex 3 F AGCTGTCACATCCAGGGTCC PLCδ in 3--ex 4 F CTTGGTAGGTTCCAGGGTTCCT β-actin in 2-ex 3 FR CCGATCCACACGGAGTACTT PLCδ in 3--ex 4 R TTGTCAGCTTTTCGCAAGCA PRPF8 ex 25--26 F AGGCAAGAGGCCATTGCTC STK11 ex 3--4 F TGAGGAGGTTACGGCACAAAA PRPF8 ex 25--26 R GATTCGAGGAATGCCACGAT STK11 ex 3--4 R GCCACACACGCAGTACTCCAT PRPF8 ex 25--in 25 F AAGAGCCCATTGCTCAGAACA STK11 in 3--ex 4 F AGCTGTGTGTCCTTAGCGCC PRPF8 ex 25--in 25 R ACTCCACACGGTTCAAAGGC STK11 in 3--ex 4 R GCACACTGGGAAACGCTTCT RPGR ex 10--11 F TCTATCAGCACGTATGCGGC C3G ex 15--16 F TCCCCAGAGGAGCTCATCAA RPGR ex 10--11 R AAACAAGCAGAAAGGCCAAGAG C3G ex 15--16 R TTGCTGACGCGCTTCTTG RPGR in 9--ex 10 F TTCTGTGGATTTATGCTGCAGG C3G in 15--ex 16 F GCCGACCACATGGCTATTTC RPGR in 9--ex 10 R TTTTGCCACACCACGATGAG C3G in 15--ex 16 R GACGCGCTTCTTGAATGTGTC ZNF198 ex 8--9 F GCGCCAAGTGATATTCAGTTGA AP2A1 ex 19--20 F CCTCGGTGCAGTTCCAGAAT ZNF198 ex 8--9 R TGCTGCAGAACTGATGCACTTT AP2A1 ex 19--20 R GCTTGGTCTGCACAGCCAG ZNF198 in 8--ex 9 F TGCTTCTGTAAAAGTGGCGTGT AP2A1 ex 19--in 19 F CTCGGTGCAGTTCCAGAATTTC ZNF198 in 8--ex 9 R TTTGCTGCAGAACTGATGCAC AP2A1 ex 19--in 19 R GACGTGGAGAGGCAGGAGG The table lists the primer name and corresponding amplified exon. Primers were designed using Primer Express 1.5 (ABI) as described in Materials and Methods. Real-time PCR ------------- Ten introns (in ten different genes) were selected for analysis to represent both U2 and U12 introns with canonical and noncanonical donor and acceptor sites. Primer Express 1.5 software (ABI Prism 7000 User's Guide, Applied Biosystems, Foster City, CA) was employed to design primer pairs. Pairs spanned two adjacent exons or an exon and adjacent intron and were used to amplify a product of 80--140 bp, with an annealing temperature between 58 and 60 °C (see [Table 2](#t2){ref-type="table"} for complete primer list). Primer pair concentrations were optimized for real-time PCR analysis according to (ABI, Applied Biosystems) recommendations. GAPDH and 18S were used as endogenous references. cDNAs of highly expressing transcripts were diluted to obtain comparable levels of cycle threshold (C~t~). Real-time PCR was performed using the SYBR Green PCR Core Reagents Kit (ABI, Applied Biosystems) as described by Ponchel et al. \[[@r50]\], and all samples were run in duplicate on an ABI PrismTM 7700 Sequence Detector System. Expression levels of each exon-exon and exon-intron were normalized to GAPDH using a standard ΔC~t~ method. Statistical analyses -------------------- The ratio of spliced to unspliced RNA was calculated for each data point as relative amount of exon-exon product divided by relative amount of exon-intron product. Descriptive statistics on the ratios for each gene and cell line were performed and graphs were produced in Excel. For each cell line the replicates were replaced by their respective means for four groups: normal cells; those from PRPF31 mutation carriers without RP; those from affected patients with PRPF31 mutations; and those with PRPF8-RP. These were compared using one-way ANOVA to detect any overall difference between the four groups, and the LSD post-hoc multiple comparison test was applied to examine individual intergroup differences, using the SPSS statistics base 17.0, SPSS inc, Chicago, IL. In addition, multivariate tests were performed in SPSS to analyze inherent group differences between the control and the mutant groups for U2 and U12 introns, and further, a one-sided bootstrap test was performed. Serum starvation ---------------- Cells were incubated in a serum-free medium for 24 h. The medium was then replaced with serum-containing medium. Samples were taken at 0, 2, 4, 6, 12, 24, and 72 h after another addition of serum. Total RNA was then extracted and cDNA synthesized for real-time PCR analysis as described in the previous section. Transfections ------------- A plasmid minigene containing mouse retinal outer segment membrane protein-1 (Rom1) intron 2 was constructed by Dr J. Wu (Saint Louis University, St. Louis, MO) and was obtained as a gift from Professor Eric Pierce, Scheie Eye Institute, Philadelphia, PA. Two donor site sequence changes were introduced using the Stratagene (La Jolla, CA) site-directed mutagenesis kit (according to manufacturer's instructions as described above), which altered the wild-type canonical GT site to AT and GA. Cells were transfected using the Effectene transfection reagent (Qiagen) following the manufacturer's protocol. Results ======= Pre-mRNA splicing in lymphoblastoid cell lines derived from patients with PRPF8 and PRPF31 adRP ----------------------------------------------------------------------------------------------- To determine whether mutations in splicing factors associated with RP cause detectable splicing defects in non-retinal tissues, we obtained lymphoblastoid cell lines from four groups of individuals: RP patients carrying the H2309P or H2309R mutation in PRPF8; those with the R372_A375delfs mutation in PRPF31 who have severe RP; those carrying the PRPF31 R372_A375delfs PRPF31 mutation but not manifesting RP \[[@r4]\]; and controls (see [Table 1](#t1){ref-type="table"}). Ten exon-intron-exon sequences were chosen for analysis. These included four "house-keeping" genes containing U2 introns with consensus splice sites, β-actin (ACTB), glyceraldehyde 3-phosphate dehydrogenase (GAPDH), the splicing factor PRPF8 gene itself, and RPGR, mutations in which also lead to RP \[[@r51]--[@r53]\]. Also included were two genes containing U2 introns with non-consensus splice sites, ZNF198 (a zinc-finger protein encoding gene), and adenine phosphoribosylase (APRT) . Finally four genes were selected containing [U12](http://www.sanger.ac.uk/Users/rd/U12) introns, one with a consensus GT/AG splice site from phospholipase C delta (PLCδ), and three with non-consensus splice sites, serine-threonine kinase 11 (STK11), ras guanine nucleotide releasing factor C3G and adaptor-related protein complex 2 α-1 (AP2A1; see [Table 3](#t3){ref-type="table"}). The U12 introns were chosen from a database of confirmed [U12](http://www.sanger.ac.uk/Users/rd/U12) introns \[[@r54]\]. ###### Genes selected for analysis of splicing Intron type Splice site sequence Gene Exons ----------------- -------------------------- ---------- -------------------- U2 GT/AG β-actin 2--3 GT/AG GAPDH 7--8 GT/AG PRPF8 25--26 GT/AG RPGR 10--11;in 9--ex 10 GC/AG ZNF198 8--9 GC/AG APRT 2--3 U12 GT/AG PLCδ 3--4 AT/AC STK11 3--4 AT/AC C3G 15--16 AT/AG AP2A1 19--20 This table summarizes gene intron type, splice site sequence and amplified exon for each gene used in analysis. The splicing efficiency of these introns in control and patient cell lines was determined by examining ratios of spliced versus unspliced transcripts using real-time PCR. Relative levels of exon-exon product were divided by relative levels of exon-intron product to produce an estimate of splicing efficiency. Values for the three groups of cell lines were then compared with values obtained in control cell lines, and results analyzed using ANOVA and LSD tests as described in the previous section. Results are shown in [Figure 1](#f1){ref-type="fig"}, [Figure 2](#f2){ref-type="fig"}, and [Table 4](#t4){ref-type="table"}. ![Splicing ratios of U2. Splicing ratios of U2 containing introns RPGR (A) and GAPDH (B) RNA in cell lines derived from control and splicing-factor mutated cell lines. Group 1 represents controls, group 2 represents PRPF31 carriers, group 3 represents PRPF31 severe, and group 4 represents PRPF8 cell lines. The splicing ratios were obtained as described in the Methods section. Means from four to six repeat experiments are shown ±SEM. For individual cell line characteristics see [Table 1](#t1){ref-type="table"}.](mv-v14-2357-f1){#f1} ![Splicing ratio of U12. Splicing ratio of U12 containing intron STK11 RNA in cell lines were derived from control and splicing-factor mutated cell lines. Group 1 represents controls, group 2 represents PRPF31 carriers, group 3 represents PRPF31 severe, and group 4 represents PRPF8 cell lines. The splicing ratios were obtained as described in the Methods section. Means from four to six repeat experiments are shown ±SEM. For individual cell line characteristics see [Table 1](#t1){ref-type="table"}.](mv-v14-2357-f2){#f2} ###### Ratios of spliced to unspliced RNA in groups of control and splicing factor mutant cell lines for selected pre-mRNA introns Intron and splice site Gene Control PRPF31 carrier PRPF31 severe PRPF8 ---------------------------- ---------- ------------- -------------------- ------------------- ------------------ U2 GT/AG ACTIN 328 406 315 402 U2 GT/AG GAPDH 1683 1385 2993 p\<0.05 1433 U2 GT/AG PRPF8 198 173 184 140 U2 GT/AG RPGR 78 105 23 p\<0.05 95 U2 GC/AG ZNF198 154 166 128 124 ↓ U2 GC/AG APRT 8 10 7 6.5 ↓ U12 GT/AG PLCd 18 19 21 9 ↓ U12 AT/AC SKT11 39 37 32 28 p\<0.05 ↓ U12 AT/AC C3G 17 24 9 13 ↓ U12 AT/AG AP2A1 36 39 40 29 ↓ Significant differences obtained by the LSD test are highlighted in bold. A significant increase in splicing efficiency was observed for GAPDH in PRPF31 severe cell lines. In contrast a significant decrease in splicing efficiency was observed for RPGR splicing in the same cell line. In PRPF8 cell lines a significant decrease in splicing efficiency was seen for SKT11, as well as an overall decline in splicing for all U12 introns and non-canonical U2 introns. U2 introns ---------- No significant differences in splicing ratios were observed between control and PRPF8 mutant cell lines for any U2 introns. There were also no significant differences in splicing ratios of U2 introns between control cell lines and those deriving from asymptomatic patients with PRPF31 mutations. In contrast, PRPF31 mutant cell lines from patients with a severe phenotype showed significantly lower splicing ratios for RPGR with respect to controls ([Figure 1](#f1){ref-type="fig"}) and significantly higher splicing ratios for GAPDH. These data are summarized in [Table 4](#t4){ref-type="table"}. U12 introns ----------- None of the splicing ratios assayed for U12 introns in PRPF31 mutant cell lines, either those from RP patients or from asymptomatic carriers, showed significant differences from those obtained in control cell lines. However, PRPF8 mutant cells showed significantly decreased splicing ratios for intron 3 of SKT11 with respect to controls ([Figure 2](#f2){ref-type="fig"}). AP2A1, PLC-delta, and C3G introns did not show statistically significant differences between control and PRPF8 mutant cell lines; however, a general trend could be seen toward a decrease of splicing ratios in PRPF8 mutants compared to controls for all U12 introns. This trend also extended to the two noncanonical U2 introns. These data are summarized in [Table 4](#t4){ref-type="table"}. Multivariate tests ------------------ Multivariate tests were performed to investigate these findings. These tests failed to ascertain the apparent decrease in splicing ratios in PRPF8 mutant cells for U12 introns. This could be due to the sample sizes being too small, or to the parametric test relying on incorrect assumptions. Therefore a bootstrap test was performed, in which the data were resampled, a test statistic was computed, and the results were compared to the data test statistic. The one-sided p-value for the U12 introns comparing controls versus PRPF8 mutants was 0.067, whereas the p-value for controls versus PRPF31 severe was much higher (0.30). This confirmed that the trend toward decreased splicing deficiency seen for PRPF8 mutant cell lines (which just fell short of significance) was not reproduced in PRPF31 mutant cell lines. Analysis of splicing ratios under high metabolic demand ------------------------------------------------------- One possible explanation for the retina-specific manifestations of splicing-factor RP is the high metabolic rate of photoreceptors \[[@r47]\] and the consequent inability of the mutated PRPF8 or PRPF31 to cope with the accompanying splicing demands; this might become especially apparent for rate-limiting U12 intron splicing \[[@r14]\]. To investigate whether splicing is affected during periods of high demand on the splicing apparatus, we starved control and patients' cell lines of serum for 24 h. The splicing efficiency was investigated by real-time PCR following the reintroduction of serum (at 0, 2, 4, 6, 12, 24, and 72 h), which would be expected to result in a substantial increase in the rate of cell division and growth. No significant differences were found (data not shown). Study of the splicing of transfected retina-specific gene, Rom-1 ------------------------------------------------------------------ In a further attempt to determine if splicing could be affected by conditions of high splicing demand, we transfected cell lines with plasmids encoding ROM-1 (exon 1 to exon 2), both in the wild-type form and with mutations introducing a noncanonical splice donor sites (canonical GT mutated to noncanonical AT and GA). The spliced versus unspliced ratios of all constructs were again measured by real-time RT--PCR. No significant differences in splicing efficiency between the control and splicing factor mutant cell lines for any of these constructs were detected (data not shown). Discussion ========== These experiments aimed to distinguish between two hypotheses: that splicing factor mutations cause RP due to a retina-specific defect, either in splicing or in some other function involving these proteins; or that splicing factor RP results from a defect in splicing that is subpathological elsewhere in the body but which leads to a cumulative defect in the metabolically active, irreplaceable retinal tissue. To distinguish between these hypotheses, we assayed the ratio of spliced to unspliced RNA as a measure of splicing efficiency in lymphoblastoid cell lines derived from splicing factor RP patients and normal control cell lines. This assay was applied to a series of introns representing a range of intron types. Results did reveal significant differences in splicing ratios in these cells, but these differences were not consistent between the different splicing factor genes. In cells from PRPF31-RP patients, we observed a decrease in the splicing efficiency of RPGR intron 9 and an increase in the splicing efficiency of GAPDH intron 7. These are both U2 introns with canonical donor and acceptor sites. In contrast, cells from patients with PRPF8-RP showed no significant differences in splicing ratios for U2 introns but showed a consistent downward trend in splicing efficiency of all U12 and noncanonical U2 introns assayed (statistical significance was only attained for intron 3 of STK11). Cells from asymptomatic carriers of PRPF31 mutations revealed no significant differences in splicing ratio compared with control cells, consistent with the findings of Vithana et al. \[[@r43]\], who showed that high expression of the second allele of PRPF31 compensated for the haploinsufficiency. The lack of consistency between findings in the cell lines from patients with the two different mutant genes for splicing factor may be the result of the small number of cell lines assayed, together with the relatively high level of individual variation in splicing efficiency between cell lines. The finding of a difference in splicing efficiency in an intron of RPGR may be significant. RPGR is a putative guanine nucleotide exchange factor that has been localized to the Golgi region \[[@r55]\], and is thought to be important for maintaining the polarized protein distribution across the connecting cilium of the photoreceptors \[[@r56]\]. Mutations in RPGR (RP3, Xp21.1) are responsible for up to 70% of X-linked retinitis pigmentosa \[[@r51]\] and also cause X-linked cone-rod dystrophy (CORDX1) \[[@r52]\] and atrophic macular degeneration \[[@r53]\]. However, RPGR is ubiquitously expressed, and several alternatively spliced isoforms of RPGR have been found in the eye and other tissues, resulting in product sizes ranging between 5 and 20 kb \[[@r57],[@r58]\]. One of these isoforms contains a retina-specific exon, ORF15, which has been found to be a mutational hotspot in RP patients \[[@r59],[@r60]\]. The intron tested herein is present in this and other isoforms expressed in the eye, but is absent from others also present in the eye. Correct splicing of RPGR may be particularly important in the retina; mutations in PRPF31 may have a negative effect on the splicing of RPGR pre-mRNA, which could potentially contribute to the RP phenotype in these patients. In the light of this finding it may now be informative to look at splicing efficiency in other ubiquitously expressed genes which are mutated in RP. The one significant difference in splicing between control and PRPF8 mutant cells was found within the U12 intron-containing gene STK11. Furthermore, although no other significant differences were observed, there was a consistent trend toward an overall decrease in splicing efficiency of all U12 introns tested. U12 splicing is thought to be the rate-limiting step in the processing of any transcript containing a U12 intron \[[@r14]\]. In the fast-metabolizing retina, with a corresponding high demand placed on the splicing machinery, a reduction in splicing efficiency of U12 intron-containing transcripts could become significant over time, while other tissues remain unaffected. Microarray analysis by Gamundi et al. \[[@r28]\] in PRPF8 mutant lymphoblastoid cell lines showed downregulation of one U-12 type gene, SLC12A7, but failed to ascertain a pattern of downregulation of the minor U12-dependent splicing genes as a group. However that study used microarray analysis to examine transcript levels, whereas this study examined intron specific splicing efficiency, potentially a more sensitive method for detecting such an effect. It is also interesting to note that the splicing ratio of GAPDH intron 7 was significantly increased in cells from PRPF31-RP patients relative to controls. GAPDH, as well as being a major enzyme of the glycolytic pathway, has several other functions, such as regulation of gene expression \[[@r61]\]. More important, GAPDH has been found to have a role in the apoptosis of retinal cells \[[@r62]\]. The endpoint in many forms of retinal degeneration is apoptosis of the retinal photoreceptors, so this difference could also contribute to the development of an RP phenotype. Given the relatively small number of cell lines investigated and the small number of splice sites chosen, the substantial variation in splicing efficiency between cell lines (control and disease groups) and the inconsistency of differences found between the different genes for RP, these data must be interpreted with caution. Nevertheless, this study has recorded statistically significant differences in splicing efficiency in cell lines from patients with two different mutated genes for splicing-factor RP. This favors the hypothesis that RP may indeed result from a global deficit in splicing, detectable outside the retina, but which only causes pathology in the retina. Further elucidation of the mechanisms underlying splicing-factor RP will play an important role in the search for therapeutic agents. We acknowledge funding from Wellcome Trust (grant 073988 to C.F.I. and grant 067311 to R.J.G.). [^1]: The first two authors contributed equally to this work.	{ "pile_set_name": "PubMed Central" }
Introduction ============ Autism spectrum disorder (ASD) is a heterogeneous group of neurodevelopmental disorders. Before the Diagnostic and Statistical Manual of Mental Disorders-5 (DSM-5), which was the manual that officially labeled this term as a diagnostic category, ASD was used as a common clinical term that referred to the pervasive developmental disorders (PDD), as described in the DSM IV-TR \[[@B1]\] and the ICD-10 (10th International Classification of Diseases). Successive redefinitions of autism diagnostic criteria have not succeeded in constraining complexity and comorbidity in autism. A multidimensional point of view should encompass the issues posed by this categorical approach. From this perspective, child development is shaped by the interaction between several dimensions (e.g., language, motor, cognition, emotion; [@B53]). In the present study, we focus on emotion recognition, taking into account uni- and multimodal stimuli as well as other developmental comorbidities that are not included in ASD criteria but that are frequently associated with ASD, specifically language, fine motor, and neuro-visual skills. Since [@B28] first clinical description of autism, problems related to emotion processing have been seen as a hallmark symptom of the disorder. However, the status of emotion impairments remains, until now, uncertain. Research that examines emotion recognition in ASD has been limited by an over-focus on the visual modality, specifically the recognition of emotion in facial expressions. Several studies have reported that children with ASD display deficits in this ability ([@B19]; [@B46]; [@B24]; [@B48]). However, several studies have failed to find significant differences in emotion recognition tasks when comparing children with ASD and controls ([@B6]; [@B39]; [@B42]; [@B8]; [@B21]; [@B40]; [@B9]; [@B27]). Additionally, the nature of the deficits in emotion processing has been discussed, but with limited agreement among experts. The difficulties that children with ASD have with emotion processing may relate to their inability to recognize specific emotions, such as fear and disgust ([@B26]; [@B50]), anger ([@B2]), or sadness ([@B16]; [@B50]). At the same time, other authors have failed to find any deficits in the recognition of negative emotions ([@B31]). Overall, emotion recognition deficits do not appear to be universal in ASD, as reflected by heterogeneous performance across children and across tasks (for a review, see [@B37]). Studies using eye-tracking technology have examined the scanning of emotional faces in children with ASD and have found that they globally spend less time on core features (i.e., eyes, nose, and mouth) as compared to typically-developing (TD) children (e.g., [@B18]). However, more recent studies have found that deficits in emotion recognition cannot be fully explained by differences in face scanning, as children with ASD showed no particular differences in time fixation on faces ([@B44]) or showed normative pupillary reactions to emotion expressed by familiar people ([@B38]). Overall, the idea of gaze abnormalities in ASD, including less time spent focusing on emotional cues (e.g., facial expressions), remains controversial (for a review, see [@B22]; [@B55]). Understanding emotional states in real life involves processing a variety of cues that include verbal content, non-verbal cues (e.g., postures), non-verbal vocalization, and affective prosody ([@B10]). The recognition of emotion from vocalizations, which appears to be the auditory equivalent of facial emotion recognition, has been less studied, with limited and inconsistent findings. A few studies have found that, compared to typically developing children, children with ASD demonstrate impaired auditory emotion expression recognition skills ([@B41]; [@B11]). On the other hand, others have found no evidence of a fundamental deficit in ASD ([@B3]; [@B27]). According to the "weak central coherence theory," which was conceptualized by [@B23] and corresponds to the difficulties that children with ASD have when synthesizing stimuli into a coherent whole, children with ASD have impairments in multisensory processing ([@B4]; [@B34]; [@B36]; [@B35]; [@B43]; [@B14]). [@B11] noted that both ASD and TD children benefited from exposure to bimodal information, but to a lesser extent in the ASD group than the TD group. [@B49] explored the processing of neutral and emotional human stimuli (in the auditory, visual, and multimodal channels) in children with PDD-Not Otherwise Specified (NOS), when compared to TD children. The PDD-NOS group experienced difficulties with processing emotional stimuli, particularly in the visual modality. They also more easily identified happy, rather than angry or neutral, faces, and vocalizations. The children with PDD-NOS used the multimodal channel to compensate for their unimodal deficits. Similarly, [@B27] studied emotion recognition abilities by using a combination of visual and auditory tasks for two groups of adolescents, with and without ASD. They found that IQ had a large and significant effect on performance, but they also found no evidence of a fundamental deficit with emotion recognition in the adolescents with ASD. The discrepancies in the conclusions of the different studies that have been described above are underpinned by several factors: (1) the use of different experimental designs (e.g., only visual tasks) and data analyses (e.g., only univariate analyses); (2) the sizes and characteristics of the samples; (3) the heterogeneity among ASD patients in terms of developmental course requiring the consideration of developmental age when comparing to TD children; and (4) the heterogeneity among ASD patients in terms of co-occurring impairments that per se could affect emotion recognition such as visual-motor impairments or developmental coordination disorder (DCD) for visual tasks and language deficits for auditory tasks. The aim of the current study was to explore unimodal and multimodal emotion processing in children with ASD by comparing them to TD children and taking into account other developmental comorbid factors such as language, fine motor, and neuro-visual impairments (using eye-tracking technology for this last dimension). In order to understand the nature of the heterogeneity and discrepancies in the results between studies mentioned above, we chose to take into account comorbid factors and to assess a potential age effect by including a sample of children with ASD in a wide age range (6--13). We hypothesized that: (1) consistent with the literature, we expect to find significant heterogeneity in children with ASD, with regard to their emotional identification skills; (2) in comparison to the TD group, children with ASD will have difficulties processing emotional information, which will be most prevalent in the unimodal channel (visual or auditory) and partially alleviated in the multimodal condition; (3) because ASD presents atypical eye movements and fine motor skills, we expect impairments in the visual modality to be larger in children with ASD; and (4) similarly, we expect that the subgroup of patients with language disorders will be more impaired in regards to unimodal auditory stimuli. Materials and Methods {#s1} ===================== Participants ------------ A total of 19 children with ASD and 19 typically developing children (14 boys and 5 girls in each group) participated in this study. ASD was used as a common clinical term that referred to the PDD that were described in the ICD-10 classification system ([@B52]). Two trained child psychiatrists (JX and DC) clinically assessed the children with ASD (mental age range = 5.8--13.3 years, M age = 7.74 years, SD = 2.51). In reference to the PDD diagnostic category, 5 children satisfied the diagnostic criteria for autism, and 14 children satisfied the criteria for atypical autism (i.e., presence of abnormal or impaired development before the age of three and abnormalities in reciprocal social interactions or in communication without fulfilling the full criteria for autism). Children with ASD were recruited via the Child and Adolescent Psychiatry department at Pitié-Salpétrière Hospital and the University Pierre-et-Marie-Curie in Paris, France. Table [1](#T1){ref-type="table"} summarizes the socio-demographic and clinical characteristics of the patients. Each child with ASD was individually matched, according to developmental age, with a healthy typically developing child using chronological age (TD: age range = 6--13 years, Mean age = 8.84 years, SD = 1.79). The 19 TD children were recruited via a local primary school. The study was conducted in accordance with the hospital's Research Ethics Board regulations. After being fully informed about the study, parents or legal caregivers provided written consent. ###### Main characteristics of children with autism spectrum disorder (ASD; N = 19). --------------------------------------------------- --------------- Chronological age, mean (±SD) 9.95 (1.75) Male/Female 14/5 Socio-Economic Status: high/middle/low 10/5/4 Children's Global Assessment Scale, mean (±SD) IQ score 78.5 (21.02) Mental age (IQ^∗^ age/100) 7.74 (2.51) ADI-R scores Social impairment 10.11 (5.70) Verbal Communication 7.06 (4.22) Repetitive interest 2.72 (2.56) VABS for Children (socialization domain) Interpersonal relationships 34.59 (8.39) Play and leisure time 21.25 (6.61) Coping skills 15.06 (5.98) Total score 70.63 (18.17) Diagnosis (ICD-10) Atypical autism 14 Autism disorder 5 Developmental disorder of speech and language 9 Developmental Coordination Disorder (DCD) 12 --------------------------------------------------- --------------- VABS, Vineland Adaptive Behavior Scales; WISC, Wechsler Intelligence Scale for Children. During a 1-week period of clinical testing, each child was given a series of clinical assessments: the Autism Diagnostic Interview-Revised (ADI-R) was used to score autism core symptoms ([@B32]); the Children's Global Assessment Scale (CGAS; DSM-5, 2013) was used to score global severity; and Vineland Adaptive Behavior Scales were used to score psychosocial/adaptive functioning ([@B47]). The cognitive quotient was ascertained by using the WISC-IV (Wechsler Intelligent Scale for Children-IV), the WPPSI (Wechsler Preschool and Primary Scale of Intelligence), or the KABC-II (Kaufman Assessment Battery for Children, second edition). Developmental age was calculated on this basis to match patients with TD children while taking into account possible intellectual deficits. To evaluate the existence of clinical developmental comorbidities, patients also received (i) a speech and language assessment using the ELO Battery (Evaluation du Langage Oral: Assessment of Oral Language; [@B29]) by a speech therapist, and (ii) a psychomotor assessment using the M-ABC (Movement Assessment Battery for Children; [@B25]) performed by an occupational therapist. Stimuli ------- As described by [@B33], the stimuli used in the present study were categorized as visual, auditory, or bimodal (Figure [1A](#F1){ref-type="fig"}). Visual stimuli consisted of pictures of facial expressions that were obtained from the FACES database ([@B20]). The faces of six actors (three men, three women) who expressed six facial expressions (joy, fear, anger, sadness, disgust, and neutral) constituted a set of 36 visual stimuli. Auditory stimuli included non-verbal affective vocalizations from the Montreal Affective Voices database ([@B5]), in which actors produced emotional interjections by using the vowel /a/ (cry, laugh, etc.). The voices of six actors (three men, three women) who expressed six vocal expressions (joy, fear, anger, sadness, disgust, and a neutral) constituted a set of 36 auditory stimuli. Bimodal stimuli consisted of congruent combinations of an emotional face and an affective vocalization (36 bimodal stimuli). ![Experimental setting. (A) Examples of facial and vocal emotional stimuli presentations for each of the six primary emotions (e.g., "Joy" with the outline of a prosodic laugh). (B) Experimental procedure.](fpsyg-06-01954-g001){#F1} Eye-Tracking Apparatus ---------------------- During tasks with visual and bimodal stimuli, children's gazes were monitored by using an integrated Tobii T120 eye-tracker (Tobii Technology, Danderyd, Sweden). The T120 eye-tracker is a device that is built into a screen (17-inch) and does not require restriction of the children's heads, thus allowing them to look at the pictures freely and naturally. The system tracks both of the children's eyes separately at a rated accuracy of 0.5 degrees and a sampling rate of 120 Hz. A five-point infant calibration was used, and the experiment began after the points were correctly calibrated. By using the Tobii Studio Analysis Software, we evaluated the cumulative number of saccades (as an index of eye movements) and the durations of fixations (as an index of eye stability) on the face. Thus, one area-of-interest (AOI) region was defined for each face image including the whole face. A minimum of 50% valid gaze time was required for the analysis. In the ASD group, seven children were excluded from the eye-tracking analysis, as they did not yield data that were valid for any of the three tasks (e.g., due to excessive movement). A Wilcoxon test comparing the mental age of the 19 vs. 12 subjects remaining was not significant (W = 75.5, p = 0.122), so we are able to conclude that no bias was introduced by removing the seven patients from the analysis. Procedure --------- Before the experimental procedure, we ensured that all participants were able to understand each basic emotion (i.e., we asked each child to explain with examples of each emotion; for the ASD group this ability was confirmed by their language assessments). Children were tested individually in a single session that lasted approximately 30 min (Figure [1B](#F1){ref-type="fig"}). The experiment consisted of three tasks: visual (facial emotions), auditory (vocal emotions), and bimodal. The bimodal (audio--visual) stimuli consisted in the synchronous and congruent facial and vocal presentation for the same emotion. The experiment was run with E-prime software. After the statement of the set, the eye calibration and the familiarization with the four sample items, the experiment began. Each child was seated approximately 60 cm far from the screen of the eye-tracker. Each trial began with the presentation of a fixation cross (500 ms), which was followed by the presentation of the target stimulus during 3 s (i.e., temporal window for the eye-tracking recording); then, labels appeared at the bottom of the screen until the child responded. During the auditory task, no visual stimuli appears on the screen. Participants were asked to select (by clicking with the computer mouse) one label from a list choice that best described the emotion that was being expressed. The order of the three tasks was counterbalanced across children, and the order of trials was pseudo-randomized across each task. There was an inter-trial interval of 700 ms, and a resting pause was offered after every ten trials. Correct answers and the eye-tracking data (the number of saccades and fixation durations on the faces in the visual and bimodal tasks) were recorded. Statistical Analysis -------------------- The data for the present study were Analysed using the statistical program R, version 2.12.2 (R Foundation for Statistical Computing), with two-tailed tests and a 95% confidence level. Due to the repeated measures design and the forced choice paradigm (six possible answers) that was used in our experiment, we used a Generalized Linear Mixed Model (GLMM; lme4 package) to explore the data. A binomial family was specified in the GLMM model to estimate the log-odds ratio for the corresponding factors in the model. With the exception of the eye tracking experiment, we used the entire sample (19 vs. 19). In the general analysis, factors included group (ASD vs. TD), emotion (joy, neutral, anger, disgust, fear, or sadness), task (visual, auditory, or bimodal), developmental mental age (DA) and sex (model formulation: number of successes ∼ Task + Emotion + Group + DA + Sex + Random Participant factor). Then, in tasks where gaze was recorded (unimodal visual and bimodal tasks), we performed new analyses by including fixation duration and number of saccades as additional factors in the precedent model. It is not possible to compute the observed power in the case of mixed models. However, a good estimation is to compute the power for a standard regression and consider the true power to be a bit higher due to the repeated measures design. In our case, if we considered three regressors (group, stimulus, and emotion), a sample size of 38, a type I error rate of 0.05, and a large effect size of 0.35 ([@B13]), the power of the regression was 84%. A medium effect size of 0.15 gave a power of 45%, so even if the true power was \>50%, our experimental protocol did not allow us to detect small differences. A secondary multidimensional analysis was conducted in the entire sample for each task separately (i.e., visual, auditory, and bimodal), and then in each group. In the ASD subgroup, we added the following dimensional factors: Vineland total score (VABS), language disorder and DCD to assess whether such comorbidities affected participant scores during the task. Results ======= General Analysis ---------------- Figure [2](#F2){ref-type="fig"} shows the probabilities of correct answers being obtained by children with ASD and TD control children on the three tasks (visual, auditory, and bimodal) and for each emotion. For the two groups, the probability of success on the bimodal task was greater than on the unimodal visual task (estimate = -0.78, p \< 0.001), which was greater than that on the unimodal auditory task (estimate = 0.61, p \< 0.001). The GLMM model revealed that the rate of correct emotion recognition depended on specific emotions: joy was the most easily recognized by the two groups of children, when compared to the emotions of neutral (estimate = -1.96, p \< 0.001), sadness (estimate = -1.51, p \< 0.001), anger (estimate = -2.68, p \< 0.001), fear (estimate = -1.45, p \< 0.001) and disgust (estimate = -1.13, p \< 0.001). The ASD group was able to perform the multimodal task. Regardless of the emotion that was involved, their probability to succeed was over 0.7 (Figure [2A](#F2){ref-type="fig"}). Neutral and anger were the most difficult emotions to identify for the ASD group (estimate = -1.76, p \< 0.001, and estimate = -1.79, p \< 0.001, respectively). Regarding all of the tasks, we found no significant difference between the results that were obtained by the two groups (estimate = -0.31, p = 0.104), but the scores were strongly and positively associated with the children's developmental age (estimate = 0.27, p \< 0.001). ![Probability of correct emotion recognition, as a function of the emotions and tasks, in ASD (A) and TD (B) children.](fpsyg-06-01954-g002){#F2} When taking into account the combination of the two visual stimuli (visual unimodal and bimodal) and the eye-tracking variables, the ASD group performed worse than the control group (estimate = -0.75, p = 0.03). In addition, the discrepancy between the scores of disgust and joy (estimate = -1.27, p = 0.03) and between the scores of neutral and joy (estimate = -2.13, p = 0.001), was greater in the children with ASD than in the TD controls. There were no significant differences between joy and the three other emotions (anger, fear, and sadness) for the ASD group and the control group. Finally, the discrepancy between the results for the unimodal visual stimuli and the bimodal stimuli was less important in the ASD group than in the control group (estimate = 0.61, p = 0.033), which suggests that adding audio for TD children was more useful than for children with ASD. Eye-tracking data revealed that the ASD group made more saccades (estimate = 0.61, p = 0.002) and had shorter fixation durations (estimate = -0.3, p = 0.012) than the TD controls. However, the eye-tracking variables were not associated with the emotion recognition scores (for saccades: estimate = 0.0003, p = 0.60; for fixation duration: estimate = -0.37, p = 0.054). Multidimensional Analysis by Task --------------------------------- For the unimodal visual task, there was a significant correlation between the probability of success and the children's developmental age; the older children had greater scores. However, we found that there was no effect of sex (estimate = 0.24, p = 0.50), and we did not find an association between the eye-tracking data (saccades and fixation durations) and the scores of the participants (estimate = -0.0005, p = 0.52, and estimate = -0.46, p = 0.12, respectively). The children with ASD tended to perform worse than the TD group on this task (estimate = -0.59, p = 0.06). Within the ASD group, joy was better identified than the other emotions, including sadness (estimate = -2.01, p = 0.0001), anger (estimate = -1.25, p = 0.02), fear (estimate = -1.46, p = 0.008), disgust (estimate = -1.85, p \< 0.001) and neutral (estimate = -1.46, p = 0.007). Interestingly, in the ASD group, we found no correlation between the recognition scores and developmental age (estimate = 0.14, p = 0.18), the Vineland total score (estimate = 0.009, p = 0.52), or the presence of language or coordination disorders (all p \> 0.07). For the TD group, joy was better identified than anger (estimate = -2.36, p = 0.002), fear (estimate = -2.66, p \< 0.001), and sadness (estimate = -3.12, p \< 0.001). In contrast, no significant differences were found between the recognition scores for joy and the scores for neutral (estimate = -0.73, p = 0.41) and disgust (estimate = -1.46, p = 0.07). Regarding auditory stimuli processing, the ASD group did not perform significantly worse than the TD group (estimate = -0.08, p = 0.64). There was a significant effect of the children's developmental age: the older children had better scores (estimate = 0.3, p \< 0.001). For the ASD group, developmental age was strongly correlated with the probability of success in emotion recognition (estimate = 0.24, p \< 0.001) and, to a lesser extent, the Vineland total score (estimate = 0.015, p = 0.02) and the presence of language impairment (estimate = -0.48, p = 0.05). With regard to specific emotions, the two groups obtained the lowest scores, with equivalent values, for anger. Finally, during the bimodal emotional task, the ASD group did not perform significantly worse than the TD group (estimate = -0.62, p = 0.10). Joy was better recognized than anger (estimate = -1.62, p \< 0.001) and neutral (estimate = -0.93, p = 0.03). These results were similar for the ASD group. The probability of success was strongly associated with the children's developmental age: the older children had significantly higher scores (estimate = 0.6, p \< 0.001). By contrast, we found that there was no significant association between the probability of success in the emotion recognition task and the Vineland total score or the presence of language or coordination disorders (all p \> 0.51). Moreover, the probability of emotion recognition was not associated with the eye-tracking data. Discussion ========== Our study aimed to reveal the specific patterns of emotional processing in children with ASD (according to the PDD diagnostic category of the ICD-10), when compared to normally developing children, through a multimodal identification task and the assessment of developmental comorbidities. As a whole, the children with ASD performed well on the emotional multimodal tasks. Unlike [@B49] study, we did not find a significant difference between the scores that were obtained by the two groups. We also did not find a significant difference between the scores of the two groups on the auditory and bimodal tasks. The discrepancy between [@B49] results and our results may be explained by their sample, which included patients with PDD-NOS who also met the Multiple Complex Developmental Disorder criteria (i.e., patients also exhibited emotional/anxiety symptoms; [@B12]; [@B7]; [@B54]). Similar to [@B27], who tested adolescents with and without ASD by using facial and vocal emotional tasks, we found that the probability of correct emotion recognition was strongly correlated with the developmental age of the participants for each of the tasks. Interestingly, these results were confirmed in the ASD group, except for the visual task. In the ASD group, this may be explained by two factors: (1) for the unimodal visual modality, this group tends to perform worse than the TD group, but most of all, (2) the auditory modality seems to be discriminant in this association, with the strongest estimate between emotion recognition scores and developmental age being found for this task (estimate = 0.24, p \< 0.001) and being higher than that of the bimodal task (estimate = 0.52, p \< 0.001). However, [@B27] described a similar emotion processing style in the ASD and non-ASD groups, except for a difficulty in recognizing surprise. On the contrary, we found, consistent with the literature, an important heterogeneity in our results, which could be considered to be a key issue. For the entire multimodal task, we found a strong contrast between joy recognition (the emotion that was most easily identified) and the neutral emotion or anger (the most difficult to identify) for the ASD group. These results, which are congruent with the data of [@B49], are also found in the bimodal and auditory tasks' secondary analyses. The results concerning the neutral emotion could be due to abnormalities in the recognition of specific emotions in ASD children, based on a greater discrepancy being found between neutral and joy recognition in the ASD group than the TD group. Conversely, for anger in the auditory task, both groups obtained the lowest scores, which had equivalent values. The ASD group performed significantly worse than the TD group in the case of the visual and bimodal stimuli. The lowest scores were obtained by both of the groups in the auditory task; multisensory processing allowed children with ASD to partially compensate for the difficulties that were experienced in the visual modality, which confirms the results of [@B49]. This result seems to be incongruent with the impairments in multimodal processing that are described in individuals with ASD ([@B23]; [@B14]). However, according to [@B11], the discrepancy between the visual and bimodal tasks is significantly more important in the TD group than in the ASD group. These results do not necessarily contradict the "weak central coherence theory" in regard to people with ASD ([@B23]); this partial improvement in accuracy on the bimodal task could be better explained by the addition of redundant (visual + auditory) targets, rather than the multisensory integration of visual and auditory cues into a unified percept. A few studies of young adults have demonstrated that congruent emotional information that is processed via multisensory channels optimizes behavioral responses, which results in enhanced accuracy and a faster response time (RT; see [@B17]; [@B30]; [@B33]). Multisensory enhancement is sometimes explored in behavioral studies that use RT by comparing the observed RT distribution to the distribution that is predicted by a 'race model' (e.g., [@B15]). Multisensory integration occurs when the reaction time for bimodal trials is faster than what is predicted by the race model (e.g., [@B11]; [@B33]). Testing this hypothesis by using RTs in our study was not possible because the RTs were not recorded due to the particular eye-tracking procedure (i.e., forced stimuli exploration over 3 s) and the studied population (i.e., TD children and children with ASD). The eye-tracking data revealed interesting differences between the two groups (i.e., an important number of saccades and shorter fixation durations for the children with ASD than TD children). These results are consistent with studies that describe gaze abnormalities in ASD, including a shorter time being spent on emotional cues (for a review, see [@B22]; [@B55]). However, these different gaze patterns were not associated with the scores of the participants and were not able to explain the discrepancy between the performances of the two groups. It is, therefore, difficult to conclude by stating that autistic children process faces in a holistic fashion. These abnormalities could be, in part, explained by our experimental design and the eye-tracking data analysis (see limitations). Because children with DCD have poor cross-modal integration (see [@B51], for a meta-analysis) and frequent neuro-visual impairments, we expected that children with ASD and comorbid DCD would preferentially fail on the unimodal visual or bimodal visuo-auditory tasks. Our results were not consistent with this hypothesis; no association was found between the presence of a DCD and the scores on the visual or bimodal tasks. We assumed that children in the ASD group who had comorbid language disorders would be particularly impaired concerning unimodal auditory stimuli. Our hypothesis was partially correct: we found a statistical trend in this direction (estimate = -0.48, p = 0.05) regarding this modality. Furthermore, we did not find any association between the presence of this comorbidity and the scores for emotion recognition on the other tasks (visual and bimodal; all p \> 0.78). Finally, a unique significant correlation between the scores on Vineland and the performances in emotion recognition was found for the auditory stimulus. There are several limitations that warrant consideration. First, different metrics were used to match the ASD and TD groups based on age. Cognitive assessments were performed on the ASD group to assess developmental age but not on the TD group, for which chronological age was used instead. In addition, the lack of some significant effects may be due to the power of the study. Given the small sample size of our groups and considering the clinical heterogeneity of the ASD group, we were only able to detect major effect sizes and unable to detect possible subtle impacts of co-occurring factors (i.e., comorbidities) or potentially clinically meaningful effects (i.e., Vineland Adaptative Coordination Scale). Regarding the eye-tracking data analysis, we consider faces as a whole, rather than more precisely analyzing the gaze upon certain areas of interest (e.g., eyes, nose, and mouth). This choice is a limitation for the fine comparisons between children with ASD and TD children when exploring faces during an emotional task. However, our analysis was based on current, state-of-the-art machine learning methods to recognize facial emotion. Machine learning has shown that facial emotion classifiers use Action Units that are distributed all over the face to achieve the best performance ([@B45]). Conclusion ========== Children with ASD demonstrate rather high performance on emotion recognition, particularly for multimodal stimuli. The difficulties that are experienced for visual stimuli are partially alleviated when using bimodal stimuli. Developmental age plays a major role for TD children, whereas its role is limited to the multimodal task for children with ASD. However, performances in emotion recognition in ASD are heterogeneous and do not simply correlate with comorbidities. The existence of a language disorder seems to have an impact on the performances of the ASD group in the auditory modality. Conversely, in the visual or bimodal (visuo-auditory) tasks, the impact of a DCD or gaze impairments has not been demonstrated. Future studies with larger samples will allow researchers to confirm and refine these data. Given the heterogeneity of the results that are found in the literature, we wonder whether impaired emotion processing may constitute a dimension that should be explored in ASD so that we can define subgroups in this condition that are more homogeneous and tailor interventions. Conflict of Interest Statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. We are grateful to the children and their parents for generously giving their time to participate in this study. [^1]: Edited by: Rosario Cabello, University of Granada, Spain [^2]: Reviewed by: Cheryl Dissanayake, La Trobe University, Australia; Ruth Ford, Anglia Ruskin University, UK [^3]: This article was submitted to Developmental Psychology, a section of the journal Frontiers in Psychology	{ "pile_set_name": "PubMed Central" }
![](edinbmedsurgj71964-0042){#sp1 .36} ![](edinbmedsurgj71964-0043){#sp2 .37} ![](edinbmedsurgj71964-0044){#sp3 .38} ![](edinbmedsurgj71964-0045){#sp4 .39} ![](edinbmedsurgj71964-0046){#sp5 .40} ![](edinbmedsurgj71964-0047){#sp6 .41} ![](edinbmedsurgj71964-0048){#sp7 .42}	{ "pile_set_name": "PubMed Central" }
Background ========== The distal nephron plays an important role in the maintenance of sodium balance, extra cellular fluid volume and blood pressure \[[@B1]\]. It is well known that inappropriate water and sodium retention is thought to be a key factor in several forms of hypertension, and that aquaporin2 waterchannels (AQP2) play a key role in several water balance disorders \[[@B2],[@B3]\]. Animal models have shown a reduced AQP2 expression in conditions with acquired nephrogenic diabetes insipidus such as lithium treatment and an increased expression in diseases with water retention such as congestive heart failure \[[@B4]\]. Gain-of-function mutations in the epithelial sodium channels (ENaC) cause inappropriate renal sodium retention and consequent increases blood pressure \[[@B5]\]. Because AQP2 and ENaC plays such an important role in water and sodium balance and associated disorders, it is important to identify factors involved in the reabsorption of water and sodium by the kidneys in order to study these disorders in the future. The exact role of AQP2 and ENaC has never been examined under volume expansion in healthy humans. The effect of an acute intravenous volume load, with isotonic and hypertonic saline and glucose, on urinary excretion of AQP2 (u-AQP2), urinary excretion of ENaC gamma subunit fractions (u-ENaCγ) and its relationship to vasopressin (AVP) and kidney function in healthy humans has not been studied by simultaneous measurement of other important regulatory hormones of water and sodium homeostasis such as the renin-angiotensin-aldosterone system (RAAS). In this present study we wanted to study the sodium and water transport in the distal nephron by measuring 1) the excretion of u-AQP2 and u-ENaCγ 2) associated regulating hormones and 3) the renal response after volume expansion in healthy humans. In order to analyse these physiological mechanisms, we performed a randomized, crossover study in healthy subjects. We investigated the effects of infusion with isotonic- and hypertonic saline and isotonic glucose on urinary excretion of AQP2 and ENaCγ corrected for creatinine (u-AQP2~CR~ and u-ENaCγ~CR~), renal function and sodium handling, vasoactive hormones and systemic blood pressure. Methods ======= Participants ------------ ### *Inclusion criteria* Healthy non-smoking men and women with age between 18 -- 45 years were included in this study. ### *Exclusion criteria* Subjects with clinical signs or history of heart, lung, kidney, endocrine or malignant disease; abnormal findings in ECG, urine dipstick or biochemistry (blood cell count, plasma concentrations of haemoglobin, sodium, potassium, creatinine, albumin, glucose, bilirubin, alanine aminotransferase, alkaline phosphatase and cholesterol); arterial hypertension (ambulatory BP \>130/80 mmHg); medical treatment; alcohol and substance abuse; present smoking; pregnancy; breast feeding; donation of blood within one month prior to the study and obesity (BMI \> 32 kg/m^2^) were excluded from this study. ### *Withdrawal criteria* Subjects who developed the condition given in exclusion criteria during the course of the experiment, who withdrew their informed consent, and who had a poor compliance were withdrawn from this study. Ethics ------ This study was approved by the Regional Committee on Health Research Ethics (j. no. M-2011003) and carried out in accordance with the Helsinki Declaration. Written informed consent was obtained from all subjects. Recruitment ----------- Healthy male and female volunteers were recruited through advertisement at public institutions in Holstebro, Denmark. Design ------ The study was conducted as a randomized, placebo-controlled, crossover study. On three different occasions separated by at least two weeks, subjects were randomised to 0.9% isotonic saline (0.9% NaCl), 3.0% hypertonic saline (3% NaCl) or 5% glucose (glucose), which was administered as a sustained infusion over 60 minutes. The amount of fluid used in this study was 0.9% NaCl: 23 ml/kg, 3% NaCl: 7 ml/kg and glucose: 23 ml/kg. Effect variables ---------------- The main effect variables were u-AQP2, as well as u-ENaCγ. Other effect variables were diastolic (DBP) and systolic blood pressure (SBP), plasma concentration of renin (PRC), angiotensin II (AngII), aldosterone (Aldo), vasopressin (AVP), free water clearance (C~H2O~)~,~ glomerular filtration rate (GFR), fractional excretion of sodium (FE~Na~) and potassium (FE~K~), plasma sodium (p-Na), urine osmolality (u-osm), plasma osmolality (p-osm) and plasma albumin (p-alb). Number of subjects ------------------ Using a significance level of 5% and a power of 90% it was calculated that the number of subjects needed were 21, when the minimal relevant difference in U-ENaCγ was 100 ng./min and SD was 95 ng./min. Incomplete voiding during examination days was expected in some subjects, therefore 26 subjects were included in the study. Experimental procedures ----------------------- ### *Experimental procedure prior to the study day* Four days prior to each study day, subjects consumed a standardized diet regarding calories, sodium and fluid. The diet consisted of 11,000 kJ/day with an energy distribution of 55% carbohydrates, 30% fat and 15% protein in accordance to general dietary guidelines. The sodium content was 120 mmol pr. day. The subjects were asked to drink exactly 2500 ml/day. No alcohol, coffee, tea or soft drink consumption was allowed while on the standardized diet. Subjects were instructed to keep their physical activity as usual during the experiments and to abstain from hard training. A 24-hour urine collection, ending at 7:00 AM on the examination-day, was used to assess water and sodium balance. ### *Experimental procedure on the study day* After an overnight fast, subjects arrived at our facility at 8:00 AM. Two indwelling catheters for blood sampling and administration of ^51^Cr --EDTA and fluid were placed in both cubital veins. Every 30 minutes, starting at arrival, participants received a 175 ml oral water load of tap water. Urine was collected in standing or sitting position. Otherwise subjects were kept in supine position in a quiet temperature-controlled room (22-25°C). Three 30 minutes baseline clearance periods were obtained from 9:30 AM to 11:00 AM. These were followed by one clearance period from 11:00 AM to 12:00 PM during which a sustained infusion of either glucose, 0.9% NaCl or 3% NaCl was administered. The post infusion period consisted of three 30-minute periods from 12:00 PM to 1:30 PM. Blood and urine samples were collected every 30 minutes from 9:30 AM to 1:30 PM and analysed for ^51^Cr-EDTA, electrolytes and osmolality. Analysis of plasma concentrations of PRC, Ang II, Aldo and AVP were conducted from blood samples drawn at 11:00 AM (baseline), 12:00 AM (cessation of fluid infusion), 12:30 PM (30 min after cessation of fluid) and 1:30 PM (90 min after cessation of fluid). For data analysis, the 30-minute periods were subdivided into: baseline (0--90 min), infusion (90--150 min) and post infusion (150--180 min, 180--210 min, 210--240 min). Measurements ------------ ### *Renal function* Glomerular filtration rate was measured by the constant infusion clearance technique using ^51^Cr-EDTA as reference substance. A priming dose of ^51^Cr-EDTA was given followed by sustained infusion that was kept stable using a volume-controlled infusion pump. More than 15% variation in GFR between the three baseline periods led to exclusion of analysis. ### *Blood samples* Were centrifuged for 10 minutes at 2200 × g at 4°C. Plasma hormone samples were kept frozen at -20°C (AngII) and -80°C (PRC, Aldo, and AVP) until assayed. Renin in plasma was determined using an immunoradiometric assay from and a kit from CIS Bio International, Gif-Sur-Yvette Cedex, France. Minimal detection level was 1 pg./mL the coefficients of variation were 14.5% (interassay) and 4.5% (intra assay). Aldosterone in plasma was determined by radioimmunoassay using a kit from Demeditec Diagnostics Systems Laboratories Inc. (Webster, TX, USA). Minimal detection level was 22 pmol/L. The coefficients of variation were 8.2% (inter-assay) and 3.9% (intra-assay). Arginine vasopressin and Angiotensin II were extracted from plasma with C~18~ Sep-Pak (water associates, Milford, MA, USA) and subsequently measured using radioimmunoassay as previously described. The antibody against angiotensin II was obtained from the Department of Clinical Physiology, Glostrup Hospital, Glostrup, Denmark. Minimal detection level was 2 pmol/L. The coefficients of variation were 12% (inter-assay) and 8% (intra-assay). The antibody against AVP was a gift from Professor Jacques Dürr (Miami, FL, USA). Minimal detection level was 0.2 pmol/L. The coefficients of variation were 13% (inter-assay) and 9% (intra --assay). ### *Urine samples* Were kept frozen at -20°C until assayed. U-AQP2 was measured by radioimmunoassay as previously described \[[@B6],[@B7]\]. Antibodies were raised in rabbits to a synthetic peptide corresponding to the 15 COOH-terminal amino acids in human AQP2 to which was added an NH~2~-terminal cystein for conjugation and affinity purification. Minimal detection level was 34 pg/tube/tube. The coefficients of variation were 11.7% (inter-assay) and 5.9% (intra-assay). U-ENaCγ was measured by radioimmunoassay as previously described \[[@B8],[@B9]\]. Antibodies were raised against the synthetic ENaCγ peptide in rabbits and affinity purified as described previously \[[@B10]\]. Minimal detection level was 48 pg/tube. The coefficients of variation were 14% (inter-assay) and 6.7% (intra-assay). ### *Blood pressure measurement* Brachial blood pressure was recorded using a semiautomatic oscillometric devise (Omron 705IT, Omron Matsusaka, Japan). Plasma and urine concentrations of sodium, potassium, creatinine and albumin were measured using routine methods at the Department of Clinical Biochemistry, Holstebro Hospital. Fractional excretion of sodium was calculated as \[sodium clearance (C~Na~)/GFR × 100%\]. Fractional excretion of potassium was calculated as \[potassium clearance (C~K~)/GFR × 100%\]. Free water clearance was calculated as \[urine output (UO) -- osmolar clearance (C~OMS~)\]. C~OSM~ was calculated as \[urine osmolarity/plasma osmolarity × UO\]. Statistics ---------- Statistical analyses were performed using IBM SPSS statistics version 20.0.0 (IBM Corp.; Armonk, NY, USA). Single baseline values were obtained by taking the weighed average of the measurements from the three baseline periods. Parametric data are presented as means ± standard deviation (SD) and nonparametric data as medians with interquartile ranges. General linear model (GLM) with repeated measures was performed, with time as within-subject factor and intervention as between subject factor, to test for differences within and between groups. One-way ANOVA was used for comparison of means between groups when differences were found. For non-parametric data related samples Friedman's two-way analysis (FM) was used. For comparison within groups at baseline and post infusion period 210--240 minutes, a paired t-test was used when data were parametric and Wilcoxon's signed rank test was used when data were nonparametric. Statistical significance was defined as p \< 0.05 in all analyses. Results ======= Demographics ------------ A total of 31 healthy women and men were enrolled in the study. Five subjects were excluded due to: abnormal blood samples (1), 24-h BP above 130/80 mmHg (1), non-compliance (1) and withdrawal of informed consent (2). Thus, 26 persons completed the study. Three were not able to void satisfactorily during clearance experiments and were excluded from analysis. One was not able to void in two post intervention periods after 3% NaCl and was excluded in channel analysis only. The remaining 23 males (n = 9) and women (n = 14) had a median age of 26 years (range 18--42) and a mean BMI of 24.4 ± 2.3 kg/m^2^. Mean ambulatory blood pressure was 119/70 ± 8/4 mmHg. Screening blood values were b-haemoglobin 8.5 ± 0.7 mmol/L, p-sodium 139 ± 2, p-potassium 3.9 ± 0.4 mmol/L, p-creatinine 74 ± 9 μmol/L, p-albumin 42 ± 3 g/L, p-glucose 5.1 ± 0.6 mmol/L, p-alanine transaminase 25 ± 9 U/L and p-cholesterol 4.5 ± 0.5 mmol/L. Twenty-four-hour urine collection --------------------------------- Table [1](#T1){ref-type="table"} shows the results of the 24-h urine collection in 23 healthy subjects after 4 days of standardized diet. Mean u-AQP2, u-ENaCγ, urinary sodium, urine osmolarity, C~H2O~ and urine volume were the same in all three examination days indicating that the subjects had kept their supplied diets and fluid intake. ###### Urine output, urine osmolarity (u-osm), free water clearance (C~H2O~), urinary AQP2 excretion per minute (u-AQP2), urinary excretion of ENaCγ per minute (u-ENaCγ), urinary sodium excretion (u-Na) and fractional excretion of sodium (FE~Na~) during 24-hours urine collection with fluid deprivation (12 PM to 8.00 AM) in a randomised, crossover study of 23 healthy subjects Examination day: P (ANOVA) ---------------------------- ---------------------- --------------- ------------- ------ Urine Output (ml/24 h) 2306(559) 2120(695) 2281(650) 0.56 u-osm (mosm/24 h) 848(231) 797(207) 833(239) 0.73 C~H2O~(ml/min) -0.46(0--61) -0.47(0.57) -0.44(0.64) 0.99 u-AQP2 (ng/min) 0.93(0.24) 0.91(0.22) 0.93(0.24) 0.89 u-ENaCγ (pg/min) 405(163) 389(188) 385(164) 0.92 u-Na (mmol/24 h) 124(52) 119(41) 120(56) 0.93 FE~Na~(%) 0.49(0.15) 0.47(0.13) 0.45(0.18) 0.67 Values are means with SD in brackets. One-way ANOVA was used for comparison between groups. Water excretion, u-AQP2, u-osm ------------------------------ Table [2](#T2){ref-type="table"} shows the absolute values of UO, C~H2O~, u-AQP2~CR,~ u-AQP2 excretion rate and u-osm during the baseline period, the infusion period and the post infusion period. ###### Effect of 0.9% isotonic saline (0.9% NaCl), 3% hypertonic saline (3% NaCl) and isotonic glucose (Glucose) on urinary output (OU), free water clearence (C~H2O~), urine osmolarity (u-osm), urinary aquaporin2 excretion rate (u-AQP2) and urinary aquaporin2 corrected for creatinine (u-AQP2~CR~) in a randomized, crossover study of 23 healthy subjects Periods Baseline Infusion Post infusion p (GLM-within) --------------------------------- -------------- -------------- ------------------- -------------------- ----------------- ---------- UO (ml/min) 0.9% NaCl 6.90(1.33) 8.75(1.76) 7.95(2.31) 7.39(2.36) 8.12(2.26) \<0.0001 3% NaCl 7.12(1.35) 2.75(0.73) 1.81(0.64) 1.98(1.07) 2.05(1.09)\\ Glucose 6.83(1.58) 11.84(2.28) 16.05(2.82) 15.28(2.47) 13.45(2.47)\\ p (GLM between) \<0.0001 p ANOVA 0.77 \<0.0001 \<0.0001 \<0.0001 \<0.0001 C~H2O~ 0.9% NaCl 3.87(1.55) 5.29(1.44) 4.07(2.09) 3.01(2.04) 3.62(1.59) \<0.0001 3% NaCl 3.97(1.21) -0.35(0.73) -1.87(0.83) -2.22(0.88) -2.26(0.96) Glucose 3.84(1.32) 6.38(1.77) 11.60(0.52) 12.47(2.35) 11.17(2.12) p (GLM between) \<0.0001 p ANOVA 0.95 \<0.0001 \<0.0001 \<0.0001 \<0.0001 U- osm (mosmol/kg) 0.9% NaCl 140(62) 115(21) 149(58) 176(50) 161(33) \<0.0001 3% NaCl 136(37) 342(96) 606(114) 637(105) 640(111)\* Glucose 136(41) 133(21) 79(21) 52(12) 48(7)\* p (GLM between) p \<0.0001 p ANOVA 0.95 \<0.0001 \<0.0001 \<0.0001 \<0.0001 u-AQP2 (ng/min) 0.9% NaCl 1.03 (0.25) 1.02(0.22) 1.22(0.34) \- 1.26(0.32)\* \<0.0001 3% NaCl 1.02(0.22) 0.99(0.31) 1.28(0.30) \- 1.37(0.34)\* Glucose 1.01(0.22) 1.26(0.25) 1.19(0.29) \- 0.78(0.18)\* p (GLM between) 0.274 u-AQP2~CR~(ng/mmol) 0.9% NaCl 102.5(19.9) 110.8(18.1) 119.8(25.4) \- 125.0(21.0)\\ \<0.0001 3% NaCl 105.9(15.7) 108.6(27.2) 125.1(31.4) \- 132.7(26.5)\* Glucose 105.7(19.1) 137.1(22.3) 118.1(26.5) \- 87.9(15.4)\* p (GLM between) 0.565 Values are mean with SD in brackets. General linear model (GLM) with repeated measures was performed for comparison within the group and intervention as between subjects factor. One way ANOVA was performed when differences were found between interventions. Paired t-test was used for comparison within treatment group at baseline and post infusion period 210 -- 240 minutes. \p \< 0.0001; \\p \< 0.001. UO increased significantly after 0.9% NaCl and glucose. The 3% NaCl infusion induced a significantly and sustained decrease in UO. The relative changes in UO were significantly different between the three interventions. C~H2O~ increased during the infusion with 0.9% NaCl, and decreased slightly, although significant in the postinfusion period. At the end of the examination-day C~H2O~ increased towards baseline levels with an over all relative change of -10%. There was a pronounced increase in C~H2O~ after glucose, whereas C~H2O~ decreased after 3% NaCl and changed from positive values at baseline to negative values after infusion. Thus, indicating a change from free water excretion to water reabsorption (Table [2](#T2){ref-type="table"}). U-AQP2~CR~ increased by 27% (p \< 0.001) in response to 0.9% and by 26% (p \< 0.0001) after 3% NaCl and reached maximum at 240 min after baseline. During glucose infusion (90--150 min) there was a primary increase in u-AQP2~CR~ after which u-AQP2~CR~ decreased and reached a minimum of - 16% (p \< 0.0001) at 210--240 min (Figure [1](#F1){ref-type="fig"}A). The excretion of u-AQP2 divided by gender, showed that u-AQP2~CR~ tended to be higher in women than in men, but there was no statistical significant difference. This was due to a lower creatinine concentration in women's urine (data not shown). U-AQP2 excretion rate followed the same pattern (Table [2](#T2){ref-type="table"}). The relative changes in u-AQP2 did not differ between 3% NaCl and 0.9% NaCl, but both were significantly different from the relative change in u-AQP2 after glucose infusion. ![Effects of isotonic 0.9% saline (■), hypertonic 3% saline (♦) and isotonic glucose (●) on urinary excretion of A) u-AQP2 and B) u-ENaCγ adjusted to creatinine, C) plasma concentration of vasopressin (AVP) and D) plasma osmolality in 23 healthy subjects.* Values are means ± SEM. Paired t-test was used for comparison of post infusion period 210--240 min vs. baseline. \* p \< 0.01; \\ p \< 0.001; \\\* p \< 0.0001.](1471-2369-14-202-1){#F1} U-Osm decreased during 0.9% NaCl infusion with minimum after infusion ended at 150 minutes, after which u-osm increased, coherent with the changes seen in C~H2O~. U-osm increased significantly in response to 3% NaCl and lasted throughout the experiment. During glucose infusion u-osm remained constant for 60 minutes until glucose infusion was completed, after which u-osm declined and reached minimum at 210--240 min (Table [3](#T3){ref-type="table"}). ###### Effect of 0.9% isotonic saline (0.9% NaCl), 3% hypertonic saline (3% NaCl) and isotonic glucose (Glucose) on urinary sodium excretion (u-Na), fractional excretion of sodium (FE~Na~), urinary potassium wxcretion (u-K), fractional excretion of potassium (FE~K~), urinary gamma fraction of the epithelial sodium channels excretion rate (u-ENaCγ) and urinary gamma fraction of ENaC corrected for creatinine (u-ENaCγ~CR~) Periods Baseline Infusion Post infusion p (GLM within) ---------------------------------- -------------- -------------- ------------------- -------------------- -------------------- ---------- u-Na (mmol/min) 0.9% NaCl 1.24(0.55) 1.83(0.56) 2.51(1.05) 2.78(0.99) 3.07(0.98)\* \<0.0001 3% NaCl 1.38(0.58) 1.86(0.75) 2.65(1.06) 2.94(1.44) 3.17(1.38)\* Glucose 1.23(0.51) 1.38(7.3) 0.94(0.43) 0.83(0.35) 0.72(0.35)\* p (GLM between) \<0.0001 p (ANOVA) 0.599 0.028 \<0.0001 \<0.0001 \<0.001 FE~Na~ 0.9% NaCl 1.26(0.53) 1.93(0.61) 2.35(0.72) 2.67(0.76) 2.80(0.75)\* \<0.0001 3% NaCl 1.44(0.61) 2.08(0.81) 2.66(1.07) 3.07(1.27) 3.01(1.26)\* Glucose 1.27(0.53) 1.36(0.64) 0.88(0.37) 0.81(0.37) 0.74(0.37)\* p (GLM between) \<0.0001 p (ANOVA) 0.465 0.002 \<0.0001 \<0.0001 \<0.0001 U-K (mmol/min) 0.9% NaCl 26.5(9.1) 22.5(7.6) 21.1(8.9) 22.7(12.2) 23.5(10.2)\\\* \<0.0001 3% NaCl 25.4(9.2) 17.3(7.3) 16.9(8.1) 20.1(10.5) 21.1(9.9)\\\* Glucose 24.9(8.7) 17.2(5.8) 10.8(3.7) 9.6(3.2) 11.23(4.4)\* p (GLM between) \<0.0001 p (ANOVA) 0.827 0.016 \<0.0001 \<0.0001 \<0.0001 FE~K~ 0.9% NaCl 28.2(9.9) 23.3(7.4) 20.4(7.8) 21.5(8.2) 21.5(7.9)\\\* \<0.0001 3% NaCl 26.3(9.0) 18.5(6.8) 17.0(7.5) 20.9(9.3) 20.0(8.9)\\\* Glucose 25.7(8.8) 16.7(5.5) 9.8(3.1) 9.3(3.4) 11.5(4.4)\* p (GLM between) \< 0.0001 p (ANOVA) 0.624 \<0.01 \<0.0001 \<0.0001 \<0.0001 u-ENaCγ (ng/min) 0.9% NaCl 351.2(155.0) 302.3(108.5) 341.2(123.0) \- 327.2(111.3) \<0.0001 3% NaCl 358.8(160.0) 375.9(173.7) 462.6(203.4) \- 435.9(168.7)\\ Glucose 338.3(150.0) 301.7(128.8) 335.2(185.1) \- 283.7(164.4)\\\* p (GLM between) 0.076 u-ENaCγ~CR~(ng/mmol) 0.9% NaCl 36.25(3.63) 32.98(3.50) 34.40(3.63) \- 32.56(3.57) \<0.001 3% NaCl 37.56(3.70) 43.29(3.58) 45.83(3.72) \- 44.46(3.65)\\\* Glucose 35.33(3.62) 32.78(3.50) 31.68(3.63) \- 31.71(3.57) p (GLM between) 0.091 Values are means with SD in brackets. General linear model (GLM) with repeated measures was performed for comparison within the group and intervention as between subjects factor. One way ANOVA was used for comparison of means between subjects when differences were found. Paired t-test was used for comparison within treatment group at baseline and third post infusion period \p \< 0.0001, \\p \<0.001, \\\ p \< 0.01. Sodium excretion, u-ENaCγ, u-Na, FE~Na~, u-K and FE~K~ ------------------------------------------------------ Table [3](#T3){ref-type="table"} shows the absolute values of u-Na, FE~Na~, u-K, FE~K~, u-ENaCγ~CR~ and u-ENaCγ excretion rate during the baseline period, the infusion period and the post infusion period. Infusion with 0.9% NaCl and 3% NaCl were accompanied by a significant and similar increase in u-Na and FE~Na~ that lasted throughout the experiment. There were no significant differences between 0.9% NaCl and 3% NaCl infusions. In contrast, U-Na and FE~Na~ decreased after glucose infusion. The relative changes in u-Na and FE~Na~ were significant lower after glucose compared to both saline infusions. U-K and FE~K~ decreased significantly after all three infusions, but with the greatest extend after glucose infusion. In the post infusion period (150--240 min) the excretion of potassium in urine increased slightly more after 3.0% NaCl than 0.9% NaCl, but did not reach baseline levels. U-ENaCγ~CR~ decreased slightly, but non-significantly during 0.9% NaCl and glucose infusions. A significant increase was seen in u-ENaCγ~CR~ in response to 3% NaCl (p \< 0.01) (Figure [1](#F1){ref-type="fig"}B) and the relative increase in u-ENaCγ~CR~ were significantly higher in response to 3% NaCl compared to 0.9% NaCl and glucose. Divided by gender the differences in u-ENaCγ~CR~ showed no statistical significant difference, although u-ENaCγ~CR~ tended to be higher in women due to the lower urine creatinine (data not shown). U-ENaCγ excretion rate followed the same pattern with regard to saline infusions, whereas a significantly lower u-ENaCγ excretion rate occurred after glucose infusion (Table [3](#T3){ref-type="table"}). Vasoactive hormones ------------------- PRC, Ang II and Aldo were suppressed to the same extent in all three parameters in response to 0.9% NaCl and 3% NaCl with no significant difference between interventions. There was a primary decrease during glucose infusion (90--150 min), but when infusion ceased values returned to baseline levels with no overall significant change (Figure [2](#F2){ref-type="fig"}). ![Effects of isotonic 0.9% saline (■), hypertonic 3% saline (♦) and isotonic glucose (●) on plasma renin (A), plasma angiontensin II (B) and plasma aldosterone (C) concentrations. Values are expressed as mean ± SEM. General linear model (GLM) with repeated measures within subjects was significant for all three variables. Paired t-test was used for comparison within treatment groups at postinfusion 240 min vs. basal. \* p \< 0.0001.](1471-2369-14-202-2){#F2} AVP did not change in response to 0.9% NaCl and glucose, but increased significantly after 3% NaCl with a maximum at 150 minutes and a steady fall during the post infusion period (Figure [1](#F1){ref-type="fig"}C). Blood pressure, pulse rate, GFR, p-Na, p-alb and p-osm ------------------------------------------------------ Table [4](#T4){ref-type="table"} shows the absolute values of systolic and diastolic blood pressure, pulse rate, GFR, plasma sodium and plasma albumin during the baseline period, the infusion period and the post infusion period. ###### Effect of 0.9% isotonic saline (0.9% NaCl), 3.0% hypertonic saline (3% NaCl) and 5% glucose (Glucose) on^51^Cr-EDTA-clearance, plasma sodium, plasma albumin, systolic blood pressure (SBP), diastolic blood pressure (DBP) and pulse rate in a randomized, placebo-controlled, crossover study of 23 healthy subjects Periods Baseline Infusion Post infusion p (GLM-within) ---------------------------------------------------------- -------------- -------------- ------------------- -------------------- ------------------- ---------- ^51^Cr-EDTA-clearance (ml/min/1.73 m^2^) 0.9% NaCl 98.1(10.1) 97.2(9.9) 100.0(12.4) 106.4(19.5) 106.1(13.3)\\ 0.002 3% NaCl 95.7(10.4) 93.0(11.4) 96.6(14.8) 97.0(14.8) 101.8(13.9)\\\* Glucose 96.9(10.3) 100.0 (9.9) 105.9(10.5) 104.0(13.9) 99.1(9.7) p (GLM between) 0.244 p-Sodium (mmol/l) 0.9% NaCl 138.1(2.1) 139.4(2.1) 139.6(2.4) 139.1(2.4) 138.7(2.3)\\ \<0.0001 3% NaCl 138.0(2.0) 142.5(1.8) 141.7(1.7) 141.1(2.1) 139.7(2.0)\* Glucose 138.1(2.0) 128.7(3.7) 134.1(3.6) 135.7(3.7) 136.9(2.9)\* p (GLM between) \<0.0001 p (ANOVA) 0.94 \<0.0001 \<0.0001 \<0.0001 =0.001 p-albumin (g/L) 0.9% NaCl 40.3(2.6) 40.1(2.4) 35.3(2.0) 35.8(2.1) 36.0(2.2)\* \<0.0001 3% NaCl 40.1(2.6) 39.8(2.6) 35.2(2.2) 36.0(2.3) 36.2(2.7)\* Glucose 40.9(2.2) 40.3(2.8) 36.1(2.5) 39.1(2.7) 39.7(2.9)\* p (GLM between) \<0.05 p (ANOVA) 0.510 0.765 0.328 \<0.0001 \<0.0001 SBP (mmHg) 0.9% NaCl 114.5(10.6) 117.7(10.4) 116.6(11.0) 116.9(11.6) 117.7(11.5) 0.439 3% NaCl 114.0(9.5) 117.9(9.1) 117.2(9.2) 115.5(10.0) 116.4(9.7) Glucose 115.5(8.4) 118.2(9.8) 117.7(9.1) 115.5(8.2) 117.0(9.1) p (GLM between) 0.975 DBP (mmHg) 0.9% NaCl 62.9(4.3) 64.0(6.1) 62.2(5.7) 63.3(5.6) 63.4(5.1) 0.038 3% NaCl 62.9(4.2) 62.1(5.0) 61.0(5.4) 61.4(4.9) 62.9(4.9) Glucose 64.4(4.2) 65.8(5.6) 65.3(6.3) 65.5(5.3) 63.6(7.5) p (GLM between) 0.12 Pulse rate 0.9% NaCl 54.1(11.0) 57.2(11.9) 56.4(11.6) 58.0(12.1) 57.4(11.9)\* \<0.0001 3% NaCl 54.0(10.5) 58.2(11.2) 57.0(11.2) 56.9(11.3) 57.7(11.5)\* Glucose 54.7(10.5) 57.4(12.5) 57.1(12.0) 60.3(11.7) 61.3(12.3)\* p (GLM between) 0.917 Values are mean with SD in brackets. General linear model (GLM) with repeated measures was performed for comparison within the group and intervention as between subjects factor. One way ANOVA was performed for comparison of means when differences were found between intervensions. Paired t-test was used for comparison within treatment groups at baseline and post infusion period 210 -- 240 minutes. \p \< 0.0001; \\p \< 0.001; \\\p \< 0.05. Systolic BP was the same after all three infusions. There was a small difference in diastolic BP pattern during the examination day, but the changes were very small and might be by chance. During the examination day pulse rate increased slightly in response to 0.9% NaCl and 3% NaCl, while the heart rate increased to a higher extent in response to the glucose infusion (Table [4](#T4){ref-type="table"}). The increase in pulse rate did not differ between 0.9% saline and 3% saline, but there was a difference in the relative increase in pulse rate between saline and glucose infusion (p \< 0.01). GFR increased slightly, although significantly, on the examination day. However the changes were very small (Table [4](#T4){ref-type="table"}). P-Na increased in response to both 0.9% NaCl and 3% NaCl with maximum after 150 minutes. In response to glucose p-Na decreased markedly after 150 minutes to a mean of 128.7 mmol/l (Table [4](#T4){ref-type="table"}). The increase was higher after 3% NaCl compared to 0.9% NaCl and accordingly the changes after glucose were lower compared to saline. P-alb decreased significantly in response to 0.9%, 3% NaCl and glucose infusions. The decline was significantly lower and sustained after both saline infusions compared to glucose, which is related to an expected increase in extracellular fluid. P-osm increased slightly during 0.9% NaCl infusion, but remained unchanged at the end of the examination day. P-osm increased significantly in response to 3% NaCl, with a maximum of 293 mosm/kg and decreased significantly after glucose to 280 mosm/kg at 150 min. The changes in p-osm indicated that isotonic, hypertonic and hypotonic conditions were established (Figure [1](#F1){ref-type="fig"}D). Fluid, sodium balance and body weight during the examination days ----------------------------------------------------------------- The average fluid administered intravenous was 1749 ml of 0.9% NaCl (SD 270), 555 ml of 3% NaCl (SD 90) and 1736 ml of glucose (SD 282). The cumulative water input was 3674 ml (SD 270), 2480 ml (SD 90) and 3661 ml (SD 282) respectively, as participants drank an additional 1925 ml of tap water each examination day. During the examination days the average total urine output was 1858 ml (SD 246) in subjects who received 0.9% NaCl, 984 ml (SD 202) in subjects who received 3% NaCl and 2682 ml (SD 351) in subjects who received glucose. The fraction of water excreted after 240 min was 51% when 0.9% NaCl was infused, 40% when 3% NaCl was infused and 73% when glucose was infused. The total amount of sodium infused was 269 mmol (SD 42) of 0.9% NaCl and 285 mmol (SD 46) of 3.0% NaCl. The cumulative sodium output at 240 min was 50 mmol (SD 16) after 0.9% NaCl, 54 mmol (SD 21) after 3% NaCl and 21 mmol (SD 9) after glucose. The fraction of sodium excreted after 240 min was 19% after both 0.9% and 3% NaCl infusions. This was accompanied by a significant increased bodyweight in response to 0.9% NaCl from 73.2 kg (SD 11.3) at baseline to 74.3 kg (SD 11.4) at the end of the study day \[+1.1 kg (SD 0.39); p \< 0.0001\], in response to 3% NaCl from 73.3 kg (SD 11.6) at baseline to 74.1 kg (SD 11.7) at the end of the study day \[+0.8 kg (SD 0.39); p \< 0.0001\] and to a smaller extent in response to glucose from 72.8 kg (SD 11.8) at baseline to 73.1 kg (SD12.0) at the end of the study day \[+0.3 kg (SD 0.5); p \<0.05\]. Discussions =========== In the present study we examined the effect of an acute intravenous volume load of 0.9% saline, 3% saline and isotonic glucose infusions on u-AQP2 and u-ENaCγ in a randomized, crossover study of healthy subjects. The purpose was to evaluate the transport activity via the aquaporin 2 water channels and the epithelial sodium channels in the principal cells in the distal part of the nephron. During infusion and in the period immediately after, adaptive physiological changes take place in renal function and vasoactive hormones. Thus, the main changes in the effect variables could be expected to occur after the infusion. In the present study, we paid special attention to changes in the effect variables in the last post infusion period (Post infusion 210--240), i.e. 60--90 minutes after infusion had ceased. During this period, u-AQP2 increased after hypertonic and isotonic saline infusion and decreased after glucose infusion. At the same time, u-ENaCγ increased after hypertonic saline infusion and remained unchanged after isotonic saline and glucose infusion. U-AQP2 after infusion with hypotonic and isotonic saline and isotonic glucose ----------------------------------------------------------------------------- Aquaporin-2 (AQP2) is located in the collecting duct principal cells \[[@B11]\] and is expressed in the apical plasma membrane \[[@B12]\]. Vasopressin (AVP) regulates AQP2 by binding to V2 receptors in the basolateral membrane, \[[@B11],[@B13]\]. Short term exposure to AVP causes trafficking and insertion of the intracellular vesicles, containing AQP2, to the apical membrane and increases the water permeability and absorption \[[@B11]-[@B14]\]. Long-term regulation occurs over a period of hours to days, and is caused by AVP-regulated gene transcription resulting in an increase in AQP2 whole-cell abundance \[[@B4],[@B13]\]. Experiments in rats showed that infusion of dDAVP increased u-AQP2 \[[@B15]\]. This is consistent with the view that increased delivery of AQP2 channels to the apical membrane results in increased excretion of AQP2 after stimulation with AVP \[[@B6],[@B15]-[@B21]\]. Approximately 3% of AQP2 in the collecting duct are excreted into urine \[[@B20]\], but the underlying mechanisms are unknown. Volume expansion with 3% hypertonic saline increases plasma osmolarity beyond the threshold of the hypothalamic osmoreceptors, triggering release of AVP and a subsequent increase in u-AQP2. Saito et al found a significant relationship between urinary excretion of AQP2 and p-AVP in healthy subjects after 5% hypertonic saline infusion \[[@B17]\]. Pedersen et al found a positive correlation between u-AQP2 and p-AVP during 24 h of water deprivation and after 3% hypertonic saline infusion \[[@B6]\]. Thus, previous studies in humans have demonstrated that the activity of the AQP2 water channels can be determined by measuring u-AQP2 \[[@B6]-[@B8],[@B17],[@B18]\]. Surprisingly, Baumgartner et al found no change in u-AQP2 after infusion of 2.5% NaCl in healthy volunteers, despite a significant rise in both urine osmolarity and AVP \[[@B22]\]. However, the oral water load was 3--4 times higher prior to infusion compared to our study. Thus, the large water load before infusion might have overruled the stimulatory effects of hypertonic saline. As expected, our study showed that u-AQP2 increased after 3% NaCl with a corresponding rise in urine osmolarity and a reduction in C~H2O~. Thus, our findings indicate an increased water reabsorption via the aquaporin-2 water channels in the distal tubules. Prior to the increase in u-AQP2, there was an abrupt rise in p-osm and p-AVP induced by the hypertonic saline infusion. Animal studies have shown that hypertonicity can cause an up regulation of AQP2 expression in the apical membrane comparable with that achieved by AVP alone \[[@B23],[@B24]\]. It cannot be excluded that this might play an active part in the increased excretion of u-AQP2. Most likely, the increased water reabsorption was mediated by an increase in p-AVP. U-AQP2 continued to rise throughout the examination day, suggesting that AQP2 channels remained inserted and active in the apical membrane due to actions of elevated p-AVP. Infusion of isotonic saline depresses the fractional water and salt reabsorption in the proximal tubules in animals \[[@B25]\]. In the present study, infusion with 0.9% NaCl caused the same response in u-AQP2, u-osm and C~H2O~ as 3% NaCl infusion, albeit to a lesser extent. There was a small rise in p-osm to a maximum level of 286 mosmol/kg corresponding to a rise of 0.5%. This increase is below the osmoreceptor threshold, and we did not see, nor expect, any significant change in p-AVP. Therefore AVP could not be the main regulator of AQP2 during 0.9% NaCl. Most likely, the increased water transport via AQP2 is a compensatory phenomenon to antagonize a decrease in the renal water absorption in the proximal tubules, which occurs after isotonic volume expansion. The mechanism might be due to an increased activity in the natriuretic peptide system \[[@B26],[@B27]\]. Infusion of 5% glucose causes a volume expansion distributed throughout the fluid phases in the body with only a very small increase in plasma volume. This is illustrated in the measurements of plasma albumin, where concentrations at 240 min were virtually equal to baseline (Table [4](#T4){ref-type="table"}), indicating no change in extracellular fluid. According to our knowledge, no study has measured u-AQP2 after glucose infusion. A study of healthy subjects showed, that after an oral water load of 20 mL/kg for 15 minutes (mean intake 1605 ml) u-AQP2 decreased 17% after 210 minutes \[[@B21]\]. In a recent study, subjects received an oral water load of 20 ml/kg for 15 min (mean intake 1389 ml) with a subsequent 27% decrease in u-AQP2 after 240 minutes \[[@B28]\]. Both plasma osmolarity and p-AVP decreased. Thus, it has been showed that u-AQP2 is reduced during water diuresis after oral water intake \[[@B21],[@B28],[@B29]\]. In our study, subjects received a mean of 1736 ml glucose IV. In the last post infusion period the expected aquaretic response occurred, with a 16% decrease in u-AQP2~cr~, a decrease in u-osm and an increase in UO and C~H2O~. Plasma osmolarity decreased from 285 mosm/kg to 280 m0sm/kg, i.e. a 2% decline, but with no accompanying reduction in p-AVP. Our findings indicate a reduced reabsorption of water via the aquaporin-2 water channels in the distal tubules after isotonic glucose infusion. The lack of change in p-AVP could firstly be explained by the fact that the subjects had received 1225 ml of oral water load prior to the infusion start, and this could have suppressed AVP in baseline periods beforehand. Secondly, the measurements of p-AVP concentration may not be sensitive enough to detect a small decrease. The recent discovered peptide Apelin, may also play a role. Apelin is colozalized with AVP in magnocellular neurons of hypothalamus \[[@B30],[@B31]\]. In healthy male volunteers decreasing plasma osmolarity by waterloading reduced p-AVP modestly but p-Apelin increased rapidly \[[@B32]\]. Apelin regulation is opposite to that of AVP and data suggests that Apelin, like AVP may participate in regulating water homeostasis \[[@B32]\]. We did not measure p-Apelin, but it could have been of interest to investigate plasma apelin in parallel with p-AVP in conditions of different volume expansions. Thus, in the last post-infusion period, u-AQP2 increased approximately to the same extent after hypertonic and isotonic saline infusions, whereas a marked fall was seen after isotonic glucose infusion. A possible explanation for the delay in changes of u-AQP2 could be that it takes few minutes for changes in AVP to act on the principal cell, either by insertion or removal of AQP2 from the apical membrane, but it takes several minutes before the effect is seen in the excretion of u-AQP2 in the urine. U-ENaCγ after infusion with hypotonic and isotonic saline and isotonic glucose ------------------------------------------------------------------------------ Sodium transport across the collecting duct occurs through the epithelial sodium channel and is responsible for reabsorption of 3--5 % of filtered sodium \[[@B33]\]. ENaC is composed of three distinct subunits: α, β and γ and localized at the apical plasma membrane of principal cells \[[@B34],[@B35]\]. ENaC is a target of aldosterone that acts on the mineralocorticoid receptor. Aldosterone increases sodium transport by redistributing ENaC subunits from intracellular locations to the apical membrane as well as altering gene transcription \[[@B33],[@B36],[@B37]\]. While the action of aldosterone occurs over hours or days, another synergistically pathway involves AVP \[[@B35],[@B37]-[@B41]\]. In the cortical collecting ducts in rats, AVP binds to the V2 receptors, stimulates cAMP and increases sodium reabsorption by promoting trafficking and insertion of ENaC into the apical membrane inducing a rapid change in channel activity \[[@B34],[@B35],[@B40]\]. Recent studies in humans demonstrated that AVP, via V2 receptors, stimulates ENaC mediated sodium reabsorption across principal cells \[[@B39],[@B42],[@B43]\]. Fractions of ENaC are normally excreted into the urine. The amount of ENaC-fractions is supposed to reflect the activity of the sodium transport via the epithelial sodium channels just as u-AQP2 reflects the functional status of the AQP2 water channels. Recently, our group introduced a new method to evaluate sodium reabsorption in the principal cells in the distal tubules. Lauridsen et al demonstrated a significant correlation between changes in urinary sodium excretion and changes in urinary excretion of the beta fraction (u-ENaC~β~) in healthy humans \[[@B44],[@B45]\]. Apparently, u-ENaC~β~ can be used as a biomarker for the transport of sodium via ENaC. In the present study, we measured the gamma fraction of the protein of the epithelial sodium channels to evaluate the up-and down regulation of γ-ENaC expression and sodium transport via ENaC as previously reported from our group \[[@B9],[@B46]\]. The sodium-chloride symporter (NCC) in the distal convoluted tubules (DCT) is as another major sodium reabsorbing pathway. Sodium reabsorption in DCT is essential to define the amount of sodium delivery to the principal cells in the collecting duct. It is widely accepted that NCC is regulated by Ang II and aldosterone \[[@B47],[@B48]\]. Studies have also shown that high AVP increase phosphorylation of NCC and presumably result in greater sodium reabsorption \[[@B49]\]. Experimental animal-studies have demonstrated that isotonic and hypertonic saline IV reduced reabsorption of sodium in the proximal tubules, and thereby increased the amount of sodium in the urine \[[@B25],[@B50]\]. Andersen LJ et al studied the effects of hypertonic and isotonic saline in healthy subjects on a controlled diet. The subjects received an IV sodium load of either 25 ml/kg isotonic saline or 4.5 ml/kg 3% hypertonic saline for 90 minutes \[[@B51]\]. Urinary sodium excretion increased in both isotonic and hypertonic saline, with natriuresis after hypertonic saline exceeding that after isotonic saline. Plasma sodium and plasma osmolarity increased substantially after hypertonic saline, as did p-AVP. Our study showed that 3% NaCl infusion increased u-ENaCγ, FE~Na~, p-Osm, p-Na and p-AVP. Thus, our findings reflect an increased sodium reabsorption via ENaC in the principal cells, and furthermore confirmed the results by Andersen et al \[[@B51]\]. The increased u-ENaCγ could partly be explained by a considerable decrease in the renal sodium absorption proximal in the nephron, compensated for and adjusted by an increase in absorption in the distal part. However, the rise in p-AVP seen immediately after 3% NaCl infusion could also indicate that the increased u-ENaCγ is caused by actions of AVP. An increased sodium movement from the lumen to the cell via ENaC would theoretically drive potassium secretion through the ROMK channels \[[@B52],[@B53]\]. Surprisingly we measured a fall in excretion of potassium in the urine. This could argue against a major role of ENaC mediated sodium transport. If NCC increased sodium reabsorption, both to compensate for a decrease in proximal reabsorption and due to high p-AVP, then less sodium would to be transported by ENaC and thus potassium secretion would not take place. A possible role of NCC after infusion with hypertonic saline is purely speculative as we did not measure the activity of NCC. Perhaps we did not see the positive effect on potassium secretion within our time limits. However, the potassium transport is complex and factors modulating potassium transport, such as altered tubular flow and aldosteron, are many. After volume expansion with isotonic saline the oncotic pressure is slightly reduced, which leads to an immediately increase in GFR and smaller reabsorption of water in the proximal tubule. We measured a small increase in GFR and UO output. Sodium excretion increased, but u-ENaCγ, p-Na, p-osm and p-AVP remained unchanged, thus the findings were as we expected. Regarding NCC, one would not expect any change in NCC mediated sodium reabsorption during isotonic saline. No study has ever evaluated u-ENaCγ during water diuresis. In our study, we measured a trend towards a reduction in u-ENaCγ after glucose infusion reflecting a small reduction of sodium reabsorption via ENaC in the principal cell. As previously mentioned, we measured a 2% fall in p-osmolality after glucose infusion, which theoretically should trigger a decrease in AVP. We did not detect a fall in p-AVP, presumably due to a low p-AVP caused by oral water loading in advance or the fact that the measurements of p-AVP concentration may not be sensitive enough to detect small changes. It could be hypothesized that the decrease in u-ENaCγ could be due to a lack of AVP binding to V2 receptors in the basolateral membrane of the principal cell. Lack of AVP stimuli leads to an increased endocytosis of ENaC channels from the membrane surface into recycling vesicles, there by decreasing reabsorption of sodium \[[@B54],[@B55]\]. Thus, in the last post infusion period u-ENaCγ increased markedly after hypertonic saline infusion, was approximately at the same level after isotonic saline and decreased or tended to decrease in response to glucose infusion. The rise in p-osm and p-AVP was seen immediately after 3% NaCl infusion stopped. The delay and constant level of u-ENaCγ after hypertonic saline could be explained by the fact that it takes few minutes to increase trafficking of intracellular depots of ENaC channels into the apical membrane but several minutes to excrete ENaC into the urine after stimulation with AVP. Vasoactive hormones ------------------- In addition to AVP, the renin-angiotensin-aldosterone system (RAAS) is a key regulator of renal sodium excretion and thereby of body fluid volume. It is well known that sodium depletion activates and that chronic sodium load reduces the RAAS \[[@B56]\]. In vitro and in vivo studies have shown that aldosterone stimulates the mineralocorticoid receptor to an increased transcription of genes-encoding proteins involved in sodium transport i.e. ENaC and Na,K-ATPase \[[@B57]\]. Numerous studies of changes in blood volume have demonstrated that acute changes are associated with inverse adjustments of the renin-angiotensin-aldosterone system \[[@B21],[@B27],[@B51],[@B58]\]. In the present study, volume expansion with 3% and 0.9% saline resulted in a similar and significant reduction in PRC, p-AngII and Aldo consistent with an increase in extracellular volume. This is in agreement with previous studies \[[@B51],[@B58]\]. After glucose infusion, we measured no significant change in PRC, p-AngII or p-Aldo. This was expected, as glucose infusion does not cause any marked change in extracellular volume. Our study was not designed to allow any regulatory effects of aldosterone as the action of aldosterone occurs over hours or days. Therefor other factors must be implicated in the regulation of ENaC. Strengths and limitations ------------------------- The major strength of this study was the design as a randomized crossover study with a homogenous group of healthy young men and women. The test conditions were very well defined regarding diet, sodium and fluid intake. Thus, the results are not confounded by different salt or water balance. This study explored only the acute effects of volume expansion. No doubt, we could have gained further information regarding the long-term effects of volume expansion and the urinary excretion of AQP2 and ENaCγ if the post infusion period had been longer. In addition, the study was not placebo-controlled, by means of infusion with a negligible amount of 0.9% saline. This could have distinguished the effects of volume expansion from the overall variability of water and salt reabsorption. In this study it was not possible to perform ANP measurements. It could have made a positive contribution to our results. Conclusions =========== In conclusion volume expansion with 3% and 0.9% saline clearly increased u-AQP2, while isotonic glucose decreased u-AQP2. Infusion of hypertonic saline increased u-ENaCγ, whereas u-ENaCγ was not significantly changed after isotonic saline and decreased or tended to decrease after glucose. Thus, the transport of water and sodium changed, both via the aquaporin 2 water channels and the epithelial sodium channels, during all three types of volume expansion in order to regulate and maintain water- and sodium homeostasis in the body. Changes in the renin-angiotensin-aldosterone system did not seem to bear a causal relationship with the changes in u-AQP2 or u-ENaCγ. Competing interests =================== The authors declare that they have no competing interests. The authors alone are responsible for the content and writing of the paper. Authors' contributions ====================== All authors have contributed to the manuscript. JMJ, FHM and EBP designed the project. JMJ and FHM performed the experiments and statistical analyzes. JMJ, FHM, JNB, SN and EBP wrote and edited the manuscript. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2369/14/202/prepub> Acknowledgements ================ The authors greatly acknowledge the skilful assistance of our laboratory technicians: Lisbeth Mikkelsen, Anne Mette Ravn, Kirsten Nygaard and Henriette Vorup Simonsen. Funding ======= The study was supported by grants from Region Midt's Research Foundation for Health Science, The Lundbeck Foundation and Helen and Ejnar Bjornows Foundation.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec0005} =============== The mediastinum consists of anatomical structures that can give rise to both benign and malignant tumors, or can be the seat of metastases of malignancies that arise elsewhere in the body. The diagnostic approach to patients with mediastinal pathology is not always simple and an improper diagnostic work-up can lead to significant diagnosis delay. This work is reported in accordance with the SCARE criteria \[[@bib0005]\]. 2. Case presentation {#sec0010} ==================== A 34 year-old woman was admitted to the Emergency Department (ED) complaining of thoracic pain lasting from 6 days, dyspnoea, fever, and productive cough. She had a history of smoke and a major thoracic trauma due to a car accident that occurred one year before. A chest X-ray performed on that occasion had ruled out a chest wall injury. At admission to the ED her physical examination showed a painful swelling over the sternum's upper left margin, which had become evident 4 months earlier. A chest X-ray revealed a homogeneous opacity behind the sternum. Blood tests were irrelevant, except for a mildly elevated C-reactive protein (CRP 3.62 mg/dl). To better characterize the lesion, a chest Computer Tomography (CT) scan was performed. This evaluation showed a retrosternal oval lesion (5.5 x 4 cm) with an intra- and extra-thoracic growth and a multifocal lymphadenopathy ranging between 7 mm and 4 cm in the prevascular area, aortopulmonary window, and right paratracheal station ([Fig. 1](#fig0005){ref-type="fig"}).Fig. 1Preoperative CT scans showing the mediastinal mass and the involvement of the sternum.Fig. 1 The new findings, together with the young age of the patient, sped up our investigations, and we performed an ultrasound guide biopsy of the retrosternal mass. At histology, the specimen consisted of "fibroconnectival tissue with necrotic areas, lymphocytic, eosinophilic, and granulocytic infiltrate". No significant results emerged from cultures. All the additional imaging procedures, including abdominal ultrasound and skeletal scintigraphy, proved inconclusive. As anti-inflammation therapy was ineffective on dyspnoea and chest pain, a new CT scan was performed 30 days later. The size and characteristics of the retrosternal lesion had not changed, but the scan revealed a new paracardiac lesion showing similar features, a multifocal bone involvement with osteolytic and osteosclerotic areas and multiple, bilateral, pulmonary nodules, suspicious for malignancy. A bilateral, axillary lymphadenopathy was also described. The retrosternal lesion maintained the same size and features, but there was a deeper bone structural alteration. In the meantime the patient complained of the persistence of pain and dyspnoea, so that a Positron Emission Tomography (PET)/CT scan was performed. It showed a high metabolic activity lesion in the mediastinum, involving the sternum. These data suggested a malignant aetiology for the mediastinal mass, and therefore a surgical biopsy was performed. Pathological analysis of the surgical specimen showed a lymphoid lesion surrounded by dense fibrotic tissue. Within the lesion a mixed infiltrate was evidenced, composed of small lymphocytes, histiocytes, plasmacells, and several eosinophils. Furthermore, several mono- or plurinucleated lymphoid cells were present, reminiscent of Reed-Sternberg cells, sometimes surrounded by a clear halo. At immunohistochemistry these cells showed cytoplasmic (often membranous) positivity for CD30 ([Fig. 2](#fig0010){ref-type="fig"}), and negativity for CLA, CK, CD79 and CD3. The final diagnosis was Hodgkin's lymphoma.Fig. 2Pathological images of the lesion: A) lymphoid nodule surrounded by fibrotic tissue (H&E, 20x), B) intralesional infiltrate composed of small lymphocytes, histiocytes, plasmacells, and numerous eosinophils (H&E, 200x), C) several mono- or plurinucleated lymphoid cells are evident, reminiscents of Reed-Sternberg cells, sometimes surrounded by a clear halo (H&E, 400x), D) these cells showed cytoplasmic (often membranous) positivity for CD30 (CD30, 600x).Fig. 2 The patient underwent chemotherapy in the Haematology-Oncology Department. The chest CT follow-up showed a great improvement: there was a reduction of the size of the primary lesion and the lymphadenopathy, and a great increase in the degree of mineralization of the sternum. 3. Discussion {#sec0015} ============= A sternal swelling associated to thoracic pain and fever can be due to several conditions, such as neoplasms or inflammatory and infective processes. In the present case, the gender, age, and clinical history of the patient, as well as the inflammatory pattern observed, could suggest an inflammatory condition, namely the Tietze's syndrome \[[@bib0010]\]. The Tietze's syndrome was firstly described in 1921 as a benign, non-suppurative, painful swelling of the superior chondrosternal joints \[[@bib0010]\]. It is a quite common pathology \[[@bib0015]\], especially amongst adult women, and it may also follow a thoracic trauma \[[@bib0020]\]. The clinical manifestation is that of a generalized inflammatory state with a huge chondrosternal joints' oedema \[[@bib0025]\]. In particular, the personal history of a thoracic trauma due to a car accident could have explained the sub-acute onset of the symptoms. In fact, the oedema had appeared four months before the ED access. Moreover, the radiological aspects of an aggressive lesion, and the presence of an important mediastinal lymphadenopathy, confirmed by the PET/CT scan, ruled out the Tietze\'s syndrome \[[@bib0030]\]. A malignant tumour with high cellularity \[[@bib0035]\] or a tuberculous osteomyelitis \[[@bib0040],[@bib0045]\] with lung involvement was also suspected. At first, any infectious etiology was excluded, with negative microbiological tests and cultures, including those for Mycobacteria Tuberculosis. Unfortunately, the ultrasound guide biopsies were inconclusive in relation to the clinical development, so the surgical biopsy became mandatory \[[@bib0050]\]. Hodgkin's lymphoma (HL), formerly called Hodgkin\'s disease, arises from germinal centre or post-germinal centre B cells \[[@bib0055]\]. HL has a unique cellular composition, containing a minority of neoplastic cells (Reed-Sternberg cells and their variants) in an important acute or chronic inflammatory background \[[@bib0055]\]. Because of the morphological characteristics of HL the diagnosis is not always straight forward \[[@bib0055],[@bib0060]\]. Very few neoplastic diagnostic cells (Reed Sternberg cells) \[[@bib0065],[@bib0070]\] are present within inflammatory tissues containing a variable number of small lymphocytes, eosinophils, neutrophils, macrophages, plasma cells, fibroblasts, and collagen fibers. This case points out how challenging the diagnosis of Hodgkin's disease can be, and the impact that a diagnostic delay can have in the prognosis \[[@bib0075],[@bib0080]\] of a life-threatening condition. The acute and atypical symptoms (pain, dyspnoea, sternal oedema) misled us, letting to the suspicious of a benign disease. Therefore, the correct diagnosis of Hodgkin's lymphoma was delayed as a result. Fortunately, in this case, the response to the specific chemotherapy has been good despite the delay in treatment. 4. Conclusions {#sec0020} ============== In conclusion, mediastinal masses with an aggressive behaviour, confirmed at the chest CT scan or PET CT, should always be considered to be potentially malignant even if needle biopsy proves negative for neoplastic cells. Surgical biopsy can be the only way to correctly diagnose the pathological process, especially in the case of Hodgkin's lymphoma in which few diagnostic cells (Reed-Sternberg cells) are generally embedded in an abundant inflammatory background tissue. Conflicts of interest {#sec0025} ===================== The authors have no conflicts of interest to declare Funding {#sec0030} ======= This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Ethical approval {#sec0035} ================ This study is exempt from ethical approval in our institution. Informed consent {#sec0040} ================ The patient has given informed consent for publication of this manuscript. Author contribution {#sec0045} =================== Alessandro Fois wrote the paper. The other authors collected and analysed data and contributed to the final drafting of the manuscript. Registration of research studies {#sec0050} ================================ Our study does not require registration. Guarantor {#sec0055} ========= Dr. Alessandro Giuseppe Fois.	{ "pile_set_name": "PubMed Central" }
GENOME ANNOUNCEMENT {#s1} =================== Microcystis aeruginosa is one of the most harmful bloom-forming cyanobacteria in freshwater ([@B1]). It is a unicellular colony-forming cyanobacterium distributed worldwide in eutrophic freshwater environments. During summer, Microcystis blooms result in serious environmental problems, such as bad odor releases, bottom-layer anoxia, and production of hepatotoxic cyanotoxins called microcystins. Microcystins are the only cyanotoxins for which the World Health Organization has set drinking and recreational water standards ([@B2]). M. aeruginosa NIES-87 was isolated from Lake Kasumigaura, Japan. This strain is available from the Microbial Culture Collection at the National Institute for Environmental Studies, Japan (<http://mcc.nies.go.jp>). It is an axenic culture and does not produce microcystin; however, it releases a bad odor, the origin of which is unknown. This strain also produces an antialgal peptide kasumigamide ([@B3]). A recent multilocus phylogenetic study suggested that M. aeruginosa is genetically divided into at least eight clades (groups A to G and X) ([@B4]). In that study, the phylogenetic position of NIES-87 was not well resolved. Tominaga et al. ([@B5]) reported that this strain harbors two plasmids with sizes of 2.3 and 5 kb. This strain has been widely used in various studies; however, to date, whole-genome sequencing of the strain has not been conducted. Here, we report the draft genome sequence of M. aeruginosa NIES-87. This information would be useful for unveiling genome evolution in M. aeruginosa. DNA extraction was performed using 10 ml of NIES-87 axenic culture using the DNeasy plant minikit (Qiagen), following the manufacturer's instructions. Subsequently, DNA was fragmented to approximately 550 bp using Covaris M220. Genomic libraries of paired ends and mate pairs were constructed using the TruSeq Nano DNA library prep kit (Illumina) and the Nextera mate pair library preparation kit (Illumina), respectively. Whole-genome sequencing was performed using the MiSeq platform (Illumina) employing the 600-cycle MiSeq version 3 reagent kit. The resulting paired-end and mate pair reads were 1,576,274 and 1,403,390 in number, respectively. Low-quality reads/bases were filtered using Trimmomatic version 0.36 ([@B6]), and de novo assembly was performed using SPAdes version 3.7.1 ([@B7]). The resulting genome comprised 206 contigs of 4,947,287 bp. The average genome coverage of the paired-end reads was 95.3×. The maximum contig length was 652,566 bp. Genome annotation was performed using the DDBJ Fast Annotation and Submission Tool (DFAST) ([@B8]). The genome comprised 4,355 predicted protein-coding sequences, one set of rRNAs, and 41 tRNAs. The G+C content of the genome was 42.9%. The genome was scanned using antiSMASH ([@B9]), and microviridin and other secondary metabolite gene clusters were predicted. The 16S rRNA sequence of NIES-87 was subjected to a BLAST search, and the top hits obtained were those of M. aeruginosa NIES-2481 and NIES-2549 (99.8% identity each). We used OrthoFinder ([@B10]) to identify orthologous genes among the three genomes, and 56.9% and 57.1% of the genes in NIES-2481 and NIES-2549, respectively, were identified as orthologous to those of NIES-87. Accession number(s). {#s1.1} -------------------- This whole-genome shotgun project has been deposited in GenBank under the accession no. [BFAC01000001](https://www.ncbi.nlm.nih.gov/nuccore/BFAC01000001) to [BFAC01000206](https://www.ncbi.nlm.nih.gov/nuccore/BFAC01000206). Citation Yamaguchi H, Suzuki S, Kawachi M. 2018. Draft genome sequence of Microcystis aeruginosa NIES-87, a bloom-forming cyanobacterium from Lake Kasumigaura, Japan. Genome Announc 6:e01596-17. <https://doi.org/10.1128/genomeA.01596-17>. We thank Nobuyoshi Nakajima for technical assistance. This study was supported by the Japan Agency for Medical Research and Development (AMED).	{ "pile_set_name": "PubMed Central" }
Abbreviations used in this paper: DCT, distal convoluted tubule; DVF, divalent-free; TRP, transient receptor potential; WMN, wortmannin. INTRODUCTION ============ Transient receptor potential (TRP) channels are widespread and play many important functions, ranging from thermal, tactile, taste, osmolar and fluid flow sensing, and embryo development to epithelial Ca^2+^ and Mg^2+^ transport ([@bib12]; [@bib22]; [@bib2]). They are classified into TRPC, TRPV, TRPM, TRPP, TRPML, TRPA, and TRPN subfamilies ([@bib12]; [@bib22]; [@bib2]). The TRPV subfamily is named after the first mammalian member of the subfamily, vanilloid receptor 1 ([@bib22]; [@bib2]). TRPV5 and TRPV6 are highly Ca^2+^-selective TRP channels that mediate trans-epithelial Ca^2+^ transport in kidney and intestine ([@bib12]). Intracellular Mg^2+^ is a cofactor for many enzymes and controls inward rectification of ion channels by causing voltage-dependent block of outward currents ([@bib33]; [@bib9]). Recently, intracellular Mg^2+^ has also been reported to regulate the activity of several TRP channels including TRPM6, TRPM7, and TRPV6 via a mechanism different from voltage-dependent block ([@bib23]; [@bib34]; [@bib42], [@bib43]). The intracellular free Mg^2+^ concentration in most mammalian cells is between 0.5 and 1 mM and changes only slightly in response to physiological stimuli ([@bib33]). TRPM7 is permeable to Mg^2+^ ([@bib37]). It is believed that Mg^2+^ ions entering cells through TRPM7 feedback on the channel to regulate its activity ([@bib37]). TRPV6 does not conduct Mg^2+^ in physiological conditions. The physiological role of intracellular Mg^2+^ regulation of TRPV6 is unknown. PIP~2~ directly regulates inward rectifying K^+^ and other channels ([@bib11]). How membrane lipids alter channel function is an area of intense research interests. For many ion channels, direct interaction with PIP~2~ stabilizes channels in certain conformations ([@bib11]). By stabilization in one conformation, PIP~2~ modulates the responses of ion channels to regulators, such as GTP binding proteins, intracellular pH, Na^+^, ATP, etc. ([@bib14]; [@bib38]; [@bib39]; [@bib17]). In this report, we show that intracellular Mg^2+^ causes a fast voltage-dependent block and a slower inhibition of TRPV5. We further show that PIP~2~ activates TRPV5 and decreases sensitivity of the channel to the Mg^2+^-induced slow inhibition. Hydrolysis of PIP~2~ via receptor activation of phospholipase C increases the sensitivity of TRPV5 to the Mg^2+^-induced slow inhibition. These results provide a novel mechanism for regulation of TRP channels by PLC-activating hormones by increasing the sensitivity to intracellular Mg^2+^. MATERIALS AND METHODS ===================== Molecular Biology and Cell Culture ---------------------------------- Nucleotide coding sequence of cDNAs for rabbit TRPV5 in pCDNA3 mammalian expression vector ([@bib47]) was used as template for site-directed mutagenesis using a commercial mutagenesis kit (QuickChange; Stratagene). Mutations were confirmed by sequencing. CHO-K1 clone (from American Type Culture Collection) were cultured in F12-K medium (GIBCO-BRL) containing 10% FCS. Cells (at ∼50% confluence) were cotransfected with cDNA for pEGFP (1 μg) plus cDNAs for wild-type or mutant TRPV5 (10 μg per 35 mM dish for cell-attached and inside-out recording; 3 μg per dish for whole-cell recording) using lipofectamine-plus transfection kits (GIBCO-BRL) as previously described ([@bib47], [@bib48]). About 24--48 h after transfection, cells were dissociated by limited trypsin treatment and placed in a chamber for recording. Transfected cells were identified using epi-fluorescent microscopy for recordings. Electrophysiological Recordings ------------------------------- Cell-attached, inside-out, and whole-cell recordings were performed using an Axopatch 200B patch-clamp amplifier (Axon Instruments) as described previously ([@bib47], [@bib48]). For cell-attached recordings, the pipette and bath solution contained (in mM) 140 NaAsp (sodium aspartate), 10 NaCl, 1 EDTA, 10 HEPES (pH 7.4), and 140 KAsp, 10 NaCl, 1 EDTA, 10 HEPES (pH 7.4), respectively. For cell-attached recordings shown in [Fig. 8](#fig8){ref-type="fig"} A, the pipette solution contained (in mM) 140 NaCl, 1 MgCl~2~, 1 CaCl~2~, and 10 HEPES (pH 7.4). For inside-out recordings, membrane patches were excised into a bath solution containing (in mM) 140 NaAsp, 10 NaCl, 10 HEPES (pH 7.4), 10 EDTA, and various amounts of MgCl~2~ titrated to the desired concentrations of ionized Mg^2+^ as specified. Bath solution was perfused at ∼200 μl per second. The volume of perfusion chamber and connecting tubing is ∼100 μl. Thus, it takes 2--3 s to exchange the bath solution. Concentrations of ionized Mg^2+^ were calculated using a computer program based on algorithms by Fabiato ([@bib6]) ([www.stanford.edu/\~cpatton/webmaxc/webmaxcE.htm](http://www.stanford.edu/~cpatton/webmaxc/webmaxcE.htm)). For whole-cell recordings, the initial pipette and bath solution contained (in mM) 140 NaAsp, 10 NaCl, 10 HEPES (pH 7.4), 10 EDTA, and 140 NaAsp, 10 NaCl, 1 EDTA, 10 HEPES (pH 7.4), respectively. For whole-cell recordings shown in [Fig. 9](#fig9){ref-type="fig"} B, the bath solution contained (in mM) 140 NaCl, 1 MgCl~2~, 1 CaCl~2~, and 10 HEPES (pH 7.4). Intracellular Mg^2+^ in whole-cell recording was altered by intrapipette perfusion of solutions containing (in mM) 140 NaAsp, 10 NaCl, 10 HEPES (pH 7.4), 10 EDTA, and various amount of MgCl~2~ titrated to the desired concentrations of ionized Mg^2+^ as specified. The intrapipette perfusion was performed via a capillary quartz tubing (flamed and pulled manually to outer and inner diameter ∼50 and 20 μm, respectively) placed within ∼200 μm of the tip of the opening ([@bib10]). Voltage protocol for each experiment was described in the individual figure. Currents were low-pass filtered at 1 kHz using an 8-pole Bessel filter, sampled every 0.1 ms (10 kHz) with Digidata-1300 interface and stored directly onto computer hard disk using pCLAMP9 software. Data were transferred to compact discs for long-term storage. Data Analysis ------------- Dose--response curves for inhibition of TRPV5 by intracellular Mg^2+^ and concentration of Mg^2+^ for half-maximal inhibition (IC~50~) were obtained by fitting relative currents (normalized to maximal currents; I/Imax) at different Mg^2+^ concentrations using the Sigma-Plot program according to equation I/Imax = IC~50~/IC~50~ + \[Mg^2+^\]. The fractional electrical distance of the Mg^2+^ binding site from the outer surface of the membrane (δ) was calculated according to equation IC~50~(V) = IC~50~(0)exp (2δFV/RT) ([@bib26]), where IC~50~(V) and IC~50~(0) represent IC~50~ at the test voltage V and at 0 mV, respectively, and R, T, and F have their usual meanings. Current traces were fitted to a single or double exponential curve (as indicated) to analyze time constant (τ) for inhibition of currents by Mg^2+^. The on-rate constant (K~on~) for slow inhibition by Mg^2+^ was calculated according to the equation K~on~ = 1/τ\[Mg^2+^\]. Data are shown as mean ± SEM of number of observations. Statistical comparison was made using unpaired Student\'s t test. RESULTS ======= Multiple Effects of Intracellular Mg^2+^ on TRPV5: Slow Reversible Inhibition, Fast Voltage-dependent Block, and "Run-down" via Reduction of PIP~2~ --------------------------------------------------------------------------------------------------------------------------------------------------- ### Effect One: Slow Reversible Inhibition We studied TRPV5 channels expressed in Chinese hamster ovary (CHO) cells by cell-attached and inside-out patch-clamp recording ([Fig. 1](#fig1){ref-type="fig"} A). In cell-attached recording, TRPV5-expressing CHO cells exhibited the characteristic strongly inwardly rectifying Na^+^ currents ([@bib27]; [@bib13]; [@bib47]) ([Fig. 1](#fig1){ref-type="fig"} B, C/A). Inward (closed circles) and outward currents (open circles) both increased markedly when inside-out patch membranes containing TRPV5 channels were excised ([Fig. 1](#fig1){ref-type="fig"} B, I/O) into divalent-free (DVF) bath solutions. Exchanging DVF solutions with one that contains 1 mM Mg^2+^ caused inhibition of TRPV5 currents over tens of seconds ([Fig. 1, B and C](#fig1){ref-type="fig"}). Currents recovered upon reintroduction of DVF solutions. TRPV5 currents increased by a similar extent when inside-out membranes were excised into Mg^2+^-free bath solutions containing 0, 200, or 1,000 nM Ca^2+^ (unpublished data). Thus, most (if not all) of the increase in currents upon excision of inside-out membranes into DVF solutions is due to loss of inhibition by intracellular Mg^2+^. We refer to this inhibition by intracellular Mg^2+^ occurring over tens of seconds as "slow reversible inhibition" to distinguish it from the much faster (also reversible) voltage-dependent block described below. ![Multiple effects of intracellular Mg^2+^ on TRPV5. (A) Configuration and voltage protocol of recording. Voltage stimuli were applied every 10 s. Black and gray arrows indicate time points when inward and outward currents shown in [Fig. 1](#fig1){ref-type="fig"} B were taken, respectively. (B) Intracellular Mg^2+^ causes a fast reversible voltage-dependent block, a slow reversible inhibition, and an irreversible run-down of TRPV5 in inside-out membranes. See text for details. Inside-out patches were bathed in either DVF or 1 mM Mg^2+^-containing solution. C/A and I/O indicate cell-attached and inside-out. Application of PIP~2~ and 1 mM La^3+^ (La) in DVF solution are indicated. Time interval between data points (shown in circles) is 10 s. As described previously by us ([@bib47], [@bib48]), expression of TRPV5 in CHO cells is robust. We estimated that under our experimental condition each excised patch contains ∼50--150 channels. PIP~2~ reactivated run-down TRPV5 channels (i.e., currents recovered to \>50% of the level before run-down) in 54 out of 62 recordings. We did not observe significant background currents in inside-out membranes of mock-transfected cells either before or after PIP~2~ (inward \[−100 mV\] and outward \[+100 mV\] currents: −87 ± 32 pA and 117 ± 28 pA, respectively, before PIP~2~ vs. −135 ± 41 pA and 156 ± 48 pA, respectively, after PIP~2~; mean ± SEM, n = 6). (C) Ramp I-V curves of currents at indicated time points in B. Scale bars for X (40 ms) and Y axis (0.2 nA) are shown. Dotted line indicates 0 current. In the X axis, the time points corresponding to ramp potentials −100, 0, and +100 mV (from left to right) are indicated by three downward arrows, respectively. Inset shows I-V curves scaled up to the same peak current level.](200509314f1){#fig1} ### Effect Two: Voltage-dependent Block and Distinction from the "Slow Reversible Inhibition" Intracellular free Mg^2+^ also causes reversible voltage-dependent block of TRPV5 occurring over a relatively much faster time course (in millisecond time scale; see [@bib13]; [@bib42]). The distinction between the above slow reversible inhibition occurring over tens of seconds and the voltage-dependent block by Mg^2+^ is rather evident from our results (see ramp I-V curves in [Fig. 1](#fig1){ref-type="fig"} C). First, as mentioned above, inward currents shown in [Fig. 1](#fig1){ref-type="fig"} B were measured at the end of 40-ms hyperpolarization (−100 mV) step (indicated by black arrow in voltage ramp protocol in [Fig. 1](#fig1){ref-type="fig"} A). At these time points, Mg^2+^ ions were completely dissociated (from channels) by the hyperpolarization ([Fig. 1](#fig1){ref-type="fig"} C) and thus did not exert voltage-dependent block on channels. Second, the kinetics of voltage-dependent block distinguishes itself from the slow reversible inhibition. The Mg^2+^-induced voltage-dependent block of TRPV5 was evident by the characteristic unblock during the −100 mV hyperpolarization step (indicated by "a" in [Fig. 1](#fig1){ref-type="fig"} C), block of outward current ("b"), shift of reversal potential toward more negative membrane potential ("c"), and steeper I-V near the reversal potential ("d") in the current--voltage (I-V) relationship curves ([Fig. 1](#fig1){ref-type="fig"} C). The scaled-up I-V curves of currents from 10, 20, and 30 s after application of Mg^2+^ (time point 2, 3, and 4, respectively; time point 1 is immediately before application of Mg^2+^) were superimposable ([Fig. 1](#fig1){ref-type="fig"} C, inset), indicating that voltage-dependent block reached its maximal effect immediately after application of intracellular Mg^2+^ while slow reversible inhibition gradually reached its maximum over 30--40 s ([Fig. 1, B and C](#fig1){ref-type="fig"}). Thus, intracellular Mg^2+^ causes distinct fast voltage-dependent block (occurring in ms time scale) and slower reversible inhibition of TRPV5 (in seconds to tens of seconds) (see [Fig. 6](#fig6){ref-type="fig"} for time constant measurements for both processes). ### Effect Three: "Run-down" via Reduction of PIP~2~ Membrane phospholipid PIP~2~ regulates function of many ion channels and transporters ([@bib11]). In inside-out membranes, a prolonged exposure of the cytoplasmic face to Mg^2+^ activates Mg^2+^-dependent lipid phosphatases and depletes PIP~2~ ([@bib14]; [@bib18]). Depletion of PIP~2~ causes an irreversible loss of activity ("run-down") of channels that are regulated by PIP~2~ ([@bib14]). For TRPV5, though each short exposure to Mg^2+^ solution caused reversible inhibition, repetitive exposures to 1 mM Mg^2+^ solutions led to progressive irreversible run-down of currents eventually ([Fig. 1](#fig1){ref-type="fig"} B). Application of exogenous PIP~2~ to inside-out membranes reactivated TRPV5 after its irreversible run-down by Mg^2+^ ([Fig. 1](#fig1){ref-type="fig"} B), suggesting that irreversible run-down of TRPV5 by intracellular Mg^2+^ is due to loss of PIP~2~ in the membrane and mechanistically different from the reversible voltage-dependent block and "slow inhibition" by Mg^2+^. Activation of a Ca^2+^-sensitive PLC by intracellular Ca^2+^ depletes PIP~2~ and leads to desensitization of TRPM8 channels ([@bib32]). We found that U-37122 (an inhibitor of PLC, 5--10 μM) did not prevent run-down of TRPV5 caused by prolonged exposure to Mg^2+^ (unpublished data), supporting that Mg^2+^-induced run-down is caused by activation of lipid phosphatase(s) rather than by activation of some Mg^2+^-sensitive PLC. A monoclonal antibody against PIP~2~ ([@bib7]) interferes with PIP~2~ regulation of ion channels presumably by sequestering PIP~2~ in the membranes ([@bib14]). Application of anti-PIP~2~ antibody to the cytoplasmic face of inside-out membranes decreased TRPV5 currents ([Fig. 2](#fig2){ref-type="fig"} A). Currents recovered partially after extensive washout of the antibody ([Fig. 2](#fig2){ref-type="fig"} B). Polylysine, with its positively charged residues, has also been used to inhibit the activity of PIP~2~-activated ion channels through competing PIP~2~--channel interaction ([@bib14]; [@bib28]). We found that application of polylysine (average molecular weight 1 kD; 30 μg/ml) to the inside-out membranes inhibited TRPV5 by 89 ± 9% in 3 min (n = 5; unpublished data). Similar to application of PIP~2~, application of Mg-ATP after irreversible run-down in Mg^2+^ solutions reactivated TRPV5 ([Fig. 2](#fig2){ref-type="fig"} B). Mg-ATP presumably works by activating lipid kinases, including phosphoinositide 4-kinase to regenerate PIP~2~ in the membranes ([@bib14]; [@bib11]). Consistent with this idea, the reversal by Mg-ATP was blunted by an inhibitor of phosphoinositide 4-kinase wortmannin ([@bib24]) ([Fig. 2](#fig2){ref-type="fig"} B, WMN). Furthermore, Na-ATP did not reactivate TRPV5 (unpublished data). [Fig. 2](#fig2){ref-type="fig"} C shows I-V curves of currents at several indicated time points. These results provide compelling evidence for a critical role of membrane PIP~2~ in the function of TRPV5. ![PIP~2~ is critical for TRPV5 channel activity. (A) Anti-PIP~2~ antibody reduces TRPV5 currents in inside-out patches. Experiments were repeated five times with similar results. The average (mean ± SEM) inhibition by anti-PIP~2~ antibody was 88 ± 13% and 79 ± 14% for outward and inward currents, respectively. (B) A representative experiment showing long exposure of inside-out patches to Mg^2+^ causes run-down of TRPV5, which can be subsequently reactivated by Mg-ATP. // indicates 10 min break. The inhibition by Mg^2+^ shown here appears to be slower compared with that in [Fig. 1](#fig1){ref-type="fig"} B. This variability is likely due to differences in the configuration and accessibility of excised membranes to perfused bath solutions. Closed and open circles are currents at −100 and +100 mV, respectively. Similar results were observed in six separate experiments. In some experiments, Mg-ATP was applied together with WMN (10 μM). Gray triangles illustrate one such recording receiving Mg-ATP + WMN superimposed on one receiving the Mg-ATP alone. Similar results were observed in four separate experiments. Dimethyl sulfoxide (DMSO, vehicle for WMN, 0.5%) did not prevent reactivation by Mg-ATP (n = 4, not depicted). Of note is that wortmannin did not prevent recovery of currents from transient inhibition by Mg^2+^ (not depicted), indicating that the slow reversible inhibition is not due to regeneration of PIP~2~ from lipid kinases. (C) I-V curves of currents at indicated time points in B. In this experiment, voltage ramp from −100 to +100 mV was applied without a 40-ms step at −100 mV as indicated in [Fig. 1](#fig1){ref-type="fig"} A. Dotted line indicates 0 current. In the X axis, the time point corresponding to ramp potential 0 mV is indicated by upward arrow. The characteristic voltage-dependent block in trace 3 is due to free (ionized) Mg^2+^ (0.13 mM) present in the Mg-ATP solution.](200509314f2){#fig2} Application of PIP~2~ to Inside-out Membranes Releases Mg^2+^-induced Slow Reversible Inhibition ------------------------------------------------------------------------------------------------ Activation of inward-rectifier K^+^ and other channels by PIP~2~ modulates the regulation of channels by other signaling pathways ([@bib11]). We examined whether PIP~2~ activation of TRPV5 alters its sensitivity to Mg^2+^-induced slow inhibition. We examined dose--response relationships for the intracellular Mg^2+^-induced slow reversible inhibition of TRPV5 in excised inside-out membranes before and after application of PIP~2~. As before, after reaching maximal currents in DVF solutions in inside-out membranes, TRPV5 currents were completely inhibited by a solution containing 1 mM Mg^2+^ ([Fig. 3](#fig3){ref-type="fig"} A). Currents recovered in DVF solutions. The concentration of Mg^2+^ for half-maximal inhibition (IC~50~) of the inward currents of TRPV5 was 0.11 ± 0.03 mM (mean ± SEM, n = 7; [Fig. 3](#fig3){ref-type="fig"} C, closed circles). Application of exogenous PIP~2~ to inside-out membranes in DVF solutions increased TRPV5 currents by 2.4 ± 0.3--fold (n = 8; see [Fig. 3](#fig3){ref-type="fig"} A for representative experiment), indicating a submaximal concentration of PIP~2~ present in the excised membranes. The application of PIP~2~ reduced the sensitivity of TRPV5 to the slow reversible inhibition by intracellular Mg^2+^. After PIP~2~, the inward currents of TRPV5 were partially inhibited by 1 mM Mg^2+^ but completely inhibited by 10 mM Mg^2+^ ([Fig. 3](#fig3){ref-type="fig"} A). Currents recovered in the DVF solution ([Fig. 3](#fig3){ref-type="fig"} A). The IC~50~ for the slow inhibition by Mg^2+^ was increased by ∼28-fold to 3.11 ± 0.25 mM (n = 8; [Fig. 3](#fig3){ref-type="fig"} C, open circles) after application of exogenous PIP~2~. The Mg^2+^-induced voltage-dependent block remains distinguishable from the slow reversible inhibition even after application of exogenous PIP~2~. As shown by the scaled-up I-V curves ([Fig. 3](#fig3){ref-type="fig"} B), voltage-dependent block was complete in 1 mM Mg^2+^ in the presence of PIP~2~ (i.e., I-V curves at time point 2 and 3 are superimposable, [Fig. 3](#fig3){ref-type="fig"} B). Yet, Mg^2+^-induced slow inhibition requires 10 mM Mg^2+^ over 30--40 s to reach its maximum (i.e., more inhibition occurred from time point 2 to time point 3 in [Fig. 3](#fig3){ref-type="fig"} A). ![PIP~2~ decreases the sensitivity to Mg^2+^-induced slow inhibition but not the voltage-dependent block. (A) Application of PIP~2~ in inside-out patches increases TRPV5 currents and reduces the sensitivity to inhibition by Mg^2+^. Voltage ramp protocol is same as in [Fig. 1](#fig1){ref-type="fig"}. Holding potential between ramps is 0 mV. (B) Scaled-up I-V curves of currents at indicated time points in A. (C) Dose--response curves of inhibition by Mg^2+^ in inside-out patches before (no PIP~2~) and after application of PIP~2~ (+PIP~2~). IC~50~ was obtained by fitting relative inward currents (inward currents at −100 mV normalized to maximal currents; I/Imax) at different Mg^2+^ concentrations according to the equation I/Imax = IC~50~/IC~50~ + \[Mg^2+^\].](200509314f3){#fig3} Effect of Intracellular Mg^2+^ on TRPV5 in Whole Cell ----------------------------------------------------- The concentration of PIP~2~ in the excised membranes is likely considerably lower than in the plasma membranes in vivo ([@bib25]). We further examined regulation of TRPV5 by intracellular Mg^2+^ using whole-cell recording ([Fig. 4](#fig4){ref-type="fig"} A), which resembles physiological plasma membrane conditions better than the inside-out membranes. Strongly inwardly rectifying Na^+^ currents were observed immediately after establishment of whole-cell recording in TRPV5-expressing cells (time point 1 in [Fig. 4](#fig4){ref-type="fig"} B and trace 1 in [Fig. 4](#fig4){ref-type="fig"} C) but not in mock-transfected CHO cells ([Fig. 4](#fig4){ref-type="fig"} D). Dialysis of the intracellular space with a DVF solution led to a large increase of whole-cell TRPV5 currents ([Fig. 4, B and C](#fig4){ref-type="fig"}). As reported previously ([@bib13]; [@bib47]), whole-cell TRPV5 currents exhibited slight outward rectification at membrane potentials \>+50 mV in the presence of extracellular EDTA ([Fig. 4](#fig4){ref-type="fig"} C, trace 3). Perfusion of the intracellular space with a solution containing 1 mM Mg^2+^ caused a large reversible inhibition on whole-cell TRPV5 currents ([Fig. 4, B and C](#fig4){ref-type="fig"}). Unlike in the inside-out recordings, run-down of TRPV5 by Mg^2+^ is rather infrequent in whole-cell recordings (\>90% of experiments show \>80% recovery upon changing from Mg^2+^ to DVF solutions) despite an average longer exposure of the cytoplasmic face of plasma membrane to Mg^2+^. Thus, run-down of TRPV5 via activation of lipid phosphatases (as observed in excised inside-out patches) is likely not a physiological action of intracellular Mg^2+^ (as in the conditions of whole-cell recording). Intracellular Mg^2+^ at concentrations ranging from 0.1 to 3 mM caused a dose-dependent inhibition of whole-cell TRPV5 currents ([Fig. 4](#fig4){ref-type="fig"} E). The IC~50~ for intracellular Mg^2+^ inhibition of TRPV5 channels in whole-cell recordings was 0.29 ± 0.02 mM (n = 11; [Fig. 4](#fig4){ref-type="fig"} F). This IC~50~ is between that for inside-out membranes before PIP~2~ (IC~50~: 0.11 mM) and after PIP~2~ (IC~50~: 3.11 mM), suggesting that PIP~2~ content in whole cell membranes is intermediate between the latter two inside-out membrane conditions. ![Regulation of TRPV5 by intracellular Mg^2+^ in whole-cell recording. (A) Alteration of intracellular Mg^2+^ in whole-cell recording using intrapipette perfusion. (B) Intracellular perfusion of Mg^2+^ (1 mM) in whole-cell recording inhibits TRPV5 reversibly. The Y axis is current density (nA/pF). Time bar for the X axis represents 4 min. (C) I-V curves of currents at indicated time points in B. (D) Time control of intracellular Mg^2+^ perfusion in mock-transfected CHO cells (transfected with GFP plasmid alone). The Y axis is current density (pA/pF). Time bar for the X axis represents 4 min. (E) Dose-dependent inhibition of whole-cell TRPV5 currents by Mg^2+^ (0--3 mM). Each experiment at a given Mg^2+^ concentration was performed as in [Fig. 4](#fig4){ref-type="fig"} B and inward currents were scaled to the same maximal (current at DVF solution; I/I~max~ = 1) and minimal level (current after La^3+^; I/I~max~ = 0). (F) Dose--response curve of Mg^2+^ inhibition in whole-cell recording (closed squares) vs. in inside-out patches before (closed circles; no PIP~2~) and after PIP~2~ (open circles; +PIP~2~). IC~50~ was calculated according to the equation I/Imax = IC~50~ /IC~50~ + \[Mg^2+^\]. I/Imax = relative currents normalized to maximal currents. The two curves for inside-out patches before and after PIP~2~ are from [Fig. 3](#fig3){ref-type="fig"} C.](200509314f4){#fig4} Regulation of ion channels including inward rectifier K^+^ and TRP channels by PIP~2~ involves direct interaction with positively charged amino acids in the channels ([@bib11]; [@bib30]; [@bib32]). Recently, Rohács et al. reported that arginine-606 of the rabbit TRPV5 (corresponding to arginine-599 of the rat TRPV5) is critical for interaction with PIP~2~ ([@bib32]). We found that R606Q mutant was more sensitive to inhibition by intracellular Mg^2+^ in whole-cell recordings (IC~50~: 0.03 ± 0.01 mM for R606Q \[n = 4\] vs. 0.29 ± 0.02 mM for wild type, P \< 0.05; not depicted in [Fig. 4](#fig4){ref-type="fig"}). These results further support that PIP~2~ activation of TRPV5 releases intracellular Mg^2+^ inhibition of the channel. Mg^2+^-induced Slow Reversible Inhibition Is Voltage Dependent -------------------------------------------------------------- To examine membrane voltage sensitivity, whole-cell TRPV5 currents were measured in intracellular DVF solution and in a solution containing 0.3 mM Mg^2+^ at different holding potentials. The fraction of TRPV5 currents in 0.3 mM intracellular Mg^2+^ relative to maximal currents in DVF solution (I/Imax) increased with increasing membrane hyperpolarization ([Fig. 5](#fig5){ref-type="fig"} A). Voltage sensitivity for the intracellular Mg^2+^-induced slow reversible inhibition of TRPV5 is further supported by the finding in experiments using whole-cell recording that hyperpolarization of membrane holding potentials from 0 to −50 mV increased IC~50~ for Mg^2+^ from 0.29 to 0.79 mM ([Fig. 5](#fig5){ref-type="fig"} B). Voltage dependency of inhibition by Mg^2+^ can be viewed as Mg^2+^ binds to inhibit the channel within the membrane electrical field. The fractional electrical distance of the binding site from the outer surface of the membrane (δ) can be calculated according to equation IC~50~(V) = IC~50~(0)exp(2δFV/RT) ([@bib26]), where IC~50~(V) and IC~50~(0) represent IC~50~ at the test voltage V and at 0 mV, respectively, and R, T, and F have their usual meanings. Our results reveal that Mg^2+^-induced slow inhibition of TRPV5 is owing to Mg^2+^ binding inside the channel at δ = 0.26. ![Voltage dependency of the Mg^2+^-induced slow reversible inhibition. (A) Whole-cell currents were recorded in DVF and 0.3 mM intracellular Mg^2+^ at different holding potentials (from +50 to −100 mV) as in [Fig. 4](#fig4){ref-type="fig"} B. Relative currents (I/I~max~, current at 0.3 mM Mg^2+^/current at DVF) were plotted against holding potentials (V~h~). (B) Dose--response curve of inhibition by Mg^2+^ at 0 and −50 mV holding potentials. Dose--response curves were performed as in [Fig. 4](#fig4){ref-type="fig"} B.](200509314f5){#fig5} Time Constant Measurement Provides Further Support for Voltage Dependency for Mg^2+^-induced Slow Inhibition and its Distinction from Voltage-dependent Block ------------------------------------------------------------------------------------------------------------------------------------------------------------- Up to now, our studies examine the regulation of TRPV5 by intracellular Mg^2+^ by perfusion of solutions containing different concentration of Mg^2+^. To avoid variability in the rate of perfusion and measure time constants for the Mg^2+^-induced voltage-dependent block and the slow inhibition, we recorded whole-cell currents using voltage jump protocol at fixed intracellular Mg^2+^ concentration (1 mM) ([Fig. 6](#fig6){ref-type="fig"} A). Membrane potential was held at −100 mV. At this hyperpolarized potential, TRPV5 is fully open (not inhibited by intracellular Mg^2+^) and conducts inward current (see [Fig. 1](#fig1){ref-type="fig"}). Stepping from −100 to +100 mV resulted in a fast (i.e., occurring in ms and reaching maximal inhibition in ∼50 ms) time-dependent inhibition of the channel (shown as decrease of outward current in [Fig. 6](#fig6){ref-type="fig"} A). This fast inhibition of currents (known as the Mg^2+^-induced voltage-dependent block) represents the time course of binding of intracellular Mg ^2+^ into the pore and contributes to inward rectification of currents through TRPV5 and TRPV6 channels ([@bib13]; [@bib42]). As shown by others ([@bib42]), the time course of voltage-dependent block could be fitted by two exponential terms. [Fig. 6](#fig6){ref-type="fig"} B shows the two exponential time constants for voltage-dependent block at +25, +50, and +100 mV, respectively. Beyond the fast voltage-dependent block occurring within the first 50 ms, outward currents further decreased in a much slower time course (occurring in seconds and reaching minimal current level in \<60 s) ([Fig. 6](#fig6){ref-type="fig"} C). The mean exponential time constant (τ) for the "Mg^2+^-induced slow inhibition" at +100 mV was 8.3 ± 1.1 s (mean ± SEM, n = 4) ([Fig. 6](#fig6){ref-type="fig"} D). The on-rate constant (K~on~) for slow inhibition by Mg^2+^ at +100 mV (K~on~ = 1/τ \[Mg^2+^\]) was estimated at 120 ± 14 s^−1^M^−1^ (mean ± SEM, n = 4). Outward currents remained relatively stable beyond the first 50 ms in the absence of intracellular Mg^2+^ (current level at 60 s = 91 ± 8% of that at 100 ms, n = 4; not depicted in [Fig. 6](#fig6){ref-type="fig"}). Consistent with voltage dependency of Mg^2+^-induced slow inhibition, time constant at +50 and +25 mV were 13.4 ± 1.2 s and 16.2 ± 0.4 s, respectively ([Fig. 6](#fig6){ref-type="fig"} D, each P \< 0.05 vs. time constant holding at +100 mV). These time constant values at +100, +50, and +25 mV are in agreement with the earlier observation that Mg^2+^-induced slow inhibition occurred over 30--40 s at 0 mV (see [Fig. 1](#fig1){ref-type="fig"}; taking into consideration reduced PIP~2~ content in inside-out patches and probably a few seconds of delay in delivery of Mg^2+^ by perfusion). ![Kinetics of Mg^2+^-induced voltage-dependent block (A and B) and slow inhibition (C and D). (A) Whole-cell recording was performed at fixed intracellular Mg^2+^ (1 mM). Holding potential was −100 mV. Test pulse to +100 mV is illustrated. Time course of outward current (at +100 mV) from peak current (∼1 ms after test potential) to the current level at 50 ms is shown. Dotted line represents double exponential fit to the current. (B) Fast (squares) and slow (circles) exponential time constant of decrease of outward current from peak current (∼1 ms) to 50 ms of test potentials (+25, +50, +100 mV, respectively). (C) Time course of decrease of outward current from 100 ms to 60 s after test potential to +100 mV. Dotted line represents single exponential fit to the current. (D) Exponential time constants of decrease of outward current between 100 ms and 60 s of test potentials (+25, +50, and +100 mV, respectively).](200509314f6){#fig6} Mutation of Aspartate-542 Abolishes Mg^2+^-induced Slow Reversible Inhibition ----------------------------------------------------------------------------- Acidic amino acids located in solvent-accessible region between the second intracellular loop (2^nd^ IC loop) and the proximal region of the intracellular COOH terminus (CT) of TRPV5 may be involved in the binding and slow reversible inhibition by intracellular Mg^2+^ ([Fig. 7](#fig7){ref-type="fig"} A). We mutated seven acidic amino acids within this region, one at a time, to neutral amino acid asparagine or glutamine and tested the sensitivity of each mutant to inhibition by Mg^2+^ in whole-cell recordings. We found that neutralization of aspartate-542 (D542N) in the putative pore region completely abolished the slow reversible inhibition of TRPV5 by intracellular Mg^2+^ ([Fig. 7, B and C](#fig7){ref-type="fig"}). In voltage jump experiment, currents through D542N mutant remained unchanged over 60 s at +100 mV (current level at 60 s = 89 ± 15% of that at 100 ms, n = 3; not depicted in [Fig. 7](#fig7){ref-type="fig"}), confirming that aspartate-542 is critical for Mg^2+^-induced slow inhibition. Neutralization of other aspartate or glutamate residues in this region had no significant effects on the slow reversible inhibition of TRPV5 by Mg^2+^ ([Fig. 7](#fig7){ref-type="fig"} D). ![Identification of amino acid of TRPV5 involved in Mg^2+^-induced slow inhibition. (A) Membrane topology of amino acids between the fourth transmembrane (TM4) segment and the proximal COOH terminus (Ct) of TRPV5. Amino acids in the preselectivity filter region likely form a pore helix structure similar to that of KcsA ([@bib5]; [@bib44]). The location of amino acids mutated is shown. (B) A representative experiment of D542N mutant (in whole-cell recording) showing no inhibition by 1 mM Mg^2+^. (C) Dose--response curves of inhibition by Mg^2+^ for D542N mutant and wild-type (WT) TRPV5. (D) Average inhibition by Mg^2+^ for WT and each mutant (mean ± SEM, n = 5--11 each). \* indicates P \< 0.05 vs. WT.](200509314f7){#fig7} Aspartate-542 is believed to be a part of the selectivity filter of TRPV5 and a binding site for Mg^2+^-induced voltage-dependent block ([@bib27]; [@bib13]). Our findings indicate that Mg^2+^ binding to aspartate-542 not only causes voltage-dependent block but also the slow inhibition. This conclusion is further supported by the fact that the fractional electrical distance for the slow inhibition we measured in this study (δ = 0.26) is similar to that measured for voltage-dependent block (δ = 0.31) ([@bib26]). Reduction of PIP~2~ Enhances Mg^2+^-induced Slow Reversible Inhibition of TRPV5 without Affecting Mg^2+^ Binding ---------------------------------------------------------------------------------------------------------------- To investigate the mechanism for PIP~2~ modulation of Mg^2+^-induced slow reversible inhibition of TRPV5, we tested whether PIP~2~ affects Mg^2+^ binding. The kinetics of Mg^2+^-induced slow inhibition and voltage-dependent block were studied using voltage jump experiments as in [Fig. 6](#fig6){ref-type="fig"}. Whole-cell recordings were performed in the presence of intracellular Mg^2+^ (1 mM). The extracellular bath solution contained thrombin + wortmannin (WMN) or vehicle. WMN inhibits phosphoinositide 4-kinase (a rate-limiting enzyme for synthesis of PIP~2~ from phosphatidylinositol and phosphatidylinositol-4-phosphate) to allow reduction of membrane PIP~2~ following receptor-mediated hydrolysis of PIP~2~ ([@bib24]). [Fig. 8](#fig8){ref-type="fig"} A shows representative single exponential fit to current traces in response to voltage jump from −100 to +100 mV in the presence of thrombin + WMN (labeled as Thr) or vehicle (Control). As expected for PIP~2~ releasing Mg^2+^-induced slow inhibition, reduction of membrane of PIP~2~ by incubating with thrombin and WMN decreased the time constants for inhibition at depolarized membrane potentials ([Fig. 8](#fig8){ref-type="fig"} B, P \< 0.05, +Thr vs. Control for all potentials). Interestingly, reduction of membrane PIP~2~ had no effects on the time constants for Mg^2+^-induced voltage-dependent block to the channel ([Fig. 8](#fig8){ref-type="fig"} C; time constants measured as in [Fig. 6](#fig6){ref-type="fig"} B). ![Effects of PIP~2~ hydrolysis on Mg^2+^-induced slow inhibition (A and B) and voltage-dependent block (C). (A) Whole-cell currents (between 100 ms and 60 s; at 1 mM intracellular Mg^2+^) in response to voltage jump from −100 mV holding potential to depolarized test potentials were measured as in [Fig. 6](#fig6){ref-type="fig"} C. Cells were incubated with vehicle (Control) or thrombin + WMN (+Thr) in the extracellular bath for 10 min before patch-clamped for whole-cell recordings. Shown are representative single exponential fits to currents for Control (black curve) or +Thr (gray curve). Relative current level corresponding to 1/e (e, the base of natural logarithms) of the maximal current is shown. (B) Time constants of Mg^2+^-induced slow inhibition at +25, +50, +100 mV for Control and +Thr. \* indicates P \< 0.05 +Thr vs. Control. (C) Time constants of Mg^2+^-induced voltage-dependent block (as in [Fig. 6, A and B](#fig6){ref-type="fig"}) at +25, +50, +100 mV for Control and +Thr.](200509314f8){#fig8} Enhancement of Mg^2+^-induced Slow Inhibition as a Mechanism for PLC-coupled Receptor Regulation TRPV5-mediated Ca^2+^ Entry ---------------------------------------------------------------------------------------------------------------------------- The physiological function of TRPV5 is to allow Ca^2+^ entry through the apical membrane for transepithelial transport of Ca^2+^ in kidney and intestine ([@bib12]). We next examined the effects of PIP~2~ hydrolysis induced by receptor activation of PLC on TRPV5 functioning as a Ca^2+^-permeable channel. In the following cell-attached recordings, the pipette solution contained (in mM) 140 NaCl, 1 MgCl~2~, and 1 CaCl~2~. Because of the anomalous mole fraction behavior, TRPV5 conducts Ca^2+^ exclusively in this physiological solution ([@bib41]). CHO cells express endogenous receptors for thrombin ([@bib4]). Activation of PLC by endogenous thrombin receptors in CHO cells was verified by the release of Ca^2+^ from intracellular stores (unpublished data). Thrombin was applied to the patch membrane via intrapipette perfusion ([Fig. 9](#fig9){ref-type="fig"} A). The bath solution contained vehicle or wortmannin. ![Effect of PIP~2~ hydrolysis on the Mg^2+^-induced slow inhibition. (A) Effect of thrombin plus/minus wortmannin on TRPV5 in cell-attached recording. The top left panel shows cell-attached configuration for recording TRPV5-mediated Ca^2+^ currents. See text for the pipette solution. Voltage stimuli (step to −100 mV for 40 ms and ramp from −100 to +100 mV ramp over 200 ms) were applied every 2 s from +50 mV holding potential. Holding at +50 mV prevents Ca^2+^ entry and the Ca^2+^-dependent inactivation. Thrombin (Thr, 1 U/ml) was applied to the patch membrane by intrapipette perfusion. WMN (10 μM) or vehicle (0.5% DMSO) was applied to the extracellular bath in cells expressing TRPV5 5--10 min before addition of thrombin. The bottom panel shows inward currents (at −100 mV) normalized to the initial current (I/Imax). The top right panel shows ramp I-V curves at indicated time points. (B) Dose--response curve of inhibition by Mg^2+^ before and after membrane PIP~2~ hydrolysis. Whole-cell TRPV5-mediated Ca^2+^ currents were recorded in a bath solution containing (in mM) 140 NaCl, 1 MgCl~2~, 1 CaCl~2~, and 10 HEPES (pH 7.4). Intracellular Mg^2+^ (0 to 3 mM) was altered by intrapipette perfusion. Cells expressing TRPV5 were incubated with a bath solution containing either vehicle (closed circles) or WMN + Thr (open circles) for 5--10 min before patching for whole-cell recording. Voltage protocol and holding potential are as in the cell-attached experiment.](200509314f9){#fig9} In cell-attached recording, application of thrombin (Thr) in the absence of WMN decreased TRPV5 Ca^2+^ currents transiently (by 43 ± 8% in 90 s, n = 5, P \< 0.05 vs. control without thrombin) ([Fig. 9](#fig9){ref-type="fig"} A). Presumably due to desensitization of receptors and de novo synthesis of PIP~2~ via phosphoinositide 4-kinase, TRPV5 currents recovered at least partially over 3--4 min in the absence of wortmannin. In the presence of wortmannin, thrombin caused a persistent reduction of TRPV5 currents by 85 ± 5% in 90 s (n = 6, P \< 0.05 vs. thrombin alone). As controls, TRPV5 currents did not change significantly without thrombin (decreased by 8 ± 7% over 90 s, n = 4; WMN alone in [Fig. 9](#fig9){ref-type="fig"} A). The effects of PIP~2~ hydrolysis on dose-dependent inhibition by Mg^2+^ of TRPV5-mediated Ca^2+^ currents were examined by whole-cell recording ([Fig. 9](#fig9){ref-type="fig"} B). The IC~50~ for Mg^2+^ inhibition of TRPV5-mediated Ca^2+^ currents were 0.61 ± 0.12 mM (n = 5) and 0.11 ± 0.08 (n = 5) for control cells (treated with vehicle) and cells treated with thrombin and wortmannin, respectively (P \< 0.05). The concentration of the intracellular free Mg^2+^ in the cell-attached recording is estimated at ∼0.5 mM ([@bib33]). These results support the hypothesis that an increase in sensitivity to intracellular Mg^2+^ is important for physiological regulation of TRPV5-mediated Ca^2+^ entry by PLC-activating hormones. DISCUSSION ========== Many TRP channels are downstream of G protein--coupled receptors and PLC (for review see [@bib2]; [@bib31]). Thus, TRP channels may be regulated by G protein--coupled receptors via inositol-1,4,5-trisphosphate and/or diacylglycerol, products of PIP~2~ breakdown catalyzed by PLC ([@bib2]; [@bib31]). Several TRP channels are also directly regulated by PIP~2~ ([@bib3]; [@bib8]; [@bib35]; [@bib19]; [@bib18]; [@bib32]). Therefore, an additional mechanism for PLC to control TRP channels is via reduction of PIP~2~ content ([@bib3]; [@bib35]; [@bib18]). However, the reduction of PIP~2~ in the plasma membrane from physiological stimulation of PLC may not be sufficient to alter the activity of channels if the affinity of channels for PIP~2~ is relatively high ([@bib15]). In the present study, we find that PIP~2~ activates TRPV5 and that activation of TRPV5 by PIP~2~ decreases the sensitivity of TRPV5 to inhibition by intracellular Mg^2+^ (see working model in [Fig. 10](#fig10){ref-type="fig"}). Hydrolysis of PIP~2~ by receptor activation of PLC increases the sensitivity to the inhibition by Mg^2+^. These results provide a novel mechanism for regulation of TRP channels by PLC-activating hormones by increasing the sensitivity to intracellular Mg^2+^. ![A working model showing the relationships of three different effects of Mg^2+^ on TRPV5. Binding of intracellular Mg^2+^ to D542 of TRPV5 causes a fast (in ms time scale) voltage-dependent block. Prolonged binding to the same site by Mg^2+^ (e.g., 10 s) leads to slow inhibition, which is likely due to conformational changes of the channel. The slow inhibition is also voltage dependent. Membrane PIP~2~ stabilizes TRPV5 in open conformation. PIP~2~ interaction with TRPV5 does not affect Mg^2+^ binding but prevents Mg^2+^-induced conformational changes. Mg^2+^ causes run-down by activating lipid phosphatases. However, run-down is probably not a physiological effect of Mg^2+^ as it occurs predominantly in excised inside-out membranes. Hydrolysis of PIP~2~ by activating PLC-coupled receptors sensitizes TRPV5 to inhibition by intracellular Mg^2+^. \* indicates that time constant for Mg^2+^-induced slow inhibition of TRPV5 is shorter in PIP~2~-reduced (e.g., 5 s) than in PIP~2~-repleted state (e.g., 10 s). For comparison, Mg-induced slow inhibition has also been reported for TRPV6, TRPM6, and TRPM7 ([@bib23]; [@bib34]; [@bib42], [@bib43]). Mg^2+^-induced voltage-dependent block contributes to inward rectification of many channels, including inward-rectifier K^+^ channels and TRPV5 and 6. Mg^2+^ would presumably cause run-down on all PIP~2~-regulated ion channels.](200509314f10){#fig10} The cytosolic free Mg^2+^ concentration in most mammalian cells is between 0.5 and 1 mM ([@bib33]). The IC~50~ for the slow inhibition (0.79 mM) at −50 mV (near the physiological resting membrane potential) suggests that intracellular Mg^2+^ is an important physiological regulator of TRPV5 function in vivo. The intracellular free Mg^2+^ concentration changes only slightly in response to physiological stimuli ([@bib33]). However, the dramatic shift in Mg^2+^ sensitivity by PIP~2~ will provide amplification for regulation of TRPV5 by intracellular Mg^2+^ and PLC-activating hormones. One such example is the inhibition of calcium transport in the distal renal tubules by prostaglandin E~2~ through activation of phospholipase C via EP~1~ receptors ([@bib12]). Besides TRPV5 and TRPV6, intracellular Mg^2+^ has recently been recognized as an important regulator of TRPM6 and TRPM7 channels ([@bib23]; [@bib34]; [@bib43]). TRPM7 (also known as magnesium-inhibited cation channel \[MIC\]; [@bib16]; [@bib29]) is a ubiquitous channel important for cellular uptake of magnesium and cell viability ([@bib37]). Magnesium ions entering cells through TRPM7 feedback on the channel to regulate its activity ([@bib37]). TRPM7 is also regulated by PIP~2~ ([@bib35]). Whether PIP~2~ also modulates Mg^2+^ sensitivity for TRPM7 is unknown at present. TRPV5 does not conduct Mg^2+^ in physiological conditions. However, TRPM6 (a close homology of TRPM7) is a Mg^2+^-permeable channel responsible for magnesium transport in epithelial tissues ([@bib36]; [@bib45]; [@bib43]). Both TRPV5 and TRPM6 are present in the distal convoluted tubule (DCT) cells ([@bib12]; [@bib36]; [@bib45]; [@bib43]), which are key cells responsible for transcellular reabsorption of magnesium and calcium in the kidney ([@bib40]). Many diseases with reduced Mg^2+^ transport in the DCT exhibit increased Ca^2+^ transport. These diseases include genetic diseases such as familial hypomagnesemia with hypocalciuria and Gitelman\'s disease ([@bib46]) and acquired diseases such as cis-platinum toxicity ([@bib21]). Our present finding that intracellular Mg^2+^ negatively regulates TRPV5 raises the possibility that a decrease in Mg^2+^ influx through TRPM6 leads to an increase in Ca^2+^ influx through TRPV5/TRPV6 in DCT cells, providing a possible molecular explanation for this disease phenotype. It should be noted that patients with mutations of TRPM6 develop hypocalcemia ([@bib36]; [@bib45]), rather than hypercalcemia as would be expected from an increase in Ca^2+^ reabsorption in DCT cells alone. Hypocalcemia in these patients is believed secondary to parathyroid failure from severe magnesium wasting and hypomagnesemia ([@bib1]). Intracellular Mg^2+^ regulates inward rectification of ion channels including inward rectifier K^+^ channels and TRPV5 and TRPV6 channels via voltage-dependent block ([@bib20]; [@bib13]; [@bib42]). In the present study, we report that intracellular Mg^2+^ regulates TRPV5 additionally via a mechanism likely involving conformational change of the protein ([Fig. 10](#fig10){ref-type="fig"}). We refer to this additional inhibition by Mg^2+^ as slow reversible inhibition to distinguish from the faster voltage-dependent block. A similar dual regulation of TRPV6 by intracellular Mg^2+^ via voltage-dependent block and a slower inhibition has recently been reported ([@bib42]). Aspartate-542 of TRPV5 is a part of the selectivity filter critical for permeation of Ca^2+^ ions and a binding site for Mg^2+^-induced voltage-dependent block of Ca^2+^ and Na^+^ currents ([@bib27]; [@bib13]). Our findings suggest that binding of intracellular Mg^2+^ to aspartate-542 in the selectivity filter inhibits TRPV5 via at least two separate but interrelated mechanisms. The first mechanism is by voltage-dependent block. In this mechanism, Mg^2+^ ions bind and occlude the channel pore, preventing ion permeation. This inhibition by blocking of channel pore occurs as soon as Mg^2+^ enters and binds the selectivity filter driven by membrane depolarization and reverses quickly by membrane hyperpolarization. This is similar to voltage-dependent block of inwardly rectifying K^+^ channels and other channels by Mg^2+^, polyamines, and other charged blockers ([@bib9]). Binding of Mg^2+^ to the selectivity filter of TRPV5 for a longer duration, however, leads to the second mechanism of inhibition: the slow reversible inhibition. The slow reversible inhibition develops over 30--40 s and is not reversed by transient unbinding of Mg^2+^ ions during intermittent membrane hyperpolarization from voltage ramp applied every 10 s The exact mechanism for the slow inhibition remains unknown, but likely involves conformational change(s) of the channel protein. Our results also shed light on the mechanism by which PIP~2~ alters the sensitivity of TRPV5 to slow inhibition by Mg^2+^. PIP~2~ activates many ion channels directly by affecting their structure ([@bib11]). We find that PIP~2~ activates TRPV5 in the absence of intracellular Mg^2+^ ([Fig. 1](#fig1){ref-type="fig"} B). PIP~2~ also regulates D542N, a Mg^2+^ binding site mutant (unpublished data). Thus, activation of TRPV5 by PIP~2~ does not require its regulation by intracellular Mg^2+^. However, activation of TRPV5 by PIP~2~ decreases the sensitivity of TRPV5 for intracellular Mg^2+^-induced slow inhibition. PIP~2~ does so without affecting Mg^2+^ binding to TRPV5 ([Fig. 8](#fig8){ref-type="fig"} C). We suggest that PIP~2~ desensitizes TRPV5 to Mg^2+^-induced slow inhibition by stabilizing channel in an open conformation and prevents the conformational change induced by binding of Mg^2+^. It is known that ion binding may alter the structure of selectivity filter of ion channels. For example, occupancy of K^+^ ions in the selectivity filter slows the rate of C-type inactivation of voltage-gated K^+^ channels ([@bib49]). The crystal structure of the bacterial KcsA K^+^ channel reveals that presence of multiple K^+^ ions in the selectivity filter stabilizes the full open conformation ([@bib50]). Our present study provides an example of conformational changes of ion channels caused by binding of a pore blocker. We thank Dr. D. Hilgemann for discussion and for help in setting up intrapipette perfusion and Drs. Hilgemann, Baum, and Igarashi for critical reading of the manuscript. This work is supported by National Institutes of Health grants DK-20543, DK-54368, and DK-59530 and an Established Investigator Award (0440019N) from American Heart Association. C.L. Huang is holder of the Jacob Lemann Professorship in Calcium Transport of the University of Texas Southwestern Medical Center at Dallas. Olaf S. Andersen served as editor. [^1]: Correspondence to Chou-Long Huang: <chou-long.huang@utsouthwestern.edu>	{ "pile_set_name": "PubMed Central" }
![](transedinobsoc81623-0097){#sp1 .81} ![](transedinobsoc81623-0098){#sp2 .82} ![](transedinobsoc81623-0099){#sp3 .83} ![](transedinobsoc81623-0100){#sp4 .84} ![](transedinobsoc81623-0101){#sp5 .85} ![](transedinobsoc81623-0102){#sp6 .86} ![](transedinobsoc81623-0103){#sp7 .87} ![](transedinobsoc81623-0104){#sp8 .88} ![](transedinobsoc81623-0105){#sp9 .89} ![](transedinobsoc81623-0106){#sp10 .90} ![](transedinobsoc81623-0107){#sp11 .91} ![](transedinobsoc81623-0108){#sp12 .92} ![](transedinobsoc81623-0109){#sp13 .93} ![](transedinobsoc81623-0110){#sp14 .94}	{ "pile_set_name": "PubMed Central" }
1. Introduction =============== Coronaviruses (CoVs) are enveloped viruses with large positive-sense single-stranded RNA genomes. CoVs infect a variety of mammalian and avian species, and can cause serious disease in humans, as exemplified during the 2002--2003 outbreak of the severe acute respiratory syndrome (SARS). The CoV E protein has a well-established role in the assembly of virions where it may induce membrane curvature or aid in membrane scission. Recent studies have expanded the role of CoV E beyond assembly. CoV E has ion channel activity in vitro. CoV E also is critical for the efficient trafficking of virions through the secretory pathway, a function that may be related to its ion channel activity. The CoV E protein has recently been shown to inhibit the host cell stress response, implicating it in pathogenesis. New interacting partners for E have been identified that expand the role of the protein during infection. How all of these potential properties of CoV E fit together to impact its function(s) is the focus of this review. 2. Background ============= The E protein, along with N, S, and M, are the major coronavirus structural proteins ([Figure 1](#viruses-04-00363-f001){ref-type="fig"}A,B). The N protein is a soluble protein and packages the RNA genome to form the nucleocapsid. The S protein has a single transmembrane domain, is found in the virion envelope, and serves as the attachment and fusion protein. The M protein has three transmembrane domains, is the most abundant protein in the virion envelope, and directs the assembly process through interactions with the other structural proteins (reviewed in \[[@B1-viruses-04-00363],[@B2-viruses-04-00363]\]). CoV E is a small (76--109 amino acids) integral membrane protein, and has a single predicted hydrophobic domain (HD). It is usually encoded as the second or third message in a bi- or tri-cistronic mRNA \[[@B3-viruses-04-00363],[@B4-viruses-04-00363]\]. The E protein is targeted to the Golgi region in infected cells and also when expressed from cDNA \[[@B5-viruses-04-00363],[@B6-viruses-04-00363],[@B7-viruses-04-00363],[@B8-viruses-04-00363],[@B9-viruses-04-00363]\]. The membrane topology of CoV E is of considerable debate, having been reported as transmembrane or a membrane hairpin (see below) \[[@B5-viruses-04-00363],[@B10-viruses-04-00363],[@B11-viruses-04-00363],[@B12-viruses-04-00363]\]. There are conserved membrane proximal cysteine residues that are targets for palmitoylation \[[@B7-viruses-04-00363],[@B13-viruses-04-00363],[@B14-viruses-04-00363],[@B15-viruses-04-00363]\]. There are also conserved proline residues in the C-terminal tail ([Figure 1](#viruses-04-00363-f001){ref-type="fig"}C,D). Other than these similarities there is large variation in the primary sequence of the E proteins, which differ in size and sequence among virus groups often having \<30% identity ([Figure 1](#viruses-04-00363-f001){ref-type="fig"}C). How or if these differences affect protein function is not known. ![Primary structure of the Coronavirus (CoV) E protein. (A) A cartoon depicting a CoV virion. The structural proteins are labeled. (B) The three major CoV structural proteins in the virion envelope. Oligosaccharides are shown on S and M. A single topology is shown for E, see below for discussion on E protein topology. (C) A multiple sequence alignment of several different CoV E proteins. The hydrophobic domain (HD) is bracketed. The CoV genera (alpha, beta, and gamma) are denoted on the left of the multiple sequence alignment. Positively charged residues are shown in blue, negatively charged residues are shown in red, and polar uncharged residues are shown in yellow. The conserved Cys and Pro residues are labeled with a bracket or an asterisk, respectively. The multiple sequence alignment of CoV E proteins was carried out with ClusalW2 at the European Bioinformatics Institutes server, and Jalview software version 2 was used to generate the figure \[[@B16-viruses-04-00363],[@B17-viruses-04-00363]\]. (D) Cartoon depiction of the E protein with the hydrophobic domain shown as a cylinder and the conserved Cys and Pro residues labeled.](viruses-04-00363-g001){#viruses-04-00363-f001} 3. CoV E and Virus Assembly =========================== Unlike many other enveloped viruses, CoVs assemble and bud intracellularly at the ER-Golgi intermediate compartment (ERGIC) ([Figure 2](#viruses-04-00363-f002){ref-type="fig"}A) \[[@B18-viruses-04-00363],[@B19-viruses-04-00363]\]. One of the early discoveries in CoV assembly was that formation of the virion envelope required only expression of M and E and not N. Originally observed for mouse hepatitis virus (MHV) \[[@B20-viruses-04-00363]\], this property has been observed for infectious bronchitis virus (IBV) \[[@B5-viruses-04-00363]\], transmissible gastroenteritis virus (TGEV) \[[@B21-viruses-04-00363]\], and bovine coronavirus (BCoV) \[[@B21-viruses-04-00363]\]. There has been considerable debate about the requirements for SARS-CoV envelope formation, with reports that M and N \[[@B22-viruses-04-00363]\], M and E \[[@B23-viruses-04-00363]\], and even M alone \[[@B24-viruses-04-00363]\] can drive production of released vesicles. These results raise an interesting point about the efficacy of measuring virus-like particle (VLP) production in different cell types and expression systems. It is important to note that membrane proteins that form multimers can be secreted from cells in microvesicles \[[@B25-viruses-04-00363]\]. Thus, overexpression of viral membrane proteins may lead to release in microvesicles, complicating the interpretation of VLP experiments. Additionally, the original reports for MHV, IBV, TGEV, and BCoV all used vaccinia-based expression systems. What has recently become apparent, at least for MHV \[[@B14-viruses-04-00363]\], SARS CoV \[[@B26-viruses-04-00363]\], and IBV \[[@B27-viruses-04-00363]\], is that when using transient transfection to express the proteins from plasmids, the presence of N can greatly increase VLP yield. This result likely means that while not necessarily required for envelope formation, N plays an important role in forming a complete virion. This makes intuitive sense, and explains why empty virions are not readily purified from infected cells. However, the fact that M and E are sufficient for envelope formation is important information when considering the mechanism of assembly. ![CoVs assemble and bud intracellularly at the ERGIC. (A) Newly formed virions bud into the lumen of the ERGIC and traverse the secretory pathway for egress. (B) Potential roles for E in assembly. The E protein is shown in orange and the M protein is shown in green. CoV E could help to bend membranes or play a role in membrane scission.](viruses-04-00363-g002){#viruses-04-00363-f002} The ability of E and M to drive VLP formation clearly shows that E is important for assembly, but the mechanism is not well understood. The IBV E and M proteins interact via their cytoplasmic tails \[[@B28-viruses-04-00363],[@B29-viruses-04-00363]\]. These interactions may be important for particle assembly \[[@B21-viruses-04-00363]\]. Additionally, mutations introduced into the C-terminal tail of MHV E produced virions that were unstable, elongated, and may have resulted from failed scission events \[[@B30-viruses-04-00363]\]. However, a version of IBV E with a heterologous HD was able to drive VLP formation and produce infectious virus, suggesting that the sequence of the HD was not important for assembly \[[@B29-viruses-04-00363],[@B31-viruses-04-00363]\]. The E protein may also promote membrane rearrangements. MHV E can drive the intracellular formation of electron dense membranes derived from the ERGIC when expressed alone in BHK-21 cells \[[@B6-viruses-04-00363]\]. Additionally, the HD of SARS-CoV E can drive the in vitro tubulation of dimyristoyl phosphatidylcholine (DMPC) membranes \[[@B32-viruses-04-00363]\], although a version of the HD that was flanked with lysine residues did not have the same effect \[[@B33-viruses-04-00363]\]. All of these observations point to a pivotal role for E in assembly, possibly in the scission of particles at the ERGIC, or in inducing membrane curvature ([Figure 2](#viruses-04-00363-f002){ref-type="fig"}B). Thus, it was surprising when versions of MHV, SARS-CoV, and TGEV lacking the E gene were shown to assemble virions, albeit to a lower degree than the corresponding wild-type viruses \[[@B34-viruses-04-00363],[@B35-viruses-04-00363],[@B36-viruses-04-00363]\]. Furthermore, the E proteins from IBV, SARS-CoV, and BCoV could functionally replace MHV E in the context of infection, and TGEV E, which cannot substitute for MHV E, needed only a single amino acid change to complement MHVΔE \[[@B37-viruses-04-00363]\]. This suggests that sequence specific protein-protein interactions between M and E are not required for assembly. One final piece of information came from the study of MHVΔE. Characterization of this virus showed that the production of infectious particles was severely compromised. However, after serial passage, revertants arose with a partial duplication of the M gene consisting of the N-terminus and all three transmembrane domains, but lacking most of the C-terminal cytoplasmic tail \[[@B38-viruses-04-00363]\]. How a truncated version of M can functionally replace E is unknown, but it could allow for spacing between M proteins and disrupt lateral interactions with the M tails. Thus, the exact role of the E protein in assembly is not clear. It may be important for membrane curvature and/or scission directly, or it might alter the spacing of the M protein, which in turn is important for these effects. It is also worth noting the varied requirement for E in virion morphogenesis as revealed by the deletion studies in TGEV, MHV, and SARS-CoV \[[@B34-viruses-04-00363],[@B35-viruses-04-00363],[@B36-viruses-04-00363],[@B39-viruses-04-00363],[@B40-viruses-04-00363]\]. The E protein is essential for TGEV production, but not necessary for MHV or SARS-CoV production. Why different viruses have a varied requirement for the E protein is not understood. One possibility is that a specific accessory protein could complement the assembly process in the absence of E for some CoVs. 4. Post-Translational Modifications =================================== The best characterized post-translational modification (PTM) on E is the addition of palmitic acid onto membrane proximal cysteine residues. This modification has been reported for the E proteins of IBV \[[@B13-viruses-04-00363]\], SARS-CoV \[[@B7-viruses-04-00363]\], and MHV \[[@B14-viruses-04-00363],[@B15-viruses-04-00363]\]. It is likely that palmitoylation alters the conformation of the tail in relation to the membrane. Palmitoylation is not important for proper targeting of the E protein \[[@B13-viruses-04-00363],[@B14-viruses-04-00363],[@B15-viruses-04-00363]\]. The functional significance of palmitoylation has been well studied for MHV E, where it is important for assembly as judged both by VLP production and production of infectious virus \[[@B14-viruses-04-00363],[@B15-viruses-04-00363]\]. These studies suggest that palmitoylation of E affects how it interacts with the M protein, possibly by allowing the E protein to gain access to specific lipid microdomains at the site of assembly \[[@B14-viruses-04-00363]\]. Two other PTMs on the E protein have been reported, but not studied in depth. One study demonstrated that transiently expressed SARS-CoV E is N-glycosylated on asparagine 66 \[[@B10-viruses-04-00363]\]; however the functional relevance of this is not known since the residue is on a portion of the protein reported to be in the cytoplasm (see discussion on topology below). Two studies have shown that SARS-CoV E can be ubiquitinated \[[@B41-viruses-04-00363],[@B42-viruses-04-00363]\]. There is currently no known functional role for the ubiquitination of CoV E. 5. CoV E and the Cell Stress Response ===================================== There are varying reports on the role of the E protein in apoptosis. It was shown that overexpression of MHV E and epitope tagged SARS-CoV E induced apoptosis in some cultured cell lines \[[@B43-viruses-04-00363],[@B44-viruses-04-00363]\]. However, a recent study compared the stress response of cells infected with SARS‑CoV to cells infected with SARS-CoVΔE. Using a microarray-based approach, it was shown that the virus lacking E induced a much more robust stress response than the wild-type virus. The virus lacking E also caused a greater degree of apoptosis compared to the wild-type virus \[[@B45-viruses-04-00363]\]. This result, while contrary to the studies using overexpression, shows that SARS-CoV E may be anti-apoptotic during infection. The suppression of the host stress response by E may be important for down regulating the immune response and promoting pathogenesis. 6. Protein-Protein Interactions =============================== While the interaction between E and M has been established (reviewed in \[[@B2-viruses-04-00363]\]), two recent studies have identified novel protein interactions with the SARS-CoV E protein. In the first study, cells were infected with SARS-CoV that had the E gene replaced with a C-terminally tagged version of E. Tandem affinity purification coupled with tandem mass spectrometry was carried out on the infected cells to find interacting proteins. Several candidates were identified, and the interaction of SARS-CoV E with the N-terminal ubiquitin-like domain-1 of SARS-CoV nsp3 protein was characterized. Nsp3 and CoV E colocalize during infection and nsp3 may be responsible for the ubiquitination of SARS‑CoV E \[[@B41-viruses-04-00363]\]. Another report identified the tight junction protein PALS1 as an interactor with the C-terminal domain of SARS-CoV E \[[@B46-viruses-04-00363]\]. The authors speculate that this interaction is important in the pathogenesis of SARS-CoV by promoting disassembly of tight junctions in lung epithelium after primary infection \[[@B46-viruses-04-00363]\]. Understanding the protein-protein interactions of CoV E is an important step in elucidating the functions of the protein. Since the sequences of the E proteins are quite divergent, it will be interesting to determine if different CoV E proteins interact with the same host proteins, or are able to elicit a similar effect through interactions with other proteins. 7. The Ion Channel Activity of CoV E ==================================== Many viruses encode small proteins that have ion channel activity including Hepatitis C virus (HCV) p7 \[[@B47-viruses-04-00363],[@B48-viruses-04-00363]\], influenza M2 \[[@B49-viruses-04-00363]\], and picornavirus 2B \[[@B50-viruses-04-00363]\]. All of these proteins are small (63--97 amino acids) and contain one or two transmembrane domains, making oligomerization a requirement for channel activity \[[@B51-viruses-04-00363]\]. The role of these proteins varies between viruses, but all of them affect the secretory pathway. The M2 protein of influenza virus is the best characterized ion channel. M2 is a 97 amino acid type III membrane protein that forms a tetrameric pH activated proton channel \[[@B49-viruses-04-00363]\]. The first described role for this activity was in the entry of the virus. Influenza virus is endocytosed after binding to a susceptible cell, and upon endosome acidification M2 facilitates the transfer of protons into the virion interior to aid in uncoating of the genome \[[@B52-viruses-04-00363]\]. It was later appreciated that the M2 ion channel is active in infected cells, where M2 has an effect on the secretory pathway by raising the pH of the trans-Golgi network \[[@B53-viruses-04-00363]\]. For some strains of influenza virus, this activity prevents the premature activation of the fusion protein \[[@B53-viruses-04-00363],[@B54-viruses-04-00363],[@B55-viruses-04-00363]\]. HCV p7 is not as well studied as M2, but is relevant to the discussion because like CoVs, HCV assembles intracellularly and must navigate the host secretory pathway. HCV p7 is targeted to the ER and Golgi complex \[[@B56-viruses-04-00363]\], where it is thought to oligomerize and act as an proton channel \[[@B48-viruses-04-00363],[@B57-viruses-04-00363]\]. p7 is important for the assembly and release of HCV virions \[[@B58-viruses-04-00363],[@B59-viruses-04-00363]\]. Recently, it was shown that p7 alkalinizes acidic organelles in the secretory pathway, and that M2 can complement a version of HCV lacking p7 \[[@B57-viruses-04-00363]\]. How this activity promotes release of particles is not known. It may create an environment that protects virions from damage during exocytosis, or promote enlargement of Golgi cisternae to accommodate large cargo (see discussion of CoV E and release below). The channel activity of E was first demonstrated for SARS-CoV E in planar lipid bilayers, where it was found that a synthetic peptide corresponding to the protein could permeabilize bilayers to Na^+^ and K^+^, with a 10-fold preference for Na^+^ \[[@B60-viruses-04-00363]\]. Further study generalized this property to other CoV groups, as IBV E, MHV E, and HCoV-229E E all have ion channel activity for monovalent cations \[[@B61-viruses-04-00363]\]. Furthermore, the channels formed by MHV E and HCoV 229E E were inhibited by the broad spectrum Na^+^/H^+^ exchanger inhibitor hexamethylene amiloride (HMA) \[[@B61-viruses-04-00363]\]. Since the E protein only has a single HD, formation of an ion channel would require oligomerization. This possibility was originally addressed using computational modeling, which predicted that the E protein HD could form stable dimers, trimers, and pentamers \[[@B62-viruses-04-00363]\]. This prediction was further bolstered by several spectroscopic studies, which showed that the HD of SARS-CoV E forms a pentamer \[[@B63-viruses-04-00363],[@B64-viruses-04-00363],[@B65-viruses-04-00363]\]. However, this result has not been confirmed for the other E proteins, or for the full length SARS-CoV E protein in a natural membrane. Nonetheless, predictions can be made from these structural data. For one, it appears that Asn15 in SARS-CoV E, and possibly the equivalent residue in other E proteins, is likely important either for oligomerization or ion conductance. Indeed, when alanine is substituted for this residue in a lysine-flanked peptide of the SARS-CoV E HD, conductance is largely inhibited \[[@B63-viruses-04-00363]\]. Additionally, the equivalent residue in IBV E is required for perturbing the secretory pathway, suggesting a role for the ion channel in this effect \[[@B66-viruses-04-00363]\]. The role of the CoV E ion channel in infection is not entirely clear, but studies using inhibitors offer some insight into its function. When HMA is added to the inoculum of either HCoV 229E or MHV, replication is inhibited. Adding more weight to this result, when HMA is added to the inoculum of MHVΔE no further growth defect is observed, showing that the inhibitory effect of HMA is dependent on the presence of the E protein \[[@B61-viruses-04-00363]\]. While these studies suggest an important role for the ion channel activity of the E protein, it is not clear what step of viral replication is blocked by the drug. Two other studies have linked the putative ion channel activity with virus replication and release. A recombinant version of MHV carrying mutations in the HD was shown to have a defect in both assembly and release of particles \[[@B67-viruses-04-00363]\]. A version of IBV carrying an a mutated version of E where the HD was replaced with that of an unrelated protein showed a defect in the release of infectious particles \[[@B31-viruses-04-00363]\]. These results imply that the ion channel activity of E may be important in the release of virions from cells, and are discussed in the section below. 8. Role of CoV E in Virion Release ================================== 8.1. The Mammalian Secretory Pathway and Large Cargo ---------------------------------------------------- Enveloped viruses that assemble intracellularly must navigate through the secretory pathway for release ([Figure 3](#viruses-04-00363-f003){ref-type="fig"}). Whether these viruses simply follow the canonical trafficking pathway, or modify it to handle the flux of large cargo is not known. To better understand this question, it is important to appreciate how the compartments used by CoVs are linked. The ERGIC is the intermediate compartment between the ER and Golgi. Its major function is to sort and concentrate cargo along the biosynthetic pathway. The ERGIC may mature as a unit to form the cis-Golgi, or it may be a more stable compartment \[[@B68-viruses-04-00363]\]. From the ERGIC, cargo traffics to the Golgi complex. The Golgi complex is made up of polarized stacks of cisternae that are connected laterally into a ribbon. Golgi cisternae are generally between 700--1100 nm in diameter and \<20 nm thick \[[@B69-viruses-04-00363],[@B70-viruses-04-00363]\]. The polarity of the Golgi from cis-, medial-, to trans-Golgi is important for sequential processing and proper sorting of cargo. The mechanism by which cargo proteins move through the Golgi complex is controversial. It was originally thought this occurred by the maturation of cisternae \[[@B71-viruses-04-00363]\]. This model was challenged when intra-Golgi transport was reconstituted in vitro \[[@B72-viruses-04-00363],[@B73-viruses-04-00363]\]. However, some large cargo proteins such as procollagen I are too big to fit into normal transport vesicles. It was shown that procollagen I does not leave the Golgi cisternae during its transport, and instead the cisternae mature, as Golgi resident proteins are moved in a retrograde direction \[[@B74-viruses-04-00363],[@B75-viruses-04-00363],[@B76-viruses-04-00363]\]. However, a different large cargo complex (reversible protein aggregates) is transported between the Golgi cisternae in "megavesicles" \[[@B77-viruses-04-00363]\]. Thus, it is possible that both vesicle transport and cisternal maturation can move cargo through the Golgi complex in mammalian cells. Since virions are large cargo they would likely move through the Golgi complex via a cisternal maturation mechanism or in megavesicles. However, differences between trafficking of small and large cargo are not well understood \[[@B78-viruses-04-00363]\]. Some insight into the trafficking of virions can be gained by examining how cells normally traffic very large cargo. Procollagen I forms 300 nm triple helical bundles in the ER, which move through the secretory pathway. As mentioned above, the procollagen filaments do not appear to move between the Golgi cisternae in vesicles, implying that they move through the stack via a cisternal maturation mechanism \[[@B74-viruses-04-00363]\]. When procollagen bundles are present in the Golgi complex, the cisternae appear distended \[[@B79-viruses-04-00363],[@B80-viruses-04-00363]\]. Additionally, the trans-Golgi network (TGN) expands in cells producing collagen, and this expansion is correlated with efficient release of procollagen \[[@B81-viruses-04-00363]\]. The trafficking of lipoproteins induces similar changes in Golgi complex morphology. Chylomicrons are large (400 nm) complexes made up of apoB-48 bound to triacylglycerides \[[@B82-viruses-04-00363]\]. These complexes are formed in the ER of intestinal epithelial cells and traffic through the Golgi complex during their biogenesis \[[@B83-viruses-04-00363]\]. It has long been known that the induction of chylomicron production demands a substantial, but reversible, enlargement of the Golgi complex, as seen in electron micrographs of rat enterocytes after the animals were fed a fatty meal \[[@B84-viruses-04-00363]\]. The changes induced in Golgi complex morphology for both procollagen and chylomicrons seem to underlie a necessary step in their trafficking. How the expansion of the Golgi complex aids in their secretion, and the mechanism used to induce the reversible change(s) are not known. However, it seems likely that enveloped viruses that assemble intracellularly may have evolved to exploit this mechanism. 8.2. Impact of CoV Infection on the Secretory Pathway ----------------------------------------------------- Like the trafficking of large cargo, CoV infection leads to a rearrangement of endomembranes ([Figure 3](#viruses-04-00363-f003){ref-type="fig"}). It was initially observed that during MHV infection the Golgi complex is dispersed from its juxtanuclear position \[[@B85-viruses-04-00363]\]. This result has recently been expanded upon using quantitative immune‑EM, where it was shown that several different types of membrane rearrangements occur during MHV infection. Most relevant to the discussion here, virions were seen in large carriers derived from Golgi/ERGIC membranes, suggesting that the rearrangement of the Golgi complex may be important for virion trafficking \[[@B86-viruses-04-00363]\]. However, the mechanism underlying the change in Golgi complex morphology is not currently known. These changes appear similar to those observed for the large cargo described above, but may constitute a more dramatic change. ![CoVs use the secretory pathway for egress. (A) A cartoon depicting virions within the Golgi complex. The Golgi cisternae are enlarged and fragmented during infection, possibly to aid in the trafficking of virions. (B) Transmission electron micrographs of uninfected or IBV infected Vero cells. (1) The Golgi complex of an uninfected cell; (2) The Golgi complex of an IBV infected cell with enlarged Golgi cisternae; (3) A putative virion carrier. Arrows denote virions. Scale bar is 500 nm. For images (2) and (3) Vero cells were infected at an moi of 0.1, and samples were fixed and processed 14 hrs post infection as described in \[[@B31-viruses-04-00363]\].](viruses-04-00363-g003){#viruses-04-00363-f003} Besides CoVs, several other enveloped viruses assemble intracellularly. If modifying the secretory pathway is a requirement for virion trafficking, one would predict that an alteration of secretory compartments would be a general feature of cells infected with these viruses. Indeed, many of them do modify the secretory pathway in striking ways. Flaviviruses assemble and bud into the ER \[[@B87-viruses-04-00363]\]. Interestingly, during flavivirus infection Golgi proteins localize to the ER at the site of assembly during infection, but the function of this is not clear \[[@B88-viruses-04-00363],[@B89-viruses-04-00363]\]. Bunyaviruses assemble by budding into the Golgi lumen \[[@B90-viruses-04-00363],[@B91-viruses-04-00363]\]. During bunyavirus infection, the ERGIC and Golgi complex appear fragmented and vesiculated \[[@B92-viruses-04-00363]\], but the underlying mechanism is not known. While it may not be surprising that viruses rearrange cellular membranes for their own use, it is puzzling that these viruses seemingly disrupt the very organelles that they depend on for egress. However, the apparent disruption may represent an exaggeration of the mechanisms used by the large cargo described above. It seems possible, or maybe even plausible, that these viruses are exploiting a pre-existing mechanism to facilitate the release of large cargo from the secretory pathway. 8.3. Effect of CoV E on the Secretory Pathway and Role in Release of Virions ---------------------------------------------------------------------------- The results discussed above suggest a connection between the morphological changes during CoV infection and virion trafficking. However, the relationship between the two has been elusive. Recent studies of CoV E have shed light on this issue, and it is now apparent that a link exists between the E protein, virion trafficking, and morphological changes in the Golgi complex. Several studies have shown that the E protein is important for virion release. The E protein of TGEV is essential for the propagation of the virus \[[@B39-viruses-04-00363],[@B40-viruses-04-00363]\]. However, further characterization of TGEVΔE showed that virions were assembled in infected cells, but appeared arrested in the secretory pathway, apparently unable to traffic properly \[[@B36-viruses-04-00363]\]. Analysis of a version of SARS-CoV lacking the E gene showed that virions accumulated intracellularly with aberrant material, suggesting that they may be sorted improperly in the absence of E \[[@B35-viruses-04-00363]\]. As mentioned above, mutations introduced into the HD of MHV or IBV compromised the release of infectious particles from cells \[[@B31-viruses-04-00363],[@B67-viruses-04-00363]\]. Additionally it was shown that expression of IBV E altered the secretory pathway in a manner dependent on a single residue within its HD, an effect that correlates with the release of particles \[[@B31-viruses-04-00363],[@B66-viruses-04-00363]\]. These results have led to speculation that the E protein acts as an ion channel in the secretory pathway, driving the rearrangement of secretory organelles through the alteration of lumenal environments. This in turn leads to efficient trafficking of virions. This process may be analogous to what occurs for the large protein cargo mentioned above and should be further examined. 9. Topology of CoV E ==================== The topology of the CoV E protein is debated in the literature. The E protein does not have a canonical cleaved signal sequence \[[@B6-viruses-04-00363]\], suggesting that the E protein is likely a type II or type III membrane protein. However, that is where the consensus ends. The C-terminus of TGEV E was detected at the surface of non-permeabilized infected cells, suggesting it adopted an N~cyto~C~exo~ topology ([Figure 4](#viruses-04-00363-f004){ref-type="fig"}(3)) \[[@B12-viruses-04-00363]\]. IBV E has been reported as a transmembrane protein with the opposite topology, N~exo~C~cyto~ ([Figure 4](#viruses-04-00363-f004){ref-type="fig"}(1)) \[[@B5-viruses-04-00363]\] The C-terminus of MHV E is present in the cytoplasm of infected cells, and on the inside of the virion \[[@B6-viruses-04-00363]\]. Later it was shown, using an N-terminally FLAG tagged version of MHV E, that the N-terminus is present in the cytoplasm \[[@B11-viruses-04-00363]\]. These results suggested an N~cyto~C~cyto~ topology, which is very strange for a protein with a single predicted HD, and suggests that the HD forms a membrane hairpin ([Figure 4](#viruses-04-00363-f004){ref-type="fig"}(2)). Multiple topologies have been reported for SARS-CoV E. Using transient expression of an N- or C-terminally FLAG tagged version of SARS-CoV E, both termini were present in the cytoplasm, supporting the membrane hairpin topology \[[@B10-viruses-04-00363]\]. Additionally, this study also found that at least a portion of the expressed protein has its C-terminus modified with N-linked oligosaccharides, which would require the C-terminus to be in the ER lumen as a type II membrane protein \[[@B10-viruses-04-00363]\]. More recently, an untagged version of SARS CoV E was shown to largely adopt an N~exo~C~cyto~ topology in infected cells and in cells transiently expressing the protein \[[@B9-viruses-04-00363]\]. These results, along with those reported in \[[@B66-viruses-04-00363]\] demonstrate the danger of using epitope tags to determine topology. Complicating matters, when prediction programs are used to model the topology of the E protein the results are not consistent with, and often are in direct opposition to, what has been observed experimentally ([Table 1](#viruses-04-00363-t001){ref-type="table"}). ![Topologies of the CoV E protein. The three proposed topologies of CoV E. (1) shows a type III membrane protein; (2) shows a membrane hairpin; and (3) shows a type II membrane protein with a putative N-linked oligosaccharide.](viruses-04-00363-g004){#viruses-04-00363-f004} viruses-04-00363-t001_Table 1 ###### TM Pred \[[@B93-viruses-04-00363]\], HMMTop \[[@B94-viruses-04-00363]\], TMHMM 2.0 \[[@B95-viruses-04-00363]\], MEMSAT3 \[[@B96-viruses-04-00363]\], and MEMSAT-SVM \[[@B97-viruses-04-00363]\] were used to predict the topology of four different CoV E proteins. The location of the N- and C-termini, as well as the number of predicted transmembrane passes are shown. Program IBV E MHV E SARS E TGEV E ------------ ------- ------- -------- -------- ------- --- ------- ------- --- ------- ------- --- TM Pred lumen lumen 2 lumen cyto 1 lumen cyto 1 lumen cyto 1 HMMTop lumen lumen 2 cyto cyto 2 lumen cyto 1 cyto cyto 2 TMHMM 2.0 lumen lumen 2 lumen cyto 1 cyto lumen 1 lumen cyto 1 MEMSAT-SVM lumen lumen 2 lumen lumen 2 lumen lumen 2 cyto lumen 1 MEMSAT3 cyto cyto 2 lumen cyto 1 lumen lumen 2 lumen cyto 1 It seems possible that the E protein may adopt multiple topologies during infection. Certainly the putative ion channel activity of E would require a transmembrane protein, but another function, such as membrane bending in assembly, could require a membrane hairpin. To address this possibility, versions of IBV E were developed that adopted either a transmembrane (by adding a canonical cleaved signal sequence) or a membrane hairpin (by putting a FLAG tag on the N-terminus). The results showed that only the transmembrane protein could alter the secretory pathway (possibly through ion channel activity), but was not as efficient at forming particles as judged by VLP production \[[@B66-viruses-04-00363]\]. This may mean that a small portion of wild-type IBV E protein exists as a membrane hairpin and plays a role assembly. If CoV E is inserted into the membrane both in a transmembrane and membrane hairpin orientation, one question that arises is how a single protein might adopt multiple topologies. Certainly a type II or type III membrane protein could be generated using the signal recognition particle (SRP)-translocon pathway. However, can a membrane hairpin topology be achieved using SRP-translocon mediated membrane insertion? Caveolin is between 18--24 kD depending on the isoform, and has a \~32 amino acid long HD which is inserted as a hairpin \[[@B98-viruses-04-00363]\]. The membrane insertion of caveolin is dependent on Sec61, suggesting that the ER translocon can generate proteins with a membrane hairpin topology \[[@B99-viruses-04-00363]\]. Thus, it is possible that SRP-mediated membrane insertion could produce multiple topologies, including a membrane hairpin and a transmembrane protein. Nonetheless, the E protein is very different than caveolin, both in overall size and in the hydrophobicity of its HD. It is important to consider alternate membrane insertion pathways. Tail-anchored proteins are generally considered to have fewer than 30 amino acids following their transmembrane domain \[[@B100-viruses-04-00363]\]. Many CoV E proteins are just outside of this range, but no studies have investigated the role of the GET pathway (guided entry of tail anchored proteins) in CoV E membrane insertion. Finally, the possibility exists that the E protein may be inserted into the membrane via a novel mechanism of post‑translational insertion \[[@B101-viruses-04-00363]\]. It is interesting to speculate that CoV E may be inserted into the membrane via multiple mechanisms, resulting in distinct topologies. One last point is that the topology of CoV E proteins may differ for different CoVs. In addition to the variability shown in the prediction programs, there are also differences in the overall hydrophobicity and length of the transmembrane segments. How this impacts membrane topology is unknown, but it is worth noting that IBV E is the only E protein predicted to have two transmembrane domains. It is also possible that differences in the hydrophobic domains could affect the ion channel activity of the proteins. 10. Perspectives on CoV E ========================= The CoV E protein is an enigmatic protein. There is a high degree of variability in the behavior of the E proteins from different CoVs, including their requirement for assembly, virion trafficking, ion channel function, and method of expression in the genome. Yet, this protein is present in all known CoVs, suggesting it has a conserved role. It is interesting to speculate that the E protein from different CoVs has evolved to perform different functions. This could be due to cell-type specific requirements for each virus, or the ability of an accessory protein to complement a function normally carried out by E. Understanding how the E proteins from different CoVs vary is a crucial step in elucidating the mechanism of CoV assembly and egress. While the results reviewed here have better characterized the E protein, much work still remains to understand the mechanism of E protein function. The putative ion channel activity of E needs to be further examined. To this end, comprehensive electrophysiological analysis of the various E proteins in their natural setting (inserted into a Golgi membrane with normal post-translational modifications) needs to be carried out. It should also be noted that the E protein from different CoVs show variability in ion specificity, conductivity, and sensitivity to HMA. The reasons for this are not clear. Understanding how the ion channel activity of E affects replication is an important step in elucidating the function(s) of the protein with particular attention to cell types infected in the natural host. Recombinant viruses should be developed that encode other small ion channels along with E proteins that can support assembly but have no ion channel activity to determine if other channels can substitute for E. Additionally, analysis of the ion composition of the disrupted secretory pathway in IBV E expressing cells needs to be carried out to confirm that altered ion balance is responsible for the disruption. The idea that the E protein alters the secretory pathway to allow for virion trafficking is intriguing. If this modification is required for efficient trafficking of virions due to their large size, one would predict that the E protein should facilitate the trafficking of other large cargo such as procollagen I and chylomicrons. To this end, trafficking assays should be developed which allow for the quantitative monitoring of procollagen I and chylomicrons through the Golgi complex in the presence or absence of the E protein. It is also possible that the morphological changes in the Golgi complex are the end result of microenvironment alterations required to protect virions from proteolytic damage during egress. To test this, the lumenal ion concentration of both cells expressing E and infected cells should be investigated using ratiometric imaging. The topology of E appears to be predominately N~exo~C~cyto~; however, the existence of mixed membrane topologies cannot be ruled out. One possibility is that a small portion of the E protein is inserted as a membrane hairpin, and that this conformation facilitates particle assembly, but not efficient trafficking of virions. To address this possibility, recombinant viruses expressing topologically constrained versions of E (like those developed in \[[@B66-viruses-04-00363]\]) should be developed, and particle production as well as infectious particle release monitored. Additionally, in vitro membrane translocation assays should be carried out to determine the components necessary for membrane integration of the E protein (SRP mediated, GET mediated, or a novel route). Finally, because there is variability in the functions of E protein in different viruses, experiments to elucidate which functions are conserved and which are not should be performed. The ability of the different CoV E proteins to rescue the E deletions of other CoVs should be tested much the same way as they were for MHV \[[@B37-viruses-04-00363]\]. In addition, membrane associated accessory proteins from different CoVs should be examined in this same assay to determine if any of them can substitute for E. Finally, the tandem-affinity approach used to identify interacting proteins for SARS-CoV \[[@B41-viruses-04-00363]\] should be extended to other CoVs. It is now clear that the CoV E protein has multiple functions during infection, although they are not clearly understood. Also, the role(s) for the E protein in different CoVs may not be identical. A comprehensive model of E protein function will provide a better understanding of CoV biology and the mammalian secretory pathway, and may provide a target for therapeutic intervention during CoV infection. Acknowledgments =============== We would like to thank Andy Pekosz for thoughtful discussion on the E protein. The work in the authors' laboratory was funded by the NIH. TRR was supported by the Isaac Morris and Lucille Elizabeth Hay Graduate Fellowship. Conflict of Interest ==================== The authors declare no conflict of interest.	{ "pile_set_name": "PubMed Central" }
Introduction {#s1} ============ Unstructured Regions Define a New Heterogeneous Structural Reality {#s1a} ------------------------------------------------------------------ One central paradigm of structural biology is that the intricate details of 3-D protein structures determine protein function \[[@pcbi-0030140-b001],[@pcbi-0030140-b002]\]. In the last few years, many studies have shown that often the lack of a unique, native 3-D structure in physiological conditions can be crucial for function \[[@pcbi-0030140-b003]--[@pcbi-0030140-b021]\]. Such proteins are variously called disordered, unfolded, natively unstructured, or intrinsically unstructured proteins. A typical example is a protein that adopts a unique 3-D structure only upon binding to an interaction partner and thereby performs its biochemical function \[[@pcbi-0030140-b003]--[@pcbi-0030140-b006]\]. The better our experimental and computational means of identifying such proteins, the more we realize that they come in a great variety: some adopt regular secondary structure (helix or strand) upon binding, and some remain loopy. Some proteins are almost entirely unstructured, and others have only short unstructured regions. The more we can recognize short unstructured regions, the more we realize that the term "unstructured protein" would be misleading, as most unstructured proteins have relatively short unstructured regions. There is no single way to define unstructured regions. Here, we define an unstructured region as that which lacks unique 3-D structure by one of the following experimental techniques: circular dichroism (CD) spectroscopy, nuclear magnetic resonance (NMR) spectroscopy, X-ray crystallography, or proteolysis experiments \[[@pcbi-0030140-b007]--[@pcbi-0030140-b009]\]. Thanks to the outstanding data collection by the Dunker group, we could also describe this as regions that are the minimal common denominator between all proteins collected in DisProt \[[@pcbi-0030140-b010]\]. However, as we learned from prediction methods, DisProt and similar databases cover only a small fraction of all unstructured regions ([Figure 1](#pcbi-0030140-g001){ref-type="fig"}), and as we learned from recent experiments \[[@pcbi-0030140-b011]--[@pcbi-0030140-b013]\], there are many unstructured regions covered neither by these databases nor by existing prediction methods. ![Putative "Map" of Unstructured Regions\ Proteins with unstructured regions are likely to occupy large portions of sequence space \[[@pcbi-0030140-b007],[@pcbi-0030140-b024],[@pcbi-0030140-b027],[@pcbi-0030140-b042]\] as sketched by the light-gray inner rectangle. The space of all proteins with unstructured regions is likely to be considerably larger than what today\'s experimental techniques capture. The rounded darker gray rectangle labeled experiment sketches proteins for which some experimental method annotated natively unstructured regions. While most NORS regions (predicted long loops, striped gray ellipse) are likely to be natively unstructured, many unstructured regions are not NORS; i.e., they contain helices and strands even in their native form. Previous methods for the prediction of unstructured regions (left lens) are optimized to somehow reflect today\'s experiments. In contrast, the method introduced here (NORSnet, right lens) is developed based on predictions. This is an advantage because it avoids the bias of today\'s experimental techniques in a field that is just beginning to grasp its own dimensions, and it is a disadvantage because performance on today\'s datasets appears somehow limited.](pcbi.0030140.g001){#pcbi-0030140-g001} Unstructured Regions Can Be Defined and Recognized in Many Ways {#s1b} --------------------------------------------------------------- Methods that predict unstructured regions from sequence are mushrooming. Fast methods identify regions with high net charge and low hydrophobicity \[[@pcbi-0030140-b014],[@pcbi-0030140-b015]\], monitor the differences in amino acid propensities between unstructured and other regions (GlobPlot) \[[@pcbi-0030140-b016]\], or identify motifs associated with regions depleted of regular structure \[[@pcbi-0030140-b017],[@pcbi-0030140-b018]\]. Most methods are based on a different definition of disordered region that has been introduced by the Dunker group \[[@pcbi-0030140-b019]\]: residues for which X-ray structures do not have coordinates are considered as disordered. Methods based on this concept used neural networks \[[@pcbi-0030140-b019]--[@pcbi-0030140-b023]\] or support vector machines \[[@pcbi-0030140-b024]\]. The meetings for the Critical Assessment of Structure Prediction (CASP) have exclusively assessed disorder predictions on subsets of the "noncoordinate" data \[[@pcbi-0030140-b025],[@pcbi-0030140-b026]\]. The major drawback of this approach is that the Protein Data Bank (PDB) is biased toward proteins for which structures can be determined; natively unstructured proteins are underrepresented in the PDB \[[@pcbi-0030140-b005],[@pcbi-0030140-b010],[@pcbi-0030140-b024],[@pcbi-0030140-b027]\]. This may be one reason why most prediction methods tested by Oldfield et al. \[[@pcbi-0030140-b011],[@pcbi-0030140-b012]\] missed a substantial number of the proteins with unstructured regions identified in a large-scale NMR study spinoff from structural genomics. Other sequence features are predictive of disorder. For example, functionally flexible regions are identified from known structures through molecular dynamics simulation and can be generalized through machine learning. The Wiggle method provides predictions that overlap with unstructured regions even though it is focused on a different aspect of protein flexibility \[[@pcbi-0030140-b028]\]. Regions with No Regular Secondary Structure Provide Alternative {#s1c} --------------------------------------------------------------- Our group identified long regions with no regular secondary structure (NORS), which are stretches of 70 or more sequence-consecutive surface residues with few or no predicted helices and strands \[[@pcbi-0030140-b027]\]. NORS regions showed considerable overlap with proteins predicted to have long unstructured regions by various disorder predictors. NORS regions are overrepresented in eukaryotes (over five times more than in prokaryotes), overrepresented in regulatory and interacting proteins \[[@pcbi-0030140-b027],[@pcbi-0030140-b029]\], and share biophysical properties with unstructured regions. In addition, when natively unstructured regions are cocrystallized with their binding partner, they are still enriched in nonregular structure compared with globular proteins; ∼45% and ∼31% of the residues are in coils, respectively \[[@pcbi-0030140-b004]\]. Somewhat surprisingly, the method for predicting regular secondary structure in NORS regions, PROFsec (a profile-based neural network secondary structure predictor) \[[@pcbi-0030140-b030]--[@pcbi-0030140-b032]\], accurately predicts the secondary structure state in unstructured regions \[[@pcbi-0030140-b004]\]. NORS regions capture only one particular aspect of unstructured regions ([Figure 1](#pcbi-0030140-g001){ref-type="fig"}). The major advantages of our focus on NORS regions are that this definition implies a simple structural interpretation, and that we can reliably identify thousands of such regions by scanning entire organisms. The thresholds for the minimal length (70 residues) and for the definition of "largely loop" were optimized in order to minimize the identification of any of these stretches in the PDB \[[@pcbi-0030140-b027]\]. This procedure does not explicitly use any information about a protein other than its prediction of secondary structure and solvent accessibility. Thus, it mainly identifies extreme cases (e.g., highly exposed and long loop regions). Since many unstructured regions are shorter, one of our objectives was to capture much shorter NORS-like regions while ascertaining that we would not confuse long, well-structured loops with unstructured regions. One disadvantage of our focus on NORS was that some unstructured regions contain secondary structure elements (helix or strand) \[[@pcbi-0030140-b004]\]; i.e., not all unstructured regions are captured by NORS ([Figure 1](#pcbi-0030140-g001){ref-type="fig"}). Eukaryotic Disordered Regions Challenge Structural Genomics {#s1d} ----------------------------------------------------------- One goal of structural genomics is the determination of a 3-D structure representative for every protein family \[[@pcbi-0030140-b033],[@pcbi-0030140-b034]\]. Unstructured regions have not impeded structural genomics so far because almost all consortia have focused on bacterial proteins in order to increase the structure-to-clone ratio. However, consortia that focus on eukaryotes, such as the Northeast Structural Genomics (NESG) Consortium, or the Center for Eukaryotic Structural Genomics (CESG) have to carefully exclude such problematic targets \[[@pcbi-0030140-b035],[@pcbi-0030140-b036]\]. More than 10,000 proteins have been cloned and more than 3,000 proteins have been purified by NESG. Many of these did not adopt regular structure, possibly because they have unstructured regions that were not filtered out by our original filter, which discarded targets containing NORS regions \[[@pcbi-0030140-b029]\]. To speed up structure determination we need to increase the sensitivity in identifying unstructured regions \[[@pcbi-0030140-b011]\] (i.e., one goal of the development was to end up with a method that would be complementary to existing methods for the identification of unstructured regions). Our first hypothesis was that NORS regions share commonalities that distinguish such long unstructured loops from well-structured loops. If so, we should be able to somehow distinguish between the two types of loops at least in the sense that all loops predicted to be unstructured by our method ought to have different average features from other loops. We assumed that the neural network would pick up local correlations in amino-acid preferences for the different structural states. Our second hypothesis was that what distinguishes NORS regions from regular loops is exactly what makes regions become unstructured. If so, our method for the identification of NORS regions would also accurately predict unstructured regions. Here, we describe NORSnet, a new method that extends our NORS concept to also detect shorter (30--70 residues) NORS-like regions. The method was developed without ever using proteins with experimentally known unstructured regions. Instead, it was optimized to distinguish predicted NORS from all other regions. This unique approach, unprecedented in any machine learning method competing in a real-life application with other methods, has three important advantages. First, the data used for development and testing do not overlap. Since NORS regions were predicted from sequence, we can identify thousands of such regions. Our dataset was "dirty" in the sense that it contained many false negatives (all residues in PDB were considered to be well-structured during training) as well as some false positives (incorrect NORS predictions). This was the second major advantage: the positives (unstructured regions) sampled entirely sequenced organisms without any major bias with respect to this particular flavor of unstructured regions. Thereby, we identified unstructured regions that were missed by methods trained on more specialized datasets. The third advantage was that the resulting method explicitly focused on one feature of unstructured regions with a structural interpretation, namely that they are loops. Although we could have assessed NORSnet on any existing dataset due to the lack of overlap, we added a new set with experimental data about unstructured regions different from existing data. Note that both sets differed from each other as well as from the set used for development. Our three major results confirmed our hypothesis: (1) training on predictions succeeded in developing a powerful prediction method; (2) long loops are a major component of what is picked up by existing methods predicting unstructured regions; and (3) well-ordered and unstructured loops differ. In conjunction with existing methods, the one that we introduce here will allow the focus on particular structural aspects. Results/Discussion {#s2} ================== Accurate Distinction between Unstructured and Regular Loop Regions {#s2a} ------------------------------------------------------------------ We trained our system on NORS regions that had been predicted by our previous high-accuracy/low-coverage method \[[@pcbi-0030140-b027],[@pcbi-0030140-b029]\] for the identification of very long regions depleted of predicted helices and strands (NORSp; see [Methods](#s3){ref-type="sec"}). Technically, the task was to separate between all residues predicted to be in a NORS region and all residues in the PDB. As we used neural networks for this task, the typical assessment of accuracy usually involves a cross-validation experiment. For the first time in our work, we did not do this. In fact, we completely ignored the performance of the network on the task it optimized. Our hypothesis simply was that the only aspect that consistently separates extreme NORS regions from all residues in the PDB are the building blocks for a particular type of unstructured regions, namely the NORS-like loopy ones. Therefore, we measured performance on rather different datasets and separation tasks. First, we established success by predicting well-structured loops and NORS-like loops for DisProt, which consists of proteins with experimentally characterized unstructured regions. A total of 88% of the residues predicted by NORSnet were also predicted to be loops by PROFsec, while only 51% of the residues predicted as loops in DisProt also appeared NORS-like. In other words, most regions identified by NORSnet appeared to be in loops. Conversely, many loops were not predicted by NORSnet. Since residues in loops were identified through prediction, this difference may have been caused by prediction mistakes. To rule this out, we collected a set of 45 sequence-unique proteins that had been added to the PDB after we had completed developing our method (September 2005 to June 2006). We found that NORSnet classified only 1% of loop residues (Dictionary of Secondary Structure of Proteins states T, S, L) \[[@pcbi-0030140-b037]\] as natively unstructured regions. In other words, NORSnet largely succeeded for these new proteins. In fact, it predicted only one region in these structures to be unstructured, namely a stretch in the HIV type 1 P6 protein of 52 residues \[[@pcbi-0030140-b038]\], the NMR structure of which indicated depletion of regular secondary structure. This protein has been shown to undergo conformational changes \[[@pcbi-0030140-b038]\], suggesting that our method correctly identified it as unstructured. Very long NORS regions differ statistically from regularly structured or well-ordered loops \[[@pcbi-0030140-b027]\]. In general, unstructured regions that are not NORS-like tend to be more loopy than well-structured regions \[[@pcbi-0030140-b004]\]. Here, we showed that our ability to distinguish between well-ordered and unstructured loops was also successful for much shorter loops. Medium-length (30--70 residues) unstructured loops differed from well-structured loops ([Figure 2](#pcbi-0030140-g002){ref-type="fig"}). ![Regular, Flexible, and Predicted-To-Be Unstructured Loops Differed\ We compared the amino acid compositions between four different subsets representing four types of "loops" (nonhelix/nonstrand): loops from regular, well-ordered structures; i.e., from proteins without natively unstructured regions (states T, S, L from the Dictionary of Secondary Structure of Proteins; in blue); unstructured loops as predicted by NORSnet (in green); "flexible loops" from regular structures (TSL states with normalized B-factors ≥1 \[[@pcbi-0030140-b082]\]; in red); and unstructured regions as predicted by DISOPRED2 (in orange). The sign of the bar corresponds to overrepresentation (positive) or underrepresentation (negative) of amino acids in a subset with respect to the PDB. The NORS and DISOPRED2 residue subsets were taken from the worm genome (from the IntAct database \[[@pcbi-0030140-b067]\]) and were predicted to be unstructured by NORSnet and DISOPRED2. Flexible loops were enriched in amino acids with net charges such as lysine and glutamate (as described before \[[@pcbi-0030140-b016],[@pcbi-0030140-b039]\]). Predicted unstructured regions by NORSnet, however, differed in their composition from regular loops, flexible loops, and from any type of disorder that has been described previously (unpublished data) \[[@pcbi-0030140-b039],[@pcbi-0030140-b044]\]. Cysteines were not overabundant in the unstructured regions predicted by DISOPRED2. Overall, these data suggested that NORSnet captured something other than just "loop" and other than what is captured by methods such as DISOPRED2.](pcbi.0030140.g002){#pcbi-0030140-g002} NORSnet precisely distinguished between unstructured and well-structured loops. Although the amino acid composition of unstructured loops was similar to that in long disordered regions \[[@pcbi-0030140-b039]\], it was unique ([Figure 2](#pcbi-0030140-g002){ref-type="fig"}). For instance, the regions identified by our method contained significantly more cysteines than other PDB proteins and, within these, more than the set of residues unresolved in electron density maps. Thus, methods trained on unresolved residues, such as DISOPRED2, are likely to miss these regions. Furthermore, methods using pairwise energy potentials, such as IUPred, to predict unstructured regions are also likely to miss these regions, as many cysteines typically coincide with many paired cysteine bonds that significantly contribute to protein stability \[[@pcbi-0030140-b040],[@pcbi-0030140-b041]\]. Proteins with Unstructured Regions Accurately Identified {#s2b} -------------------------------------------------------- About 30%--60% of all eukaryotic proteins have been estimated to contain unstructured regions \[[@pcbi-0030140-b024],[@pcbi-0030140-b042]\]. However, DisProt \[[@pcbi-0030140-b010]\], the largest resource of experimentally verified unstructured regions, contains only a few hundred eukaryotic proteins, and thus covers a small fraction of sequence space ([Figure 1](#pcbi-0030140-g001){ref-type="fig"}). Moreover, this small fraction is not representative, as many unstructured regions described experimentally are missing from existing databases and are not identified by prediction methods \[[@pcbi-0030140-b011]\]. NORSnet attempted to solve both problems by sampling sequence space exhaustively (trained on all positives from entirely sequence organisms) and focusing on unstructured loops. To assess the accuracy of NORSnet and to estimate to what extent unstructured loops dominate our current identification of unstructured regions, we investigated two different datasets. The first was built around the DisProt database used previously in the literature; the second originated from careful NMR measurements and has not been used in many previous analyses. ### DisProt dataset. {#s2b1} The first set included proteins with unstructured regions from DisProt as positives and 173 PDB structures from EVA (a server for assessing protein structure prediction servers) as negatives (see [Methods](#s3){ref-type="sec"}). NORSnet correctly identified half of the DisProt proteins without false positives ([Figure 3](#pcbi-0030140-g003){ref-type="fig"}A). DISOPRED2 \[[@pcbi-0030140-b024]\] was ranked as one of the best three methods for predicting residues that are missing in electron density maps from X-ray crystallography at CASP6 \[[@pcbi-0030140-b026]\] and CASP7 (L. Bordoli, unpublished data). Many other studies corroborated the leading role of DISOPRED2 \[[@pcbi-0030140-b022],[@pcbi-0030140-b026],[@pcbi-0030140-b041],[@pcbi-0030140-b043],[@pcbi-0030140-b044]\]. Overall, NORSnet performed almost on a par with DISOPRED2 for the DisProt dataset ([Figure 3](#pcbi-0030140-g003){ref-type="fig"}A). Simply taking the average over the outputs of DISOPRED2 and NORSnet (DISOPRED2 + NORSnet) outperformed both individually. The improvement was particularly important for the realm of very high accuracy ([Figure 3](#pcbi-0030140-g003){ref-type="fig"}A). IUPred predicts unstructured regions based on a statistical potential optimized for this purpose \[[@pcbi-0030140-b041],[@pcbi-0030140-b045]\]. In our hands, IUPred clearly and consistently outperformed the other methods tested, including the averaged DISOPRED2/NORSnet output ([Figure 3](#pcbi-0030140-g003){ref-type="fig"}A). IUPred is optimized to identify all unstructured regions in DisProt \[[@pcbi-0030140-b041],[@pcbi-0030140-b045]\], but it cannot distinguish between unstructured regions dominated by loops and those dominated by regular secondary structure (as are often found in unstructured regions \[[@pcbi-0030140-b004]\]). ![Predictions for DisProt\ (A) ROC-like curve for NORSnet (green), DISOPRED2 (orange), and their combination (through arithmetic average; gray). While the performance of NORSnet and DISOPRED2 were similar, the combined method seemed to outperform both methods. Particularly, at accuracy = 100% (inset), the combined method covers significantly more sequences than each one of the methods individually. IUPred (purple) outperformed all other methods on this dataset. Note that IUPred was optimized on a set similar to the one used in this study. In contrast, NORSnet and DISOPRED2 were optimized on different sets defining disorder differently.\ (B) Venn diagram of overlap between very accurate predictions by NORSnet, DISOPRED2, and the combined method. The numbers in the circles are mutually exclusive; for instance, two proteins were identified only by DISOPRED2 to have an unstructured region, and 17 proteins were identified by both NORSnet and by the combined method to have an unstructured region.](pcbi.0030140.g003){#pcbi-0030140-g003} NORSnet predictions were not superior to those from DISOPRED2. However, the performance of these two methods was surprisingly similar despite the fact that NORSnet was not trained on a single experimentally verified unstructured region. Did the similarity in performance indicate that both methods picked up the same signal, i.e., that DISOPRED2 largely captured unstructured loops? If two prediction methods are based on very different information, their combination typically improves performance over any one of them \[[@pcbi-0030140-b044],[@pcbi-0030140-b046]\]. A more explicit way to demonstrate that methods focus on different aspects is the analysis of their predictions by Venn diagrams. We picked points for which each of the three methods (DISOPRED2, NORSnet, and DISOPRED2 + NORSnet) yielded 100% accuracy and compared the true positives predicted at those thresholds. DISOPRED2 and NORSnet identified the same 73 proteins, but each correctly identified proteins that the other missed ([Figure 3](#pcbi-0030140-g003){ref-type="fig"}B). This agreement supported our initial hypothesis that many unstructured regions are loopy (considerable overlap in true positives). But the most important result was that the two methods complemented each other. At the same 100% accuracy threshold, the combined method (DISOPRED2 + NORSnet) identified more proteins than any of the two individual methods and missed only two proteins that DISOPRED2 correctly identified. Although not surprising given the differences in training set and underlying optimizations, this result highlighted the difference in the types of unstructured regions identified. The combination of DISOPRED2 and NORSnet by averaging their outputs was better than either method alone. This did not work with IUPred and either of the two methods. This might suggest that IUPred covers the same aspects as the other two. However, this notion proved to be incorrect: IUPred missed proteins in the NESG dataset that the others captured ([Figure 4](#pcbi-0030140-g004){ref-type="fig"}B). Therefore, a beneficial combination of different methods predicting unstructured regions may require a more sophisticated algorithm. ![Predictions for NESG Data\ (A) The NESG set contains many proteins with unstructured regions that are not in DisProt and have never been used for method optimization. We compared NORSnet (in green), DISOPRED2 (in orange), their combined method (in gray), and IUPred (in purple) on these proteins. While DISOPRED2 performed better than all other methods in the low accuracy/high coverage region (top left), the combined method, NORSnet, and IUPred individually excelled in the high accuracy/low coverage region (lower right).\ (B) Venn diagram of overlap between very accurate predictions by NORSnet, DISOPRED2, and IUPred. The numbers in the circles are mutually exclusive. Note that five proteins were identified only by NORSnet to have an unstructured region.](pcbi.0030140.g004){#pcbi-0030140-g004} ### Unstructured regions from the NESG dataset. {#s2b2} Many prediction methods were optimized or benchmarked on datasets overlapping with DisProt. In contrast, the dataset from the NESG contained proteins with unstructured regions that have not been used for training existing methods yet. The NESG set was collected with a unified definition of unstructured regions based on 2-D NMR experiments \[[@pcbi-0030140-b047]\]; it included 30 proteins with unstructured regions as positives and 170 regular structures solved by NESG as negatives (Methods and [Table S1](#pcbi-0030140-st001){ref-type="supplementary-material"}). In the high accuracy region, NORSnet captured a considerable fraction of the positives (40% coverage at 100% accuracy; [Figure 4](#pcbi-0030140-g004){ref-type="fig"}A). The performance of DISOPRED2 was clearly lower than that of NORSnet for high accuracy/low coverage ([Figure 4](#pcbi-0030140-g004){ref-type="fig"}A, lower right), while the inverse was true for low accuracy/high coverage ([Figure 4](#pcbi-0030140-g004){ref-type="fig"}A, upper left). False positives from NORSnet (unstructured regions predicted and not observed) were almost equally divided between X-ray and NMR structures, while DISOPRED2′s false positives were predominantly from NMR structures. The most extreme examples for this were the ordered structures of Methanobacterium thermoautotrophicum 1615 \[[@pcbi-0030140-b048]\] and the conserved domain common to the transcription factors TFIIS, elongin A, and CRSP70 \[[@pcbi-0030140-b049]\]. ### Case study: NORSnet differed from other predictions. {#s2b3} As demonstrated above, NORSnet and other predictors give similar predictions with some exceptions. For instance, we applied NORSnet and two other prediction tools (DISOPRED2 and FoldIndex) on the Kappa-casein precursor protein that is found in milk and stabilizes micelle formation by preventing casein precipitation. Raman optical activity and thermal stability experiments revealed the protein as entirely unstructured in isolation \[[@pcbi-0030140-b050]\]. Secondary structure prediction methods such as PROFsec or PSIPRED \[[@pcbi-0030140-b051]\] predicted the protein to be highly enriched in loops ([Figure S1](#pcbi-0030140-sg001){ref-type="supplementary-material"}). We may therefore expect that the prediction of the Kappa-casein precursor as unstructured will be a simple task. However, the distinction between natively unstructured and well-structured loops is not trivial: DISOPRED2 did not identify the long loopy segment to be part of a natively unstructured region ([Figure 5](#pcbi-0030140-g005){ref-type="fig"}A). In contrast, NORSnet identified most of this protein to be unstructured in its strictest cutoff (corresponding to 100% accuracy on the DisProt dataset; [Figure 5](#pcbi-0030140-g005){ref-type="fig"}B). FoldIndex, a method that uses only amino acid composition and calculates the hydrophobicity/net charge within a given window, predicted only short segments of this protein to be unstructured ([Figure 5](#pcbi-0030140-g005){ref-type="fig"}C). ![Different Prediction Method Outputs for Kappa-Casein Precursor\ Kappa-casein precursor has been shown to be unstructured by different experiments \[[@pcbi-0030140-b050]\]. Despite its low content in predicted helices and strands, not all prediction methods identify it as unstructured. We compared outputs of DISOPRED2 (A), NORSnet (B), and FoldIndex (C) for this protein. For DISOPRED2 and NORSnet, higher values indicate unstructured regions; for FoldIndex, low values indicate unstructured regions (red). Note that FoldIndex and DISOPRED2 do not use any explicit information about secondary structure. DISOPRED2 disorder probability, however, is somewhat correlated with coil predictions ([Figure S1](#pcbi-0030140-sg001){ref-type="supplementary-material"}). DISOPRED2 was not able to distinguish these loops from structured loops. Only NORSnet clearly picked up the strong signal for unstructured regions for most of the protein.](pcbi.0030140.g005){#pcbi-0030140-g005} This example reveals that NORSnet and DISOPRED2 outputs are rather correlated. However, the signal from NORSnet clearly indicated unstructured regions, while the one from DISOPRED2 did not. One reason for this drastic difference may have been that NORSnet correctly captured some global feature from its global input units (see [Methods](#s3){ref-type="sec"}). ### Natively unstructured loops are elements of domain boundaries. {#s2b4} Although NORSnet was designed to identify all regions in any PDB structure as well-structured, the editor of this manuscript, Phil Bourne, suspected that NORSnet predictions of disorder might more often be in domain boundaries than expected at random and than expected for loop residues in general. To address this, we started with a sequence-unique subset of all PDB proteins considered to be multidomain by SCOP \[[@pcbi-0030140-b052]\] (set taken from \[[@pcbi-0030140-b053]\]). Although a much more comprehensive answer will remain the subject for future investigation, we clearly confirmed this assumption ([Figure S4](#pcbi-0030140-sg004){ref-type="supplementary-material"}); i.e., the regions in otherwise well-structured proteins that most resemble unstructured regions are domain linkers. ### Case study: DFF correctly identified despite being a tough case. {#s2b5} The DNA fragmentation factor (DFF) 45 must bind to DFF40 so that DFF40 can execute its catalytic function required for the onset of caspase-mediated apoptosis \[[@pcbi-0030140-b054]\]. The N-terminal domain (NTD) of DFF45 is natively unstructured: its folding is induced upon binding to DFF40 NTD \[[@pcbi-0030140-b055]\] ([Figure 6](#pcbi-0030140-g006){ref-type="fig"}). Methods that only use amino acid composition to predict unstructured regions are likely to perform better on such proteins than more complex prediction methods, since these proteins often have a high net charge which is neutralized upon binding to the target. For example, FoldIndex \[[@pcbi-0030140-b015]\] identified about a third of DFF45 as unstructured. ![NORSnet Captured Unstructured Regions Related to High Net Charge/Low Hydrophobicity\ DFF45 (white, yellow, and red) becomes structured upon complex formation with DFF40 (purple; \[[@pcbi-0030140-b055]\]). The interface includes a buried hydrophobic patch surrounded by hydrophilic interactions. Usually, charged residues disrupt the formation of tertiary structure; in this case, however, when the complex is formed, the negative charge of the Asp groups in DFF45 is cancelled out, with the positive charges of DFF40 allowing the protein to be folded. Visualization was done using GRASP2 \[[@pcbi-0030140-b085]\]. Since DFF45 has high secondary structure content, it is a relatively hard target for NORSnet prediction. However, NORSnet correctly identified its unstructured region at a rather stringent cutoff.](pcbi.0030140.g006){#pcbi-0030140-g006} Secondary structure-prediction methods, such as PSIPRED and PROFsec, usually predict the secondary structure of these regions the way they appear in substrate-bound form. Therefore, methods that use this type of information might be fooled by the rigidity and stability that are associated with regular secondary structure segments and identify these regions as well-structured. Since NORSnet uses secondary structure predictions as input, it may mispredict unstructured regions that become helices and strands upon binding. However, despite the fact that DFF45 NTD is enriched in regular secondary structure ([Figure S2](#pcbi-0030140-sg002){ref-type="supplementary-material"}), NORSnet identified NTD as an unstructured region at a rather stringent cutoff (the cutoff corresponded to 100% and 97.2% accuracy in the NESG and the DisProt sets, respectively). DISOPRED2 also identified NTD as unstructured, albeit at a less-stringent cutoff (corresponding to 72.2% and 94.2% accuracy). The unstructured regions in DFF45 are correctly identified by many prediction methods. NORSnet, DISOPRED2, and FoldIndex are only three of those. This example was one of 24 proteins with unstructured regions that become structured upon binding and were extensively analyzed in a recent study \[[@pcbi-0030140-b004]\]. NORSnet identified 14 of these proteins to have unstructured regions in its strictest cutoff. Again, this underlines the surprising finding that methods based on loop predictions can capture unstructured regions of this type. DFF45 and similar proteins are just some of many examples for unstructured regions involved in protein--protein interactions. How representative are they? Predicted Unstructured Regions Are Abundant in Protein--Protein Network Hubs {#s2c} ---------------------------------------------------------------------------- The structural plasticity of proteins with unstructured regions may enable its binding to many proteins, i.e., may typify a protein--protein interaction hub (a protein with many binding partners in an interaction network) \[[@pcbi-0030140-b006],[@pcbi-0030140-b056]--[@pcbi-0030140-b059]\]. Several detailed studies have specifically identified unstructured regions in hub proteins that are involved in signaling \[[@pcbi-0030140-b003],[@pcbi-0030140-b005],[@pcbi-0030140-b006],[@pcbi-0030140-b060]--[@pcbi-0030140-b062]\]. Natively unstructured regions are also predicted to be abundant in other regulatory processes (e.g., alternative splicing \[[@pcbi-0030140-b063]\] and transcription \[[@pcbi-0030140-b064]\]) and in cancer-associated signaling proteins \[[@pcbi-0030140-b065]\]. We addressed this point by correlating sustained large-scale datasets of physical protein--protein interactions (see [Methods](#s3){ref-type="sec"}) with predictions for unstructured regions. We applied NORSnet, DISOPRED2, and IUPred to all proteins in the worm (Caenorhabditis elegans) proteome and considered only predictions at thresholds corresponding to 100% accuracy. The subset of interacting proteins resulted from the high-throughput experiment by Vidal et al. \[[@pcbi-0030140-b066]\] and from IntAct \[[@pcbi-0030140-b067]\]. Predictions for unstructured regions for all three methods correlated with the average number of interacting partners; in other words, proteins with more unstructured regions had more binding partners ([Figure 7](#pcbi-0030140-g007){ref-type="fig"}). Since we used two different datasets to determine the thresholds for what constituted reliable predictions (DisProt and NESG), we also obtained two different thresholds for each method. For the purpose of fishing for hubs in protein--protein networks, we counted the number of proteins with unstructured regions according to any of those thresholds. Using DisProt to tune thresholds, DISOPRED2 predicted more proteins with unstructured regions than did NORSnet (1279 ± 88 versus 899 ± 76); using the NESG dataset, NORSnet predicted many times more proteins with unstructured regions than did DISOPRED2 (1282 ± 87 versus 321 ± 46; [Figure 7](#pcbi-0030140-g007){ref-type="fig"}). These results agreed with recent studies that estimated hub proteins to be enriched in unstructured regions \[[@pcbi-0030140-b057]--[@pcbi-0030140-b059]\]. However, could NORSnet identify any new unstructured regions in hub proteins? ![Unstructured Regions Overrepresented in Protein--Protein Hubs of the Worm\ We ran both NORSnet and DISOPRED2 on worm proteins that are involved in protein--protein interactions (as identified by yeast two-hybrid \[[@pcbi-0030140-b066]\]). The number of proteins that are predicted to be either unstructured or well-structured is plotted against the number of interacting partners for two different thresholds of reliability of the two methods: (A) and (B) were compiled for thresholds at which both methods maintained 100% accuracy for the NESG data ([Figure 4](#pcbi-0030140-g004){ref-type="fig"}), while (C) and (D) were compiled for 100% accuracy on DisProt ([Figure 3](#pcbi-0030140-g003){ref-type="fig"}). Since the number of observed interaction partners falls off dramatically, we had to group the data into bins of roughly equal sizes (x-axes). (A) and (C) show the results for the number of proteins predicted in each bin of interaction partners, while (B) and (D) show the normalized ratios to zoom into the difference between unstructured and structured proteins in each bin. These ratios were compiled as Ratio(bin) = {\#unstructured(bin) / \#structured(bin)} / {\#unstructured(1) / \#structured(1)}. As all ratios are greater than 1, proteins with more than one interaction partner have more unstructured regions than proteins with one partner.\ (A) These graphs were compiled with the reliability threshold at which each method achieved 100% accuracy by the NESG data ([Figure 4](#pcbi-0030140-g004){ref-type="fig"}). Overall, this threshold resulted in NORSnet (filled bars) predicting many more proteins with unstructured regions than DISOPRED2 (hatched bars). The difference was particularly relevant for proteins with more interacting partners.\ (B) NORSnet (filled, dark green) predicted many more unstructured regions in proteins with seven or more interaction partners than did DISOPRED2 (hatched, light green).\ (C) For the thresholds at which both methods achieved 100% accuracy on the DisProt dataset, DISOPRED2 identified more proteins with unstructured regions than did NORSnet. In contrast to the situation for the NESG set (A), the difference was not as significant for promiscuous proteins (ten or more partners).\ (D) Although NORSnet (filled, dark green) predicted as many unstructured as structured regions in hubs (seven or more), this ratio was significantly smaller than the one for proteins with a single interaction partner. In other words, even on this dataset NORSnet picked up a much stronger overrepresentation of unstructured regions in hubs than did DISOPRED2 (hatched, light green).](pcbi.0030140.g007){#pcbi-0030140-g007} We chose the cutoff that yielded the highest number of unstructured regions (NORSnet, 1,279; DISOPRED2, 1,282) for each method and checked whether the two methods predicted unstructured regions in the same hub proteins. Both methods predicted unstructured regions in most (74) of the proteins observed with more than ten partners (140). DISOPRED2 predicted unstructured regions in another 13 of the promiscuous proteins, and NORSnet in another 21 proteins. If the reliable predictions of both methods are correct, 77% of all promiscuous proteins in the worm (74 + 13 + 21 = 108 of 140) have unstructured regions. While these data do not suffice to identify hubs from sequence, we undoubtedly showed that methods such as NORSnet and DISOPRED2 clearly have some capability in the identification of unstructured regions that will adopt 3-D structures upon binding. While this finding was not new, our particular perspective was that the differences between DISOPRED2 and NORSnet resulted from the difference in the focus of the two. NORSnet focuses more on loopy regions than DISOPRED2, and it also identified more hub proteins. Similar results were obtained when we compared NORSnet and IUPred predictions on the same dataset. Again, IUPred identified the hub signal but much less clearly than did NORSnet ([Figure S3](#pcbi-0030140-sg003){ref-type="supplementary-material"}). All these observations suggested that the aspect of unstructured regions most relevant to hubs might actually be the unstructured loops. While NORSnet has some ability to identify unstructured regions that are often involved in binding ([Figure 6](#pcbi-0030140-g006){ref-type="fig"}), it may miss many of these regions due to their enrichment in regular secondary structure (helix, strand) in their bound form. We may therefore wonder why NORSnet identified so many worm hub proteins to have unstructured regions in the first place. Interestingly, many of the hubs had several modules/domains, some of which were predicted not to contain unstructured regions. Some of these modules were DNA-binding domains (such as Homeobox domains) or protein--protein interaction binding motifs (such as EGF repeats). The majority of the unstructured regions predicted by NORSnet in these hubs bridged connections between well-structured domains: these bridges were often on the surface (unpublished data). At first glance, the fact that these regions were predicted to be unstructured might seem biologically unimportant. However, there are several possible biological consequences of the abundance of hubs with unstructured loops. These exposed unstructured/loopy regions might serve as sites for proteolysis, allowing some parts of the protein to undergo proteolytic degradation under different cellular conditions. Such differential degradation could allow different modules of the same protein to be functional under different conditions. Alternatively, these long connecting loops might function as extremely flexible connecting linkers that facilitate the modules to adopt different orientations, thereby allowing the binding of different targets or binding similar targets in different fashion. Each of these alternatives could be at the heart of a different function. These two hypotheses may explain some of the regulatory characteristics of hub proteins. ### Mapping the sequence space of proteins with unstructured regions. {#s2c1} Most likely, unstructured regions and NORS regions occupy slightly different parts in sequence space ([Figure 1](#pcbi-0030140-g001){ref-type="fig"}). Indications for the overlap between NORS and unstructured regions are that both are enriched in proline and both depleted of glycine (\[[@pcbi-0030140-b039]\] and [Figure 2](#pcbi-0030140-g002){ref-type="fig"}). Some experimentally observed unstructured regions have been shown to contain cysteines. For instance, Zinc fingers often become structured only upon binding zinc. Nevertheless, most previous studies of unstructured regions did not find cysteines to be overrepresented with respect to well-structured regions in the PDB. This may be due partially to the fact that in well-structured proteins cysteines often stabilize disulfide bonds. Methods optimized to identify regions missing in electron density maps from X-ray crystallography are therefore likely to miss many of the cysteines in unstructured regions. In contrast, NORSnet captured cysteines in unstructured regions ([Figure 2](#pcbi-0030140-g002){ref-type="fig"}). In addition, the differences between DISOPRED2 and NORSnet that were revealed both by our head-to-head comparison on different sets of proteins with unstructured regions and by our analysis of protein hubs pointed to the different types of unstructured regions that we may have to separate ([Figure 1](#pcbi-0030140-g001){ref-type="fig"}). To complicate matters further, some proteins with unstructured regions may look just like any regular protein, while others may be generically different. Consequently, some of the proteins with unstructured regions may be missed by any prediction method. ### Refining target selection for structural genomics. {#s2c2} One goal of structural genomics projects is to contribute considerably to the increase in the fraction of proteins with known 3-D structures. To achieve this goal, 3-D structures are experimentally determined for representatives of as many large families as possible \[[@pcbi-0030140-b033],[@pcbi-0030140-b034],[@pcbi-0030140-b053],[@pcbi-0030140-b068]--[@pcbi-0030140-b070]\]. In particular, the large structural genomics initiatives financed by the Protein Structure Initiative (PSI) from the US National Institutes of Health (NIH) systematically target the experimental determination of structures for large families without representatives of known structure. Structural genomics also aims at making 3-D structures more readily accessible to nonstructural molecular biology and at reducing the costs and difficulty of determining structures. All of these goals require high-throughput determination of 3-D structures. This implies that experimental high-throughput pipelines have to move on if structure determination fails for some families, and that targets are also chosen with the objective to increase the throughput. This does not imply that PSI consortia "go for the low-hanging fruits." Quite to the contrary, they have succeeded where many small-scale studies have failed. Membrane proteins and proteins with unstructured regions are the two major types of proteins that are not only avoided by conventional small-scale structural biology but also by structural genomics efforts. Due to the fact that proteins with unstructured regions are much more abundant in eukaryotic organisms, consortia that focus on eukaryotes, such as NESG and CESG, have to carefully avoid such difficult targets. In the last six years, thousands of proteins have been cloned, expressed, and purified by NESG. Although the NESG target selection filtered out many domains with strong predictions for the presence of unstructured regions \[[@pcbi-0030140-b035],[@pcbi-0030140-b036]\], many were left for which biophysical data suggested that they contain unstructured regions \[[@pcbi-0030140-b013]\]. We applied NORSnet to 11,587 putative NESG targets that had already passed our previous and cruder NORS filter ([Table 1](#pcbi-0030140-t001){ref-type="table"}). Using two different cutoffs, NORSnet predicted that 13%--20% of the previously filtered targets have unstructured regions. Although NORSnet was not optimized to identify very short unstructured regions (≤30 residues), NORSnet predicted 47%--58% of the proteins to contain such regions. The same filter would not have excluded any of the proteins that succeeded in the experimental pipeline, suggesting that the application of NORSnet could have increased the structure--clone ratio. However, the ultimate proof for this assumption will have to wait until another hundred or so experimentally determined structures are added by NESG to the PDB in the next year(s). ###### NORSnet Predictions for Structural Genomics Targets ![](pcbi.0030140.t001) Conclusions {#s2d} ----------- The intricate details of protein 3-D structures are crucial for their functional role; i.e., structure determines function. Natively unstructured regions do not necessarily contradict this structure--function paradigm. Nevertheless, a variety of proteins require unstructured regions in order to function as domain linkers, filling material, and detergents. For other proteins with unstructured regions, changes in the environment (e.g., pH change, presence of target) or posttranslational modifications can trigger the formation of a regular 3-D structure that will then again determine function. In an evolutionary sense, the required changes/modifications constitute an integral part of the function and are therefore likely to be somehow encoded in the sequence of such proteins. The unusual aspect is that the key structural feature of these proteins is to keep regions natively unstructured or adaptable. The experimental and in silico identification and characterization of proteins with unstructured regions is evolving into an increasingly important challenge for structural biology. In facing this challenge, it becomes increasingly clear that the term "unstructured" describes a rather mixed bag of phenomena from regions that alter between different conformations to those that remain molten globule-like, and from regions that adopt regular helices and strands to those that remain intrinsically loopy. Here, we present NORSnet, a neural network--based method that revisited the task of identifying unstructured regions from a different angle than that taken by other methods. It focuses on the distinction between unstructured and well-structured loops. The success in this undertaking confirmed our initial hypothesis, namely that short unstructured loops resemble very long unstructured loops (NORS regions). Our application of machine learning was rather unconventional in two ways. First, we trained on positives (predicted NORS) that did not contain the feature we sought to predict (short unstructured loops) and on negatives (all regions in the PDB) that contained regions that we wanted the method to predict as positives; i.e., we implicitly hoped that our development would fail for many cases. Second, we did not optimize any parameters on the dataset used for assessing the performance of our method. Due to the difference in our approach, NORSnet complemented existing methods that optimize on previous datasets of unstructured regions. Consequently, NORSnet will enable the application of additional filters for structural genomics. Last, through a comparison between our new and other prediction methods, we confirmed the importance of unstructured regions for protein--protein interactions. Moreover, we specifically touched on the importance of unstructured loops for network complexity. Materials and Methods {#s3} ===================== Dataset of NORS regions. {#s3a} ------------------------ We created our dataset of residues in natively unstructured regions ("positives") in the following way. We grouped all proteins from 62 entirely sequence proteomes into domain-like families using CHOP and CLUP \[[@pcbi-0030140-b035],[@pcbi-0030140-b071],[@pcbi-0030140-b072]\]. We identified proteins with long NORS regions by the application of NORSp; i.e., all residues that are located in a stretch of \>70 consecutive residues with \<12% predicted helix or strand \[[@pcbi-0030140-b027],[@pcbi-0030140-b029]\] by PROFsec \[[@pcbi-0030140-b030]--[@pcbi-0030140-b032]\] and have at least one contiguous segment longer than ten residues predicted to be on the protein surface \[[@pcbi-0030140-b073]\]. The hope was that all residues in this pool have commonalities that we could extract through machine learning, and that will also be shared by proteins with unstructured regions much shorter than 70 residues. Due to the fact that PROFsec is especially accurate for natively unstructured regions \[[@pcbi-0030140-b004]\], the noise in these data that originated from the prediction mistakes was likely very low. To distinguish between proteins with and without unstructured regions, we needed a set of "negatives" (i.e., residues that are well-structured). For this, we chose a sequence-unique subset of globular protein structures from the PDB. Technically, this sequence-unique subset was taken from the EVA server \[[@pcbi-0030140-b074],[@pcbi-0030140-b075]\]. Specifically, the sequence redundancy was removed above HSSP (a measure for sequence-proximity) similarity values of 0 \[[@pcbi-0030140-b076],[@pcbi-0030140-b077]\] (corresponding to \<22% pairwise sequence identity for long alignments). Any pair of sequences between training and testing sets that could be aligned at PSI--BLAST \[[@pcbi-0030140-b078]\] E-values of \<10^−3^ according to our standard procedure of three automated iterations \[[@pcbi-0030140-b079]\] was also removed. To further amplify the signal from well-structured regions in the negative set, we also excluded all loops longer than 30 residues. Our datasets were not fully clean in the sense that our negative set of well-structured PDB proteins certainly contained some residues that did not appear in the X-ray structure (which were implicitly treated as well-structured), and that the positive set (predicted NORS) might contain some regular, ordered regions. However, due to the immense size of both datasets and to our use of neural networks, we did not worry about such outliers. In fact, our particular generation of a prediction-based training set that is more than ten times larger, and certainly more representative, than sets used previously might be the most important difference to all previous methods. In the context of a different problem, we showed how beneficial the use of prediction-based sets with errors might be \[[@pcbi-0030140-b080]\]. Training and testing set. {#s3b} ------------------------- To optimize the parameters of the method, we trained the network on 90% of the sequences and tested it on the remaining 10%. Note that these data were only used for the development of the method. We never reported the performance of the method on these data. The datasets on which we did assess NORSnet had no overlap (HSSP-value \<0; i.e., \<22% pairwise sequence identity for 250 aligned residues) with any of the proteins used for development. In particular, NORSnet was not optimized in any way on DisProt and the NESG dataset, as these were solely used to assess its performance. DisProt data. {#s3c} ------------- After optimizing our method to predict NORS regions (as described below in the prediction method section), we assessed NORSnet performance on different sets without any further optimization. In the first benchmark, we used DisProt proteins that have unstructured regions longer than 30 residues as positives and a sequence-unique subset of 173 PDB X-ray structures as negatives. The latter subset was taken from the EVA server \[[@pcbi-0030140-b074],[@pcbi-0030140-b075]\], and did not include sequences that were in the original training set. One particular advantage of testing our method on DisProt was that we did not have to run any additional cross-validation experiment since we used different proteins; respectively, the same proteins with different labels (all residues from PDB in DisProt were explicitly treated as "well-structured" by our training procedure). NESG dataset. {#s3d} ------------- To further validate the method, we tested it on a set of proteins from the NESG consortium. The positive set included 30 proteins that were identified to have unstructured regions ("NESG unfolded"), and the negative set included 170 recently determined protein structures. Both sets were identified as such by the NESG consortium. The definition of "unstructured region" was as follows: (1) HSQC (heteronuclear single quantum correlation) was high signal to noise and very low dispersion; and (2) hetNOE (heteronuclear Overhauser effect) data was clean negative (G. T. Montelione, personal communication). Using this set contributed to the removal of two types of biases in DisProt and similar databases. (1) Structure determination method: the negative set was almost equally divided between X-ray and NMR structures. (2) Length bias: while usually sequences selected for NMR structure determination are shorter than for X-ray determination, the NESG consortium reduced this artifact by using both methods in parallel to determine the structures of the same sequences. Thus, the length distribution of the NESG unfolded set is similar to the one of the folded set, in contrast to DisProt database, which consists of some much longer sequences (see [Table S1](#pcbi-0030140-st001){ref-type="supplementary-material"}). Protein--protein interaction set. {#s3e} --------------------------------- For the large-scale predictions of proteins that are involved in protein--protein interactions, we used the IntAct database (<http://www.ebi.ac.uk/intact>). IntAct includes both large- and small-scale experiments for different organisms \[[@pcbi-0030140-b067]\]. Specifically, we used proteins from interactions that were detected in a large-scale yeast two-hybrid screen of C. elegans (worm) proteins \[[@pcbi-0030140-b066]\]. The set included 2,622 proteins that participate in 4,039 interactions. Prediction method. {#s3f} ------------------ We used a standard feed-forward neural network described elsewhere in more detail \[[@pcbi-0030140-b030],[@pcbi-0030140-b032],[@pcbi-0030140-b073],[@pcbi-0030140-b081]\] The crucial novelty for the given task was the choice of input information. This choice was largely influenced by what we found to succeed in different contexts, namely for the prediction of normalized B-values \[[@pcbi-0030140-b082]\] and protein--protein interfaces \[[@pcbi-0030140-b083]\]. Local input information was taken from a sliding window of 13 sequence-consecutive residues (the prediction was for the central residue in that window). For each residue, we used the evolutionary profile (from PSI-BLAST alignments according to our standard protocol \[[@pcbi-0030140-b079]\]), the three-state secondary structure predicted by PROFsec \[[@pcbi-0030140-b030]--[@pcbi-0030140-b032]\], the two-state solvent accessibility state predicted by PROFacc (a profile-based neural network predictor of solvent accessibility) \[[@pcbi-0030140-b073]\], and the two-state flexibility prediction by PROFbval \[[@pcbi-0030140-b082],[@pcbi-0030140-b084]\]. Global input information was represented by the global amino acid composition (20 units), the composition in predicted secondary structure (three units), and solvent accessibility (two units), as well the length of the protein/domain-like fragment (three units as in \[[@pcbi-0030140-b082]\]), and the mean hydrophobicity divided by the net charge as was first suggested by Uversky et al. \[[@pcbi-0030140-b014]\]. DISOPRED2, FoldIndex, and IUPred. {#s3g} --------------------------------- We downloaded the DISOPRED2 package from <http://bioinf.cs.ucl.ac.uk/disopred> and installed it locally. The package included DISOPRED2 V0.2 and PSIPRED Version 2.45 (from November 2003). To assess its performance on our datasets, we ran the program using the default parameters. The prediction for casein precursor in [Figure 5](#pcbi-0030140-g005){ref-type="fig"}A was taken from the DISOPRED2 server. We ran FoldIndex using the server at <http://bip.weizmann.ac.il/fldbin/findex> (in September 2006) with default parameters. We ran IUPred using the server at <http://iupred.enzim.hu/index.html> (in December 2005 and January 2006) with default parameters. Supporting Information {#s4} ====================== ###### PSIPRED Prediction for Kappa-Casein Precursor The protein is predicted to have several long loops (residues 24--42, 89--125, and 130--171). Note that the location of the loops is correlated with high scores predicted by NORSnet and DISORPED2 that use this information. (7.1 MB TIF) ###### Click here for additional data file. ###### Secondary Structure Predictions of the N-Termini Domains of DFF45 Despite the fact that the N-term domain of DFF45 is unstructured, PSIPRED predicts secondary structure elements within that region. (5.0 MB TIF) ###### Click here for additional data file. ###### Unstructured Regions Overrepresented in Protein--Protein Hubs of Worm Similarly to [Figure 7](#pcbi-0030140-g007){ref-type="fig"}, we ran IUPred on worm proteins that are involved in protein--protein interactions. NORSnet data are identical to those presented in [Figure 7](#pcbi-0030140-g007){ref-type="fig"}. The number of proteins that are predicted to be either unstructured or well-structured is plotted against the number of interacting partners for two different thresholds of reliability of the two methods: (A) and (B) were compiled for thresholds at which both methods maintained 100% accuracy for the NESG data ([Figure 4](#pcbi-0030140-g004){ref-type="fig"}), while graphs (C) and (D) were compiled for 100% accuracy on DisProt ([Figure 3](#pcbi-0030140-g003){ref-type="fig"}). (A) and (C) show the results for the number of proteins predicted in each bin of interaction partners, while (B) and (D) show the normalized ratios to zoom into the difference between unstructured and structured proteins in each bin. These ratios were compiled as Ratio(bin) = {\#unstructured(bin)/\#structured(bin)} / {\#unstructured(1)/\#structured(1)}. As all ratios are greater than 1, proteins with more than one interaction partner have more unstructured regions than proteins with one partner. For the thresholds at which both methods achieved 100% accuracy on the DisProt dataset, both IUPred and NORSnet identified unstructured regions in 98 proteins that interact with seven partners or more. IUPred predicted 37 proteins with unstructured regions that NORSnet did not identify, and NORSnet predicted 17 proteins with unstructured regions that IUPred had missed. (7.0 MB TIF) ###### Click here for additional data file. ###### NORSnet Captures Domain Boundaries The domain boundaries of 524 multidomain proteins were marked in a procedure described in Liu and Rost \[[@pcbi-0030140-b053]\]. Due to the fact that NORSnet is optimized to identify unstructured stretches that are longer than 30 (and SCOP domain boundaries are often shorter), we used the raw score by NORSnet rather than the filtered output. NORSnet did considerably better than random (in red) and yielded area under the ROC curve (AUC) of 0.672 (in blue). Morever, according to our gold-standard set, termini residues are never defined as domain borders. In "NORSnet no term" (in green), we treated NORSnet outputs of the 60 termini residues in each protein as negatives, assessing only NORSnet predictions for the middle of the chain. The new method was more accurate in distingushing domain boundaries from other residues (AUC = 0.715). (5.2 MB TIF) ###### Click here for additional data file. ###### Synopsis for Supporting Online Material (624 KB DOC) ###### Click here for additional data file. ###### Dataset of Unstructured Proteins from Northeast Structural Genomics Consortium \(A\) NESG id refers to identifiers given by the NESG consortium. \(B\) Disorder signal referred to different levels of signal of a protein to be unstructured from NMR experiments. Largely marked largely unstructured proteins; e.g., (1) their HSQC has high signal to noise and very low dispersion and (2) their HetNOE data is clear negative. Partly marked partly unstructured proteins, which have some local structure but overall obey the same criteria. A total of 20 proteins were identified as largely unstructured and ten proteins were identified as partly unstructured. (63.5 KB DOC) ###### Click here for additional data file. ###### PDB Identifiers Used as a Negative Set in [Figure 3](#pcbi-0030140-g003){ref-type="fig"}A (74.5 KB DOC) ###### Click here for additional data file. Accession Numbers {#s4a} ----------------- The Protein Data Bank (<http://www.rcsb.org/pdb>) accession numbers for the structures discussed in this paper are HIV type 1 P6 protein (2c55_A), Methanobacterium thermoautotrophicum 1615 (1eij), the conserved domain common to the transcription factors TFIIS, elongin A, and CRSP70 (1eo0), and DFF40 (1ibx). The DisProt (<http://www.disprot.org>) accession number for bovine Kappa-casein precursor is DP00192. Thanks to Dariusz Przybylski and Guy Yachdav (Columbia University, United States) for providing preliminary information and programs, to Andrew Kernytsky and Marco Punta (Columbia University) for valuable discussions, and to Kazimierz Wrzeszczynski and Henry Bigelow (Columbia University) for helpful comments on the manuscript. Thanks to Jonathan Ward and David Jones (University College London, United Kingdom) for making DISOPRED2 and PSIPRED available, to Jaime Prilusky and Joel Sussman (Weizmann Institute, Rehovot, Israel) for making FoldIndex available, and to Zsuzsanna Dosztányi and István Simon (Institute of Enzymology, Hungary) for making IUPred available. Particular thanks to Guy Montelione and colleagues (Rutgers University, United States) for creating and providing the NESG datasets. Thanks to the constructive criticism of two anonymous reviewers and to those from the editor, Phil Bourne. Last, not least, thanks to all those who deposit their experimental data in public databases, and to those who maintain these databases, in particular to Keith Dunker and his colleagues for the maintenance of DisProt. The work of BR was also supported partially by grant U54-GM072980 from the US National Institutes of Health. Author contributions. AS and BR conceived and designed the experiments, analyzed the data, and wrote the paper. AS and JL performed the experiments. All authors contributed reagents/materials/analysis tools. Funding. The work was supported by grants U54-GM074958--01 from the Protein Structure Initiative of the US National Institutes of Health to the Northeast Structural Genomics Consortium and 2R01-LM07329--01 from the National Library of Medicine. Competing interests. The authors have declared that no competing interests exist. A previous version of this article appeared as an Early Online Release on June 5, 2007 (doi:[10.1371/journal.pcbi.0030140.eor](10.1371/journal.pcbi.0030140.eor)). DFF : DNA fragmentation factor NMR : nuclear magnetic resonance spectroscopy NORS : no regular secondary structure NTD : N-terminal domain PDB : Protein Data Bank PSI : Protein Structure Initiative	{ "pile_set_name": "PubMed Central" }
Introduction {#s1} ============ Clathrin is a three-legged molecule with a central hub domain from which three ∼190 kDa heavy chains are extended, each ending in an N-terminal seven-bladed β-propeller domain that allows for multiple protein interactions with various specificities between its blades [@pone.0003115-terHaar1]. A single clathrin heavy chain (CHC) molecule contains in addition eight CHC repeat segments, a proximal hairpin, a tripod region believed to be responsible for trimerisation, and a variable C-terminal segment [@pone.0003115-Fotin1]. Each CHC is furthermore associated with a ∼25 kDa clathrin light chain (CLC). This building block of a cage structure is known as a triskelion, and during endocytosis the legs of neighbouring triskelia twist around each other to form a curved lattice that self-polymerizes around invaginated pits, stabilizing them as they bud from the major sites of formation within the cell--plasma membrane, trans-Golgi network and endosomes [@pone.0003115-Crowther1]--[@pone.0003115-Keen1]. Recently, research has focussed on changes in the rates of endocytosis during cell cycle progression and in the distribution of trafficking proteins. This has resulted in some controversy in the literature over whether endocytosis is inhibited during mitosis or is maintained. An original study showed that in a broken assay mitotic cytosol could inhibit endocytosis when compared to interphase material [@pone.0003115-Pypaert1]. Latterly, single-cell imaging has been used to determine that whilst endocytosis is maintained during all phases of the cell cycle, recycling of internalised membrane is inhibited during mitosis [@pone.0003115-Boucrot1], [@pone.0003115-Boucrot2]. Clathrin has also been found at the mitotic spindle both through confocal imaging and proteomic analysis of enriched spindle fractions [@pone.0003115-Andersen1], [@pone.0003115-Okamoto1]. Knockdown of the heavy chain of clathrin in HEK293 and NRK cells using siRNA results in mitotic defects and this has led to the suggestion that clathrin may have a trafficking-independent function in mitosis [@pone.0003115-Royle1]. By contrast, components of the AP-1, AP-2 and AP-3 adaptor complexes did not colocalise with the spindle apparatus [@pone.0003115-Royle1]. Past controversy on changes in the endocytic rates during cell cycle progression suggests that it will prove important to explore the role of clathrin at the spindle in multiple cell-lines using multiple approaches. Consequently, we have used a chicken pre-B lymphoma cell line DT40, which was generated with endogenous alleles for CHC replaced by human CHC under the control of a tetracycline-regulatable promoter, in order to investigate the role of clathrin in the evolutionarily conserved process of clathrin-mediated endocytosis [@pone.0003115-Wettey1], [@pone.0003115-Wettey2]. Following repression of clathrin expression, receptor-mediated and fluid-phase endocytosis were significantly inhibited in a surviving sub-line (DKO-R) [@pone.0003115-Wettey1]. We have now used this well-characterised model of membrane trafficking to quantitatively test for the first time, using flow cytometry, the impact of clathrin knockout on cell cycle progression in a suspension cell-line. We found no difference in the cell cycle distribution of the knockout cells versus the wild-type. Additionally, we observed that the ploidy and recovery kinetics following cell cycle arrest with the microtubule-depolymerising agent nocodazole were unchanged by knocking out clathrin. Consequently, whilst clathrin is an important component of the trafficking machinery and colocalises with the mitotic spindle, in these cells, its contribution to mitosis is not appreciable. Results {#s2} ======= We first monitored the spatial distribution of CHC and alpha-tubulin during cell cycle using confocal microscopy, confirming their colocalization at the spindle apparatus during mitosis in the DKO-R line ([Figure 1A](#pone-0003115-g001){ref-type="fig"}). We then treated the line with doxycycline (50 ng/ml) to fully repress CHC expression, which was achieved after 48 hours ([Figure 1B](#pone-0003115-g001){ref-type="fig"}). Royle et al. had shown an increase in the mitotic index of cells treated with siRNA against CHC using confocal microscopy. To test this more quantitatively, we sought to analyse steady-state cell cycle distribution in the presence and absence of CHC using a propidium iodide flow cytometric assay. Intriguingly, we found no change in the G2/M phase proportion of CHC-depleted cells relative to wild-type, or evidence of changes in their ploidy status ([Figure 1C](#pone-0003115-g001){ref-type="fig"}). Nevertheless, the absence of steady-state changes would not preclude differences in the transition rate through cell cycle checkpoints. To investigate this, we added nocodazole (500 ng/ml) for 11 hours to induce full spindle disruption and thus arrest clathrin-expressing and knockout cells at the G2/M checkpoint[@pone.0003115-Sudo1]. We then sampled the cells over a time course (1, 2, 4, 8 and 24 hours) having washed out the drug, and measured the proportions in each phase of the cell cycle by flow cytometry. There were no statistically significant differences in the recovery rates between clathrin depleted and control cells ([Figure 1D](#pone-0003115-g001){ref-type="fig"}). Furthermore, an apoptosis assay conducted on these cells over the same time course showed no significant differences in the level of cell death between the recovering populations ([Figure 2](#pone-0003115-g002){ref-type="fig"}). Finally, we confirmed that the clathrin-depleted cells are able to assemble mitotic spindles and are mononuclear ([Figure 3](#pone-0003115-g003){ref-type="fig"}). On the other hand, the same flow cytometric approach combined with targeting clathrin for knockdown using siRNA in a HEK293 cell-line resulted in an increase in the G2/M population from 15.66% to 24.54% despite the presence of detectable residual clathrin (∼10%) in total cell lysates, as detected using the same clathrin antibody ([Figure 4](#pone-0003115-g004){ref-type="fig"}). Consequently whilst the repression of clathrin in DKO-R cells produced a greater reduction in detectable clathrin than siRNA on HEK293 cells, the functional effects of this reduction were only measurable in HEK293 cells. This implies a difference in the mitotic sensitivity of the two cell-lines to changes in clathrin levels. ![A) Clathrin heavy chain colocalizes with alpha-tubulin at the spindle apparatus in mitotic DKO-R cells.\ Confocal immunofluorescence microscopy on an asynchronised population of DKO-R cells grown in the absence of doxycycline (upper left panel). Cells were stained with Hoechst 33258 (blue; Sigma), anti-alpha-tubulin DM1A (green; AlexaFluor 488 donkey anti-mouse conjugate, Invitrogen) and anti-CHC Ab21679 (red; AlexaFluor 594 goat anti-rabbit conjugate, Invitrogen). The asterisk indicates a mitotic cell in which CHC is targeted to the kinetochore fibres. A Z-stack projection of the indicated mitotic cell is shown (right panel). Scale bar, 10 µm. B) Repression of CHC to undetectable levels is achieved after 48 hours treatment with doxycycline. Western blot analysis of DKO-R cells treated with doxycycline for the indicated time periods. β-tubulin was used as a loading control. C) Cell cycle distribution is unaffected by CHC depletion. DKO-R cells were treated with doxycycline for 48--72 hours and then subjected to flow cytometric analysis. DNA content was monitored by PI staining. The cell cycle profiles of CHC-expressing and CHC-depleted cells are shown. The percentage of cells in each cell cycle phase was calculated by ModFit Sofware and is shown in the boxed legend. D) The recovery kinetics from cell cycle arrest at the G2/M checkpoint are unaffected by CHC depletion in DKO-R cells. Time-course analysis of DKO-R cells following release from a nocodazole-induced block. Cells were grown in the presence (CHC-depleted) or absence (CHC-expressing) of doxycycline for 48--72 hours and then synchronized at metaphase by treatment with 500 ng/ml of nocodazole for 11 hours. A representative vehicle-treated asynchronous sample is indicated for each population ('Asynch' bar). At the indicated time points post-release from nocodazole block, cells were harvested and subjected to propidium iodide flow cytometric analysis. Data represent the mean values from three independent experiments with error bars denoting standard errors. The post-arrest recovery kinetics were compared for the clathrin-expressing and clathrin-depleted cells at each phase in the cell cycle using the analysis of variance (ANOVA). A p-value of 0.05 or less would be required to declare a significant difference between the two populations. P-values were 0.20 for G1-phase, 0.16 for S-phase and 0.09 for G2/M phase, indicating that there is no evidence of a difference between the recovery kinetics of the two populations over time following a nocodazole-induced block.](pone.0003115.g001){#pone-0003115-g001} ![Time course analysis of cell death levels after release from a nocodazole-induced metaphase block.\ Asynchronous and nocodazole-arrested CHC-expressing and -depleted DKO-R cells were harvested at the indicated time points and stained with Annexin V and 7-aminoactinomycinD (7-AAD) followed by flow cytometry analysis. Although there is an increase in the levels of cell death during the recovery period, which has been well-established to be an effect of microtubule-targeting agents [@pone.0003115-Jordan1], there are no significant difference in the proportion of dead cells between the two populations. Data represents values from one experiment.](pone.0003115.g002){#pone-0003115-g002} ![CHC-depleted DKO-R cells preserve the ability to assemble a mitotic spindle.\ Confocal immunofluorescence microscopy on an asynchronised population of DKO-R cells grown in the presence of 50 ng/ml doxycycline for 72 hours. Cells were stained with Hoechst 33258 (blue; Sigma), anti-alpha-tubulin DM1A (green; AlexaFluor 488 donkey anti-mouse conjugate, Invitrogen) and anti-CHC Ab21679 (red; AlexaFluor 594 goat anti-rabbit conjugate, Invitrogen). The asterisk indicates a mitotic cell. Scale bar, 10 µm.](pone.0003115.g003){#pone-0003115-g003} ![Clathrin knockdown in HEK293 cells leads to an increase in the G2/M phase content.\ A) Western blot analysis of HEK293 cells transfected with siRNA against CHC17 or with scrambled siRNA. β-actin was used as a loading control and two different clathrin heavy chain antibodies raised against the N-terminal domain (TD.1; Abcam) and the C-terminus (Ab21679; Abcam) were used on cell lysates harvested at the time of cell cycle analysis. B) The cell cycle profiles of scrambled siRNA treated and CHC siRNA treated cells are shown. The percentage of cells in each cell cycle phase was calculated by ModFit Sofware and is shown in the boxed legend.](pone.0003115.g004){#pone-0003115-g004} Discussion {#s3} ========== Given that clathrin does not seem critical for the function of the mitotic spindle in the DKO-R line, an alternative explanation for its localisation at the spindle apparatus remains that clathrin redistribution to the spindle is a by-product of Golgi disassembly which occurs during mitosis. Clathrin has been implicated in a number of nonendocytic functions including the maintenance of basolateral polarity involving the regulation of protein exit from the Golgi [@pone.0003115-Deborde1], and post-mitotic Golgi reassembly [@pone.0003115-Radulescu1]. A number of proteins required for the maintenance of the Golgi stack are known to associate with the spindle [@pone.0003115-Efimov1] and clathrin redistribution may be a marker for their recruitment. Clathrin is a stable protein with a slow turnover [@pone.0003115-Acton1] and consequently degrading and resynthesising this protein within the timeframe of M phase is not a feasible means of disassembling and regenerating subcellular organelles and vesicles during cell cycle progression. Our work strongly suggests that cell type and context are important determinants of the significance of the contribution that clathrin makes to mitosis, as has long been known for the role that it plays in endocytosis. For example, B cell antigen receptor internalisation was shown to be inhibited 70% by clathrin depletion in another DT40 variant cell-line (DKO-S), but to be unimpaired by depletion in the DKO-R line [@pone.0003115-Stoddart1]. Clathrin has also been shown to contribute to the delivery of Lamp1 and lysosomal proteins to the lysosome through the application of siRNA to knockdown expression in HeLa cells [@pone.0003115-Janvier1]. By contrast, no lysosomal defects or abnormalities in the distribution of lysosomal proteins have been observed upon clathrin depletion in the DKO-R cell-line [@pone.0003115-Wettey1]. DKO-R cells are not normally grown in the absence of clathrin, there is merely inducible depletion. However, there remains the possibility of transient adaptive changes to clathrin depletion in these cells over the 72-hour timecourse used in the DKO-R experiments,as there might be in other cells treated with siRNA for up to 72 hours. An expression array study is currently underway to address this point further. Nonetheless, just as in other cell-lines, aspects of receptor-mediated and fluid-phase endocytosis are impaired by clathrin depletion in the DKO-R cell-line. There is additionally a second clathrin heavy chain variant, CHC22, which is predominantly expressed in muscle and is highly similar in sequence to the ubiquitous CHC17 form introduced into the DKO-R cell-line. Whilst there is also a small amount of CHC22 detectable in this cell-line, it is not upregulated in response to doxycycline treatment and small levels are also found in HEK293 and HeLa cells (Dr. A. P. Jackson; Personal communication). Furthermore, CHC22 has not been detected at the mitotic spindle (Professor Frances Brodsky, UCSF; Personal communication). Consequently, we do not feel that compensation or adaptation to clathrin depletion through CHC22 is likely to happen in this case. In conclusion, further detailed study is required to better understand at the molecular level the recruitment of clathrin to the spindle, as this represents a common feature of chicken B-cells, mammalian [@pone.0003115-Okamoto1] and plant cells [@pone.0003115-Tahara1], and yet not of apparently equivalent functional significance for mitosis. Just as clathrin-independent or compensating endocytic mechanisms have been characterised, perhaps in the future both clathrin-dependent and clathrin-independent (or compensating) mechanisms for spindle assembly and chromosome segregation will be unearthed. Materials and Methods {#s4} ===================== Cell Culture {#s4a} ------------ DKO-R cells were cultured in RPMI media (Invitrogen) supplemented with 10% Foetal Bovine Serum (FBS; Invitrogen), 1% chicken serum (Invitrogen) and 10^−5^ M β-mercaptoethanol (Sigma) at 40°C in a humidified incubator supplied with 5% CO~2~ [@pone.0003115-Buerstedde1]. HEK293 cells were cultured in Dulbecco Modified Eagle\'s Medium (DMEM) containing 10% FBS at 37°C in a humidified incubator supplied with 5% CO~2~. Antibodies {#s4b} ---------- All of the antibodies used in this study were obtained from Abcam (Cambridge, UK). Three clathrin antibodies were used; TD.1 (N-terminal domain epitope; Ab24578), X22 (C-terminal domain epitope; Ab2731), and Ab21679 (C-terminal epitope). Loading control antibodies were against beta-tubulin (D66; Ab11307) and actin (AC-40; Ab11003). In addition an alpha-tubulin antibody (DM1A; T9026) obtained from Sigma was used to stain microtubules. siRNA knockdown of clathrin in HEK293 cells {#s4c} ------------------------------------------- The procedure used has been adapted from Motley et al., [@pone.0003115-Motley1]. Briefly HEK293 cells were seeded in full-serum medium into 100 mm dishes, at a density of 3×10^6^ cells per dish. The first transfection was carried out when cells had reached 80% confluency. For each 100 mm dish, 39 µl of DMRIE-C transfection reagent (Invitrogen) were added to 6 ml of serum-free DMEM, to which 70 nM of a set of 4 siRNA CLTC Duplexes (Dharmacon ON-TARGETplus Set of 4, Catalog No. LQ-004001-01) were added. Control siRNA was supplied by Qiagen (AllStars Negative Control siRNA (5 nmol Catalog No. 1027280). The transfection mixture was added to the cells after rinsing them once with serum-free DMEM, and the cells were left in this mixture until they were trypsinized the following day into two 100 mm dishes. 24 hours later a second transfection was carried out in a similar manner, and cells were equally trypsinized the following day and plated onto 100 mm dishes for protein extraction and cell cycle analysis 24 hours later. Western Blotting {#s4d} ---------------- Total cell extracts were prepared by lysing cells in Lysis Buffer containing 50 mM Tris-HCl pH 7.8, 150 mM NaCl, 5 mM EDTA, 15 mM MgCl~2~, 1% IgaPal, 0.75% sodium deoxycholate, 1 mM DTT, complete protease and phosphatase inhibitors (Roche) and 1 mM NEM. The extracts were cleared by centrifugation for 10 min at 14,200 rpm, followed by collection of the supernatant. Protein quantification was performed using Coomassie reagent and a BSA serial dilution for standard curve calculation. Cell lysates containing 15 µg of protein were analysed by 8% SDS-PAGE. Transfer to nitrocellulose membranes was carried out using the i-blot Dry Transfer System (Invitrogen), and membranes were immunoblotted with primary antibodies against clathrin heavy chain and beta-tubulin, followed by the appropriate HRP-conjugated secondary antibodies (Dako). Immunorreactive signals were visualized by ECL Plus (GE Healthcare). Apoptosis detection {#s4e} ------------------- DMSO- (vehicle) or nocodazole-treated wild-type and CHC-depleted DKO-R cells were stained with Annexin V-FITC (Cat\# TA5532, R&D Systems, Inc.) and 7-actinomycinD (7-AAD; Invitrogen) according to the manufacturer\'s instructions, and subsequently analysed by flow cytometry on a four laser LSR II SORP flow cytometer (BD Biosciences, San Jose, CA). The fluorescence emitted by Annexin V-FITC and 7-AAD was collected using a 530/30 and 575/26 bandpass filter. Data collected from the experiment was analysed with FlowJo. V 8.3 (Treestar, Ashland, OR). Immunofluorescence and Confocal Microscopy {#s4f} ------------------------------------------ DKO-R cells were prepared for imaging as described previously[@pone.0003115-Wettey1] with the following exceptions. Coverslips were mounted using ProLong Gold antifade reagent (Invitrogen) with 1 µg/ml Hoechst 33258 (Sigma), and the following primary antibodies were used: anti-alpha-tubulin (DM1A, Sigma) and anti-CHC (ab21679, Abcam). Images were acquired on a Nikon Eclipse C1si confocal microscope system (Nikon UK Limited, Surrey, UK), and processed using Volocity image analysis software (Improvision, Coventry, UK). Flow Cytometric Analysis {#s4g} ------------------------ DMSO- (vehicle) or nocodazole-treated wild-type and CHC-depleted DKO-R cells were stained with Propidium Iodide (Sigma) as described elsewhere[@pone.0003115-Hawkins1], followed by analysis on a four laser LSR II SORP flow cytometer (BD Biosciences, San Jose, CA). The fluorescence emitted by the Propidium Iodide was collected using a 575/26 bandpass filter. Data collected from the experiment was analysed with ModFit LT V3.0 (Verity Software House, Topsham, ME). We would like to acknowledge the wider support of The University of Cambridge, Cancer Research UK and Hutchison Whampoa Limited. The authors would like to acknowledge the support of the National Cancer Research Institute (NCRI) formed by the Department of Health, the Medical Research Council and Cancer Research UK. The authors would also like to acknowledge the helpful input of Heather Zecchini and to thank members of the CRUK Uro-Oncology Research Group and Fanni Gergely for helpful discussions during the preparation of this manuscript. Competing Interests:The authors have declared that no competing interests exist. Funding:JB is the recipient of an EU/Marie Curie Cancer Cure Early Stage Research Training (CANCURE) Fellowship. GV and IGM are supported by core CRUK funding. [^1]: Conceived and designed the experiments: JB IGM. Performed the experiments: JB. Analyzed the data: JB GV IGM. Contributed reagents/materials/analysis tools: APJ. Wrote the paper: JB IGM. Contributed to the preparation of the manuscript: APJ GV.	{ "pile_set_name": "PubMed Central" }
Introduction ============ It is estimated that above 30% of population and 65% of older population have hypertension worldwide ([@B43]). Previous studies suggest that hypertensive individuals have a greater chance of occurring the dementia and physical disability than normotensive individuals ([@B32]; [@B20]; [@B21]). Cognitive functions, especially the executive functions ([@B9]), are widely reported of being impaired in hypertensive individuals ([@B60]; [@B24]; [@B27]). Individuals with hypertension have higher chances of occurring executive dysfunction earlier than individuals with normotension, which indicates a potential vascular-cognitive association ([@B44]). However, there are very limited neuroimaging studies that investigated the neural bases of hypertension in cognitive decline ([@B35]). Former non-neuroimaging case studies further report that the African--Americans bear the greater burden of hypertension in the United States and have earlier onset of hypertension and larger hypertension-associated cognitive symptomatology and mortality than the other racial groups, including the Caucasian--Americans ([@B47]; [@B27]). The neural mechanisms underlying this racial disparity is largely under-investigated so far. A triple brain network model---a model consisting of the brain's default-mode, salience, and central-executive networks and their interactions---has been recently employed in the neuroimaging field to elucidate the differences in the connectivity patterns associated with the different levels of cognitive impairments such as higher versus lower impairment ([@B39]; [@B56]). We therefore use this model to investigate whether there is a difference in the impairment at the nodes of triple network between the African--Americans and the Caucasian--Americans hypertensive cognitively impaired individuals. Resting state functional MRI (rsfMRI) has been used to investigate the functional brain areas or neurocognitive networks ([@B4]; [@B46]). Recent neuroimaging investigations suggest that the neural basis of cognitive activity is related to a dynamically modulating interaction between multiple networks, including the salience, default-mode, and central-executive networks ([@B6]; [@B40]; [@B56]; [@B8]). The key nodes of the default-mode network include the posterior cingulate and the ventromedial prefrontal cortices, the salience network encompasses the insula and the dorsal anterior cingulate cortices, and the central-executive network comprises the posterior parietal and the dorsolateral prefrontal cortices ([@B13]). It has been demonstrated that the insula and dorsal anterior cingulate of salience network are anatomically connected ([@B5]; [@B31]) and consist of a special type of neurons named von Economo neurons that relay information processed within these regions to other brain regions, including the nodes of default-mode and central-executive networks ([@B2], [@B1]; [@B62]; [@B51]). This control signal by the insula and the dorsal anterior cingulate cortex has been suggested to be crucial for cognitive maintenance, including a rest, in cognitively normal individuals ([@B51]; [@B25]; [@B8]). Alternation in insula connectivity has been consistently implicated in diseases, including autism, frontotemporal dementia, and schizophrenia ([@B39]; [@B56]), but it has not been elucidated in cognitively impaired hypertensive patients. Literature suggests that the insula and anterior cingulate cortex---the key regions of salience network---respond as the racially biased brain regions ([@B7]), such as the greater activity of insula to faces of foreign races than faces of the same race of the subject ([@B37]; [@B38]). However, the difference in information flow---a measure from information theory that can be quantified using Granger Causality analysis ([@B17]; [@B10])---in these regions between the African--Americans and the Caucasian--Americans themselves has not been previously investigated. Previous investigations consistently report that the functional changes of brain regions (or networks) are associated with the underlying structural changes of those regions (or networks) with the progression of diseases ([@B63]; [@B40]). Specifically, recent studies suggest that the insula thickness decreases with the progression of cognitive decline in mild cognitive impairment (MCI) patients ([@B28]; [@B41]). However, whether there is a difference in insula thickness of the cognitively impaired hypertensive patients between the African--Americans and the Caucasian--Americans races has not been reported. Here, we seek to examine the difference in information flow using Granger causality ([@B17]; [@B10]) at the default-mode, salience and central-executive nodes between the African--Americans and the Caucasian--Americans hypertensive cognitive impaired individuals. As the African--Americans have higher hypertension-associated cognitive symptomatology and mortality than the other racial groups ([@B47]; [@B27]), we hypothesized that (1) the control signal of the insula of salience network over the default-mode and central-executive nodes is more impaired (more negative value) in the African--Americans than in the Caucasian--Americans. We further seek to examine the structural difference that could substrate this racial disparity by comparing the insula thickness between the two racial groups. To test this, we further hypothesized that (2) the insula thickness is lower in the African--Americans than in the Caucasian--Americans, and finally (3) lower insula thickness is associated with poorer cognitive performance. Materials and Methods {#s1} ===================== Participants ------------ This study was carried out in accordance with the recommendations of "Institutional Review Board (IRB) of Emory University" with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The study and the protocol were reviewed and approved by Institutional Review Board of Emory University. The informed written consent was provided by the participants before data collection. We recorded magnetic resonance imaging (MRI) data from 78 individuals who had hypertension and MCI. The inclusion criteria were: (a) age ≥ 55 years, (b) hypertension defined by systolic blood pressure ≥ 140 mm Hg or diastolic blood pressure ≥ 90 mm Hg and (c) MCI was assessed based on previously defined Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria ([@B12]; [@B45]): Montreal cognitive assessment (MoCA) ≤ 26, clinical dementia rating score of 0.5, minimal functional limitation as reflected by the functional assessment questionnaire ≤ 7, and cognitive performance at the 10th percentile or below till the 2nd percentile on at least one of four screening tests---trail marking test B ([@B48]; [@B54]), Stroop interference ([@B53]; [@B33]), digit span forward and digit span backward ([@B30]; [@B45]), verbal fluency and abstraction ([@B29]; [@B55]). Trail marking test B screens the participant's executive ability to draw a line from a 'number' to a 'letter' in ascending order such as '1' to 'A', 'A' to '2', '2' to 'B', and so on. Strop interference effect measures the interference of predominant response in the reaction time of a task such as when the name of a word (say 'red') is printed in a different color (say 'blue'), it takes longer time to name the color of that word compared to when the word ('red') matches the name of color ('red'). Digit span forward and digit span backward test consists of two parts: first, the participant listens to and repeats a sequence of numbers, and second, the participant listens to a sequence of numbers and repeats those numbers in reverse order. The former part screens the short-term auditory memory while the latter part screens the participant's ability to manipulate the verbal information based on the auditory information. The verbal fluency screens the participant's fluency such as the participant is asked to tell as many words as the participant can that begin with certain letter (say 'A') in a limited time and the abstraction test screens the participant's ability to deal with ideas such as how an orange and a banana are alike. These tests have been commonly used to screen the MCI patients ([@B12]; [@B45]). The participants exclusion criteria were: (a) systolic blood pressure \> 200 mm Hg or diastolic blood pressure \> 110 mm Hg, (b) renal disease or hyperkalemia, (c) active medical or psychiatric problems, (d) uncontrolled congestive heart failure (shortness of breath at rest or evidence of pulmonary edema on exam), (e) history of stroke in the past 3 years, (f) ineligibility for MRI (metal implants or cardiac pacemaker), (g) inability to complete cognitive test and MRI scan, (h) women of childbearing potential and (i) diagnosis of dementia (self-reported or care-giver reported). In total sample, the mean age was 66.9 years (SD: 9.7), 55.1% were women, mean education was 14.9 years (SD: 2.6), mean systolic blood pressure 144.4 mm of Hg (SD: 22.6), mean diastolic blood pressure 86.4 mm of Hg (SD: 12.8), and MoCA ranged from 11 (minimum value) to 26 (maximum value) with mean score of 21.9 (SD: 3.1). Out of 78 participants, there were 50 African--Americans and 28 Caucasian--Americans. In the African--Americans, the mean age was 66.3 years (SD: 8.9), 60% were women, mean education was 14.9 years (SD: 2.5), mean systolic blood pressure 141.8 mm of Hg (SD: 21.3), mean diastolic blood pressure 85.4 mm of Hg (SD: 12.0), and mean MoCA score was 21.2 (SD: 3.2). In the Caucasian--Americans, the mean age was 68.5 years (SD: 10.7), 46.4% were women, mean education was 15.0 years (SD: 2.9), mean systolic blood pressure 147.6 mm of Hg (SD: 24.0), mean diastolic blood pressure 87.0 mm of Hg (SD: 12.7), and mean MoCA score was 23.4 (SD: 2.4). The age, sex, education year, systolic blood pressure, and diastolic blood pressure were not statistically significant different between the African--Americans and the Caucasian--Americans, but the MoCA score was significantly lower in the African--Americans compared to the Caucasian--Americans (p \< 0.003) as shown in Table [1](#T1){ref-type="table"}. ###### Mean scores (standard deviations) and statistical comparison between the African--Americans (AA) and the Caucasian--Americans (CA) regarding their age, sex, education, blood pressure (BP), and Montreal cognitive assessment (MoCA) score. Characteristic N Total sample 78 AA 50 CA 28 p-Value --------------------- ----------------- -------------- -------------- ----------- Age, year 66.9 (9.7) 66.3 (8.9) 68.5 (10.7) 0.44 Sex, women 43 (55.1%) 30 (60%) 13 (46.4%) 0.25 Education, year 14.9 (2.6) 14.9 (2.5) 15.0 (2.9) 0.54 Systolic BP, mm Hg 144.4 (22.6) 141.8 (21.3) 147.6 (24.0) 0.14 Diastolic BP, mm Hg 86.4 (12.8) 85.4 (12.0) 87.0 (12.7) 0.71 MoCA score 21.9 (3.1) 21.2 (3.2) 23.4 (2.4) 0.003 p-Value indicates the level of statistical significance of the comparisons between the African--Americans and the Caucasian--Americans using non-parametric Wilcoxon rank sum for age, education, blood pressure and MoCA, and chi-square test for sex. MRI Acquisition --------------- Magnetic resonance imaging data were acquired on a SIEMENS Trio 3-Tesla scanner available at Center for Systems Imaging of Emory University, Atlanta, GA, United States. Foam padding and ear forms were used to limit head motion and reduce scanner noise to the participants. High-resolution 3D anatomical images were acquired using sagittal T1-weighted magnetization-prepared rapid gradient echo with repetition time = 2300 ms, echo time = 2.89 ms, inversion time = 800 ms, flip angle = 8°, resolution = 256 × 256 matrix, slices = 176, thickness = 1 mm. The rsfMRI were collected axially for 170 volumes during 7.14 min by using an echo-planar imaging (EPI) sequence with repetition time = 2500 ms, echo time = 27 ms, flip angle = 90°, field of view = 22 cm, resolution = 74 × 74 matrix, slices = 48, thickness = 3 mm and bandwidth = 2598 Hz/pixel. We requested the participants to hold still, keep their eyes open and think nothing during the rsfMRI scan. Image Preprocessing and Time Series Extraction ---------------------------------------------- Images were preprocessed for slice-timing correction, motion correction, co-registration to individual anatomical image, normalization to the Montreal Neurological Institute (MNI) template, and spatial smoothing of the normalized images with a 6 mm isotropic Gaussian kernel. The SPM12 (Wellcome Trust Centre for Neuroimaging, London, United Kingdom^[1](#fn01){ref-type="fn"}^) was used to perform those steps. We defined spherical regions of interest with 6 mm radius based on MNI coordinates centered at the posterior cingulate cortex (7, -43, 33) and ventromedial prefrontal cortex (2, 36, -10) of default-mode network, the insula (37, 25, -4) and dorsal anterior cingulate cortex (4, 30, 30) of salience network, and the posterior parietal cortex (54, -50, 50) and dorsolateral prefrontal cortex (45, 16, 45) of central-executive networks (see Figure [1](#F1){ref-type="fig"}) similar to the previous studies ([@B51]; [@B8]). We selected the nodes only in the right hemisphere based on most prior neuroimaging studies that report the right-lateralized activations ([@B51]; [@B13]; [@B8]). The MarsBaR software package^[2](#fn02){ref-type="fn"}^ was used to extract the voxel time courses of those nodes. ![Selection of the (A) default-mode (VMPFC, ventromedial prefrontal cortex; PPC, posterior cingulate cortex), (B) salience (AI, insula; DACC, dorsal anterior cingulate cortex), and (C) central-executive (DLPFC, dorsolateral prefrontal cortex; PPC, posterior parietal cortex) networks (P, posterior; A, anterior; L, left; R, right).](fnagi-09-00177-g001){#F1} Granger Causality Analysis -------------------------- Multivariate analysis has become a commonplace to investigate the information flow between the brain areas and to study how such coordinated brain activity disrupts in diseases ([@B16],[@B17]; [@B14]; [@B23]). Here we used Granger causality although other methods such as dynamic causal modeling, directed transfer function, and partial directed coherence provide the similar goals and results ([@B3]; [@B11]). The main benefits of using Granger causality are that it is a data-driven method and therefore computes the information flow based on the data itself at the nodes and networks level, relies on fewer assumptions about the underlying interactions, and does not need computationally intensive time/efforts as opposed to other methods such as dynamical causal modeling ([@B52]; [@B9], [@B10]). Recent studies by our group and by other groups have successfully applied Granger causality to resting state and/or task fMRI data in both health and disease and have produced meaningful results in terms of information flow at the brain nodes and networks ([@B51]; [@B14]; [@B36]; [@B3]; [@B8]). Granger causality can be mathematically expressed by considering simultaneously measured time series. Suppose we have two simultaneously recorded time series represented as, (1) X~1~(1), X~1~(2),\..., X~1~(t),\... and (2) X~2~(1), X~2~(2),\..., X~2~(t). Granger causality analysis in the frequency (f) domain examines the strengths, directions, and frequencies of interactions between dynamic processes. Granger causality from the second time series '2' to the first time series '1' (i.e., from brain region '2' to brain region '1') is computed as ([@B16],[@B17]; [@B9]), M 2 → 1 ( f ) = − ln ( 1 − ( ∑ 22 − ∑ 12 2 / ∑ 11 ) \\| H 12 ( f ) \\| 2 S 11 ( f ) ) where H is a transfer function, f represents a frequency-domain, S is spectral power, and ∑ is noise covariance. The value of Granger causality (M) varies between 0 and +∞, representing zero connectivity strength and maximum connectivity strength, respectively. If there are 'N' numbers of brain areas, the information outflow (F) at a node i can be calculated as, F i = 1 N − 1 ∑ j N ( M i → j − M j → i ) . In our case, we have six nodes (two key nodes from each network). Therefore, an index j can be 1, 2, 3, 4, 5, and 6 nodes. The Granger causality outflow (also referred as the net information outflow) from the first node is F~1~ = \[(M~1→2~ -- M~2→1~) + (M~1→3~ -- M~3→1~) + (M~1→4~ -- M~4→1~) + (M~→5~ -- M~5→1~) + (M~1→6~ -- M~6→1~)\]/5, where M~1→2~ is Granger causality from the first node to the second node, and M~2→1~ is Granger causality from the second node to the first node. Similarly, we calculate Granger causality outflows for other nodes. If the net outflow is negative at a node instead of the previously reported positive value in healthy individuals, then that node is said to have impaired directional connections. Cortical Thickness Calculation ------------------------------ FreeSurfer version 5.3^[3](#fn03){ref-type="fn"}^ was used to calculate the cortical thickness from T1-structural images. Briefly, this technique included spatial and intensity normalization, skull stripping, and an automated segmentation of cerebral white matter to locate the gray--white boundary ([@B15]). Cortical thickness was then computed from the distance between the gray--white boundary and the pial-surface ([@B22]). Cognitive Test -------------- We assessed the MoCA ([@B42]) of each participant. The MoCA is a 30-point scale test administered in 10 min and assesses the global cognitive abilities. It encompasses the following sub-tests. The short-term memory recall (five points) consists of two learning trials of five nouns and delayed recall after 5 min. Visuospatial test includes a clock-drawing (three points) and a three-dimensional cube copy (one point). Executive functions test comprises the trail-making time B (one point), a phonemic fluency (one point), and a two-item verbal abstraction (two points). Attention, concentration, and working memory include a sustained attention (one point), a serial subtraction (three points), and digit forward (one point) and digit backward (one point). Language test consists of naming animals (three points), repetition of two syntactically complex sentences and fluency (two points). Orientation test comprises orientation to time and place (six points). Moreover, if participant's formal education is 12 years or less, one point is added to his/her score. Previous studies ([@B42]; [@B26]; [@B18]; [@B49]) suggest that the MoCA is more sensitive screening tool to define the MCI as compared with other existing screening tools such as mini-mental state examination (MMSE). Statistical Analysis -------------------- We first checked whether the data are normally distributed or not using Kolmogorov--Smirnov test. The age, education, MoCA, insular thickness and net flow were negatively skewed and the systolic and diastolic blood pressures were positively skewed (asymptotic p-value in the range 6.17 × 10^-71^ to 8.42 × 10^-14^) at the 0.05 significance level. As data variables did not show normal distributions before and after applying the appropriate transformations (square root, log, and reciprocal), we therefore chose the non-parametric alternative. We compared the sample characteristics between the African--Americans and the Caucasian--Americans using non-parametric Wilcoxon rank sum and/or chi-square test for discrete variables (e.g., sex). Net information outflows between the nodes of three networks were compared using non-parametric Wilcoxon rank sum test. The correlation analysis was performed using Spearman's correlation. A p-value less than 0.05 was considered a statistically significant. MATLAB (Natick, MA, United States^[4](#fn04){ref-type="fn"}^) was used for analyzing the data. Results ======= Interactions among the Salience, Central-Executive, and Default-Mode Nodes -------------------------------------------------------------------------- We computed the Granger causality between all possible pairs of the salience, central-executive, and default-mode nodes and calculated the net information outflow from each node. Our net outflow calculation showed that the nodes of salience network have significantly lower outflow than that of the central-executive and default-mode nodes (Figure [2](#F2){ref-type="fig"}) (Wilcoxon rank sum; Z = 10.52; p \< 0.05). This negative information outflow of salience network nodes compared to previously reported positive flow in healthy individuals implied the impaired salience network nodes in our overall cohort. We compared the net information outflow between African Americans and Caucasian--Americans and found that the insula of salience network has significantly lower (negative value) in the African--Americans than in the Caucasian--Americans (Wilcoxon rank sum; Z = 2.06; p \< 0.05) as shown in Figure [3](#F3){ref-type="fig"}. ![Net flow (information outflow minus information inflow) of the key nodes of the salience (IN, insula; DACC, dorsal anterior cingulate cortex), central-executive (DLPFC, dorsolateral prefrontal cortex; PPC, posterior parietal cortex) and default-mode (VMPFC, ventromedial prefrontal cortex; PPC, posterior cingulate cortex) networks. The AI and DACC of the salience network had a significantly lower net information flow compared with the central-executive and default-mode nodes (^∗^indicates statistical significance).](fnagi-09-00177-g002){#F2} ![African--Americans (AA) versus Caucasian--Americans (CA) comparison: net flow (out-in) of the key nodes of the salience (IN, insula; DACC, dorsal anterior cingulate cortex), central-executive (DLPFC, dorsolateral prefrontal cortex; PPC, posterior parietal cortex) and default-mode (VMPFC, ventromedial prefrontal cortex; PPC, posterior cingulate cortex) networks. The AI of AA group had a significantly lower net flow (out-in) compared with that of CA group (^∗^indicates statistical significance; ns indicates not significant).](fnagi-09-00177-g003){#F3} Cortical Thickness Comparison and Correlation --------------------------------------------- To evaluate whether structural alterations explain these connectivity differences, we measured the insular cortical thickness and compared the values between African--Americans and Caucasian--Americans as displayed in Figure [4](#F4){ref-type="fig"}. Cortical thickness in the African--Americans (median = 2.797 mm) and in the Caucasian--Americans (median = 2.897 mm) was significantly different (Wilcoxon rank sum; Z = 2.09; p \< 0.05). To investigate how the thickness of impaired insular cortex relates with the performance on global cognitive testing, we performed a correlation analysis between the thickness and the MoCA scores. We found that a lower thickness was associated with lower performance reflected by MoCA scores (r = 0.30; p \< 0.05). These results are provided in Figure [5](#F5){ref-type="fig"}. Furthermore, we found that the lower cortical thickness was correlated with the lower net information outflow at the insular cortex (r = 0.31; p \< 0.05) as shown in Figure [6](#F6){ref-type="fig"}. ![Comparison of insular thickness between African--Americans (AA) and Caucasian--Americans (CA). The thickness had a significantly lower value in AA group than that of CA group (^∗^indicates statistical significance).](fnagi-09-00177-g004){#F4} ![The correlation between the thickness of insula and the Montreal cognitive assessment (MoCA) score in (A) both African--Americans (AA) and Caucasian--Americans (CA) (r = 0.30 and p \< 0.05), (B) AA (r = 0.19 and p = 0.18), and (C) CA (r = 0.40 and p \< 0.05).](fnagi-09-00177-g005){#F5} ![The correlation between the thickness and the net flow (out-in) of insula in both African--Americans (AA) and Caucasian--Americans (CA) (r = 0.31 and p \< 0.05).](fnagi-09-00177-g006){#F6} Discussion ========== Here, we investigated the pattern of connectivity among the key brain areas of the salience, central-executive, and default-mode networks in hypertensive individuals with MCI and we compared African--Americans to Caucasian--Americans within this group. We found larger impairment in the control signal of the insula (of salience network) in the African Americans than in the Caucasian--Americans, as reflected by negative net information outflow metrics measured by Granger causality analysis. Although, there are very limited neuroimaging studies that investigate the neural bases of hypertension in cognitive decline ([@B35]), our findings of greater information flow impairment in the African--Americans were in line with the previous non-neuroimaging reports that the African--Americans bear a greater risk of hypertension-associated cognitive impairment than the Caucasian--Americans ([@B47]; [@B27]). We further examined the cortical thickness of impaired insula between the African--Americans and the Caucasian--Americans and found that the insula thickness of the African--Americans is significantly lower than that of the Caucasian--Americans. The insula thickness was found to be correlated with the behavior performance and with the net information flow at the insula cortex, respectively. Those results about insula thickness were also consistent with the existing literature that insula thickness decreases with cognitive decline ([@B28]; [@B41]), however the difference in insula thickness between the African--Americans and the Caucasian--Americans has not been previously explored. Former studies consistently reported that the nodes of salience network render modulation effects over the default-mode and central-executive in healthy individuals ([@B51]; [@B25]; [@B8]). The controlling role of salience nodes over the other two networks has been argued to be structurally supported by direct white matter connections between the insula and the dorsal anterior cingulate cortex ([@B5]; [@B31]) and by their unique sharing of cytoarchitecture at neuronal level, i.e., only these regions consist of special type of neurons---von Economo neurons---that relay information processed within those nodes to other nodes, including the default-mode and executive nodes ([@B2], [@B1]; [@B62]; [@B51]). Literature also shows that the insula of salience network is functionally connected to the central-executive network ([@B61]), and has direct white matter connections to the other areas, including the inferior parietal lobe ([@B58]), and temporo-parietal junction ([@B34]). These structural and functional settings show the great involvement of insula in many cognitive processes such as in the evaluation of task performance across varying perceptual and response demands ([@B58]), the reorientation of attention in conscious error perception (or error awareness) ([@B59]), and the switching between available cognitive resources to integrate external sensory information with internal states ([@B57]). The dorsal anterior cingulate cortex of salience network is known for enhanced cognitive control ([@B19]) such as in switching activity in association with the insula during behaviorally harder tasks ([@B8]). The above mentioned neural basis of control signal of the insula and the dorsal anterior cingulate cortex network (salience nodes) has been suggested to be crucial for cognitive maintenance in both task and resting states in cognitively healthy individuals, whereas impairment to such control activity might be caused by the underlying neuroanatomical changes, including the injuries to the highly sensitive/vulnerable von Economo neurons ([@B2], [@B1]; [@B62]; [@B51]; [@B5]). Emerging evidence suggest atypical engagement of the insula of salience network in disease, including frontotemporal dementia, autism, schizophrenia, and Alzheimer's disease ([@B40]; [@B56]). The structural changes of the cortex, including the cortical thinning of insula with disease progression to MCI and/or Alzheimer's disease, is widely reported in elderly people ([@B50]; [@B28]; [@B41]). Previous studies also suggested the link between the underlying structural changes and the corresponding functional changes of brain nodes and networks ([@B63]; [@B40]). Our connectivity findings indicated that the control mechanism of the salience network was impaired (negative value) in both the African--Americans and the Caucasian--Americans, and the insula was more impaired in the African--Americans than in the Caucasian--Americans. Although, the cortical thinning of insula has been consistently reported in cognitive impairment ([@B28]; [@B41]), there are no neuroimaging studies so far to our knowledge that report the racial disparity of insula thickness. Neuroimaging literature suggests that the insula and dorsal anterior cingulate cortex---the key regions of salience network---respond as the racially biased brain regions ([@B7]), especially the greater activity of insula to out-group race than in-group race ([@B37]; [@B38]), but there are no previous reports about the difference in connectivity patterns of those regions between the African--Americans and the Caucasian--Americans groups with hypertension and cognitive impairment. Prior non-neuroimaging studies repeatedly report that the African--Americans bear the greater risk of hypertension-associated cognitive impairment than the Caucasian--Americans ([@B47]; [@B27]). Thus, our findings and existing neuroimaging/non-neuroimaging evidence taken together suggest that the insula is crucial in racial disparity in cognitively impaired individuals with hypertension. This study can be further extended in the future by including the groups of African--Americans and Caucasian--Americans with MCI but without a history of hypertension and by also including the cognitively healthy hypertensive and normotensive individuals to better dissociate the effect of hypertension alone in the brain nodes and networks. In summary, we evaluated the patterns of interactions among the salience, default-mode, and central-executive nodes in the African--Americans and the Caucasian--Americans race groups, who had hypertension and cognitive impairment. We found that the insula of the salience network was functionally impaired greater, and had lower thickness in the African--Americans than in the Caucasian--Americans. Existing literature and our findings taken together thus suggest that the insula is potential biomarker in the cognitive disorders, including the racial disparity of cognitively impaired hypertensive population. It is worth nothing that the future research should direct toward dissociating the role of hypertension alone between the race groups in the insula and the other possible regions that are functionally and/or structurally connected to the insula. Author Contributions ==================== Conceived and designed the experiment: IH, DQ. Performed the experiment: GC, JW, DQ, IH. Analyzed the data: GC, DQ, IH. Wrote the paper: GC, DQ, IH. Participated in the discussion and provided the comments: GC, JW, DQ, IH. Conflict of Interest Statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer AM and handling Editor declared their shared affiliation, and the handling Editor states that the process nevertheless met the standards of a fair and objective review. Funding. This research was conducted with NIA/NIH grants RF1AG051633 and R01AG042127 to IH. DQ is supported by NIH grants AG25688, AG42127, AG49752, AG51633, and has received research support from Medtronic and Siemens Medical Solutions. We would like to thank the members of our team and the participants who volunteered for the study. [www.fil.ion.ucl.ac.uk/spm/software/spm12](http://www.fil.ion.ucl.ac.uk/spm/software/spm12) <http://marsbar.sourceforge.net> <https://surfer.nmr.mgh.harvard.edu/> <https://www.mathworks.com> [^1]: Edited by: Ana B. Vivas, CITY College, International Faculty of the University of Sheffield, Greece [^2]: Reviewed by: Anna Emmanouel, CITY College, International Faculty of the University of Sheffield, Greece; Luis J. Fuentes, Universidad de Murcia, Spain [^3]: ^†^These authors have contributed equally to this work.	{ "pile_set_name": "PubMed Central" }
All relevant data are within the paper and its Supporting Information files. Introduction {#sec005} ============ Breast cancer is by far the most commonly diagnosed malignancy and second leading cause of cancer death in women globally. Estimated 232,670 new cases and 40,000 deaths occurred in the United States in 2014 according to American Cancer Society \[[@pone.0143537.ref001]\], which places heavy a burden on both patients and healthcare system. Owing to the breakthrough in the treatments of breast cancer, including surgical resection, adjuvant chemotherapy, radiotherapy, endocrine therapy and targeted therapy, the overall survival improved and tumor recurrence reduced during the last decades \[[@pone.0143537.ref002]\]. However, recurrence and metastasis remain the main challenge for cure \[[@pone.0143537.ref003]\]. Up to now, useful prognostic factors can identify groups of patients with various prognoses. Clinical and pathological parameters have been used as prognostic factors to predict the outcome of breast cancer patients and help to develop appropriate adjuvant treatments for high-risk patients. The American Joint Committee on Cancer (AJCC) tumor-node-metastasis (TNM) staging system and the molecular subtypes classification are important prognostic predictors \[[@pone.0143537.ref004],[@pone.0143537.ref005]\]. Clinically, the breast cancer is classified into several molecular subtypes based on the estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor-2 (HER2) and Ki-67 index status. Patients with HER2 over-expressing subtype (ER-, PR- and HER2+) and triple-negative subtype (ER-, PR- and HER2-) suffer more recurrence, distant metastasis and drug resistance than the luminal subtype \[[@pone.0143537.ref006]\]. In addition to the tumor stage and molecular subtypes, various clinical tools were developed to predict the prognosis of breast cancer patients, such as 21-gene RT-PCR assay recurrence score, Nottingham Prognostic Index (NPI) and Adjuvant Online (Adjuvant!)\[[@pone.0143537.ref007]\]. The NPI combines three prognostic factors: nodal status, tumor size and histological grade \[[@pone.0143537.ref008]\]. Since the three components of the NPI formula reflect the metastatic potential and genetic instability of tumor cells, higher NPI generally indicates worse clinical outcomes. Patients were classified into four subsets with different probability of cancer-related death: excellent, good, moderate and poor groups \[[@pone.0143537.ref009]\]. Recent studies had investigated the association between laboratory parameters and the cancers. Being members of blood lipid profiles, total cholesterol level and triglyceride level were significantly higher in breast cancer patients \[[@pone.0143537.ref010],[@pone.0143537.ref011]\]. Elevated cholesterol level and triglyceride level were risk factors for poor prognosis in lung cancer and prostate cancer \[[@pone.0143537.ref012],[@pone.0143537.ref013]\]. In addition, previous studies demonstrated that fibrinogen participated in the tumorigenesis and progression \[[@pone.0143537.ref014],[@pone.0143537.ref015]\] and elevated serum fibrinogen level predict poor clinical outcomes in several cancers \[[@pone.0143537.ref016],[@pone.0143537.ref017],[@pone.0143537.ref018]\]. In this study, we investigated the laboratory parameters routinely tested prior to the surgical treatment and set up a prognostic index model based on the independent prognostic factors. The prognostic model was assessed and compared with the NPI to estimate the clinical value in breast cancer patients. Patients and Methods {#sec006} ==================== Study population {#sec007} ---------------- We retrospectively reviewed the medical records of consecutive patients with primary breast cancer from January 1, 2002 to December 31, 2008 in Sun Yat-sen University Cancer Center (SYSUCC). Pathological diagnosis was carefully confirmed by pathologists. Other inclusion criteria included: (1) received surgical treatment; (2) female; (3) diagnosed as invasive ductal carcinoma or invasive lobular carcinoma. We excluded patients who (1) received neoadjuvant chemotherapy before surgery; (2) had surgical treatment before admission; (3) with previous or coexisting cancers other than breast cancer; (4) confirmed metastasis. All patients were followed up until December 31, 2014 or date of cancer-related deaths. Routine tests and telephone counseling were performed to understand the patients' condition during follow-up time. Clinical data collection {#sec008} ------------------------ Patient characteristics and clinicopathological factors like age, menopausal status, pathological diagnosis, tumor size, axillary lymph nodal status, histologic grade, hormonal receptor and HER2 status, date of last follow-up or cancer-related death were collected. Laboratory parameters investigated as potential prognostic factors in the current study included albumin, globulin, lactate dehydrogenase (LDH), total bilirubin (TB), uric acid, cholesterol, triglyceride and fibrinogen. The blood samples were collected and measured by the automatic biochemical analyzer in the central laboratory before surgery. The tumor stages were classified according to the AJCC TNM staging system (the 7th edition). The molecular subtypes were as follow: Luminal A (ER +, PR +, HER2 -and Ki-67≤ 14%), Luminal B (ER + and HER2 + or Ki-67\> 14%), HER2 over-expressing (ER -, PR -, HER2 +) and triple-negativer breast cancer (ER -, PR -, HER2 -). HER2 positive was defined as "3+" in immunohistochemical test or "positive" in HER2 fluorescence in situ hybridization test. The NPI was calculated by the formula of lymph node stage (1--3) + histologic grade (1--3) + 0.2×tumor size (cm) and patients were stratified into four groups as excellent (≤ 2.40), good (2.41--3.40), moderate (3.41--5.40) and poor (\>5.4)\[[@pone.0143537.ref008],[@pone.0143537.ref009]\] Statistical analyses {#sec009} -------------------- The primary end point was cancer-specific survival (CSS) calculated from the date of diagnosis to the date of cancer-related deaths or the last follow-up. Means and standard deviations were calculated, and differences were identified by t test. Difference between categories were evaluated using chi-square test. Receiver operating characteristics (ROC) curve analysis were performed to assess the prognostic value of each laboratory parameter and to set up the optimal cut-off points for potential prognostic factors. Univariate analysis and multivariate analysis (Cox regression model) were used to confirm the independent prognostic variables associated with CSS. The prognostic index (PI) model was established based on independent variables and enrolled patients were stratified into low-, moderate- and high-risk groups. Kaplan---Meier method was performed for survival analysis and compared by log-rank test. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated from the Cox regression model, and a two-tailed p value \<0.05 was considered statistically significant. The comparison between PI and NPI was made using the method of DeLonget al \[[@pone.0143537.ref019]\]. All statistical analyses were performed using SPSS 19.0 (SPSS Inc., Chicago, IL, USA). Ethics Statement {#sec010} ---------------- The study protocol was approved the by independent ethical committee/institutional review board of Sun Yat-sen University Cancer Center (SYSUCC), and written informed consent about the scientific research was obtained from each participant prior to surgery. Patient records were anonymized and de-identified prior to analysis. Results {#sec011} ======= There were 1661 female patients with primary non-metastatic invasive breast cancer enrolled in this study. The mean survival time of all participants was 123.6 months and patient characteristics were shown in [Table 1](#pone.0143537.t001){ref-type="table"}. The mean age was 42.2 years old (range 22--74 years), and 150 (9.0%) patients were under the age of 35. Invasive ductal carcinoma was the predominant type (97.2%). Tumor sizes of T1, T2 and T3 were observed in 698 (42.0%), 876 (52.7%) and 87 (5.2%) of the patients respectively, and 893 (53.8%) patients suffered regional lymph node metastasis. 129 (7.7%), 374 (22.5%), 836 (50.4%) and 322 (19.4%) were classified as excellent, good, moderate and poor NPI groups respectively. Luminal subtype comprised 77.4% of total participants, and 183 (11.0%) and 193 (11.6%) were HER2 over-expressing subtype and triple-negative subtype respectively. 89.3% (1484/1661) of enrolled participants received chemotherapies and all were performed within 45 days after the surgeries. The anthracycline and/or taxane-containing chemotherapy was the main option for adjuvant cytotoxic chemotherapy. Radiotherapies were given sequentially to the patients with more than 3 metastatic lymph nodes or tumor \>50 mm in greatest dimension or breast conserving surgery. Patients with ER/PR positivity received endocrine therapies for more than 4 years. Target therapy was not performed due to the unavailability of trastuzumab at that time. 10.1371/journal.pone.0143537.t001 ###### Clinicopathological characteristics and laboratory parameters of patients. ![](pone.0143537.t001){#pone.0143537.t001g} Characteristic All patients (n = 1661) Low-risk group (n = 1168) Moderate-risk group (n = 410) High-risk group (n = 83) p value --------------------------------- ------------------------- --------------------------- ------------------------------- -------------------------- ----------- Age \<0.001 ≤ 35 150(9.0) 126(84.0) 24(16.0) 0(0) \> 35 1511(91.0) 1042(69.0) 386(25.5) 83(5.5) Menopause \<0.001 Yes 433(26.1) 264(61.0) 134(30.9) 35(8.1) No 1228(73.9) 904(73.6) 276(22.5) 48(3.9) Tumor type 0.481 IDC 1615(97.2) 1136(70.3) 400(24.8) 79(4.9) ILC 46(2.8) 32(69.6) 10(21.7) 4(8.7) Histologic grade 0.886 G1 562(33.8) 396(70.5) 135(24.0) 31(5.5) G2 664(40.0) 465(70.0) 165(24.8) 34(5.1) G3 435(26.2) 307(70.6) 110(25.3) 18(4.1) Tumor size 0.805 T1 698(42.0) 799(71.5) 163(23.4) 36(5.2) T2 876(52.7) 607(69.3) 227(25.9) 42(4.8) T3 87(5.2) 62(71.3) 20(23.0) 5(5.7) Lymph node status 0.127 N0 768(46.2) 545(71.0) 188(24.5) 35(4.6) N1 448(27.0) 327(73.0) 98(21.9) 23(5.1) N2 259(15.6) 179(69.1) 70(29.1) 10(3.9) N3 186(11.2) 117(32.9) 54(29.0) 15(8.1) ER 0.550 Positive 1047(63.0) 743(71.0) 256(24.5) 48(4.6) Negative 614(37.0) 425(69.2) 154(25.1) 35(5.7) PR 0.123 Positive 1162(70.0) 831(71.5) 280(24.1) 51(4.4) Negative 449(30.0) 337(67.5) 130(26.1) 32(6.4) HER-2 0.042 Positive 393(23.7) 889(70.1) 324(25.6) 55(4.3) Negative 1268(76.3) 279(71.0) 86(21.9) 28(7.1) NPI 0.652 Excellent group 129(7.7) 91(70.5) 32(24.8) 6(4.7) Good group 374(22.5) 269(71.9) 86(23.0) 19(5.1) Moderate group 836(50.4) 593(70.9) 199(23.8) 44(5.3) Poor group 322(19.4) 215(66.8) 93(28.9) 14(4.3) Cancer-specific survival \<0.001 Alive 1365(82.2) 992(72.7) 319(23.4) 54(4.0) Death 296(17.8) 176(59.5) 91(30.7) 29(9.8) Albumin level (g/L,mean±SD) 43.8±4.1 Globulin level (g/L,mean±SD) 28.3±4.5 LDH (U/L,mean±SD) 161.2±55.0 TB (umol/L, mean+ SD) 13.5±9.3 Uric acid (umol/L, mean±SD) 283.0±80.4 Cholesterol (mmol/L,mean±SD) 5.09±1.05 Triglyceride (mmol/L,mean±SD) 1.37±0.94 Plasma fibrinogen(g/L, mean±SD) 2.95±0.71 Abbreviation: SD standard deviation, IDC Invasive ductal carcinoma, ILC Invasive lobula carcinomar, ER Estrogen receptor,PR Progesterone receptor, HER2 Human epidermal growth factor receptor-2, LDH lactate dehydrogenase, TB total bilirubin ROC curve analysis was performed and the AUCs were shown in [S1 Table](#pone.0143537.s001){ref-type="supplementary-material"}. The AUCs for albumin, globulin, LDH, triglyceride and fibrinogen are all statistically significant (all p \< 0.05), and potential prognostic effects of total bilirubin and uric acid were not proved (p = 0.311 and 0.326 respectively). The optimal cut-off points for each potential prognostic factors were set up with the highest Youden index. Enrolled patients were stratified into high- or low-level by various variables. In the multivariate analysis, globulin (\>30.0g/L), triglyceride (\>1.10mmol/L) and fibrinogen (\>2.83g/L) were identified as independent prognostic factors for poor cancer-specific survival in breast cancer patients (all p\< 0.05, [Fig 1](#pone.0143537.g001){ref-type="fig"}). ![Cancer-specific survival of patients according to globulin level (A), triglyceride level (B) and fibrinogen level (C).](pone.0143537.g001){#pone.0143537.g001} The preoperative globulin, triglyceride and fibrinogen identified as independent prognostic factors in multivariate analysis were used to establish the prognostic index model (PI). The criteria of prognostic index model were as follows: globulin level \>30.0g/L, triglyceride level \>1.10mmol/L and fibrinogen level \>2.83g/L were allocated 1 point each; globulin level ≤30.0g/L, triglyceride level ≤1.10mmol/L and fibrinogen level ≤2.83g/L were allocated 0 point each. The total score ranging from 0 to 3 was categorized into three prognostic index risk groups defined as: low-risk group, 0 or 1 point; moderate-risk group, 2 point; high-risk group, 3 point. There were 1168 (70.3%) patients allocated to low-risk group, 410 (24.7%) allocated to moderate-risk group and 83(5.0%) patients were categorized as high-risk group. The mean cancer-specific survival time for the low-risk group was 126.8 months (95% CI 124.5--129.2), which was significantly longer than that of 115.5 months (95% CI 111.1--112.0) in moderate-risk group and 99.7 months (95% CI 121.5--125.7) in high-risk group (both p\< 0.05). The estimated 10-years survival rates for low-risk group, moderate-risk group and high-risk group were 80.5%, 72.0% and 58.8% respectively ([Fig 2](#pone.0143537.g002){ref-type="fig"}). Univariate analysis and multivariate analysis verified that the PI was significantly associated with CSS (HR = 1.513 for moderate-risk group and 2.481 for high-risk group, both p\<0.05). In addition, menopausal status, tumor size, lymph node status, ER and HER2 status were identified as independent factors associated with CSS (all p\< 0.05, [Table 2](#pone.0143537.t002){ref-type="table"}). ![Cancer-specific survival of patients by prognostic index risk group.](pone.0143537.g002){#pone.0143537.g002} 10.1371/journal.pone.0143537.t002 ###### Univariate and multivariate analysis of PI for CSS in breast cancer. ![](pone.0143537.t002){#pone.0143537.t002g} Variable Univariate analysis multivariate analysis ---------------------- --------------------- ----------------------- --------- ------- --------------- --------- Age 0.733 0.511--1.053 0.093 0.765 0.521--1.123 0.172 ≤ 35 1 (reference) 1 (reference) \> 35 0.733 0.511--1.053 0.093 0.765 0.521--1.123 0.172 Menopausal status Yes 1 (reference) 1 (reference) No 0.755 0.591--0.965 0.025 0.690 0.531--0.896 0.005 Tumor type 0.813 0.384--1.720 0.588 0.742 0.642--1.608 0.449 Histologic grade G1 1 (reference) 1 (reference) G2 0.711 0.536--0.943 0.018 0.781 0.522--1.170 0.231 G3 1.206 0.915--1.589 0.184 1.097 0.637--1.888 0.739 Tumor size T1 1 (reference) 1 (reference) T2 1.429 1.110--1.840 0.006 1.100 0.792--1.527 0.570 T3 3.830 2.620--5.601 \<0.001 2.095 1.280--3.429 0.003 Lymph node status N0 1 (reference) 1 (reference) N1 2.043 1.462--2.854 \<0.001 1.966 1.258--3.075 0.003 N2 3.646 2.600--5.111 \<0.001 3.484 1.907--6.366 \<0.001 N3 7.610 5.508--10.516 \<0.001 6.080 3.410--10.842 \<0.001 ER Negative 1 (reference) 1 (reference) Positive 0.577 0.460--0.725 \<0.001 0.681 0.514--0.904 0.008 PR Negative 1 (reference) 1 (reference) Positive 0.612 0.485--0.773 \<0.001 0.902 0.685--1.187 0.461 HER2 Negative 1 (reference) 1 (reference) Positive 1.961 1.544--2.490 \<0.001 1.335 1.018--1.751 0.036 PI Low-risk group 1(reference) 1(reference) Moderate-risk group 1.537 1.193--1.980 0.001 1.513 1.169--1.959 0.002 High-risk group 2.696 1.820--3.994 \<0.001 2.481 1.653--3.724 \<0.001 Abbreviation: CCS Cancer-specific survival, CI confidence interval, ER Estrogen receptor,PR Progesterone receptor, HER2 Human epidermal growth factor receptor-2, PI prognostic index Under stratifications based on the molecular subtypes, the PI was independently associated with CSS in patients with Luminal A, Luminal B subtypes and triple-negative breast cancer (all p\< 0.05, [Table 3](#pone.0143537.t003){ref-type="table"}). However, no significant difference was observed in HER2 over-expressing among different PI groups. 10.1371/journal.pone.0143537.t003 ###### Multivariate analysis of PI and NPI based on various molecular subtypes. ![](pone.0143537.t003){#pone.0143537.t003g} Molecular subtypes PI NPI[^2^](#t003fn002){ref-type="table-fn"} ------------------------------- ---------------------- ------------------------------------------- ---------------------- --------- Luminal A 1.583 (1.135--2.207) 0.007 2.491 (1.859--3.338) \<0.001 Luminal B 1.885 (1.431--2.482) \<0.001 2.035 (1.557--2.660) \<0.001 HER2 over-expressing 1.352 (0.906--2.015) 0.139 1.260 (0.886--1.834) 0.227 Triple-negative breast cancer 1.579 (1.076--2.318) 0.020 1.631 (1.074--2.479) 0.022 All patients 1.570 (1.316--1.872) \<0.001 2.019 (1.724--2.365) \<0.001 ^1^PI and NPI were processed as a continuous variable ^2^tumor size and lymph node status were in included in NPI multivariate analysis Abbreviation: PI prognostic index model, NPI Nottingham Prognostic Index, HR Hazard ratio CI confidence interval, HER2 human epidermal growth factor receptor-2 As showed in [Table 3](#pone.0143537.t003){ref-type="table"}, NPI was significantly predictive of CSS in all patients and in patients with Luminal A, Luminal B subtypes and triple-negative breast cancer, which was similar with the prognostic implication of PI. The prediction efficiency of PI was compared with that of NPI. The AUC of PI was 0.571 (95%CI:0.534--0.609) and that of NPI was 0.643 (95%CI:0.609--0.677). The difference between the two prognostic models was statistically significant (Z = 3.279, p = 0.001). Discussion {#sec012} ========== Breast cancer is the most commonly diagnosed cancers among women in the United States \[[@pone.0143537.ref001]\]. Despite newer therapies in the recent years, recurrence and metastasis remains the main challenge to the cancer management. About 30% to 40% of invasive breast cancer patients will eventually develop into metastatic breast cancer \[[@pone.0143537.ref020],[@pone.0143537.ref021]\], who suffer low 5-year survival rate as 24%\[[@pone.0143537.ref022]\]. Generally, risk factors for poor prognosis include large tumor size, positive lymph node status, negative hormone receptor status and HER2 overexpression \[[@pone.0143537.ref023]\]. Moreover, previous studies have reported several factors related to poor outcomes of breast cancer patients, such as albumin, LDH \[[@pone.0143537.ref024]\], bilirubin and cholesterol \[[@pone.0143537.ref025]\]. In the present study, we retrospectively analyzed the preoperative parameters in breast cancer patients and identified globulin (\>30.0g/L), triglyceride (\>1.10mmol/L) and fibrinogen (\>2.83g/L) as independent unfavourable prognostic factors for CSS. The possible mechanisms of the above factors in tumorigenesis and progression were undefined. Globulins, generally including immunoglobulins and other acute-phase proteins such C-reactive protein (CRP), serum amyloid A, are the major component of serum proteins. An increased globulin level has been reported to reflect the presence of continuous systemic inflammation \[[@pone.0143537.ref026]\], which induces the increased levels of cytokines and promotes the tumor proliferation, progression, invasion and metastasis \[[@pone.0143537.ref027],[@pone.0143537.ref028]\]. It was found that elevated alpha and gamma globulins were linked with poor prognosis in lung cancer and colorectal cancer patients \[[@pone.0143537.ref029],[@pone.0143537.ref030]\]. Alpha-1 globulin could inhibit natural killer-cell activity and T cell-mediated cytotoxicity by suppressing cytotoxic reactions of lymphocytes \[[@pone.0143537.ref031]\]. Furthermore, COOH-terminal fragment, the degradation product of alpha-1 globulin by matrix metalloproteinase, could improve the tumor growth and invasion potential in vivo \[[@pone.0143537.ref032]\]. These biological behaviors regulate host anti-tumour defense mechanisms and promote tumor development \[[@pone.0143537.ref033]\]. Several studies have found a significant association between elevated serum triglyceride level and risk of breast cancer development \[[@pone.0143537.ref034],[@pone.0143537.ref035]\]. Besides the active glycolysis, the lipogenesis is greatly increased in tumor cells \[[@pone.0143537.ref036]\]. The triglyceride/free fatty acid (TG/FFA) cycling plays an important role in multiple signaling pathways \[[@pone.0143537.ref037]\]. The TG/FFA cycling may promote cell survival through the activation of NF-κB and thus improve the expression of anti-apoptotic protein Bcl-2 and Bcl-x \[[@pone.0143537.ref038]\]. Hypoxic condition, which commonly occurred in the centre of solid tumors, could cause an elevation of triglyceride and expression of hypoxia inducible factor (HIF) that promotes metastasis of tumor \[[@pone.0143537.ref039],[@pone.0143537.ref040]\]. The prognostic association between fibrinogen and cancer was reported by several studies. Fibrinogen can deposited around solid tumors and act as a stable framework to combine growth factors, such as fibroblast growth factor-2 (FGF-2) and vascular endothelial growth factor (VEGF), with tumor cells to increase the tumor proliferation and stimulate angiogenesis \[[@pone.0143537.ref014],[@pone.0143537.ref041]\]. Moreover, fibrinogen can improve the adhesion of tumor cells to platelets, protecting tumor cells from the innate immune system and leading to an increase of metastatic cells \[[@pone.0143537.ref042]\]. Based on these three factors, we devised a Prognostic Index (PI) model and classified the enrolled patients into low-, moderate- and high-risk groups. Independently of the tumor size, lymph node status, hormonal receptor status and HER2 status, the PI was significantly associated with the overall postoperative survival, indicating that patients with higher PI experienced poorer prognosis. Moreover, after stratification by molecular subtypes, preoperative PI remained significantly prognostic in Luminal A, Luminal B subtypes and triple-negative breast cancer. Generally, breast cancer patients with luminal subtype, especially Luminal A subtype, had relatively better clinical outcomes. Based on the preoperative PI, clinicians could identify patients with high risk of poor prognosis, and additional adjuvant treatment might be suggested beyond endocrine treatment. Moreover, both the PI and NPI could not effectively predict the prognosis in HER2 over-expressing subtype. That may be due to the carcinogenesis and the proliferation promotion effect of HER2 protein \[[@pone.0143537.ref043],[@pone.0143537.ref044]\], which induce the unique characteristics of HER2 over-expressing breast cancer subtype. The NPI combines three prognostic factors: nodal status, tumor size and histological grade, and the former two were found to be independently associated with survival in the present study. Patients with different NPI were considered to suffer distinguished prognosis. In the current study, the discriminatory ability of PI was compared with NPI, and the AUC value of NPI was significantly higher than that of PI (p\< 0.05). Axillary lymph node status is the most important prognostic factor in the staging of the breast cancer patient \[[@pone.0143537.ref045],[@pone.0143537.ref046]\]. The occurrence of axillary nodal metastases has been found to significantly decrease the 5-year survival rate up to 40%\[[@pone.0143537.ref047]\]. In the multivariate analysis, the hazard ratio of lymph nodal status was higher than that of PI, which may explain better discriminatory ability of NPI. Thus PI may act as complement to further improve the discriminatory ability of NPI. The present study is limited to the retrospective nature. Firstly, selection bias cannot be excluded even though consecutive patients were included and eligibility criteria were performed to minimize the bias. Secondly, specific quality control analysis was performed by different quality inspectors since the laboratory parameters were obtained as routine clinical tests before surgery. To sum up, our study identified three independent prognostic factors, globulin, triglyceride and fibrinogen, and set up a prognostic index (PI) based on these factors to stratify breast cancer patients into low, moderate and high risk of poor prognosis. These parameters are low-cost and routinely measured in clinical practice. Thus the PI may assist clinicians to identify high-risk patients and make personalized therapeutic approaches independently or along with other prognostic models. Supporting Information {#sec013} ====================== ###### The optimal cut-off points and multivariate analyses of potential prognostic parameters. (DOC) ###### Click here for additional data file. ###### Relevant data underlying the findings described in manuscript. (XLS) ###### Click here for additional data file. [^1]: Competing Interests:The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: JHW XMX. Performed the experiments: FY SJL. Analyzed the data: XJH LY. Contributed reagents/materials/analysis tools: XJH XSX. Wrote the paper: JHW FY. Approved the manuscript: XMX.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1-ijerph-17-05150} =============== According to the World Health Organization (WHO), the number of people over the age of 60 will reach 2 billion by 2050, which is about a fifth of the population worldwide (22%) \[[@B1-ijerph-17-05150]\]. As a country with the largest population, China has the largest number of older adults. By the end of 2018, there were 249 million (17.9%) and 167 million (11.9%) people over the age of 60 and 65, respectively \[[@B2-ijerph-17-05150]\]. However, the aging process in China is so rapid that the social system cannot keep up \[[@B1-ijerph-17-05150],[@B3-ijerph-17-05150]\]. In order to address the aging crisis and ensure older adults' right to live a long and healthy life, "healthy aging" was put forward by WHO and the Healthy China 2030 Strategy, which aimed to create an age-friendly environment through the cooperation of the whole society \[[@B4-ijerph-17-05150],[@B5-ijerph-17-05150]\]. Older adults are in a critical stage of maintaining their health-related quality of life (HRQOL), where the leading health threat is noncommunicable diseases (NCDs) \[[@B6-ijerph-17-05150]\], mainly including cardio-cerebrovascular diseases, cancers, chronic respiratory diseases and diabetes \[[@B7-ijerph-17-05150]\]. Thus, the prevention of NCDs and the maintenance of HRQOL are major difficulties for the achivement of healthy aging. Previous studies have focused on the association between one specific NCD and HRQOL, such as hypertension \[[@B8-ijerph-17-05150]\], type 2 diabetes \[[@B9-ijerph-17-05150]\] and chronic kidney disease \[[@B10-ijerph-17-05150]\]. However, the results were not consistent, since most of these studies were cross-sectional studies. A lot of studies showed a negative effect of one specific NCD on HRQOL, while others did not find a correlation \[[@B8-ijerph-17-05150],[@B9-ijerph-17-05150],[@B10-ijerph-17-05150],[@B11-ijerph-17-05150]\]. Additionally, most studies have ignored the influence of the number of NCDs, while older adults usually have more than one NCD, referred to as multimorbidity \[[@B12-ijerph-17-05150]\]. Some international researchers investigated the HRQOL of people suffering from multiple NCDs in their countries and found a negative relationship between HRQOL and multimorbidity \[[@B13-ijerph-17-05150],[@B14-ijerph-17-05150]\]. However, few studies focused on older adults, especially Chinese older adults \[[@B15-ijerph-17-05150],[@B16-ijerph-17-05150]\]. In fact, the relationship between the HRQOL and NCDs is very complex and numerous factors could make a difference. For instance, a great number of NCDs are due to unhealthy behaviors, especially poor diets \[[@B7-ijerph-17-05150]\]. A few researchers proposed the "modernization theory" to explain the rise of unhealthy diets \[[@B17-ijerph-17-05150]\]. The theory illustrated burgeoning middle classes with more disposable income tended to consume more meat and processed foods that had close associations with weight gain. Therefore, to some extent, it is reasonable to believe people with NCDs could have a good economic status. On the other hand, social demographic determinants exert a tremendous effect on HRQOL \[[@B18-ijerph-17-05150]\]. Among these factors, socioeconomic status has been proven to be an independent predictor of HRQOL \[[@B19-ijerph-17-05150]\]. People with higher socioeconomic status often have a better HRQOL \[[@B20-ijerph-17-05150]\]. Hence, it is reasonable to argue that NCDs might exert little or even no negative effect on HRQOL on account of the good economic status of people with NCDs. The relationship between HRQOL and NCDs should be further clarified through a large-scale survey. In China, due to the rapid process of industrialization and urbanization, citizens' lifestyle and disease spectrum have changed a lot \[[@B5-ijerph-17-05150]\]. Nearly 180 million Chinese older adult people suffer from NCDs, accounting for 75% of older people in China, and the prevalence of multimorbidity is about 35% \[[@B21-ijerph-17-05150]\]. Many studies on the HRQOL and NCDs among older people in China simply classified subjects according to whether they had NCDs or not, while few studies focused on the number of NCDs. Moreover, the samples of those studies were too small to be representative \[[@B15-ijerph-17-05150],[@B22-ijerph-17-05150]\]. Therefore, we conducted the cross-sectional investigation to understand the HRQOL among people with different numbers of NCDs., and to explore the relationship between them. We argue that the HRQOL of Chinese older adults might become worse with the increase in the number of NCDs. The findings are expected to provide theoretical support for the timely advance of NCD management. 2. Materials and Methods {#sec2-ijerph-17-05150} ======================== 2.1. Study Population {#sec2dot1-ijerph-17-05150} --------------------- The cross-sectional study was based on a large-scale survey named China's Health-Related Quality of Life Survey for Older Adults 2018 (CHRQLS-OA 2018) whose chief initiator and executor was the Global Health Institute of Wuhan University \[[@B23-ijerph-17-05150]\]. Interviewers also included teachers and students from other top universities in China (such as Nanchang University). All staff received professional training before the survey started. Conducted in either online or paper format between January and March of 2018, the survey recruited older adults aged 60 years and over across the country. Through convenience sampling, 5442 valid subjects were recruited. The survey mainly collected participants' individual socio-demographic characteristics, social capital, behaviors and lifestyles, health-related quality of life, mental health and coping strategies, etc. This study aims to explore the relationship between the number of NCDs and the HRQOL of Chinese older adults. In the present study, 276 subjects (5.07%) without information on NCDs and/or HRQOL were excluded. Finally, 5166 respondents (94.93%) were included in the analysis. 2.2. Description of the Measures {#sec2dot2-ijerph-17-05150} -------------------------------- ### 2.2.1. General Demographic Characteristics {#sec2dot2dot1-ijerph-17-05150} In the present study, the following general demographic characteristics of the participants were included: age, sex, nationality, body mass index (BMI), years of education, household registration, marital status, average annual household income (CNY), new personal savings and self-rated health status (self-rated health status was based on individuals' subjective feelings). The variable was divided into three groups, which were "Good", "General", and "Poor". "General" was a status between "Good" and "Poor", and was equivalent to "middle" or "moderate". These general demographic characteristics were considered as confounding factors in the analysis of the cross-sectional association between the number of NCDs and HRQOL among Chinese older adults. ### 2.2.2. Assessment of the Number of NCDs {#sec2dot2dot2-ijerph-17-05150} In "China's Health-related older adults Quality of Life Survey 2018", the number of NCDs and the time of diagnosis were self-reported. Diseases investigated included hypertension, diabetes, coronary heart disease, tumor, chronic obstructive pulmonary disease (COPD), intervertebral disc disease, asthma, rheumatoid arthritis, gastroenteritis, cataract, cerebrovascular disease and others (openly filled in). Professional medical diagnoses can reduce certain deviation, and thus the questionnaire refered to the following questions: "where and by whom was the disease diagnosed? How long has the diagnosis been made?" ### 2.2.3. Assessment of the HRQOL {#sec2dot2dot3-ijerph-17-05150} The WHO's HRQOL Research Group has further developed the World Health Organization Quality of Life-Old (WHOQOL-OLD) module for older adults based on the original QOL scale \[[@B24-ijerph-17-05150],[@B25-ijerph-17-05150]\]. The WHOQOL-OLD module can be widely used in a variety of studies concerning crucial issues of the HRQOL in old age, such as population epidemiology, health monitoring, service development and so on. The WHOQOL-OLD module includes 24 items assigned to six aspects (each of the aspects has 4 items): the "Sensory Abilities" (SAB) aspect assesses sensory capacity and the influence of loss of sensory abilities on HRQOL; the "Autonomy" (AUT) aspect means independence in old age; the "Past, Present and Future Activities" (PPF) aspect assesses satisfaction with one's achievements in life and expectations in the future; the "Social Participation" (SOP) aspect refers to participation in daily activities; the "Death and Dying" (DAD) aspect is related to concerns, worries and fears about death and dying; the "Intimacy" (INT) describes the ability to have personal and intimate relationships. The scores of each item range from 1 to 5, thus the scores of each aspect range from 4 to 20. The scores of all the six aspects can be combined into a total ("overall") score for HRQOL in old age. Essentially, high scores represent high HRQOL, low scores represent lower HRQOL. Through the evaluation of the Chinese version of the WHOQOL-OLD module, the results showed good reliability and validity, thus WHOQOL-OLD is suitable for the HRQOL evaluation of Chinese older adults \[[@B26-ijerph-17-05150]\]. 2.3. Statistical Analysis {#sec2dot3-ijerph-17-05150} ------------------------- The Statistical Package for the Social Sciences (SPSS) version 23.0 for Windows (SPSS Inc., Chicago, IL, USA) was applied to conduct all statistical analyses, with a statistical significance level of 0.05. The data analysis consisted of three aspects. First, frequencies and proportions were reported to describe the distribution of NCDs in the population with different demographic characteristics. Second, the difference in scores of all dimensions of the WHOQOL-OLD scale was examined among groups with different numbers of NCDs by ANOVA, and the Student--Newman--Keuls (SNK) test method was used for post-hoc comparisons. Third, multiple linear regression models were used to determine the relationship between the number of NCDs and the total scores and scores of each dimension of the WHOQOL-OLD scale. We established three models in the third step, including the initial model (model 1) and the adjusted models (model 2 and model 3). The final parsimonious model adjusted for confounders that were significantly associated with the dependent variables in model 2. The unstandardized coefficients (B) with a 95% confidence interval (95% CI) obtained from the models were reported. 2.4. Ethical Statements {#sec2dot4-ijerph-17-05150} ----------------------- This study was conducted in accordance with the Declaration of Helsinki, and the study protocol was reviewed and approved by the Institutional Review Board of School of Health Science and Faculty of Medical Sciences, Wuhan University (IRB number: 2019YF2050). Informed consent information was included in each questionnaire and introduced before the surveys. The questionnaires were completed by participants themselves or their caregivers according to older adults's reading and response abilities. Surveys were only conducted when subjects were fully informed of the content and aim of this research project and agreed to participate. The survey was also conducted anonymously, and respondents' information was kept confidential and only for the use of scientific research. 3. Results {#sec3-ijerph-17-05150} ========== 3.1. Descriptions of Sample Characteristics {#sec3dot1-ijerph-17-05150} ------------------------------------------- A total of 5166 older adults was included in the study. The general demographic characteristics are shown in [Table 1](#ijerph-17-05150-t001){ref-type="table"}. In general, there were 51.3% older adults with one or more NCD(s). More than half (50.9%) of men did not have NCDs, while only 46.6% of women did not have NCDs. Most people aged 60--64 years and those aged 85 years or older did not suffer from NCDs (54.0% and 55.3%, respectively). The Han population (50.8%) had a lower proportion of NCDs than minorities (58.5%). People with a normal BMI had the highest percentage of those without NCDs (50.8%). People with NCDs accounted for 58.2% of uneducated participants, which was higher than those receiving education. The proportion of non-agricultural people who did not suffer from NCDs (53.5%) was significantly higher than that of agricultural people (46.6%). People who were married or cohabiting had a lower percentage of having any NCDs (48.6%). Nearly three fifths (59.6%) of the population with an average annual household income of 45,001--60,000 did not suffer from NCDs. About two thirds (62.4%) of people who self-rated their health status as good did not suffer from NCDs. The proportion of people with new personal savings of 100,000 CNY or more who did not suffer from NCDs reached 63.1%. 3.2. Description of the Scores of the WHOQOL-OLD Scale {#sec3dot2-ijerph-17-05150} ------------------------------------------------------ As shown in [Table 2](#ijerph-17-05150-t002){ref-type="table"}, there were differences in total score and scores of the six dimensions of the WHOQOL-OLD scale among people with different numbers of NCDs, and the differences were statistically significant (p \< 0.001). Those with the highest scores of SAB, AUT, PPF, SOP, DAD, INT and total score were people without NCDs, while those with the lowest scores were people with two NCDs or three NCDs and above. The multiple comparisons found that as the number of NCDs increased, the scores of each dimension and the overall score gradually decreased. 3.3. Relationship between the Number of NCDs and the Scores of the WHOQOL-OLD Scale {#sec3dot3-ijerph-17-05150} ----------------------------------------------------------------------------------- As shown in [Table 3](#ijerph-17-05150-t003){ref-type="table"},multiple linear regression analysis was used to identify the relationship between the number of NCDs and the scores of the WHOQOL-OLD scale, and the model 2 and model 3 were adjusted by confounding factors. In the crude and adjusted models, the number of NCDs was related to the scores of the WHOQOL-OLD scale. According to the final model 3, with no NCDs as the reference group, having one NCD had a negative impact on the three dimensions of PPF, SOP, and DAD in the HRQOL of older adults (p \< 0.05). Suffering from two NCDs had negative effects on SAB, AUT, PPF, SOP, and DAD scores of older adults (p \< 0.05). Suffering from three or more NCDs exerted negative effects on all dimensions in the HRQOL of older adults (p \< 0.05). In terms of the WHOQOL-OLD total score of people with different numbers of NCDs, as compared with the reference group (with zero number of NCDs), suffering from one NCD (B = --0.87, 95% CI = --1.67 to --0.08, p \< 0.05), two NCDs (B = --2.89, 95% CI = --3.87 to --1.90, p \< 0.001), and three or more NCDs (B = --4.20, 95% CI = --5.36 to --3.05, p \< 0.001), all had negative impacts on the HRQOL of older adults, and as the number of NCDs increased, the HRQOL of older adults became worse, after adjusting for the confounding factors that were significantly associated with HRQOL. 4. Discussion {#sec4-ijerph-17-05150} ============= In the present study, we examined the association between the number of NCDs and HRQOL among Chinese older adults. Suffering from one NCD had a slight negative effect on the HRQOL of older adults, but as the number of NCDs increased, this negative effect also gradually increased. The healthy aging proposed by the WHO is not only the extension of life, but also the improvement of the HRQOL of older adults. At the same time, NCDs serving as a negative factor of the HRQOL of older adults have received increasing attention \[[@B27-ijerph-17-05150]\]. In China, Huang's \[[@B28-ijerph-17-05150]\] survey in Heilongjiang Province found that people with NCDs had a worse HRQOL than people without NCDs. A cross-sectional study of the HRQOL of older adults conducted by Zhu Yaxin \[[@B29-ijerph-17-05150]\] in Liaoning Province concluded that NCDs contributed to the poor HRQOL of older adults. However, these studies did not consider the impact of the number of NCDs on HRQOL. Previous studies from developed countries or regions have confirmed the relationship between NCDs and the HRQOL \[[@B11-ijerph-17-05150],[@B30-ijerph-17-05150]\]. For instance, Wacker \[[@B31-ijerph-17-05150]\] evaluated the HRQOL of 2291 COPD patients, and the results showed that COPD combined with other NCDs significantly reduced HRQOL. These studies have almost proved the harmful effect of NCDs on the HRQOL. However, the relationship between the number of NCDs and the HRQOL is worth discussing. In the results of this study, suffering from one NCD had a negative impact on the HRQOL of older people. However, when compared with two or more NCDs, this negative effect was relatively minor, which was consistent with Xin Yu's survey results in southern China \[[@B16-ijerph-17-05150]\]. In fact, NCDs are generally considered as diseases with an unknown etiology, long duration, and a generally slow progression \[[@B32-ijerph-17-05150]\]. These characteristics may partly explain the smaller impact of having one NCDs on HRQOL. Moreover, related studies have found a positive correlation between economics and NCDs \[[@B33-ijerph-17-05150]\]. Therefore, it is reasonable to speculate that older adults with one NCD have a higher economic level, which may also alleviate the harmful effect of NCDs on HRQOL. By further comparing the dimensions of HRQOL, the results of this study showed that the impact of having one NCD on three dimensions (SAB, AUT, and INT) of HRQOL was not significant. SAB, AUT, and INT evaluate the sensory ability, independent ability and intimacy of older people, respectively. However, NCDs generally refer to medical diseases with a long duration and no infection \[[@B7-ijerph-17-05150]\]. Therefore, the hypothesis that one NCD has a weaker effect on the sensory ability and independent ability of older adults is reliable. Moreover, NCDs are not contagious and have almost no external manifestations, which might explain why there was no difference in intimacy between people without NCDs and with one NCD. Unlike the above three dimensions, having one NCD had significant negative effects on the other three dimensions (PPF, SOP, and DAD) of HRQOL. PPF, SOP, and DAD evaluate the satisfaction with expectations, participation in social activity, and care, worry, and fear of death of older people, respectively. Older adults with NCDs not only experience a change in their physical health, but also bear a huge psychological burden \[[@B34-ijerph-17-05150]\]. This could partly explain why older people lower their expectations for the future. Moreover, because of the existence of NCDs, older people's social participations are likely to be restricted \[[@B35-ijerph-17-05150]\]. Since older adults are at the end of the life cycle, they are already facing the fear of death \[[@B36-ijerph-17-05150]\], which will inevitably increase once they suffer from a disease. All in all, having one NCD had a slight impact on the HRQOL of older adults, which may be a potential hazard. It may cause older adults with one NCD to ignore the prevention and management of the disease, which may lead to multiple diseases and a worse HRQOL. As our research results showed, suffering from two or more NCDs had a more serious negative impact on the HRQOL of older adults. The results are consistent with other studies. Previous research has shown that NCDs can impair the HRQOL of older adults, and the impairment was more severe when suffering from multiple NCDs, which may be related to the additive or synergistic damage caused by various NCDs to the body \[[@B37-ijerph-17-05150]\]. In China, Deng \[[@B38-ijerph-17-05150]\] found that adults with NCDs had a lower HRQOL than adults without NCDs, and their HRQOL had declined more significantly as the number of NCDs increased. Multiple studies in other countries have also demonstrated the inverse relationship between the number of NCDs and HRQOL \[[@B39-ijerph-17-05150],[@B40-ijerph-17-05150]\]. Furthermore, we found that having two NCDs had a significant negative effect on the five dimensions (SAB, AUT, PPF, SOP, DAD) of the HRQOL of older adults. When older people had ≥three NCDs, all dimensions of the HRQOL of older adults were negatively affected. This further confirms the fact that, as the number of NCDs increases, the negative impact on the HRQOL of older adults also increases. Therefore, our study provided strong evidence for the fact that older adult people with multimorbidity might have a worse HRQOL. It is worth noting that NCDs are considered to be complication-prone diseases, especially among older adults \[[@B41-ijerph-17-05150]\]. Although the harm of multimorbidity to older adults' HRQOL has been proven in the previous studies, the incidence of multimorbidity has continued to rise in recent years, and the HRQOL of older adults remains pessimistic \[[@B42-ijerph-17-05150]\]. Therefore, the best option to improve the HRQOL of older adults is to attach importance to the primary prevention of NCDs, which has been proven to be effective \[[@B43-ijerph-17-05150],[@B44-ijerph-17-05150]\]. For older adults with one NCD, more attention should be paid to the treatment and management of the disease to avoid greater harm to the body caused by comorbidities, and fundamentally improves the HRQOL. Generally speaking, implementing primary prevention plays a key role in preventing the occurrence of NCDs and improving the HRQOL of older adults. However, we should also note that there was a close relationship between the number of NCDs and the HRQOL. As the number of NCDs increased, the HRQOL of older adults became worse. Therefore, we should pay attention to the prevention and management of NCDs of older adults to prevent the occurrence of multiple diseases. Some limitations of this study should be taken into consideration. First, the existence of NCDs was self-reported by the respondents, which may lead to the omission of undetected diseases. Second, the types of NCDs listed in the questionnaire were incomplete, which could result in the omission of other NCDs. At the same time, due to the lack of data, the impact of NCD duration on HRQOL was not considered. Obviously, HRQOLs of patients with different types of NCDs are different, and people with the same disease may have a different HRQOL due to different treatment stages. However, considering the complex combination of disease types, in this study, we did not condider the impact of different diseases and their severity on HRQOL. Third, other potential confounding variables, such as smoking and drinking, were not adjusted in the model. Fourth, this cross-sectional survey does not allow us to obtain the dynamic impact of a single NCD or multiple NCDs on HRQOL. 5. Conclusions {#sec5-ijerph-17-05150} ============== This study clearly showed the positive relationship between the number of NCDs and the HRQOL of older people. Compared with older adults without NCDs, suffering from one NCD had a slight negative impact on the HRQOL of older adults; however, having two or more NCDs exerted a greater negative effect on the HRQOL of older adults. As the number of NCDs increased, the HRQOL of older adults was worse. Therefore, we suggest primary prevention as the key to the improvement of HRQOL of older adults by avoiding or delaying the occurrence of one NCD. More importantly, considering the seeming harmlessness of one NCD and the negative effect of multimorbidity on HRQOL, more importance should be attached to joint interventions and programs that target the prevention and control of multimorbidity among older adults with one NCD. The lead agency of China's Health-Related Quality of Life Survey for Older Adults 2018 is the Global Health Institute of Wuhan University. We would like to express our great appreciation to those students from key universities in China for their help in data collection. Conceptualization, S.W.; methodology, J.L.; software, J.L.; validation, W.Y. and S.W.; formal analysis, J.L.; investigation, J.L., W.Y., J.Z., Y.Y., S.C. and S.W.; resources, S.W.; data curation, J.L.; writing---original draft preparation, J.L., and W.Y.; writing---review and editing, J.L., W.Y. and S.W.; supervision, S.W.; project administration, S.W.; funding acquisition, S.W. All authors have read and agreed to the published version of the manuscript. This research was funded by Wuhan University "Double First-Class (World's First-Class University &World's First-Class Disciplines)" development (Special fund \#C). No support was received from industry. All researchers acted independently of the funding bodies. The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results. ijerph-17-05150-t001_Table 1 ###### The number of noncommunicable diseases (NCDs) in different demographic groups. Variables The Number of Noncommunicable Diseases (NCDs) --------------------------------------- ----------------------------------------------- -------------- ------------- ------------- Gender Male 1299 (50.9%) 681 (26.7%) 332 (13.0%) 238 (9.3%) Female 1202 (46.6%) 668 (25.9%) 423 (16.4%) 289 (11.2%) Age 60--64 618 (54.0%) 269 (23.5%) 143 (12.5%) 114 (10.0%) 65--69 581 (48.5%) 321 (26.8%) 179 (14.9%) 118 (9.8%) 70--74 551 (46.2%) 352 (29.5%) 174 (14.6%) 116 (9.7%) 75--79 327 (44.2%) 203 (27.5%) 118 (16.0%) 91 (12.3%) 80--84 257 (45.2%) 141 (24.8%) 100 (17.6%) 71 (12.5%) ≥85 167 (55.3%) 70 (23.2%) 45 (14.9%) 20 (6.6%) Nationality Han 2369 (49.2%) 1255 (26.1%) 690 (14.3%) 498 (10.3%) Minority 102 (41.5%) 76 (30.9%) 44 (17.9%) 24 (9.8%) BMI Underweight (\<18.5) 253 (46.9%) 155 (28.7%) 81 (15.0%) 51 (9.4%) Normal (18.5--23.9) 1601 (50.8%) 806 (25.6%) 438 (13.9%) 306 (9.7%) Overweight (24--27.9) 479 (43.7%) 315 (28.7%) 177 (16.1%) 126 (11.5%) Obese (≥28) 95 (41.3%) 57 (24.8%) 40 (17.4%) 38 (16.5%) Years of education 0 471 (41.8%) 313 (27.7%) 193 (17.1%) 151 (13.4%) 1--5 715 (49.4%) 377 (26.1%) 212 (14.7%) 142 (9.8%) 6--8 461 (49.7%) 241 (26.0%) 139 (15.0%) 87 (9.4%) 9--11 331 (48.0%) 182 (26.4%) 97 (14.1%) 80 (11.6%) ≥12 300 (49.6%) 180 (29.8%) 78 (12.9%) 47 (7.8%) Household registration Agricultural 1626 (46.6%) 948 (27.1%) 534 (15.3%) 385 (11.0%) Non-agricultural 852 (53.5%) 389 (24.4%) 214 (13.4%) 137 (8.6%) Marital status Married/cohabiting 1687 (51.4%) 857 (26.1%) 433 (13.2%) 302 (9.2%) others 815 (43.9%) 492 (26.5%) 323 (17.4%) 225 (12.1%) Average annual household income (CNY) \<15,000 816 (45.0%) 494 (27.3%) 274 (15.1%) 228 (12.6%) 15,000--30,000 552 (43.3%) 352 (27.6%) 215 (16.9%) 156 (12.2%) 30,000--45,000 523 (54.6%) 234 (24.4%) 135 (14.1%) 66 (6.9%) 45,001--60,000 334 (59.6%) 127 (22.7%) 59 (10.5%) 40 (7.1%) \>60,000 254 (53.6%) 135 (28.5%) 59 (12.4%) 26 (5.5%) Self-rated health status Good 942 (62.4%) 347 (23.0%) 163 (10.8%) 57 (3.8%) General 1268 (48.8%) 724 (27.9%) 384 (14.8%) 220 (8.5%) Poor 294 (28.3%) 283 (27.2%) 211 (20.3%) 251 (24.2%) New personal savings (CNY) \<10,000 862 (43.5%) 507 (25.6%) 324 (16.4%) 288 (14.5%) 10,000--30,000 473 (45.1%) 301 (28.7%) 169 (16.1%) 106 (10.1%) 30,000--50,000 355 (53.5%) 169 (25.5%) 94 (14.2%) 46 (6.9%) 50,000--100,000 360 (49.7%) 213 (29.4%) 102 (14.1%) 49 (6.8%) ≥100,000 438 (63.1%) 157 (22.6%) 62 (8.9%) 37 (5.3%) Note: Sample sizes of the demographic characteristic variables may not sum to n = 5166 due to missing values. ijerph-17-05150-t002_Table 2 ###### Scores of World Health Organization Quality of Life-Old (WHOQOL-OLD) among the population with different numbers of NCDs. Domains Score Range The Number of Noncommunicable Diseases (NCDs) F p Multiple Comparisons ------------------------------------------- ------------- ----------------------------------------------- --------------- --------------- ---------------------- -------- -------------- -------------------------- Sensory Abilities (SAB) 4--20 13.60 ± 3.42 13.12 ± 3.56 12.10 ± 3.40 11.21 ± 3.62 89.376 p \< 0.001 g 4 \< g 3 \< g 2 \< g 1 Autonomy (AUT) 4--20 13.93 ± 3.34 13.74 ± 3.34 13.21 ± 3.21 12.91 ± 3.39 19.369 p \< 0.001 g 4, g 3 \< g 2, g 1 Past, Present and Future Activities (PPF) 4--20 13.50 ± 3.15 13.03 ± 2.94 12.57 ± 2.79 12.25 ± 2.99 36.733 p \< 0.001 g 4, g 3 \< g 2 \< g 1 Social Participation (SOP) 4--20 13.37 ± 3.04 12.78 ± 2.92 12.64 ± 2.73 12.10 ± 2.94 35.342 p \< 0.001 g 4, \< g 3, g 2 \< g 1 Death and Dying (DAD) 4--20 12.53 ± 3.20 12.09 ± 3.31 11.62 ± 3.57 11.51 ± 4.09 23.346 p \< 0.001 g 4, g 3 \< g 2 \< g 1 Intimacy (INT) 4--20 12.95 ± 3.32 12.95 ± 3.32 12.67 ± 3.39 11.74 ± 3.87 20.042 p \< 0.001 g 4 \< g 3 \< g 2, g 1 Total score 24--120 79.87 ± 13.25 77.71 ± 13.13 74.80 ± 12.49 71.73 ± 14.56 70.701 p \< 0.001 g 4 \< g 3 \< g 2 \< g 1 ijerph-17-05150-t003_Table 3 ###### Relationship between the number of NCDs and WHOQOL-OLD scale scores. Models Sensory Abilities (SAB) Autonomy (AUT) Past, Present and Future (PPF) Social Participation (SOP) Death and Dying (DAD) Intimacy (INT) Total score --------- ----------------------------- ----------------------------- -------------------------------- ----------------------------- ----------------------------- ----------------------------- ----------------------------- Model 1 0 Reference Reference Reference Reference Reference Reference Reference 1 − −0.19 (−041, 0.03) −0.47 (−0.69, −0.27) \\\* −0.59 (−0.78, −0.39) \\\* −0.45 (−0.67, −0.22) \\ −0.01 (−0.22, 0.22) −2.17 (−3.04, −1.29) \\\* 2 −1.50 (−1.78, −1.22) \\\* −0.72 (−0.99, −0.45) \\\* −0.94 (−1.18, −0.69) \\\* −0.72 (−0.96, −0.48) \\\* −0.92 (−1.19, −0.64) \\\* −0.28 (−0.56, −0.01) \* −5.08 (−6.15, −4.00) \\\* ≥3 −2.39 (−2.71, −2.06) \\\* −1.02 (−1.33, −0.71) \\\* −1.25 (−1.54, −0.97) \\\* −1.26 (−1.54, −0.99) \\\* −1.02 (−1.34, −0.70) \\\* −1.21 (−1.52, −0.89) \\\* −8.15 (−9.39, −6.91) \\\* Model 2 0 Reference Reference Reference Reference Reference Reference Reference 1 −0.01 (−0.21, 0.21) 0.12 (−0.19, 0.25) −0.29 (−0.48, −0.10) \\ −0.37 (−0.56, −0.18) \\\* −0.26 (−0.49, −0.03) \* 0.16 (−0.07, 0.38) −0.68 (−1.48, 0.12) 2 −0.85 (−1.11, −0.58) \\\* −0.43 (−0.66, −0.16) \\ −0.58 (−0.82, −0.34) \\\* −0.36 (−0.60, −0.13) \\ −0.83 (−1.12, −0.54) \\\* 0.04 (−0.24, 0.32) −2.91 (−3.91, −1.91) \\\* ≥3 −1.13 (−1.44, −0.82) \\\* −0.56 (−0.86, −0.25) \\\* −0.63 (−0.91, −0.35) \\\* −0.57 (−0.84, −0.30) \\\* −0.62 (−0.95, −0.28) \\\* −0.71 (−1.03, −0.38) \\\* −4.01 (−5.17, −2.84) \\\* Model 3 0 Reference Reference Reference Reference Reference Reference Reference 1 −0.02 (−0.23, 0.19) 0.06 (−0.15, 0.27) −0.33 (−0.52, −0.14) \\\* −0.42 (−0.61, −0.24) \\\* −0.25 (−0.48, −0.02) \* 0.18 (−0.04, 0.41) −0.87 (−1.67, −0.08) \* 2 −0.81 (−1.07, −0.55) \\\* −0.33 (−0.59, −0.07) \\ −0.58 (−0.81, −0.34) \\\* −0.38 (−0.61, −0.15) \\\* −0.80 (−1.08, −0.51) \\\* 0.04 (−0.23, 0.32) −2.89 (−3.87, −1.90) \\\* ≥3 −1.19 (−1.50, −0.89) \\\* −0.32 (−0.63, −0.02) \\ −0.65 (−0.93, −0.37) \\\* −0.59 (−0.86, −0.32) \\\* −0.62 (−0.95, −0.29) \\\* −0.72 (−1.04, −0.39) \\\* −4.20 (−5.36, −3.05) \\\* Note: \* p \< 0.05; \\ p \< 0.01; \\\* p \< 0.001. Model 1 was the crude model. Model 2 adjusted for all the potential predictors (age, sex, nationality, BMI, years of education, household registration, marital status, average annual household income (CNY), new personal savings and self-rated health status). Model 3 was the final parsimonious model, adjusting for all the predictors that were significantly associated with HRQOL in Model 2. [^1]: These authors contributed equally to this study.	{ "pile_set_name": "PubMed Central" }
1000 Genomes Project data were obtained from a third party, and are available at the following URL: <http://www.1000genomes.org/data>. COGA data are available from the dbGAP database, accession number phs000125.v1.p1. Introduction {#sec001} ============ Consequences of nicotine dependence are the leading cause of preventable death in the USA and it has been noted that some populations experience higher levels of addiction than others \[[@pone.0134393.ref001]\] but the reason for this is not understood. Multiple studies have demonstrated a genetic component to nicotine addiction \[[@pone.0134393.ref002]--[@pone.0134393.ref004]\], but little is known about the role of natural selection in shaping the genetic components of nicotine addiction. Such knowledge could help us understand the genetic and behavioral nature of addiction and ultimately facilitate the design and delivery of appropriate interventions to reduce nicotine addiction. It has been estimated that approximately 10% of the genome has been affected by linkage due to recent selective sweeps \[[@pone.0134393.ref005]\]. However, it is challenging to make direct and clear inference on the phenotypic that is the target of natural selection. This is particularly true when the phenotype being examined has no obvious evolutionary consequences, and has a disease or trait-related association with no clear reproductive consequence today, but where patterns of genetic variation are consistent with a positive selective sweep in recent human history. In this case, additional mechanisms and/or alternative explanations must be sought for the existence of selection on the gene of interest. An example of such a situation occurs in the gene encoding hemoglobin. In homozygous form, the 'sickle cell' allele, HbS, drives the formation of malformed red cells, which aggregate, blocking blood flow to numerous organs including the brain. This results in organ damage and strokes, severely shortening the lifespan of the individual. Nonetheless, the HbS allele is maintained in the gene pool in regions where malaria is endemic because in heterozygous form it provides protection against malaria (for a review see \[[@pone.0134393.ref006]\]. The case of nicotine addiction represents a similar conundrum. Several genetic variants that modify susceptibility to or protection from nicotine dependence have been identified by genome-wide association studies (GWAS) \[[@pone.0134393.ref002],[@pone.0134393.ref004],[@pone.0134393.ref007]\]. Perhaps not surprisingly, the loci identified in these studies mainly include genes encoding neuronal nicotinic cholinergic receptors (CHRNs). Neuronal cholinergic nicotinic receptors (CHRNs) are a heterogeneous class of cation (positively charged) channels expressed in the central and peripheral nervous system. There are 11 neuronal CHRN genes, each of which encodes a receptor subunit. The neuronally expressed nicotinic receptors consist of combinations of alpha and beta subunits, encoded in humans by 8 alpha (α2-α 7, α9-α10) and 3 beta (β2-β4) genes \[[@pone.0134393.ref008]\]. These subunits form homo- or hetero-pentameric subtypes, which are present in various regions throughout the nervous system. To form a receptor, five subunits must be combined within the cell and the specific combination of these subunits defines the receptor subtype. In the body, the opening of these channels is controlled by the endogenous ligand, acetylcholine, a chemical produced by neurons to activate other nearby neurons. Nicotine, the major psychoactive chemical present in tobacco smoke is a chemical present in the environment that can also stimulate the opening of these nicotinic acetylcholine receptor ion channels \[[@pone.0134393.ref009]\]. A number of GWAS studies have been performed that demonstrate an association between the nicotinic receptors and smoking. The strongest association between nicotinic receptors and nicotine addiction is a non-synonymous change (rs16969968, D398N) in the gene encoding the α5 subunit of the nicotinic receptor (CHRNA5) on chromosome 15 \[[@pone.0134393.ref007],[@pone.0134393.ref010]--[@pone.0134393.ref014]\]. When cells are made to express nicotinic receptors containing the minor allele form of this SNP (398N), agonists induce less channel opening and cell activation than in cells that express receptors containing the major allele (398D) \[[@pone.0134393.ref011]\]. Thus, the minor allele at this SNP results in a significant functional change in the behavior of this ion channel, causing more nicotine to be needed in individuals with the minor allele to produce the same effect. Additionally, SNPs within the chromosome 15 and chromosome 8 regions have been associated with alocohol and cocaine dependence in addition to their associations with smoking related traits \[[@pone.0134393.ref015]--[@pone.0134393.ref017]\]. Although the underlying functional mechanism underlying the associations at the nicotinic receptors on chromosome 8 is not known, there is nevertheless a genome-wide significant signal with the LD bin \[[@pone.0134393.ref018]\]. Here we test the hypothesis that natural selection has acted on these genes. But if the null hypothesis of neutral evolution or demographic processes is rejected, why would nature seemingly select for this trait, especially given the fact that it is believed that nicotine has not been a part of our evolutionary history long enough, and in large enough quantities, for its effects to be visible in our genomes? One hypothesis is that selection acted on another phenotype and the effect on nicotine addiction was secondary and incidental, a genetic phenomenon termed hitchhiking. Nicotine is known to have an enhancing effect on cognitive performance. It enhances the reorientation of attention in visuospatial tasks \[[@pone.0134393.ref019]\] and alters the neuronal activity responsible for increased attention and arousal \[[@pone.0134393.ref020]\]. Furthermore, several studies have found associations between SNPs in multiple nicotinic receptor genes and cognitive performance \[[@pone.0134393.ref021],[@pone.0134393.ref022]\]. However, effects on cognitive performance at SNPs related to nicotine dependence seem peculiar given the presumed acceleration in cognitive development over recent human history. Therefore, these results suggest that if selection is shaping the genetic landscape of these genes, it may be through their effect on cognitive function in the absence of drug use. Evidence from nicotinic receptor knockout mice also supports a role for nicotinic receptors in memory and learning, as well as anxiety levels. CHRNA7 knockout mice have impaired reaction times \[[@pone.0134393.ref023]\] and decreased procedural learning \[[@pone.0134393.ref024]\], while, CHRNA6 knockout mice show that this receptor plays a role in nicotinic modulation of dopaminergic transmission, an important component of learning and memory \[[@pone.0134393.ref025]\]. Based on these observations, we hypothesize that at least some of the nicotinic receptors may have been targets of recent selection and that this selection is related to the role of nicotinic receptors in memory and learning. To test this hypothesis, we use two summaries of genetic variation that have different statistical power to make inferences, depending on the model of selection and associated population demography. Two different methods were used for detecting natural selection at two loci relevant to nicotine dependence, specifically the CHRNA5-A3-B4 region on chromosome 15q25 and the CHRNB3-A6 region on chromosome 8p11. We provide strong evidence for selection in the CHRNB3-A6 region and moderate evidence for selection in the CHRNA5-A3-B4 region. However, there is only a modest correlation between nicotine dependence and score on the Wechsler Adult Intelligence Scale (WAIS) Digit Symbol test in our dataset. Overall, we suggest that one possible explanation for these results is that SNPs in these regions associated with risk of nicotine dependence are associated with natural selection acting at these loci to improve human memory and learning. Results {#sec002} ======= Tajima's D Test {#sec003} --------------- We calculated Tajima's D over the regions of the nicotinic receptors on chromosomes 8 and 15, as well as lactase on chromosome 2 as the positive control. As the program utilized does not incorporate ancestral information, all results are based on the folded frequency spectrum, i.e. the distribution of polymorphic sites according to the number of chromosomes that carry a given minor allele rather than the number of chromosomes that carry the derived, non-ancestral allele. \[[@pone.0134393.ref026]\][Fig 1](#pone.0134393.g001){ref-type="fig"} shows regional plots of the Tajima's D values for the LCT, CHRNA5-A3-B4 nicotinic receptor gene region and the CHRNA6-B3 gene region. Within each of these regions, we calculated sliding windows of 10 kb with an increment of 1 kb. We then compared the number of windows within each region with Tajima's D values above 2 or below -2 (this represents the 95% confidence interval of values in our data) to permutations of 10,000 randomly selected regions of the same size across the genome. ![Tajima's D and iHS Results.\ A-CTajima's D andD-F) iHS for the LCT region on chromosome 2, the CHRNA3-B4-A5 gene cluster on chromosome 15 and the CHRNB3-A6 gene cluster on chromosome 8 across individuals of European (red), African (Purple) or Asian (green) ancestry. Lines are the 95% confidence intervals as calculated by permutation for iHS and Tajima's D.](pone.0134393.g001){#pone.0134393.g001} We utilized the LCT gene region as a positive control. LCT encodes the protein lactase and mutations in the region give rise to lactase persistence. The lactase persistence phenotype is due to changes in a regulatory region that enhances the expression of LCT located in intron 13 of the neighboring gene, MCM6 \[[@pone.0134393.ref027]\]. [Fig 1A](#pone.0134393.g001){ref-type="fig"} shows the distribution of Tajima's D values in the sliding windows in the LCT region in AFR, ASN and EUR. As expected, both the Asian and European populations show many windows above 2 but very few exist in the African population. All p-values are derived using permutations of 10,000 regions of the same size from across the genome. For Europeans and Asians, the proportion of windows with extreme values was 8.6% and 6.5%, respectively ([Table 1](#pone.0134393.t001){ref-type="table"}). This is highly significant in EUR, when compared to the negative control regions, and modestly significant in ASN (\<p = 2x10^-7^ and p = 0.01 for EUR and ASN, respectively) and appears inconsistent with the null hypothesis of neutral evolution and demography at the LCT locus especially in the EUR populations. In contrast, the histogram for AFR shows few windows with extreme Tajima's D values. Indeed, the proportion of windows with extreme values was \<1% and was not significantly different from the negative controls, therefore we are unable to reject the null hypothesis of neutral evolution and demography at this locus in AFR. 10.1371/journal.pone.0134393.t001 ###### Summary of Extreme Tajima's D and iHS Values in Each Region in Comparison with Negative Controls. ![](pone.0134393.t001){#pone.0134393.t001g} Gene(s) Pop \# windows\\|TajD\\|\>2 Proportion p-value \# windows\\|iHS\\|\>2 Proportion p-value ---------------- --------- ----------------------- ------------------------- ---------------------------- ---------------------- ------------------------- --------------------------- CHRNA5-A3-B4 EUR 173/1273 13.60% *\<2 x 10* ^*−16^ 1/181 \<1% NS ASN 91/1273 7.10% 4 x 10* ^*−4^ 3/224 1.3% NS AFR 30/1294 2.30% NS 16/266 5.6% NS CHRNB3-A6* EUR 13/1066 1.22% NS 26/110 23.6% *4 x 10* ^*−4^ ASN 1/973 \<1% NS 3/116 2.59% NS AFR 23/1097 2.10% NS 93/357 26.1% 3 x 10* ^*−4^ LCT/MCM6* EUR 98/1136 8.63% *2 x 10* ^*−7^ 111/138 80.4% \< 1 x10* ^*-6^ ASN 76/1161 6.55% *0.01** 20/219 9.1% NS AFR 5/1240 \<1% NS 9/278 3.2% NS Proportion of extreme Tajima's D and his values in selected regions. P-values are produced by permutation of 10,000 randomly selected regions of the same size across the genome. We next measured Tajima's D in the CHRNA5-A3-B4 gene cluster on chromosome 15. In this region, we analyzed a 120 kb region that includes the nicotinic receptor gene cluster as well as \~18 kb upstream of CHRNA5. This was done to ensure that we included the large region upstream of CHRNA5 that has been associated with regulation of the level of mRNA expression for CHRNA5 \[[@pone.0134393.ref028]\]. [Fig 1B](#pone.0134393.g001){ref-type="fig"} shows the distribution of Tajima's D values in the sliding windows in the CHRNA5-A3-B4 region in AFR, ASN and EUR. Overall, we observed increased numbers of sliding windows with extreme Tajima's D values in each of the three populations, which is not consistent with a neutral evolution and demography in this region ([Table 1](#pone.0134393.t001){ref-type="table"}). Of note, several SNPs previously shown to be associated with nicotine, alcohol and/or cocaine dependence \[[@pone.0134393.ref007],[@pone.0134393.ref010]--[@pone.0134393.ref014]\] are within regions with high Tajima's D values. Additionally, within the EUR population, these regions also harbor SNPs in an LD bin tagged by rs5887655 previously shown to be significantly associated with CHRNA5 mRNA expression in the brain \[[@pone.0134393.ref028]\]. [Table 2](#pone.0134393.t002){ref-type="table"} summarizes the Tajima's D values for windows containing associated SNPs in each of these two LD bins. 10.1371/journal.pone.0134393.t002 ###### Tajima's D and iHS values for CHRNA5-A3-B4 region. ![](pone.0134393.t002){#pone.0134393.t002g} SNP Function EUR Freq Tajima's D EUR iHS EUR ASN Freq Tajima's D ASN iHS ASN AFR Freq Tajima's D AFR iHS AFR ---------------- ------------------- ---------- ------------------ --------- ---------- ------------------ --------- ---------- ------------------ --------- rs72740955 intergenic 0.37 (1.53, 2.07) -0.74 0.03 (0.72, 0.97) -1.25 0.09 (0.88, 1.00) 0.44 rs2036527 intergenic 0.37 (0.11, 0.11) -0.5 0.03 (-0.57, -0.54) -0.85 0.16 (-1.15, 0.20) 1.48 rs55853698 CHRNA5 5\'UTR 0.37 (0.56, 1.93) -0.65 0.04 (-0.02, 1.00) -0.31 0.06 (0.24, 0.56) 1.05 rs17486195 CHRNA5 intronic 0.36 (1.59, 1.64) -0.62 0.03 (-0.17, 0.14) -0.82 0.11 (-0.10, -0.07) 0.48 rs17486278 CHRNA5 intronic 0.36 (0.39, 0.70) -0.62 0.31 (0.04, 0.60) -0.45 0.28 (0.42, 0.47) 0.72 rs72740964 CHRNA5 intronic 0.36 (0.78, 1.63) -0.73 0.03 (0.21, 0.98) -0.57 0.04 (0.08, 0.56) 0.14 rs951266 CHRNA5 intronic 0.36 (-0.22, 0.22) -0.65 0.03 (-0.84, -0.37) -0.5 0.08 (-0.89, -0.74) 0.91 rs16969968 CHRNA5 missense 0.36 (2.00, 2.33) -0.71 0.03 (0.43, 1.01) -0.44 0.02 (0.29, 0.62) -0.01 rs1051730 CHRNA3 synonymous 0.36 (1.40, 1.90) -0.64 0.03 (0.41, 0.78) -0.67 0.09 (-0.29, -0.23) 0.69 rs1317286 CHRNA3 intronic 0.36 (0.32, 1.27) -0.8 0.09 (2.00, 2.56) 0.78 0.24 (0.73, 1.36) -0.43 rs12914385 CHRNA3 intronic 0.4 (2.35, 2.55) -0.39 0.32 (1.85, 2.01) 0.36 0.2 (1.13, 1.35) 0.39 rs114205691 CHRNA3 intronic 0.64 (0.80, 1.65) 1.53 0.68 (0.00, 0.81) 0.15 0.8 (1.75, 1.97) 0.23 rs8040868 CHRNA3 synonymous 0.59 (0.87, 1.37) 1.54 0.62 (1.63, 2.17) 0.25 0.63 (-0.27, 0.13) 0.67 rs55958997 intergenic 0.39 (1.03, 1.66) -1.08 0.06 (0.74, 1.23) 0.23 0.3 (-0.41, -0.38) -0.45 rs72743158 CHRNB4 intronic 0.38 (0.26, 1.27) -0.79 0.02 (0.79, 1.34) -0.83 0.03 (0.20, 0.20) 1.11 rs55988292 CHRNB4 intronic 0.39 (-1.52, -0.24) -0.26 0.02 (-0.35, 028) -0.94 0.14 (-0.93, -0.61) -0.7 rs4275821 intergenic 0.33 (2.02, 2.32) -0.39 0.14 (0.61, 088) -0.02 0.18 (0.74, 0.99) -0.13 rs588765 CHRNA5 intronic 0.4 (2.18, 2.22) -0.32 0.15 (0.00--0.26) -0.2 0.25 (-0.30, -0.30) 0.44 rs6495306 CHRNA5 intronic 0.4 (2.07, 2.32) -0.32 0.15 (-0.42, 0.06) -0.2 0.26 (-0.16, 0.31) 0.41 rs495090 CHRNA5 intronic 0.35 (1.47, 2.06) -0.66 0.2 (1.17, 1.37) 0.48 0.4 (0.77, 1.23) 0.06 rs680244 CHRNA5 intronic 0.4 (-0.77, 0.83) -0.27 0.21 (-0.60, 0.24) 0.62 0.4 (-0.63, -0.06) -0.11 rs621849 CHRNA5 intronic 0.41 (0.36, 1.35) -0.27 0.21 (2.07, 2.08) 0.57 0.4 (0.47, 1.13) -0.21 rs11637635 CHRNA5 intronic 0.65 (1.61, 2.18) 1.51 0.86 (-0.28, -0.13) 0.99 0.78 (-0.60, -0.35) 0.35 rs481134 CHRNA5 intronic 0.6 (0.14, 0.77) 1.1 0.85 (-0.48, 0.19) 1.06 0.75 (-0.87, -0.61) 0.19 rs555018 CHRNA5 intronic 0.4 (-0.84, 0.78) -0.22 0.15 (-0.39, 0.04) 0.01 0.26 (-0.77, -0.23) 0.97 rs647041 CHRNA5 intronic 0.4 (-1.06, -0.14) -0.24 0.18 (-1.05, 0.03) 0.24 0.24 (-1.08, -0.60) 0.77 rs615470 CHRNA5 3\' UTR 0.65 (2.54, 2.57) 1.46 0.83 (-1.20, -0.75) 0.66 0.67 (-1.09, -0.60) -0.08 rs6495307 CHRNA3 intronic 0.6 (-1.17, 0.14) 0.96 0.82 (0.55, 0.84) 0.71 0.66 (-0.83, -0.41) -0.07 rs62010327 CHRNA3 intronic 0.35 (0.15, 1.60) -0.74 0.14 (0.42, 0.85) -0.78 0.08 (0.65, 0.97) -0.99 rs12901300 CHRNA3 intronic 0.4 (-1.37, -0.26) -0.3 0.19 (-0.71, 0.11) 0.26 0.34 (-0.47, 0.32) 0.91 rs3743077 CHRNA3 intronic 0.4 (1.86, 2.62) -0.27 0.18 (-0.25, 0.14) 0.34 0.11 (-0.96, -0.77) 0.33 rs62010328 CHRNA3 intronic 0.34 (1.86, 1.90) -0.8 0.14 (-0.54, -0.25) -0.69 0.07 (-1.17, -0.96) -0.54 rs2869546 CHRNA3 intronic 0.36 (2.08, 2.40) -0.18 0.18 (2.31, 2.63) -0.09 0.29 (3.17, 3.41) 0.66 rs4366683 CHRNA3 intronic 0.54 (-0.72, 0.13) 0.44 0.45 (1.50, 2.00) -0.96 0.5 (0.70, 1.54) 0.02 rs58643100 CHRNA3 intronic 0.46 (-0.72, 0.13) 0.63 0.26 (1.50, 2.00) 0.84 0.41 (0.70, 1.54) 0.24 All SNPs have r^2^ value of 0.8 or greater with rs16969968 or rs588765. Significant values and SNP names are bolded. For the CHRNB3-A6 region, we examined a rather larger segment of the genome upstream of the gene cluster. This was done in order to include several upstream SNPs that have previously been shown to exhibit associations with nicotine dependence or cocaine dependence \[[@pone.0134393.ref017]\]. [Fig 1C](#pone.0134393.g001){ref-type="fig"} shows a regional plot of Tajima's D values across the sliding windows in the CHRNB3-A6 region. In AFR, there were several windows with extreme Tajima's D values upstream of CHRNB3 and in the intergenic region between CHRNB3 and CHRNA6. Among the three populations tested, none showed a significant increase in the number of windows with extreme Tajima's D values when compared to the permutations. The concentration of windows with extreme Tajima's D values upstream of CHRNB3 is noteworthy in the context of risk for nicotine addiction. Of note, a recent genome-wide association study found that a SNP in this region, rs1451240, was associated with reduced risk for nicotine dependence, measured using the Fagerstrom Test for Nicotine Dependence (FTND) \[[@pone.0134393.ref018]\]. The LD bin tagged by rs1451240 spans \~66 kb and several other alleles of SNPs in this bin have been associated with reduced risk for nicotine dependence, although no others are significant at the genome-wide level \[[@pone.0134393.ref018]\]. The data in [Table 3](#pone.0134393.t003){ref-type="table"} show that in EUR and AFR, four adjacent SNPs from this LD bin, including rs1451240, were present in sliding windows with extreme Tajima's D values in AFR. These data suggest that these nicotine dependence-associated SNPs may be undergoing balancing selection or positive selection in these two populations. 10.1371/journal.pone.0134393.t003 ###### Tajima's D and iHS values for CHRNB3-CHRNA6 region. ![](pone.0134393.t003){#pone.0134393.t003g} SNP Function EUR Freq Tajima\'s D EUR iHS EUR ASN Freq Tajima\'s D ASN iHS ASN AFR Freq Tajima\'s D AFR iHSAFR ---------------- ----------------- ---------- ------------------ ---------- ---------- ----------------- --------- ---------- ------------------ -------- rs1979140 intergenic 0.77 (-0.75, -0.34) -1.44 0.82 (-1.21, -1.15) -0.47 0.28 (-1.26, -0.99) -0.15 rs7816726 intergenic 0.77 (0.33, 1.00) -1.44 0.81 (-0.02, 0.70) -0.49 0.28 (0.08, 0.17) -0.56 rs10958726 intergenic 0.23 (0.1, 0.68) 2.46 0.18 (-0.85, -0.15) 1.06 0.63 (-0.70, 0.18) 0.65 rs7842601 intergenic 0.23 (0.17, 0.62) 2.58 0.18 (-0.33, -0.17) 1.06 0.63 (0.63, 1.33) 0.84 rs13273442 intergenic 0.23 (-0.39, 0.69) 2.32 0.18 (-0.60, -0.45) 1.02 0.63 (-0.75, 0.04) -0.21 rs9792277 intergenic 0.77 (1.56, 1.80) -1.45 0.81 (1.06, 1.17) -0.54 0.33 (1.38, 2.32) 0.77 rs1451239 intergenic 0.77 (2.41, 2.58) -1.45 0.82 (1.34, 1.40) -0.46 0.34 (2.18, 2.51) 0.89 rs1451240 intergenic 0.23 (2.41, 2.54) 2.48 0.18 (1.33, 1.41) 1.11 0.67 (2.18, 2.46) 0.09 rs1901281 intergenic 0.77 (2.37, 2.68) -1.56 0.81 (1.26, 1.33) -0.54 0.28 (2.18, 2.19) 0.37 rs4736835 intergenic 0.23 (2.48, 2.68) 2.6 0.18 (1.17, 1.26) 1.11 0.66 (1.47, 2.63) -0.08 rs1955185 intergenic 0.77 (1.49, 1.67) -1.58 0.82 (0.64, 0.69) -0.52 0.28 (1.08, 1.59) 0.54 rs13277254 intergenic 0.23 (0.89, 1.27) 2.69 0.18 (0.64, 0.97) 1.32 0.63 (1.11, 1.29) -0.6 rs13277524 intergenic 0.77 (0.64, 0.89) -1.63 0.82 (0.24, 0.97) -0.65 0.28 (0.43, 1.29) 0.61 rs6474412 intergenic 0.77 (1.09, 1.43) -1.63 0.82 (0.53, 0.66) -0.65 0.34 (1.01, 1.49) 1.25 rs6474413 intergenic 0.77 (1.86, 2.19) -1.66 0.82 (0.13, 0.22) -0.65 0.28 (1.15, 1.33) 0.62 rs7004381 intergenic 0.23 (0.68, 1.86) 2.72 0.18 (-0.15, 0.22) 1.32 0.63 (0.29, 1.15) -0.72 rs6985052 intergenic 0.77 (-0.37, 0.68) -1.66 0.82 (-0.85, -0.16) -0.65 0.29 (-1.04, -0.29) 0.71 rs4950 CHRNB3 5\'UTR 0.23 (-0.39, 0.12) 2.71 0.18 (-0.44, -0.16) 1.31 0.8 (0.38, 1.55) 0.25 rs9643891 CHRNB3 intronic 0.77 (-0.34, 0.13) -1.85 0.82 (0.60, 0.72) -1.01 0.15 (-0.65, -0.16) -0.15 rs9643853 CHRNB3 intronic 0.77 (-1.04, -0.34) -1.89 0.82 (-0.60, 0.60) -1.01 0.15 (-1.55, -0.65) -0.1 rs13280604 CHRNB3 intronic 0.23 (0.47, 1.68) 2.9 0.18 (0.07, 0.50) 1.94 0.8 (0.04, 0.83) 0.12 rs6997909 CHRNB3 intronic 0.77 (0.70, 0.70) -1.88 0.82 (-0.25, 0.11) -1.19 0.15 (-1.00, -1.00) 0 rs6474414 CHRNB3 intronic 0.77 (0.40, 0.40) -1.88 0.82 (-0.60, -0.21) -1.19 0.15 (-0.76, -0.12) 0 rs6474415 CHRNB3 intronic 0.23 (1.26, 1.35) 3.07 0.18 (-0.12, 0.22) 1.97 0.85 (-0.85, -0.52) 0.84 rs4236926 CHRNB3 intronic 0.77 (-0.04, -0.03) -1.71 0.81 (-0.14, -0.11) -1.41 0.15 (-0.54, -0.54) 0.14 rs16891561 CHRNB3 intronic 0.77 (-1.28, -0.90) -1.67 0.81 (-0.89, 0.72) -1.26 0.15 (-1.08, -0.48) 0.2 rs55828312 CHRNB3 intronic 0.24 (-(0.89, -0.74) 2.86 0.19 (-1.13, -0.31) 2.13 0.83 (-0.85, 0.67) 1.12 All SNPs have r^2^ value of 0.8 or greater with rs1451240. Significant values and SNP names are bolded. Integrated Haplotype Score (iHS) Analyses {#sec004} ----------------------------------------- [Fig 1](#pone.0134393.g001){ref-type="fig"} shows regional plots of iHS scores for the LCT, CHRNA5-A3-B4 nicotinic receptor gene region and CHRNA6-B3 nicotinic receptor gene region. Within each of these regions, we calculated the iHS score for all SNPs for which we could unambiguously determine the chimp ancestral allele. We then compared the number of SNPs within each region with iHS values above 2 or below -2, to the 10,000 equally sized permutated regions that were chosen at random. We again used the LCT/MCM6 region as a positive control. [Fig 1D](#pone.0134393.g001){ref-type="fig"} shows the regional plots of iHS values in the LCT/MCM6 region for EUR, ASN and AFR populations. As expected, the AFR population shows few extreme values (only 3.2%) and this does not differ significantly from the negative control. In EUR and ASN, the overall average proportion of extreme values for this region is 80.4% and 9.1%, respectively ([Table 1](#pone.0134393.t001){ref-type="table"}). The clustering of extreme iHS values in the genic areas of this region is consistent with what is known about large-scale positive selection at this locus in the EUR population and to a lesser extent in the ASN population \[[@pone.0134393.ref027]\]. This demonstrates the validity of this approach for identifying genes undergoing selection. [Fig 1E](#pone.0134393.g001){ref-type="fig"} shows the histogram of iHS values across the CHRNA5-A3-B4 locus on chromosome 15. The summary statistics are given in [Table 1](#pone.0134393.t001){ref-type="table"}. None of the three populations showed a proportion of extreme iHS values that was significantly different than predicted by permutation. In addition, none of the SNPs with extreme values included any of the SNPs previously found to be associated with nicotine dependence ([Table 2](#pone.0134393.t002){ref-type="table"}). The presence of a significant enrichment of extreme Tajima's D values in this region, however, suggests that if positive selection in this region occurred, it may have occurred long enough ago such that the long haplotypes required for iHS have broken down by recombination. [Fig 1F](#pone.0134393.g001){ref-type="fig"} shows a histogram of the iHS values for the CHRNB3-A6 cluster on chromosome 8. Summary statistics are shown in [Table 1](#pone.0134393.t001){ref-type="table"}. Both the AFR and EUR populations show an excess of extreme iHS values (26.1% and 23.6%, respectively). By contrast, in the ASN population, there were few windows of extreme iHS scores and the overall proportion was not significantly different from the negative control. Several SNPs with extreme iHS values in these populations are contained within bins previously shown to be associated with either nicotine or cocaine dependence phenotypes \[[@pone.0134393.ref017],[@pone.0134393.ref018]\]. [Table 3](#pone.0134393.t003){ref-type="table"} lists the SNPs from the LD bin showing genome-wide significance for reduced risk for nicotine dependence and provides the iHS value for rs1451240, the tag SNP. While there are several extreme values in AFR in the middle of this gene cluster, these values do not overlap with known SNPs related to nicotine dependence in this region. In EUR, 13 SNPs in this LD bin, including the tag SNP rs1451240, have extreme iHS values. All have positive values, indicating the presence of unusually long haplotypes containing the ancestral allele suggesting that the ancestral allele, which is associated with a greater risk of nicotine dependence is being favored by selection. An LD bin in the CHRNB3-A6 region bin that has been shown to have SNPs significantly associated with increased risk for cocaine dependence \[[@pone.0134393.ref017]\] also contained an abundance of SNPs with extreme iHS values ([S1 Table](#pone.0134393.s001){ref-type="supplementary-material"}). This bin is fairly large and spans the entire CHRNB3-A6 cluster. It contains rs4952 and rs4953, two low frequency synonymous variants in CHRNB3 that have previously been reported to be associated with lower risk for nicotine dependence \[[@pone.0134393.ref010]\]. All SNPs in the bin are present at around 10% in AFR and 4% in EUR but absent in ASN, possibly explaining the lack of extreme iHS values in the ASN population. Overall, 40% of the SNPs in this bin showed extreme values in AFR and 15% of the SNPs in this bin showed extreme values in EUR. As these SNPs are absent from the ASN population, none showed extreme iHS values. The dense clustering of extreme iHS values in AFR and EUR is rarely expected under a neutral model, and is consistent with the hypothesis of the action of recent selection. In EUR, all of the SNPs in this LD bin with extreme iHS values had positive values and all were shared with AFR. Thus, in both populations, the ancestral allele associated with increased risk for nicotine dependence and decreased risk for cocaine addiction is being favored. Nicotine addiction and cognitive function {#sec005} ----------------------------------------- The Tajima's D analysis and integrated haplotype score both indicate that the CHRNB3-A6 cluster is undergoing selection and in particular, the iHS scores suggest that it is the risk allele for nicotine dependence on chromosome 8 that is under positive selection. As it seems unlikely that risk for nicotine dependence is the phenotype undergoing selection, and because nicotinic receptors are involved in memory and learning, we hypothesized that a phenotype related to memory or learning, such as attention, might be the phenotype being selected. To test this possibility, we obtained genotype and cognitive phenotype data from the Collaborative Study of the Genetics of Alcoholism (COGA). Using this dataset, we tested the association between genotypes and three of the most relevant phenotypes, namely scores on the Wechsler Adult Intelligence Scale (WAIS) Block Design, WAIS Digit Symbol and WAIS Information tests. These tests were designed to measure aspects of perceptual organization, processing speed and verbal comprehension, respectively \[[@pone.0134393.ref029]\]. [Table 4](#pone.0134393.t004){ref-type="table"} summarizes our findings for the top SNPs associated with WAIS Digit Symbol test. No other neurocognitive phenotypes besides WAIS digit symbol had SNPs with significant values in the CHRNB3-A6 region and none of the three neurocognitive phenotypes had a significant association with SNPs in the region of CHRNA5-A3-B4 on chromosome 15 (not shown). Of the 17 SNPs in the CHRNB3-A6 region on chromosome 8, one SNP--rs7017612---passed multiple test correction (p≤0.003) for association with the score on the WAIS Digit Symbol test (β = 0.43, p = 0.003). rs7017612 lies in the intergenic region between CHRNB3 and CHRNA6. This SNP is highly correlated with rs6474413 (r^2^ = 0.75; D' = 0.95), a SNP tagging the genome-wide significant bin for decreased risk for nicotine dependence. These data suggest a modest association between genotype at these SNPs one measure of cognitive function. 10.1371/journal.pone.0134393.t004 ###### Association results with SNPs in CHRNB3-A6 region and Scores on the WAIS Digit Symbol test. ![](pone.0134393.t004){#pone.0134393.t004g} SNP β SE p-value ---------------- ---------- ---------- --------------- rs7017612 0.43 0.19 **0.003 rs6982753 0.39 0.20 0.009 rs10958725 0.25 0.18 0.035 rs13273442 0.28 0.18 0.036 rs4950 0.24 0.18 0.038 rs1530848 0.24 0.18 0.038 rs6474413 0.25 0.18 0.038** rs10107450 0.10 0.19 0.064 rs16891620 0.27 0.25 0.066 rs2196128 0.22 0.20 0.066 rs1530847 0.23 0.20 0.104 rs16891530 0.19 0.42 0.414 rs4952 0.19 0.42 0.414 rs7815274 0.19 0.42 0.426 rs10109429 0.07 0.31 0.494 rs13270610 -0.16 0.35 0.632 rs16891604 0.23 0.42 0.826 Beta, Standard Error (SE), and P-value for all SNPs in the CHRNB3-A6 region of the COGA family GWAS and their association with scores on the WAIS Digit Symbol test. Covariates used were age, sex and FTND score. For all tests, 492 individuals were used for the association test. A second SNP in the CHRNB3-A6 region--rs6982753- had a nominal p-value with the WAIS Digit Symbol phenotype before multiple test correction and almost passed the multiple test correction (p = 0.009). Interestingly, this SNP has an r^2^ of 0.91 with rs892413 (β = 0.39, p = 0.008), a SNP that has previously been associated with increased risk for cocaine dependence \[[@pone.0134393.ref017]\]. Discussion {#sec006} ========== Many studies have demonstrated that risk for nicotine addiction has a genetic component. We performed two tests of selection on chromosomal regions containing the genes encoding five nicotinic receptor subunits and each of these analyses indicate that selection likely occurred at the CHRNB3-A6 locus. Both the Tajima's D test and iHS point to an ongoing sweep in humans on chromosome 8. In the case of the CHRNB3-A6 locus, all of the extreme values in the Tajima's D analysis were positive. High positive Tajima's D values occur when there is an excess of variants in a region with intermediate allele frequencies. This can occur in either balancing selection or ongoing positive selection. We also found extreme iHS values in the CHRNB3-A6 locus. This region fulfills the criteria for a sweep laid out by Voight and colleagues \[[@pone.0134393.ref030]\], i.e. clustering of extreme iHS values. Extreme values of iHS are unlikely under simple demographic models, and thus can indicate the action of an ongoing selective sweep. Despite the fact that none of the populations had significantly different p-values from the negative control in the chromosome 8 region, a few key SNPs did have extreme values. Together, these data imply that the Tajima's D analysis is picking up on ongoing positive selection rather than balancing selection. Several SNPs in the CHRNB3-A6 locus on chromosome 8 have previously been associated with a decreased risk of nicotine dependence \[[@pone.0134393.ref018]\]. One of these, rs1451240, was present in a window that showed extreme values in both the Tajima's D test and iHS. The extreme positive iHS value in the window including rs1451240 indicates that the haplotype containing the ancestral allele is being positively selected. As the derived allele provides protection from nicotine addiction, this suggests that it is the allele that is associated with a greater risk of nicotine dependence that is being selected. Since highly concentrated sources of nicotine were not present in the ancestral environment, it seems likely that this phenotype of nicotine dependence would have hitchhiked along with a more beneficial phenotype. One challenge with this region is that it is approximately 1,500,000 bp away from the centromere of chromosome 8. This could be affecting the results by some unknown mechanism. However, the region including the nicotinic receptors on chromosome 8 was among the top 5% of iHS scores among all regions tested in the genome. Selective pressures in our ancestral environments were likely not on addiction, but rather on behaviors that were biologically rewarding (i.e. mate or food finding, avoidance of harmful stimuli). Given the role of nicotine in neurological function, it is possible that, in the case of nicotine addiction, the phenotype on which natural selection was working was related to enhancements in memory or cognition. The addiction phenotype would have hitchhiked along because it acts through the same or related mechanisms. The addiction phenotype was likely not selected against in ancestral environments because the availability and opportunity for prolonged use of purified drugs was negligible. To test this possibility, we assessed the association of SNPs in the CHRNB3-A6 locus with scores on WAIS tests of memory and cognitive function. Our analysis of the individuals in the COGA dataset suggests that one SNP, rs7017612, which lies in the intergenic region between CHRNB3 and CHRNA6, is associated with increased score on the WAIS Digit Symbol test. This test is thought to largely measure processing speed, but also, to some extent, memory. rs7017612 itself has not been reported to be associated with nicotine dependence. However, it is in moderately high LD (r^2^ = 0.75) with rs6474413, a SNP tagging the genome-wide significant bin for decreased risk for nicotine dependence. Thus, our data are consistent with the possibility that improved performance on this particular cognitive test is modestly associated with a decreased risk for nicotine dependence and that alleles of SNPs in these regions have effects on a subset of cognitive pathways best captured here by the WAIS digit symbol test. It is possible, however, that a function other than addiction or cognition is the true phenotype undergoing natural selection at these loci. Genetic studies of nicotine addiction have identified an inverse relationship between the risk for nicotine addiction and the risk for cocaine addiction. For instance, the minor allele of rs16969968, a missense variant in CHRNA5, increases risk for the development of nicotine dependence, and independently decreases risk for cocaine dependence \[[@pone.0134393.ref015]\]. One hypothesis would be that the true underlying selective pressure is on cocaine related phenotypes or a characteristic that affects cocaine related reward pathways in the brain and that the alleles' effect on nicotine dependence is merely an accidental consequence. However, caution must be used when interpreting this information, given that all drugs of addiction similarly affect the dopaminergic reward pathways. Another alternative hypothesis is that the selective pressure at this locus was on social behavior. Cocaine addiction is characterized by a dampened reward response to social interaction, meaning that it inhibits the positive emotions that accompany social interaction or feelings of belonging. A recent study demonstrated that cocaine users process social gaze (joint attention on an object) differently than controls, resulting in a reduced activation of the reward system during social interactions \[[@pone.0134393.ref031]\]. Using fMRI, these authors showed that cocaine users had decreased activation of the medial orbitofrontal cortex, a region of the brain central for reward processing. If alleles that alter cocaine dependence risk alter an individuals' natural reward system during social interactions, these observations could explain why alleles that protect against cocaine dependence could have provided advantage to carriers in the ancestral environment. Since nicotine sensitizes the animal to the effects of cocaine, which blunts the reward of social interactions, alleles that reduced the ability of nicotine to enhance the effects of cocaine would have undergone positive selection. In this scenario, the nicotine and cocaine dependence phenotypes are not hitchhiking with memory or learning, but rather with phenotypes protecting against antisocial and therefore maladaptive behavior. There was also moderate evidence for selection at the CHRNA5-A3-B4 locus. In particular, rs16969968, the SNP that encodes the missense mutation in α5, that is strongly associated with risk of nicotine dependence, lies in a sliding window exhibiting a high Tajima's D score. The iHS analysis of this locus did not provide evidence for selection. This could indicate that the selective pressure exerted on this locus is older that that seen for the CHRNB3-A6 locus and as such has allowed extended haplotypes to be broken down by recombination. In this scenario, Tajima's D would still be extreme while iHS scores in the region might be less so. This may be particularly true if the selective pressure being exerted on the CHRNB3-A6 locus is ongoing. Here we have used two statistical tests of selection and uncovered evidence of positive selection at the nicotinic receptors on chromosome 8 chromosome 15. Multiple drug-related phenotypes are associated with SNPs in or near these loci, however for several reasons it is unlikely that these phenotypes are the direct targets of this selective pressure. We have proposed two possible explanations 1) phenotypes related to memory and learning and 2) phenotypes related to social behavior. We were only able to discover a modest association with memory-related phenotypes, likely due to the small sample size. We also are as of yet unable to test this second hypothesis because we do not have data in our sample for a phenotype measuring sociality. However this work is the first to explicitly describe signs of natural selection acting on loci underlying substance dependence phenotypes. Materials and Methods {#sec007} ===================== To determine whether the nicotinic receptor loci are under selection, we used Tajima's D, and integrated haplotype score (iHS) to examine the landscape of natural selection at three loci previously demonstrated to harbor genetic variants contributing to the risk of nicotine dependence. These tests have different but complementary strengths. Tajima's D test functions best on recently completed selective sweeps. There are many variables that contribute to how far in the past a sweep can be detected, such as how extreme the sweep was in the first place. Both the mutation rate and the recombination rate affect it as well and vary widely across the genome making generalizations difficult. By contrast, integrated haplotype score iHS functions best for detecting sweeps in progress with alleles at intermediate frequencies, mainly in the range of or after the separation of European, Asian and African populations, during the agricultural phase of human evolution. We utilized 1000 Genomes data for these analyses. Data was obtained from the 1000 genomes website (<http://www.1000genomes.org/data>), a third party source for population whole genome sequence data. The populations were grouped into EUR (GBR, TSI, CEU, FIN), ASN (CHS, CHB, JPT), and AFR (YRI, LWK, ASW). All methods were calculated for the same regions: the CHRNA5-A3-B4 region on chromosome 15q25, the CHRNB3-A6 region on chromosome 8p11, the LCT region as a positive control on chromosome 2q21, and ten intergenic negative control regions where applicable. Tajima's D test {#sec008} --------------- Tajima\'s D is a method of addressing the frequencies of variant sites, based on the expectation that under neutrality, different estimates of expected diversity (θ) should be equal. Tajima's D tests for a skew in the frequency spectrum by comparing two estimates of θ -- the number of segregating sites (S), and pairwise nucleotide diversity (π) \[[@pone.0134393.ref032]\]. Extreme positive values can indicate either balancing selection or population subdivision, and extreme negative values can indicate positive selection or population growth \[[@pone.0134393.ref033]\]. If the same skew is detected across the genome, the effect is likely due to demography, whereas if the skew is localized to a few loci, selection is more likely to be occurring. In a review, Garrigan & Hammer \[[@pone.0134393.ref034]\] have combined published data for Tajima's D values from 65 autosomal loci. They find the mean value for Africans is slightly negative (-0.20) and for non-Africans is slightly positive (0.13). Overall, the values range from approximately -2 to 2. As such, we have taken Tajima's D values above 2 or below -2 to count as extreme values, as this represents the 95% confidence interval of values in our data. Tajima's D was calculated using the program Variscan \[[@pone.0134393.ref035]\]. After an exploratory data analysis of window size, we used a sliding window size of 1000 bp, and window increments of 100 bp for the analysis. Smaller window sizes resulted in too few SNPs in a window to calculate Tajima's D, while larger windows made it much harder to narrow down specific SNPs that may be the ultimate target of selection. Variscan outputs a file giving the Tajima's D value for every window of the specified bp size on the sliding scale \[[@pone.0134393.ref035]\]. These values were then superimposed onto graphs of the regions. Integrated Haplotype Score {#sec009} -------------------------- iHS is a measure of whether a SNP is on an unusually long haplotype carrying the ancestral or derived allele. It compares the rate of haplotype decay between haplotypes carrying either the ancestral or derived allele at a given site, referred to as the core SNP. Haplotypes whose core SNP is under selection will be unusually long compared to those evolving neutrally. Long haplotypes with derived alleles are indicated by negative iHS values and those with ancestral alleles are indicated by positive iHS values. Under neutrality, extreme scores are distributed throughout the genome, however under selection, they are clustered across the selected region \[[@pone.0134393.ref030]\]. iHS is a good method for detecting directional selection, especially in a sweep that is in its early phases. We used the program WHAMM to calculate this statistic \[[@pone.0134393.ref030]\]. The haplotype decay is calculated until the extended haplotype homozygosity (EHH) reaches 0.05. EHH is defined as "the probability that two randomly chosen chromosomes carrying the core haplotype of interest are identical by descent for the entire interval from the core region to point x" \[[@pone.0134393.ref036]\] (p. 833). Long haplotypes with derived alleles are indicated by negative iHS values and those with ancestral alleles are indicated by positive iHS values. Under neutrality, extreme scores are distributed throughout the genome, however under selection, they are clustered across the selected region \[[@pone.0134393.ref030]\]. First, we extracted the desired regions from the 1000 Genomes dataset. We then selected known SNPs within each region, and extracted a region of plus or minus 2000 SNPs around that SNP, except in the case of CHRNB3-A6 where we selected plus or minus 2500 SNPs. We constructed recombination maps using cM maps provided by the SHAPEIT2 program \[[@pone.0134393.ref037]\]. Ancestral alleles were determined using the latest version of Seattleseq (<http://snp.gs.washington.edu/SeattleSeqAnnotation137/>). Phased haplotypes were coded as number of copies of the derived allele. All positions in which the derived allele could not be determined unambiguously (i.e. C/G or A/T SNPs) as well as those without known chimp alleles were removed from further analyses. All analyses were run on each population separately. As iHS is greatly influenced by SNP allele frequency, iHS values from WHAMM were standardized using the average and standard deviation of all SNPs on chromosome 15 and 8 binned by allele frequency such that the average iHS value for each bin after standardization was identical. We excluded SNPs with a minor allele frequency less than 5% because low frequency SNPs are difficult to normalize accurately. After removing these SNPs, extracting just the desired gene regions, and removing those with MAF of \<0.05, there were \~150--350 SNPs per region, depending on the population. Standardization was done separately for each population using population specific averages and standard deviations. iHS values were then superimposed onto graphs of the regions. The haplotype on which a beneficial allele resides tends to be significantly longer than the other haplotypes at the same frequency in the population when adjusted for the recombination background. However, long haplotypes tend to occur in regions with low recombination, and these can be confused with genuine genomic signals of positive selection \[[@pone.0134393.ref038]\]. This is why WHAMM attempts to control for recombination by requiring the input of a cM map. The map we used here was the cM map for imputation available on the website for the program SHAPEIT2. Candidate regions of positive selection were defined as genomic regions containing an uncharacteristic clustering of SNPs with high iHS statistics. This was quantified as the proportion of SNPs with \\|iHS\\| \> 2 in the four regions of interest. Candidate regions of positive selection were identified as containing any SNP with an iHS score of \\|iHS\\| \> 2, as this corresponds to the top \~5% of all scores. The iHS value at a SNP "measures the strength of evidence for selection acting at or near that SNP" however does not provide a formal significance test (Voight et al. 2006). Association Analyses {#sec010} -------------------- COGA recruited subjects in Hartford, Connecticut; Indianapolis, Indiana; Iowa City, Iowa; New York City, New York; San Diego, California; St Louis, Missouri; and Washington, DC. For inclusion in SAGE, cases had to meet lifetime criteria for DSM-IV alcohol dependence, the majority of cases were recruited from alcoholism treatment centers. Control subjects, were both biologically unrelated to cases, and had consumed alcohol but never experienced any significant alcohol or drug-related problems, according to the Semi-Structured Assessment for the Genetics of Alcoholism (SSAGA) \[[@pone.0134393.ref018]\]. The COGA sample utilized in this study consisted of family GWAS data from 2102 European-Americans \[[@pone.0134393.ref014]\]. De-identified data from the Collaborative Studies on the Genetics of Alcoholism (COGA) were used. All participants in COGA provided written informed consent for genetic studies and agreed to share their DNA and phenotypic information for research purposes. The Washington University Human Research Protection Office granted approval for data to be used for this study. COGA administered a variety of neuropsychological tests to its subjects including the three used here: Wechsler Adult Intelligence Scale-Revised (WAIS-R) Block Design, WAIS Digit Symbol, and WAIS Information. In total there were 1247 European-Americans with these neuropsychological phenotypes. However, the overlap between this number and those with family GWAS data was 492. Therefore, our analyses were performed on 492 subjects. In WAIS Block Design, the subject replicates models or pictures of two-color designs with blocks. In WAIS Information, the subject answers a series of questions about factual information. In WAIS Digit Symbol, the subject writes down as quickly as possible the symbols that correspond to a series of numbers. SNPs in the region of the nicotinic receptor clusters on chromosomes 8 and 15 were tested for association with the scaled scores of neuropsychological phenotypes in European-Americans from the COGA study using an additive linear mixed effects (LME) model with the lmepack.batch function as implemented in the GWAF package in R using age, sex and FTND score as covariates \[[@pone.0134393.ref039]\]. Neither alcohol nor cocaine symptom count were significant covariates in the analysis and were thus not included in the analysis. The GWAF package enables association testing with the ability to include individuals from families by correcting for relatedness specified in a pedigree file. Supporting Information {#sec011} ====================== ###### Tajima's D and iHS values for SNPs correlated with rs9298626 within the CHRNB3-A6 region. All SNPs were highly correlated (r^2^\>0.9) with the SNP rs9298626. (PDF) ###### Click here for additional data file. The authors would like to thank Dr. Benjamin Voight at University of Pennsylvania for his edits and advice with regard to the use of WHAMM as well as data interpretation. The Collaborative Study on the Genetics of Alcoholism (COGA), Principal Investigators B. Porjesz, V. Hesselbrock, H. Edenberg, L. Bierut, includes ten different centers: University of Connecticut (V. Hesselbrock); Indiana University (H.J. Edenberg, J. Nurnberger Jr., T. Foroud); University of Iowa (S. Kuperman, J. Kramer); SUNY Downstate (B. Porjesz); Washington University in St. Louis (L. Bierut, A. Goate, J. Rice, K. Bucholz); University of California at San Diego (M. Schuckit); Rutgers University (J. Tischfield); Texas Biomedical Research Institute (L. Almasy), Howard University (R. Taylor) and Virginia Commonwealth University (D. Dick). Other COGA collaborators include: L. Bauer (University of Connecticut); D. Koller, S. O'Connor, L. Wetherill, X. Xuei (Indiana University); Grace Chan (University of Iowa); S. Kang, N. Manz, M. Rangaswamy (SUNY Downstate); J. Rohrbaugh, J-C Wang (Washington University in St. Louis); A. Brooks (Rutgers University); and F. Aliev (Virginia Commonwealth University). A. Parsian and M. Reilly are the NIAAA Staff Collaborators. We continue to be inspired by our memories of Henri Begleiter and Theodore Reich, founding PI and Co-PI of COGA, and also owe a debt of gratitude to other past organizers of COGA, including Ting-Kai Li, currently a consultant with COGA, P. Michael Conneally, Raymond Crowe, and Wendy Reich, for their critical contributions. The authors thank Kim Doheny and Elizabeth Pugh from CIDR and Justin Paschall from the NCBI dbGaP staff for valuable assistance with genotyping and quality control in developing the dataset available at dbGaP (Accession \#: phs000125.v1.p1). [^1]: Competing Interests:Dr. AM Goate is listed as an inventor on the patent "Markers for Addiction" (US 20070258898) covering the use of certain SNPs in determining the diagnosis, prognosis, and treatment of addiction. [^2]: Conceived and designed the experiments: BS GH AG. Performed the experiments: BS GH. Analyzed the data: BS GH. Contributed reagents/materials/analysis tools: HE JT AB JK MS JN. Wrote the paper: BS.	{ "pile_set_name": "PubMed Central" }
INTRODUCTION ============ Human beings have become more sensitive to the welfare of farm animals. For each individual animal, the quality of life contributes to productivity, with better welfare promoting increased efficiency compared to poor welfare. Animal well-being can however be depressed during aversive human-animal interactions \[[@b1-ajas-18-0151]\] that are done in the pens and crushes. These handling facilities are designed for management practices such as vaccinations that enhance animal welfare. However, they are usually associated with the isolation of an individual animal from its cohorts, for 'aversive', stressful and fearful handling \[[@b2-ajas-18-0151]\]. It is therefore maintained that in handling facilities such as the race and crush, cattle lose control and this in itself can be stressful. Hence, race and crush scores are some of the indices that have since been used to determine the extent to which animals can actively resist or be willing to be restrained. Eccentric behaviour, such as stupor, restlessness, kicking, shaking the facility and attempt to escape can be displayed by animals that are reluctant to be handled and restrained \[[@b2-ajas-18-0151]\]. Behavioural responses such as avoidance, and the distance maintained from handlers can be used to determine fear of cattle towards humans. The avoidance distance (AD) test has since been used to determine the distance at which animals withdraw from an approaching human \[[@b3-ajas-18-0151]\]. Further, excretory behaviour such as urination and defaecation can be frequent especially in fearful animals \[[@b4-ajas-18-0151]\]. Repeated exposure of an animal to the same stressor is however associated with reduced subsequent responses \[[@b5-ajas-18-0151]\], such that aggressive animals that regularly go through the same handling procedures may progressively become docile. Animal handling and restraint on the other hand trigger physiological responses such as the activation of the hypothalamic-pituitary-adrenal (HPA) axis exhibited in increased release of glucocorticoids (GCs) such as cortisol. Cortisol is therefore used to quantify stress \[[@b6-ajas-18-0151]\]; hitherto, blood cortisol is often used as a reliable stress biomarker, to determine how animals respond to different degrees of stress \[[@b7-ajas-18-0151]\]. Cook \[[@b8-ajas-18-0151]\] however highlighted that handling may have confounding effects on the measurements, which can be caused by restraint and venepuncture. In addition, Rushen \[[@b9-ajas-18-0151]\] underlined that data related to the HPA is inconsistent. To mitigate erratic measurements, less-invasive procedures which make use of saliva and faecal samples are being used. For more than a decade, the quantification of faecal GCs has become popular as a non-invasive tool in the study of adrenocortical activity during stressful situations. Faecal glucocorticoid metabolites (FGM) are used for the measurement of anthropogenic disturbances on animals \[[@b10-ajas-18-0151]\] and in cattle, the use of dung to reflect faecal GCs is validated \[[@b11-ajas-18-0151]\]. To our knowledge, the relationship between FGM and stress-related behaviour is illustrated in some zoo and wild animals such as the marmoset \[[@b12-ajas-18-0151]\], as well as the domestic horse \[[@b13-ajas-18-0151]\], but is not yet determined in cattle. Observation of animal behaviour in relation to handling stress is less-invasive. Nevertheless, accurate assessment of how cattle react to handling and restraint must employ both behavioural and physiological parameters \[[@b14-ajas-18-0151]\]. The juxtaposition of behaviour and physiological parameters can be used to assess animal welfare. For instance, rectal temperature, a more reliable parameter for measuring core body temperature, increased in cattle that exhibited adverse behavioural reactions during restraint in the chute \[[@b15-ajas-18-0151]\]. Understanding of behaviour and physiological parameters can help reduce animal stress and ultimately increase animal production efficiency. High GC values can be expected in animals that lack experience in handling stress. However, as the number of handling encounters increase, the magnitude of GC levels can be anticipated to subsequently decline \[[@b5-ajas-18-0151]\]. Salivary cortisol reflects the biologically active fraction of the total plasma steroid hormone with a positive relationship between the two media \[[@b16-ajas-18-0151]\]. Moreover, the association between the media is linear over a wide range of concentrations \[[@b8-ajas-18-0151],[@b16-ajas-18-0151]\]. Saliva samples can be collected both prior to and after an imposed stress even at fixed time intervals \[[@b17-ajas-18-0151]\], to indicate the activity of the HPA axis in response to different stressors. The correlation of plasma and saliva cortisol has been established in some domesticated ruminant animals such as goats \[[@b18-ajas-18-0151],[@b19-ajas-18-0151]\] and sheep \[[@b20-ajas-18-0151],[@b21-ajas-18-0151]\]. Studies that validate the use of saliva cortisol as an alternative to blood cortisol in cattle are however still limited. Hitherto, a hormonal challenge was done based on sampling intervals of at least 15 min \[[@b17-ajas-18-0151],[@b22-ajas-18-0151]\]. The study by Hernandez et al \[[@b23-ajas-18-0151]\] was characterised by short sampling intervals of 10 min, and, the hormonal challenge was not done. Notwithstanding, plasma cortisol levels can change due to non-aversive procedures. Thus, cattle handling prior to sampling can affect the characterisation of the relationship between plasma and saliva cortisol. Our objective was to determine the relationship between plasma and salivary cortisol levels in ACTH induced beef cows that were handled prior to sampling. Acclimation to handling was done prior to the characterisation of the relationship. MATERIALS AND METHODS ===================== Ethical approval of animal use and location ------------------------------------------- All procedures involving animals were reviewed and approved by the University of Fort Hare Animal Research and Ethics Committee (Reference Number: MUC291SDZV01). The study was conducted at a research farm under the coordinates 32°48′S (Latitude) and 26°53′E in Alice, South Africa during mid-spring (mean temperature 23°C±4.1°C). Description and management of animals ------------------------------------- Twenty-one clinically healthy non-pregnant Nguni cows aged between 48 and 144 months and weighing between 426 and 436 kg were selected at random, from the different camps at the farm. They were of different age groups, group 1 with n = 7 (5 to 7 yr), group 2 with n = 6 (8 to 10 yr), and group 3 with n = 8 (11 to 13 yr) and these had average weights of 436 kg, 429 kg and 426 kg for each respective age group. The cows were identified by ear-tagging and markings on the flanks according to the respective farm's management system of identification. The identification system made it feasible to use the same cows throughout the trial. The cows were handled separately, however, they were all exposed to handling through routine paddock rotations, health and welfare related check-ups. All the cows had access to grazing on natural pastures and water ad libitum. Measurements and data collection -------------------------------- ### Experiment 1 Data was collected from each individual animal unsystematically between 9:00 and 11:00 am from July to September. There was a seven-day interval per observation and sampling per animal. Maximum and minimum temperatures during sampling days are indicated in [Figure 1](#f1-ajas-18-0151){ref-type="fig"}. Different animal groups were separately brought into the holding pens 14 h prior to observations and samplings. The AD test was done on each randomly selected individual animal in a handling pen, based on modifications to the studies by Dodzi and Muchenje \[[@b3-ajas-18-0151]\]. The handler who wore the same clothing (green work-suit) on every occasion, approached an individual stationary cow at a rate of one step every 5 s from a distance of 3 m, whilst avoiding direct stares to the animal. The test ended when the cow showed an avoidance reaction (retreat). Scores 1 to 3 as indicated in [Table 1](#t1-ajas-18-0151){ref-type="table"} were allotted accordingly. The lower end indicated that the animal was less willing to be approached by the handler than the one which was assigned a Score 3. For cows which could not be immobile enough to be approached for the AD test, an AD Score 1 was assigned. Each individual animal was then restrained in the chute for 30 s and its behaviour observed and scored based on scores modified from Goldhawk et al \[[@b24-ajas-18-0151]\]. Thereafter, for 15 s the cow was restrained in the crush pen with its head fixed with a head gate. Immediately after fixation, behaviour was observed and evaluated with the scoring system in [Table 1](#t1-ajas-18-0151){ref-type="table"}. Rectal temperature readings were then collected on the assumption that, human contact affects cattle behaviour and this may stimulate different biological variables including the former \[[@b15-ajas-18-0151]\]. A digital thermometer (GLA M500, GLA Agricultural Electronics, San Luis Obispo, CA, USA) was used and readings recorded as done by Sánchez-Rodríguez et al \[[@b25-ajas-18-0151]\]. Ambient temperature readings were also taken. Fresh faecal samples were then extracted directly from the rectum of each cow, using a gloved hand and placed in sterile vials. The samples were placed on ice and thereafter frozen at −20°C until analysis. During handling, vocalisation, excretion of watery faeces and urine were observed. Only occurrences of at least three times during the sampling times were considered. The parameters were recorded and assigned scores as either present (1) or absent (0), thereby giving vocalisation, defecating and urinating scores accordingly. From when avoidance behaviour was observed to when the animals were released from the crush pen, total handling time for each animal was on average 120±10 seconds. #### Faecal sample analysis Faecal samples were defrosted at room temperature (20°C to 23°C) over an average period of 4 h. A well-mixed (stirred) wet faecal sample (1 g) was dispersed in 5 mL of 80% methanol and vortexed for 16 h. Of the dispersion, 50 μmL was pipetted into Eppendorf tubes, centrifuged at 1,000×g for 10 min at 4°C and mixed into a methanol 4:1 distilled water solution. The FGM concentrations were assayed using a commercial double-antibody 125I-corticosterone radioimmunoassay kit for animal testing (MP Biomedicals, 07120103, Carlsbad, CA, USA; Lot No. RCBK163) at half volume \[[@b26-ajas-18-0151]\]. The supernatant was aspirated and the precipitate counted on a PerkinElmer Wizard2 Gamma Counter. The intra-assay coefficient of variation was 7.1%. ### Experiment 2 i. Animals: Six clinically healthy non-pregnant Nguni cows were randomly selected from animals used for experiment 1. The cows had a mean weight of 407 kg (range: 336 to 506 kg) and each cow was used as an experimental unit. Samples were collected between 08:30 and 11:30 h at ambient temperatures of not more than 30°C ([Figure 1](#f1-ajas-18-0151){ref-type="fig"}). ii. Adrenocorticotropic hormone administration: A standardised dose (1 μg/kg), of ACTH (Synacthen Depot, tetracosactide 1 mg/mL, Lot S1358, Novartis, South Africa) was administered once to each of the six cows at time 0 \[[@b27-ajas-18-0151]\]. iii. Extraction of blood samples: Blood samples (approximately 6 mL) were collected whilst the animals were restrained in a crush pen with the head in a head gate through jugular venepuncture into tubes containing SST gel. A nose grip was also used for stable restraint. Baseline samples were collected 10 min prior to ACTH administration. Thereafter, samples were drawn every 10 min for 1 h \[[@b28-ajas-18-0151]\], to obtain a total of eight samples per animal. Samples were placed on wet ice (4°C) and then centrifuged for 10 min at 3,550×g (22°C). The serum was then transferred to red-topped tubes (4 mL) with a Clot Activator and stored at −20°C until analysis for cortisol. iv. Extraction of saliva samples: Saliva samples were collected immediately after blood sample collection using cotton based swabs (Salivette cortisol; Sarstedt, Nümbrecht-Rommelsdorf, Germany), which provided a method for easy and safe collection. Cotton balls were also used to collect drooling saliva to maximise the amount of saliva collected. The cotton swabs were inserted at an angle of the lips into the mouths of the cows with the help of a nose grip until well soaked \[[@b29-ajas-18-0151]\]. Each swab and the corresponding cotton ball were then placed in the salivette tube which was placed on ice (4°C). The samples were centrifuged at 1,000×g for 2 min at 20°C. The swabs and cotton balls were removed together with the inner tube of the salivettes. The saliva which collected into the outer tubes was immediately stored at −20°C until analysis. Salivary cortisol was determined by competitive enzyme-linked immunosorbent assay (ELISA). v. Cortisol analysis in plasma and saliva: Blood plasma samples were defrosted and vortexed (Vortex Genie-2, Scientific Industries Inc, New York, USA) at room temperature (24°C). Quantification of plasma cortisol was then done by ELISA using a Cortisol ELISA kit (IBL International, GmBH, Hamburg, Germany, RE52611) as described by Olbrich and Dittmar \[[@b30-ajas-18-0151]\]. Saliva samples were defrosted and centrifuged at 7°C, 1,200 ×g for 10 mins and cortisol levels were measured as explained for plasma cortisol. The intra-assay coefficient of variation (CV) ranged from 3.1% to 6.1% for saliva, and 5.9% to 9.9% for plasma. The inter-assay CV ranged from 4.2% to 17.0% for saliva and 13.0% to 20.0% for plasma. The detection limits of the saliva and plasma assays were 0.083 nmol/L and 0.14 nmol/L, respectively. For both saliva and blood plasma, optical density measurements were done at 450 nm using a photometer Multiskan Ascent with a Genesis Lite microplate computer programme (Labsystem, Finland). Statistical analysis -------------------- ### Experiment 1 Data on subjective scores were square root transformed. Thereafter, all data on FGM concentrations, rectal temperature; avoidance, chute, crush, urinating, defecating and vocalising behaviour scores, were analysed using a repeated measure PROC general linear model procedure of Statistical Analysis System (SAS Inst. Inc., Cary, NC, USA). A repeated measures analysis of covariance with weight as a covariate, was run to determine the differences within variables. Week of sampling and age were fixed factors and animal was a random variable. Significant differences among means were tested by use of the Fisher's least significant differences method at p\<0.05. The model was differences were considered statistically different at p\<0.05. The model used was as follows: Y~ijk~ = μ+α~i~+β~j~+γ(W\) +ɛ~ijk~; where; Y~ijk~ is the response variable (physiological and subjective parameters); μ is the mean; α~i~ is the effect of sampling week (i = 1, 2, 3, 4, 5); β~j~ is the effect of group (1, 2, 3); γ(W\) is the adjusted covariate mean and ɛ~ijk~ is the standard error. Pearson's correlation coefficients were also determined using SAS. ### Experiment 2 Data were analysed by means of a repeated measures general linear model of SAS. The individual cow was a random variable and time was a repeated factor. Data for plasma cortisol was adjusted for normality using the log-transformation. Significance was set at p\<0.05 and values are given as means±standard error of the mean (SEM). To assess the relationship between plasma and saliva cortisol the CORR procedure of SAS was used. RESULTS AND DISCUSSION ====================== Experiment 1 ------------ The FGM were assayed and rectal temperatures were measured in addition to observing stress-related behaviour. The FGM and rectal temperatures were different throughout the sampling weeks ([Table 2](#t2-ajas-18-0151){ref-type="table"}). Repeated handling, however affected the physiological parameters, avoidance and crush behaviour as well as excretory behaviour. The FGM can be estimated to quantify the level of stress an animal is subjected to. We anticipated that with each handling encounter, FGM values would decrease. From our findings, the values were erratic, however, they are generally comparable to those reported by Xavier et al \[[@b31-ajas-18-0151]\]. When there is an overrepresentation of faeces of some individuals, FGM values can be inconsistent \[[@b32-ajas-18-0151]\]. Our study was however characterised by individual identification of each faecal sample, making the source of variability relatively indeterminable. In the same way, we recorded variable rectal temperature values, though they were analogous to those reported by Gruber et al \[[@b15-ajas-18-0151]\] (range: 38.3°C to 40.8°C). Restlessness as a result of movements associated with the experimental procedures could have triggered irregular blood flow thereby stimulating erratic rectal temperatures. In addition, variable responses of the different individuals could have contributed to erraticism. Furthermore, Grandin \[[@b14-ajas-18-0151]\] highlighted that such inconsistencies in the physiological parameters can be attributed to fear, a psychological stressor. Fear can be regarded as a reaction to imaginary danger \[[@b33-ajas-18-0151]\], whereby an animal is fearful of that which is perhaps absent or non-existent. As cattle are gregarious animals, they are vulnerable to exhibit fear of novelty. Accordingly, the general isolation of each individual animal from its respective group for the avoidance test and restraint in the crush could have elicited such kind of fear, especially in the first weeks \[[@b4-ajas-18-0151]\]. Fear of novelty or the human-animal interaction could have simultaneously triggered excretory behaviour in the first weeks \[[@b34-ajas-18-0151]\]. Perhaps, this could be the reason why the Pearson's correlation test, indicated a significant positive relationship between AD test and urination, though it was moderately weak ([Table 3](#t3-ajas-18-0151){ref-type="table"}). However, we could not establish why excretory behaviour increased as animal behaviour became more docile while the animals were restrained in the crush. On the other hand, results imply that animals that avoid human approach at longer distances (\<3 m away but ≥2 m away) have a tendency of being aggressive in the chute and the crush. Results for the avoidance test and restraint in the crush showed some degree of animal adaptation to human approach and handling ([Table 2](#t2-ajas-18-0151){ref-type="table"}). Animals can habituate if they can control or predict the stressor. The animals might have become familiar with the handler and the handling procedure, thereby becoming more at ease with being approached as indicated by a decrease in the ADs. In addition, cows could have learned that after restraint in the crush, they were to be released, thus elucidating calmer behaviour during the last weeks of handling. We anticipated that group 1 cows would be more excitable during handling than their cohorts, assuming that they have the least human-animal and restraint experience \[[@b5-ajas-18-0151]\]. Such behaviour was only observed whilst the cows were handled in the chute. Their behaviour was however not maintained through the other procedures and therefore we assume that cows can for one reason or the other, react variably even when subjected to uniform environmental conditions. On the other hand, group 2 cows excreted urine less than (p\<0.05) the other cows ([Table 2](#t2-ajas-18-0151){ref-type="table"}). This could imply that cows in this group were generally less stressed than their cohorts. However, resolute conclusions could be made if the other parameters showed a similar trend. Experiment 2 ------------ [Figure 2](#f2-ajas-18-0151){ref-type="fig"} illustrates results for plasma and saliva cortisol. Data are presented as means±SEM. Both plasma and saliva cortisol concentrations were expected to escalate after the hormonal challenge and decrease thereafter to baseline values. Basal plasma cortisol concentrations did not significantly increase (p\>0.05) after ACTH administration ([Figure 2](#f2-ajas-18-0151){ref-type="fig"}). Endogenous secretion of ACTH could have occurred at its peak due to the handling involved in the study. In contrast, saliva cortisol concentrations significantly increased (p\<0.05) 30 min after the challenge, suggesting that there is an association between the HPA axis and saliva cortisol. Similar findings for saliva were reported for dairy cows \[[@b17-ajas-18-0151],[@b22-ajas-18-0151]\]. These early studies however observed lower baseline values for both plasma and saliva cortisol than those reported in this study. Schwinn et al \[[@b22-ajas-18-0151]\] assayed saliva cortisol concentrations of about 2.79 to 19.45 nmol/L (converted) and 6.32 to 60.31 nmol/L for plasma cortisol whereas Negrão et al \[[@b17-ajas-18-0151]\] reported baseline values of approximately 5.52 to 8.28 nmol/L for plasma cortisol. It was however stated by Dunn \[[@b34-ajas-18-0151]\] that extreme stress would cause plasma cortisol values of 256.59 nmol/L (93 ng/mL) in cattle, suggesting that values obtained in our study (552.12±106.617 nmol/L, plasma cortisol and 8.75±15.869 nmol/L, saliva cortisol) were generally high. Activation of the HPA axis begins during restraint and handling \[[@b7-ajas-18-0151]\]. From the time when the animals were restrained in the pen to when they were in the crush pen, both plasma and saliva cortisol levels could have risen. Restraint itself stimulates a dramatic increase in cortisol \[[@b5-ajas-18-0151]\], especially with respect to plasma, and this could have occurred in our study. The cows exhibited high avoidance behaviour as they at many occasions retreated from entering the crush pen; therefore, fear a psychological stressor might have contributed to elevated baseline values. High cortisol values are expected in animals that do not willingly enter a crush \[[@b5-ajas-18-0151]\]. These could be reasons why basal plasma cortisol values were already high before hormonal stimulation. Repeated handling thereafter might have then led to unrestrained stress levels. It is also important to note that adaptation to stress depends on the intensity and type of the handling experience \[[@b3-ajas-18-0151]\], hence, repeated venepuncture could have contributed to confounding effects on cortisol levels \[[@b8-ajas-18-0151]\]. Total plasma cortisol increase is an indicator that an animal is exposed to stress. However, the free plasma cortisol fraction accounts for the exponential rise, thus defining the biological activity of plasma cortisol. Saliva sampling is done to determine cortisol levels as a reflection of the free plasma state \[[@b16-ajas-18-0151]\]. In saliva cortisol, only the biologically active fraction is measured \[[@b7-ajas-18-0151]\] and the steroid concentrations only reflect those in the free fraction of plasma. Results for this study, suggest that plasma cortisol could be a reflection of the total plasma cortisol fraction. This could have been coupled with its volatility making baseline values difficult to determine. In addition, response to stress is governed by the interplay of aggressive previous handling experiences and genetic factors such as temperament. Thus, cattle that are naturally excitable, have difficulty in adapting to handling procedures. These excitable cattle are associated with increased physiological responses \[[@b14-ajas-18-0151]\]. This tallies with observations that cows used in this study displayed aggression such as shaking the chute during sampling. Salivary cortisol is associated with a pronounced activation of the HPA axis and is correlated to plasma cortisol \[[@b22-ajas-18-0151]\]. Our study, to a large extent, was consistent with the assertion indicated by the pattern for saliva cortisol. However, due to varying patterns in saliva and plasma cortisol upon ACTH challenge, we observed a weak insignificant correlation (r = 0.0131, p = 0.9310) between saliva and plasma cortisol concentrations. Thus, our results for the correlation were inconsistent to findings by other researchers \[[@b17-ajas-18-0151],[@b22-ajas-18-0151],[@b23-ajas-18-0151]\]. Weak correlations between the two media were however reported in dairy cattle subjected to feeding and drinking actions \[[@b22-ajas-18-0151]\]. In the same way, the delay to reach a steady state between plasma and saliva cortisol as suggested by Hernandez et al \[[@b23-ajas-18-0151]\] was not observed in our study, consistent with the reports by Schwinn et al \[[@b22-ajas-18-0151]\]. While the fact that the breed used is an aggressive breed, the predictor variables of the relationship between plasma and saliva cortisol could be a combination of independent and mediating factors such as causal activity. However, how animals respond and cope with handling or restraint, cannot be easily understood due to the complex physiological and biochemical responses involved \[[@b15-ajas-18-0151]\], something to be considered in our study. In conclusion, no significant relationship was observed between serum and salivary cortisol levels in beef cows that were handled prior to sampling. Cows that are fearful of being approached by handlers, can exhibit excitable behaviour in the crush, thereby influencing physiological response. This study can therefore be regarded as a preliminary for more studies in beef cows. CONFLICT OF INTEREST We certify that there is no conflict of interest with any financial organization regarding the material discussed in the manuscript. We thank the Nguni cattle Project (University of Fort Hare) for funding this project. The authors would like to thank Dr Neville Pitts at the Witwatersrand University for his assistance with ELISA and RIA analyses for this study. The University of Fort Hare Research farm staff and manager, Mr D. Potgieter as well as Mr D. Pepe are highly acknowledged for their support. Authors thank the National Research Foundation (NRF) of South Africa for providing a scholarship for the primary author. ![Line graphs showing the average maximum ( ![](ajas-18-0151i1.jpg)) and minimum ( ![](ajas-18-0151i2.jpg)) environmental temperatures during the five sampling days for Experiment 1. The temperatures increased then decreased as indicated by the patterns of graphs and error bars represent standard error of means.](ajas-18-0151f1){#f1-ajas-18-0151} ![Cortisol concentrations of cows subjected to a hormonal challenge using a standardised dose (1 μg/kg) of adrenocorticotropic hormone at time 0. Line graphs represent plasma cortisol ( ![](ajas-18-0151i1.jpg) PlCortisol) and salivary cortisol ( ![](ajas-18-0151i2.jpg) SaCortisol) concentrations. Line graphs represent cortisol concentrations which either increased or decreased as depicted by the down-up or up-down lines respectively and the error bars represent the standard error of means. Insignificant differences were observed for plasma concentrations in contrast to the saliva concentrations which were variable (p\<0.001).](ajas-18-0151f2){#f2-ajas-18-0151} ###### Avoidance distance, chute and crush scores used for this study Category Score --------------------------------------------------------------------------------------- ----------------- Responses to an approaching experimenter Avoidance score Animal avoided the approaching person at a distance \<3 m away but ≥2 m away 1 Animal avoided the approaching person at a distance \<2 m but ≥1 m away 2 Animal avoided the approaching person at a distance \<1 m but ≥0 m away 3 Behaviour displayed in the chute Chute score No resistance to movement 1 Moves entire body to and fro rhythmically at least twice 2 Uneasy, head, body, tail and feet movements, hides head under another cow's abdomen 3 Active escape behaviour 4 Rears and needs encouragement to move forward 5 Stupor/refusing to move requiring force to probe forward 6 Behaviour exhibited in the crush (modified from Geburt et al \[[@b28-ajas-18-0151]\]) Crush score Cow is calm with no movement 1 Cow is slightly excited with minimal movement 2 Cow is nervous and occasionally shakes the crush 3 Cow is agitated and continuously moves and shakes the crush 4 Cow is very agitated and exhibits escape behaviour, animals may kneel/fall 5 ###### Effect of sampling week and age category on the physiological and behavioural parameters[1)](#tfn2-ajas-18-0151){ref-type="table-fn"} Variable Sampling and observation week SEM Group p-value ---------- ---------------------------------------------------- --------------------------------------------------- ---------------------------------------------------- ---------------------------------------------------- ---------------------------------------------------- ------- -------------------------------------------------------- -------------------------------------------------------- -------------------------------------------------------- ---------- -------- -------- FGM 23.84[bc](#tfn3-ajas-18-0151){ref-type="table-fn"} 19.08[c](#tfn3-ajas-18-0151){ref-type="table-fn"} 30.57[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 26.40[ab](#tfn3-ajas-18-0151){ref-type="table-fn"} 23.44[bc](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.776 23.84±1.249 25.10±1.444 25.06±1.464 0.0008 0.5978 0.6350 Rect T 37.88[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 38.76[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 38.30[ab](#tfn3-ajas-18-0151){ref-type="table-fn"} 37.78[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 38.12[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 0.207 38.09±0.146 38.19±0.168 38.22±0.171 0.0052 0.8071 0.2400 ADS 1.20[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.21[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.38[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.47[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.51[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 0.053 1.28±0.038 1.35±0.043 1.43±0.044 \<0.0001 0.0841 0.1250 Chute 2.00 1.88 1.84 1.75 1.77 0.073 1.99[a](#tfn3-ajas-18-0151){ref-type="table-fn"}±0.052 1.78[b](#tfn3-ajas-18-0151){ref-type="table-fn"}±0.060 1.78[b](#tfn3-ajas-18-0151){ref-type="table-fn"}±0.060 0.1540 0.0084 0.3539 Crush 1.85[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.87[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.63[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.56[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.47[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 0.067 1.67±0.047 1.58±0.054 1.77±0.055 \<0.0001 0.0514 0.1829 Urin 1.02[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.13[ab](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.10[ab](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.21[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.21[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 0.040 1.11[b](#tfn3-ajas-18-0151){ref-type="table-fn"}±0.028 1.22[a](#tfn3-ajas-18-0151){ref-type="table-fn"}±0.033 1.08[b](#tfn3-ajas-18-0151){ref-type="table-fn"}±0.033 0.0065 0.0045 0.8710 Def 1.14[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.13[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.17[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.32[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.24[ab](#tfn3-ajas-18-0151){ref-type="table-fn"} 0.044 1.19±0.031 1.23±0.036 1.18±0.036 0.0085 0.4836 0.1801 Vocal 1.09 1.10 1.17 1.15 1.17 0.045 1.10±0.031 1.21±0.036 1.10±0.036 0.5806 0.0531 0.9798 SEM, standard error of means; W×G, interaction of sampling and observation week and group; FGM, faecal glucocorticoid metabolites concentrations (ng/g); Rect T, rectal temperature (°C); ADS, avoidance score; Chute, chute score; Crush, crush score, Urin, urinating score; Def, defecating score; Vocal, vocalisation score. Data are least squares (LS) means. Means within a row without a common superscript significantly differ (p\<0.05). ###### Pearson correlation coefficients for parameters Variable FGM RT Avoidance Chute Crush Urinating Defecating ------------ -------- -------- ------------------------------------------------------- ---------------------------------------------------- ------------------------------------------------------- ----------- ------------ RT 0.102 \- \- \- \- \- \- Avoidance −0.003 −0.024 \- \- \- \- \- Chute −0.011 0.135 −0.372[\\](#tfn6-ajas-18-0151){ref-type="table-fn"} \- \- \- \- Crush −0.125 −0.017 −0.351[\\](#tfn6-ajas-18-0151){ref-type="table-fn"} 0.241[\](#tfn5-ajas-18-0151){ref-type="table-fn"} \- \- \- Urinating −0.022 −0.004 0.280[\](#tfn5-ajas-18-0151){ref-type="table-fn"} −0.132 −0.335[\\](#tfn6-ajas-18-0151){ref-type="table-fn"} \- \- Defecating 0.099 −0.035 0.188 −0.117 −0.319[\*](#tfn5-ajas-18-0151){ref-type="table-fn"} 0.156 \- Vocalising −0.011 −0.077 0.067 −0.198 −0.077 0.127 0.027 FGM, faecal glucocorticoid metabolites concentrations; RT, rectal temperature; Avoidance, avoidance distance test score; Chute, chute score; Crush, crush score; Urinating, urinating score; Defecating, defecating score; Vocalising, vocalisation score. p\<0.05, p\<0.01.	{ "pile_set_name": "PubMed Central" }
Introduction {#Sec1} ============ Governments and health organizations worldwide have published dietary guidance for sugar intake \[[@CR1]\]. Despite access to the same published literature, recommendations vary greatly and create confusion for health practitioners and consumers. Since 1980, Dietary Guidelines for Americans (DGA) has recommended we "avoid too much sugar", yet dietary advice has typically recommended foods high in sugar, such as fruits and dairy products. As a way to clarify the types of sugar to avoid, the terms added sugars and free sugars are used. Added sugar recommendations have been in existence since 2002, with recent recommendations becoming progressively more restrictive over the years \[[@CR1]\]. This paper addresses current and proposed added sugar recommendations and assesses their practicality within the United States. Definition of added sugars {#Sec2} ========================== No universally accepted definition for added sugars exist (Table [1](#Tab1){ref-type="table"}). The Food and Drug Administration (FDA) classifies added sugars as, mono and disaccharides added to foods during production including sugars, syrups, fruit juice concentrates, honey, etc. This would not include sugars that naturally exist in foods, such as sugars in fruits or dairy products \[[@CR2]\]. A common point of contention between institutions is whether or not fruit juice should be included as added sugars. The proposed revisions to the Nutrition Facts and Supplements Label published by the FDA in 2014 classifies fruit juice concentrate added to food products as added sugar, while juice not from concentrate as not added sugar. In comparison, the USDA recommendations do not specify that fruit juice from concentrate contributes to added sugar totals \[[@CR2], [@CR3]\].Table 1Added sugar definitions and distinctions from various agenciesFood and Drug Agency (FDA) Proposed Nutrition Facts Label \[[@CR3]\]Mono- and disaccharides added to foods during production including: sugars, syrups, fruit juice concentrates, honey, etc.United States Department of Agriculture (USDA) Choose MyPlate \[[@CR2]\]Sugars added during processing and preparation. Includes: sugars, syrups, honey, nectars, etc.Excludes: fruit juice and fruit juice concentratesWorld Health Organization (WHO) Free Sugar Guidelines \[[@CR4]\]Mono- and disaccharides added to foods during processing or by the consumer during preparation. Includes: Sugars, syrups, honey, fruit juice and fruit juice from concentrate.Scientific Report of the 2015 Dietary Guidelines Advisory Committee \[[@CR8]\]Sugars, syrups, isolated naturally occurring sugars (ex: fruit juice concentrate) and other caloric sweeteners In addition to various definitions for the term "added sugars", the World Health Organization (WHO) utilizes the term "free sugar". Free sugar is similar to added sugars, as it includes all sugars and syrups added to foods; however, free sugar also includes sugars naturally present in fruit juices and fruit juice concentrates \[[@CR4]\]. Free sugar includes sugars naturally found in fruit juice that is consumed as a beverage as well as fruit juices added to food products. Assessing added sugar intake and compliance with recommendations would be extremely difficult without a clear and established definition of the term "added sugar" and, specifically, how fruit juice should contribute to added sugar values. Function of added sugar {#Sec3} ======================= Added sugars are chemically identical to sugar that naturally occurs in food products \[[@CR1]\]. The body cannot distinguish the source of the nutrient and processes the sugar in the same way. Sugar may be added to food products for many reasons, the most obvious reason being adding sweetness and enhancing the palatability of foods. Although this function of sugar is often opposed and criticized, many American consumers would not find a number of "healthy" foods palatable without added sugar. Some examples include cranberries, yogurt and oatmeal. Nutrition professionals often encourage clients to consume these foods as part of a healthy diet, even with some added sugar \[[@CR5]\]. Another function of sugar within food products is texture enhancement. Sugar produces a tender texture in baked products, and inhibits ice crystallization in frozen products. Sugar provides body to products and, when removed, has to be substituted with bulking agents to achieve a similar mouth feel \[[@CR6]\]. Carmelization and maillard browning are both reactions specific to sugar and provide an appearance expected in food products. Sugar also plays a role in food safety by inhibiting the growth of microorganisms at high concentrations. By binding with water molecules, sugar can maintain moisture contents in products lengthening the shelf life \[[@CR6]\]. Overall, it is important to remember that sugar functions in many capacities beyond just flavor. Added sugar intake in the American diet {#Sec4} ======================================= Added sugar intake is on average 13 % of total energy intake in adults and 16 % in children, consistently decreasing with age \[[@CR7]\]. Added sugar consumption has declined in all age groups from NHANES data taken in 2001--2004 to data from 2007--2010. Meanwhile, rates of obesity did not mimic the decline over the same time period \[[@CR8]\]. According to NHANES data from 2009--2010, 47 % added sugars in the American diet come from beverages, 31 % from snacks and sweets, 8 % from grains, and 14 % from the categories of dairy, mixed dishes, condiments, fruits and fruit juice and vegetables combined \[[@CR8]\]. While there is room for improvement in the American diet, this decrease in added sugar intake is encouraging and understanding the main sources of added sugars provides a direction to focus our efforts. Added sugar recommendations in America {#Sec5} ====================================== In 2002, the Institute of Medicine (IOM) Dietary Reference Intakes recommended that less than 25 % of total energy should come from added sugars. The recommendation is based on the concept that foods containing high amounts of added sugars are typically high in calories and low in micronutrients \[[@CR9]\]. The idea that added sugars are "empty calories" is a commonly cited reason that added sugar recommendations are necessary. Diets containing a large amount of energy as "empty calories" can lead to micronutrient malnutrition or over consumption of calories. Consuming the daily recommendation of all nutrients within an individual's estimated energy requirement is challenging when the individual is consuming a large portion of his or her calories as empty calories. Repeated consumption of empty calories without compensation from other nutrients can lead to weight gain. The current 2010 Dietary Guidelines for Americans, includes solid fats and added sugars (SoFAS) in their recommendation of 5--15 % of total energy from solid fats and added sugars \[[@CR10]\]. Minimizing SoFAS consumption is encouraged to reduce excess calorie consumption and to replace foods high in added sugars with foods lower in added sugars and greater nutrient density. SoFAS consumption above the recommendation is considered to be incompatible with the USDA Food Patterns, likely exceeding calorie limits or obtaining inadequate micronutrient intake \[[@CR10]\]. The USDA Food Patterns were created to assist the public in following Dietary Guideline recommendations, providing amounts of food from each food group to achieve optimal nutrient intake \[[@CR11]\]. The USDA Food Patterns groups added sugars and solid fats together and recommends adult females and adult males to limit "empty calorie" intake to 120--250 calories per day and 160--330 calories per day, respectively, depending on caloric needs \[[@CR8]\]. Consumption of empty calories is typically above the current recommendations in all age groups; almost 90 % of Americans exceed the USDA food pattern recommendations \[[@CR8]\]. The evolution of the concept of discretionary calories (2005 DGAs) to empty calories (2010 DGAs) is explained by Nicklas and O'Neil \[[@CR12]\]. The authors also explain that the reduction of solid fats and added sugars is to remove calories from the diet, not because solid fats and added sugars are linked to negative health outcomes \[[@CR12]\]. The World Health Organization not only cites the effects of excess calories, but also the impact that sugar can have on dental health. The current World Health Organization recommendation of fewer than 10 % of total calories from free sugars was set in 2003 \[[@CR13]\]. However in 2015, WHO set a conditional recommendation suggesting that less than 5 % of total energy should come from free sugars \[[@CR4]\]. This conditional recommendation proposed by WHO is based on a positive association between free sugar intake and dental caries among children \[[@CR4]\]. Sugar consumption has been positively associated with risk of dental disease. According to a meta-analysis published in 2014, there is moderate evidence indicating that a free sugar intake less than 10 % of total calories was associated with decreased risk of dental caries \[[@CR14]\]. Further decrease in caries was seen in Japanese surveys, taken between 1959 and 1960, when free sugar intake approached 5 % of total calories \[[@CR15]\]. The area surveyed had low fluoride exposure so it may not be an accurate model to extrapolate to areas with good fluoride exposure in the United States. Although, the WHO states that all populations, regardless of fluoridation, could possibly see improvement in dental caries with decreased free sugar intake \[[@CR4], [@CR15]\]. Additionally, the sugar consumption data was calculated by looking at sugar consumption per capita, added sugar intake compared to incidence of dental carries for each individual was not known \[[@CR15]\]. The limitations of the Japanese studies prevented the WHO from setting a strong recommendation to consume fewer than 5 % of calories from free sugars \[[@CR4]\]. However, because dental caries occur throughout the lifespan, consuming fewer free sugars is estimated to have a cumulative effect and result in decreased dental problems later in life and no evidence of harm was seen in diets containing fewer than 5 % energy from free sugars \[[@CR4]\]. Dietary Guidelines for Americans 2015 {#Sec6} ===================================== The release of the Scientific Report of the 2015 Dietary Guidelines Advisory Committee (DGAC) in February 2015 brought further attention to added sugars. The DGAC report placed a large focus on added sugars, making it one of the five "cross-cutting topics" \[[@CR8]\]. The Committee reexamined the evidence surrounding the potential health effects of added sugars. The DGAC assessed the evidence that added sugar negatively impacts the health risks for obesity, type II diabetes, cardiovascular disease and dental carries. The DGAC determined, based on the available evidence, there was a strong correlation between added sugars and negative health risks. Most of the cited evidence examines the association between sugar sweetened beverage (SSB) consumption and the health risk rather than the consumption of added sugar from variety of foods \[[@CR8]\]. It is easier to count consumption of SSBS with food frequency instruments used in epidemiologic studies than to estimate total added sugar intake since few databases included information on added sugars. While SSB consumption may be the best method available for added sugar estimates, it is not without its limitations including possible confounding variables within the population. According to a recent study of over 12,000 participants, individuals reporting to consume one or more SSB per day were significantly more likely to smoke, consume fewer fruits and vegetables and report a sedentary lifestyle \[[@CR16]\]. No discussion of if these confounding variables were considered in the DGAC report \[[@CR8]\]. After an examination of the evidence and diet modeling, the DGAC suggested an appropriate intake of calories from added sugars to be between 4--6 % and set a maximal intake of 10 % total energy from added sugars \[[@CR8]\]. After a period of time allowing for comments from the general public, the USDA and Department of Health and Human Services will assess evidence behind the recommendation from the USDA to set the Dietary Guidelines for Americans 2015 added sugar recommendation \[[@CR8]\]. With this suggested restriction on added sugars, the DGAC recognizes that the logical consequence of removing added sugars from the diet and food products would be replacing the added sugars with low calorie sweeteners. However, the DGAC report advises against this replacement due to the minimal evidence regarding long-term effect of low calorie sweeteners. Instead, the DGAC encourages the replacement of sugar-sweetened beverages with water and does not suggest a replacement in food products. Removal of sugar from products will change the taste, texture and shelf-life of products due to the functions of sugar previously discussed \[[@CR6]\]. The sugar must be replaced with other ingredients and, if not low calorie sweeteners, what would be a better alternative? Evidence exists to support the use of low-calorie sweeteners in weight reduction \[[@CR8]\] and many consumers utilize this approach to support weight loss. The FDA recognizes artificial and low-calorie sweeteners as safe for consumption, and the Academy of Nutrition and Dietetics advises that non-nutritive sweeteners can fit into a healthy diet \[[@CR17]\]. Identifying alternative sweeteners or ingredients to produce comparable food and beverage products is essential in changing the consumption patterns in Americans. Taste is consistently the most important buying factor for most Americans, and without great tasting alternatives consumers are not likely to make dietary changes \[[@CR18]\]. Proposed addition of "Added sugar" to nutrition facts panel {#Sec7} =========================================================== Currently, there is no easy way for consumers, researchers or health professionals to track added sugar consumption and assess compliance with recommendations. Very few databases exist that calculate added sugars, and, due to the various added sugar definitions, the information obtained from these databases may result in a range of added sugar values. In March 2014, the FDA proposed changes to the Nutrition Facts Panels to assist consumers in making more educated food choices that would lead to a healthy diet consistent with Dietary Guidelines for Americans. The recent proposal to update the Nutrition Facts Panel advocates for the addition of an "Added sugars" category below the "Sugars" category, that would provide a way to track and compare added sugars \[[@CR19]\]. The proposed amendments to the food labels suggest displaying added sugar in grams. The DGAC report supports such changes to the food labels and recommends displaying added sugar values in grams, teaspoons and percent daily value \[[@CR8]\]. A supplemental proposed rule regarding the Nutrition Facts Panel was published in July of 2015. The FDA proposed to establish a less than 10 % Daily Reference Value (DRV) and to include the percent Daily Value (DV) on the Nutrition Facts Panel \[[@CR20]\]. The supplemental proposed rule cites the 2015 DGAC report as their basis for instituting an added sugar DRV. The proposed rule states that the 2015 DGAC showed a "strong association between a dietary pattern of intake characterized, in part, by a reduced intake of added sugars and a reduced risk of cardiovascular disease" \[[@CR20]\]. Traditionally, DRVs and %DVs have been established for nutrients where an average dietary requirement can be determined from available scientific evidence \[[@CR21]\]. The data used to determine the \<10 % DRV for added sugar was based primarily on diet modeling conducted for the 2015 DGAC. No DRV is has been proposed for total sugars at this time due to lack of available evidence for a reference intake \[[@CR8], [@CR20]\]. The purpose of the FDA's changes to the Nutrition Facts Panel is to help consumers make choices leading to healthier diets, however; the addition of the "added sugar" category may not provide much novel knowledge to consumers. According to NHANES 2009--2010, nearly 80 % of added sugars come from sugar-sweetened beverages (47 %) and snacks and sweets (31 %) \[[@CR8]\]. The proposed changes to the food labels would require food companies to invest their resources to calculate the added sugar in their products, when the majority of added sugar consumed comes from obvious sources of sugar. Just over 20 % of added sugars consumed by Americans come from non-obvious forms where the consumer would benefit from the knowledge of added sugars on the food labels, if they choose to read the label \[[@CR8]\]. A study presented by the International Food and Information Council showed that the addition of the category "Added Sugars" to the Nutrition Facts Panel reduced the consumer comprehension of the food label. The percent of participants able to accurately identify the total grams of sugar dropped from 92 to 55 % when the added sugars category was included, with more than half the participants adding added sugars with the sugar category \[[@CR22]\]. A similar study was later conducted by the FDA, finding consistent results. Ability to accurately identify the grams of sugar per serving decreased from 81 % to 65 % when the label was updated to the proposed format \[[@CR23]\]. Other research supports that consumers are interested in added sugar labeling. Kyle & Thomas report that consumers believe Nutrition Facts labeling for added sugar will be more helpful than confusing \[[@CR24]\]. A study in European Union found that consumers expect that a reduction in free sugars in a product will be linked to a reduction in the calorie content of the food \[[@CR25]\]. Nevertheless, a consumer study with cereals found that participants rated cereals containing "fruit sugar" as healthier than cereals containing "sugar", although there were no differences in nutrient content between the cereals \[[@CR26]\]. Total sugar analysis is challenging enough. When sugar content of commercial foods targeted to infants and children was conducted by a blinded laboratory analysis of accepted chemical methods, nutrient label data underestimated or overestimated actual sugar content routinely. The authors suggest that more effort should be made to standardize methods for sugar labeling of foods, especially foods targeted to children \[[@CR27]\]. Health Canada recently removed the added sugars category from their proposed nutrition facts table and included a 20 % DV for total sugar \[[@CR28]\]. Consumer research by the Canadian government found that information about carbohydrates and total sugars was confusing when the table included added sugars. It also found the % DV approach to be useful and easy to understand. They state: "the proposal to declare the amount of added sugars was popular among consumers and health stakeholders (including health professionals). However, industry stakeholders questioned the scientific basis of requiring the declaration of added sugar given that the body metabolizes naturally occurring and added sugars in the same way. Similarly, the inability of analytical method to distinguish between naturally occurring and added sugars would contribute to significant compliance and enforcement challenges." Because added sugars are not chemically different from intrinsic sugars, there is no way to analytically determine the amount of added sugar in a food product \[[@CR1]\]. Food manufacturers would have to calculate the added sugars based on the recipe in order to determine the added sugar content in each product every time the product is reformulated. Without a clear definition of added sugars the resultant labeling will likely be inconsistent. The FDA would require food companies to document, maintain and provide records on product composition to verify the published value of added sugars \[[@CR3]\]. Due to competition within the food industry and the proprietary nature of the formulations, food manufacturers would be very resistant to release such information. Moreover, each of these steps will require additional time, money and an acquired skill set that smaller food companies may not have the resources to comply with. Conclusion {#Sec8} ========== Excess calorie consumption can lead to weight gain and increased risk of obesity and obesity-related co-morbidities \[[@CR8]\]. Empty calories which include solid fats and added sugars play a role in this when consumed in abundance. Added sugars are low in nutrient density and calories from added sugars can add up quickly if the individual is not conscious of their diet. However, there is no evidence suggesting that excess calories from added sugars specifically are worse than excess calories from any other food source. Much of the evidence linking added sugars to chronic disease is done measuring sugar sweetened beverages rather than percent calories from all added sugars \[[@CR8]\]. With nearly half of added sugar consumption in America being attributed to sweetened beverages, perhaps encouraging healthy beverage alternatives to sugar sweetened beverages should be the focus, rather than zeroing in on all added sugars. Regulatory attempts to tax sugar sweetened beverages in countries, such as Mexico and communities in the US, including Berkeley, CA may increase tax revenue, but whether these more aggressive approaches can limit calorie intake and/or improve health outcomes await clinical trial results. Recommendations as low as 5 % total energy from free sugars are likely too restrictive for most Americans to achieve \[[@CR29]\]. Added sugars should be consumed at a minimum as they are often a source for surplus calories in the American diet; however, stringent recommendations and mandatory food labeling are likely not the most effective ways to reduce added sugar and excess calorie consumption. Education on healthy beverages, snack choices and portion sizes may be a better starting point for reducing empty calorie intake. DGA : Dietary Guidelines for Americans DGAC : Dietary Guidelines Advisory Committee DRV : Daily reference value DV : daily value FDA : Food and Drug Administration IOM : Institute of Medicine SoFAS : solid fats and added sugars SSB : sugar sweetened beverage WHO : World Health Organization Competing interests Jennifer Erickson, is a PhD student in Nutrition at the University of Minnesota working with Dr. Joanne Slavin. Joanne Slavin is a professor in the Department of Food Science and Nutrition, University of Minnesota. In the past 5 years, she has given 150 scientific presentations in 13 countries. Many of these meetings received sponsorship from companies and associations with an interest in carbohydrates and nutritive sweeteners. Her research group works mostly on satiety and gut health outcomes with dietary exposures including dietary fiber and protein. Her research funding for the past 5 years has included grants from General Mills, Inc., Tate and Lyle, Nestle Health Sciences, Kellogg Company, USA Rice, USA Pears, Minnesota Beef Council, Minnesota Cultivated Wild Rice Council, Barilla Company, USDA, American Egg Board, American Pulse Association, MNDrive Global Food Ventures, International Life Science Institute (ILSI), and the Mushroom Council. She serves on the scientific advisory board for Tate and Lyle, Kerry Ingredients, Atkins Nutritionals, Midwest Dairy Association and the Alliance for Potato Research and Education (APRE). She holds a 1/3 interest in the Slavin Sisters Farm LLC, a 119 acre farm in Walworth, WI. Authors' contributions JE and JS contributed to the planning, research and writing of this paper. Both authors read and approved the final manuscript.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1-materials-13-02906} =============== Ferromagnetic shape memory alloys (FSMA) are recently one of the most extensively studied group of modern smart materials \[[@B1-materials-13-02906],[@B2-materials-13-02906],[@B3-materials-13-02906]\]. Among them, NiMnGa-based Heusler compounds stand out as the most complex alloys due to their unique magnetomechanical properties, such as magnetic field induced strains \[[@B4-materials-13-02906],[@B5-materials-13-02906],[@B6-materials-13-02906]\], pseudoelasticity/superelasticity \[[@B7-materials-13-02906],[@B8-materials-13-02906],[@B9-materials-13-02906]\], magnetoresistance \[[@B10-materials-13-02906],[@B11-materials-13-02906]\] or magneto- and mechanocaloric effects \[[@B12-materials-13-02906],[@B13-materials-13-02906],[@B14-materials-13-02906],[@B15-materials-13-02906],[@B16-materials-13-02906]\]. All above mentioned properties are associated with the stress-, magnetic field- or temperature-driven first-order reversible phase transition, and undergo a change from a high symmetry austenitic phase to low symmetry martensitic phase \[[@B17-materials-13-02906]\]. Furthermore, a low temperature martensitic phase is characterized by strong magnetocrystalline anisotropy, which helps to induce and control the twin variant reorientation within the martensitic phase \[[@B18-materials-13-02906]\]. The strong correlation between the microstructure and the magnetic properties of the austenitic and martensitic phases leads to the abrupt drop of magnetization in the vicinity of the martensitic transformation. This significant difference in magnetization strongly influences the majority of magnetomechanical properties in NiMnGa-based materials. When considering potential multifunctional applications of FSMA, it is necessary to manipulate the magnetic behavior of the martensite and austenite phases. It is well documented that the magnetism of NiMnGa-based alloys varies significantly with composition in and near stoichiometric Ni~2~MnGa samples depending on the Ni \[[@B19-materials-13-02906],[@B20-materials-13-02906],[@B21-materials-13-02906]\], Mn \[[@B22-materials-13-02906],[@B23-materials-13-02906]\] or Ga \[[@B24-materials-13-02906],[@B25-materials-13-02906]\] concentration. However, it has been recently reported that an addition of the fourth alloying element may cause a significantly change in the magnetostructural properties of the existing phases \[[@B13-materials-13-02906],[@B26-materials-13-02906],[@B27-materials-13-02906]\]. The introduction of doping elements changes the stability of the austenite and martensite phases, which shifts the structural transformation temperatures and changes the temperature dependent magnetic behavior of doped NiMnGa-based materials. Moreover, alloying elements have different magnetic moments than primary Ni, Mn and Ga, which also influence the ferromagnetic properties of the material. Crucially, even a small amount of doping element can also notably shift the temperature of the martensitic transformation \[[@B28-materials-13-02906],[@B29-materials-13-02906],[@B30-materials-13-02906],[@B31-materials-13-02906],[@B32-materials-13-02906],[@B33-materials-13-02906],[@B34-materials-13-02906]\]. The present work aimed to investigate the influence of Ti and Gd substitution for Ga in polycrystalline Ni~50~Mn~25~Ga~20~X~5~ (X = Ti or Gd) FSMAs with a magnetic behavior of the martensite and austenite phases before, during, and after phase transitions. Due to the fact that magnetocrystalline anisotropy changes during structural transformation, we estimated the temperature dependence of anisotropy energy from the series of high field magnetization curves based on the law of approach to magnetic saturation \[[@B35-materials-13-02906]\]. 2. Materials and Methods {#sec2-materials-13-02906} ======================== The polycrystalline bulk samples with the nominal composition of Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ (at.%) were produced from high purity elements via an arc-melting method under a protective argon atmosphere. The melting procedure was repeated several times to ensure a high homogeneity of the samples. Despite the fact that Mn is a highly volatile element, the weight loss after the melting procedure was less than 1% for each sample. The produced ingots were subsequently vacuum-sealed in a quartz ampule and annealed at 1173 K for 5 h, followed by water quenching. Our previous studies \[[@B30-materials-13-02906]\] showed that Gd-doped NiMnGa-based FSMA require a higher annealing temperature, so to ensure a good homogeneity in the Ni~50~Mn~25~Ga~20~Gd~5~ sample, the alloy was further annealed at 1430 K for 3 h and then quenched in water. The nominal compositions of the Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ alloys were confirmed with the help of a Scanning Electron Microscope JEOL JSM-5800LV (JEOL Ltd., Akishima, Tokyo, Japan) supported by an energy-dispersive detector (EDS). The crystal structure identification was done on the basis of powder X-ray diffraction measurements at room temperature performed in the X-ray Diffractometer Rigaku MiniFlex 600 (Rigaku, Tokyo, Japan)) with CuKα radiation, followed by subsequent Rietveld analysis. The magnetic properties characterization was carried out using the Vibrating Sample Magnetometer (VSM) module from the VersaLab System (Quantum Design, San Diego, CA, USA). In order to reduce the influence of the demagnetizing field, the thin, needle-like samples (with a length to diameter ratio of more than 10) were used for the magnetic measurements. The temperature dependence of magnetization under an external magnetic field of 250 mT was measured from 50 K up to 400 K using zero-field cooled (ZFC) and field cooled (FC) protocols with heating and cooling rates of 10 K/min. Then, the series of magnetic hysteresis loops was also recorded in the same temperature range (50--400 K) under an external magnetic field of up to 2 T. 3. Results and Discussion {#sec3-materials-13-02906} ========================= 3.1. Microstructure Analysis {#sec3dot1-materials-13-02906} ---------------------------- [Figure 1](#materials-13-02906-f001){ref-type="fig"} depicts the observed and calculated XRD patterns obtained at room temperature for the investigated Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ alloys. In order to clearly show the slight deviation of the peak locations in each pattern, only three peaks with the highest intensity are presented in [Figure 1](#materials-13-02906-f001){ref-type="fig"}. The presented reflections were assigned to a standard cubic L2~1~-type austenitic structure with an Fm3m space group (No. 225). In this case Ni, Mn and Ga occupied the 8c (0.25, 0.25, 0.25), 4a (0, 0, 0) and 4b (0.5, 0.5, 0.5) Wyckoff atomic positions, respectively. The lattice parameters calculated from the Rietveld analysis are as follows: a = 5.826 Å for Ni~50~Mn~25~Ga~25~, a = 5.848 Å for Ni~50~Mn~25~Ga~20~Ti~5~ and a = 5.832 Å for Ni~50~Mn~25~Ga~20~Gd~5~ alloys. It can be seen that an addition of doping elements (Ti or Gd) leads to the slight expansion of the austenitic unit cell. Moreover, in the case of the Ni~50~Mn~25~Ga~20~Gd~5~ sample, additional peaks were observed in the XRD pattern. These reflections were assigned to a residual hexagonal Ga~1.5~Gd~1~Ni~3.5~ phase (space group P6/mmm, No. 191) precipitated during the annealing process. The lattice parameters calculated for this phase are a = 4.980 Å, c = 4.107 Å and γ = 120°. 3.2. Temperature Dependence of Magnetization {#sec3dot2-materials-13-02906} -------------------------------------------- [Figure 2](#materials-13-02906-f002){ref-type="fig"} shows the temperature dependence of magnetization for the investigated Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ alloys. A notable change of magnetization occurred in the vicinity of the phase transition in every investigated sample, both during heating and cooling procedures. The substantial hysteresis between the temperature of the martensitic transformation (during cooling) and the austenitic transformation (during heating) was very characteristic for the first-order temperature-driven transformation, especially in polycrystalline materials where the transition takes place in every single grain. Moreover, the martensitic transformation temperature was strongly dependent from the Gibbs free energy difference between the austenite and martensite phases and the addition of alloying elements may have significantly changed these energies \[[@B36-materials-13-02906],[@B37-materials-13-02906],[@B38-materials-13-02906]\]. In order to compare the transformation temperatures between the studied alloys, the average value of all temperatures (M~s~, M~f~, A~s~, A~f~) was calculated as:$$T_{M} = \frac{1}{2}\left( {\frac{M_{s} + M_{f}}{2} + \frac{A_{s} + A_{f}}{2}} \right),$$ where T~M~ is structural transformation temperature, M~s~ and M~f~ are martensitic transformation start and finish temperature, and A~s~ and A~f~ are austenitic transformation start and finish temperature. As was expected, both Ti and Gd substitutions for Ga notably shifted the phase transformation temperatures (T~M~). Ti slightly decreased the T~M~ from 193 K to 173 K for Ni~50~Mn~25~Ga~25~ and Ni~50~Mn~25~Ga~20~Ti~5~ alloys respectively. On the other hand, the same atomic amount of Gd substantially increased the T~M~ to 279 K for the Ni~50~Mn~25~Ga~20~Gd~5~ alloy. These results suggest that the addition of Ti decreases the overall free energy difference between the martensite and austenite phases, which subsequently results in a decrease of the martensitic transformation temperature. In a similar manner, the addition of Gd notably increases the free energy difference between the austenite and martensite phases, which leads to a considerable increase in the martensitic transition temperature. Moreover, in the case of the Gd-doped sample, the temperature of the phase transition was very close to the room temperature; this may be an advantage in potential future applications of this material. Furthermore, the influence of Ti and Gd on magnetic transformation is significantly different than it is on structural transitions. The Curie temperature (calculated as an average value of the peak location on the first derivative's dM/dT curves established during the heating and cooling protocol) dropped substantially in the case of the Ti-doped sample (T~C~ = 373 K and 312 K for the Ni~50~Mn~25~Ga~25~ and Ni~50~Mn~25~Ga~20~Ti~5~ alloys, respectively). When it comes to the Ni~50~Mn~25~Ga~20~Gd~5~ alloy, the Curie temperature remained almost unchanged (T~C~ = 374 K). 3.3. Law of Approach to Magnetic Saturation {#sec3dot3-materials-13-02906} ------------------------------------------- The well-known law of approach (LoA) to magnetic saturation was used to analyze the magnetic behavior of all of the studied alloys and to estimate the temperature dependence of their effective magnetocrystalline anisotropy \[[@B35-materials-13-02906]\]. However, in order to properly use the LoA model, the authors made some important assumptions. First of all, only the high field magnetization curves (H ≫ Hc) were taken into account as initial curves. Next, after the first calculations were made considering the fitting procedure accuracy, the high field magnetization curves were limited to the last 5% of magnetization (M \> 0.95M~s~). Further, in this applied range of magnetization, recorded hysteresis loops were closed so that all hysteretic processes could be neglected, since further increase of magnetization was mainly induced by rotational processes of magnetic domains. Properly selected magnetization curves were then analyzed in terms of the law of approach to magnetic saturation \[[@B35-materials-13-02906]\]:$$M = M_{s}\left( {1 - \frac{a}{H} - \frac{b}{H^{2}}} \right) + \chi H,$$ where M~s~ is saturation magnetization, coefficient a is related to the to the structural inhomogeneity of the material, non-magnetic inclusions or internal microstresses, and coefficient b is directly connected with the existence of magnetocrystalline anisotropy. The last term of the equation χH describes the high field spontaneous magnetization. Considering the fact that coefficient a plays a significant role only in the lower fields \[[@B35-materials-13-02906],[@B39-materials-13-02906]\] and its contribution at high magnetic fields could be neglected, Equation (2) may be simplified to:$$M = M_{s}\left( {1 - \frac{b}{H^{2}}} \right) + \chi H,$$ Equation (3) was used to fit to the selected part of the magnetization curves, which is depicted in [Figure 3](#materials-13-02906-f003){ref-type="fig"}. The open circles show the measured data and the solid lines represent the best fittings according to Equation (3). The significant difference between the magnetization behavior of the low temperature martensite phase and the high temperature austenite phase is observable for the all studied alloys. As expected, even though martensite is characterized by a higher magnetization at high values of an applied magnetic field, it saturates much slower than notably softer austenite. This reflects the fact that high magnetocrystalline anisotropy of the martensite phase decreases the initial permeability of the material at low magnetic fields. Coefficient b calculated from Equation (3) was used to estimate the effective magnetic anisotropy constant (Keff) from the following relation:$$b = c\left( \frac{K_{eff}}{\mu_{0}M_{s}} \right)^{2},$$ where c is a constant dependent from the crystal structure of the material and equals c = 8/105 for cubic anisotropy \[[@B35-materials-13-02906]\]. Thus, K~eff~ can be finally estimated by utilizing the following formula:$$K_{eff} = \mu_{0}M_{s}\sqrt{\frac{105c}{8}},$$ [Figure 4](#materials-13-02906-f004){ref-type="fig"} presents the calculated effective magnetic anisotropy constant for all three Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ alloys. A significant drop of anisotropy (about 60%) in the vicinity of the phase transition was evident for every studied material. It was also clearly seen that the anisotropy energy decreased with the increase of temperature in both low temperature martensitic phases and high temperature austenitic phases (up to the Curie temperature where the material became paramagnetic). Moreover, the obtained values of anisotropy constant for the martensitic phase were the following: 9.6--8.5 × 10^5^ J/m^3^ for Ni~50~Mn~25~Ga~25~, 2.6--1.9 × 10^5^ J/m^3^ for Ni~50~Mn~25~Ga~20~Ti~5~ and 5.0--3.4 × 10^5^ J/m^3^ for Ni~50~Mn~25~Ga~20~Gd~5~ alloys; for the austenitic phase: 3.4--1.1 × 10^5^ J/m^3^ for Ni~50~Mn~25~Ga~25~, 0.7--0.3 × 10^5^ J/m^3^ for Ni~50~Mn~25~Ga~20~Ti~5~ and 1.4--0.7 × 10^5^ J/m^3^ for Ni~50~Mn~25~Ga~20~Gd~5~ alloys are in good agreement with the results reported for other NiMnGa-based materials \[[@B40-materials-13-02906],[@B41-materials-13-02906],[@B42-materials-13-02906],[@B43-materials-13-02906]\]. The relatively high values of magnetocrystalline anisotropy in the martensitic phase were essential for magnetic field induced strains. The obtained results suggested that the additions of Ti and Gd to NiMnGa-based compositions significantly decreased the effective anisotropy constant. This was particularly apparent in the case of the Ni~50~Mn~25~Ga~20~Ti~5~ alloy in which K~eff~ decreased about three times in comparison to the undoped Ni~50~Mn~25~Ga~25~ sample. [Figure 4](#materials-13-02906-f004){ref-type="fig"} also depicts the temperature dependence of a coercive field obtained upon heating for Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ alloys. In this case, the structural martensitic transition was also visible as a distinct drop in coercivity, which further showed the evident differences between the magnetic softness of the austenitic and martensitic phases. What is more in general, the coercive field was strongly coupled with the magnetic anisotropy of the material. This phenomenon is clearly observable in [Figure 4](#materials-13-02906-f004){ref-type="fig"}, as the temperature dependence of the coercive field followed the same trend as the temperature dependence of effective anisotropy. Another comparison of the temperature dependence of anisotropy is presented in [Figure 5](#materials-13-02906-f005){ref-type="fig"}a, where an effective anisotropy constant was normalized by dividing it by a maximum value of K~eff~ measured at 50 K; alternatively, the temperature axis was normalized by using the phase transformation temperature of each alloy. Thus, the dashed line at T/T~M~ = 1 on [Figure 5](#materials-13-02906-f005){ref-type="fig"}a corresponds to the structural transformation. The presented normalization procedure revealed that despite the significant differences in K~eff~ and T~M~, the overall temperature dependence of effective anisotropy behaved very similarly in all studied materials and was strongly connected with the structural phase transition. Moreover, [Figure 5](#materials-13-02906-f005){ref-type="fig"}b compares the normalized K~eff~ with the normalized coercive field (H/H~c(50K)~). It was previously shown in [Figure 5](#materials-13-02906-f005){ref-type="fig"} that K~eff~ is significantly related to the coercivity; this is even more evident now, in [Figure 5](#materials-13-02906-f005){ref-type="fig"}b. The two characteristic regions for magnetically softer austenite and significantly harder martensite were easily distinguishable for all studied alloys. 4. Conclusions {#sec4-materials-13-02906} ============== The influence of elemental doping with Ti and Gd on the temperature dependence of magnetic behavior in the Ni~50~Mn~25~Ga~20-x~Z~x~ (x = 0 or 5, Z = Gd, Ti) ferromagnetic shape memory alloys were studied in detail. The XRD analysis confirmed the cubic L2~1~ austenitic structure in all investigated materials and revealed that an addition of Ti or Gd elongates the lattice parameters of the austenite cell. Thermomagnetic measurements showed the strong influence of chemical composition on both martensitic and magnetic transformations. Following this fact, Ti addition to the Ni-Mn-Ga alloy significantly reduces the martensitic transformation temperature and Curie temperature in comparison to the Ni~50~Mn~25~Ga~25~ precursor. In the Gd-doped sample, the temperature of magnetic transition remained almost the same as in the Ni~50~Mn~25~Ga~25~ alloy, whereas the temperature of phase transition increased significantly, and came close to room temperature. Comprehensive studies in high magnetic fields based on the law of approach to magnetic saturation theory revealed the close correlation between magnetic and structural behavior in investigated samples. The abrupt drop of magnetocrystalline anisotropy at phase transformation temperatures and significant differences of anisotropies between martensitic and austenitic phases confirmed the completely different magnetic nature of the two existing phases in NiMnGa-based FSMA. What is more, the strong connection between the temperature dependence of the coercive field and the temperature dependence of magnetocrystalline anisotropy was also emphasized in this study. Conceptualization, A.Ł. and M.H.; methodology, A.Ł. and M.H.; formal analysis, A.Ł. and M.H.; investigation, A.Ł., M.H. and J.K.; data curation, A.Ł.; writing---original draft preparation, A.Ł.; writing---review and editing, A.Ł., M.H. and J.K.; supervision, J.K. All authors have read and agreed to the published version of the manuscript. This research received no external funding. The authors declare no conflict of interest. ![XRD patterns of the three peaks of the L2~1~ austenitic phase (a) (2 0 0), (b) (4 0 0) and (c) (4 2 2) obtained for the Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ alloys. The dotted line represents the measured values and the solid line shows the calculated pattern obtained from the Rietveld analysis. The additional peaks observed only in the Ni~50~Mn~25~Ga~20~Gd~5~ alloy are related to the residual Gd-rich Ga~1.5~Gd~1~Ni~3.5~ phase (space group P6/mmm).](materials-13-02906-g001){#materials-13-02906-f001} ![Temperature dependence of normalized magnetization for the (a) Ni~50~Mn~25~Ga~25~, (b) Ni~50~Mn~25~Ga~20~Ti~5~ and (c) Ni~50~Mn~25~Ga~20~Gd~5~ alloys measured under an external magnetic field of 250 mT. The arrows represent the austenitic transformation's start/finish temperatures (A~s~/A~f~) and the martensitic transformation's start/finish temperatures (M~s~/M~f~).](materials-13-02906-g002){#materials-13-02906-f002} ![High field parts of magnetization curves (M \> 0.95M~s~) for the (a) Ni~50~Mn~25~Ga~25~, (b) Ni~50~Mn~25~Ga~20~Ti~5~ and (c) Ni~50~Mn~25~Ga~20~Gd~5~ alloys measured from 50 K up to the Curie temperature. Open circles represent recorded data and solid lines show the best fitting to the law of approach to magnetic saturation (see Equation (3)).](materials-13-02906-g003){#materials-13-02906-f003} ![The temperature dependence of effective anisotropy constant K~eff~ (circles) estimated from the law of approach to magnetic saturation (see Equation (5)) and temperature dependence of coercive field (squares) for the (a) Ni~50~Mn~25~Ga~25~, (b) Ni~50~Mn~25~Ga~20~Ti~5~ and (c) Ni~50~Mn~25~Ga~20~Gd~5~ alloys.](materials-13-02906-g004){#materials-13-02906-f004} ![The normalized effective anisotropy constant (K~eff~/K~eff(50K)~) (a) as a function of normalized temperature (T/T~M~); the dashed line at T/T~M~ = 1 shows the structural transformation, (b) as a function of the normalized coercive field (H/H~c(50K)~) for the Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ alloys.](materials-13-02906-g005){#materials-13-02906-f005}	{ "pile_set_name": "PubMed Central" }
INTRODUCTION {#S1} ============ Psoriasis is a chronic immune-mediated inflammatory skin disease, affecting 1% to 2% of the Caucasian population.^[@R1]^ A pathogenic cross-talk between innate and adaptive cells, including keratinocytes, dendritic cells, and T-cells, underpins a dysregulated immune response leading to abnormal epidermal proliferation.^[@R2]^ Although psoriasis was initially considered as a Th1-mediated disease, mounting evidence from preclinical results, genetic data and clinical trials have demonstrated a key contribution of IL-17-producing T-cells to psoriasis in humans.^[@R3]^ Expression of IL-17A, IL-17F, and IL-17C is elevated in psoriatic lesional tissue compared with non-lesional tissue.^[@R4]^ IL-17A, IL-17F and IL-17C act directly on keratinocytes to stimulate the production of a number of molecules known to be elevated in psoriasis lesional tissue such as cytokines; antimicrobial peptides (AMPs); and chemokines such as IL-8, CCL20 and CCL2, thereby enabling IL-17 to bridge the innate and adaptive immune systems to sustain chronic inflammation.^[@R5],\ [@R6]^ The IL-17-signaling pathway utilizes a ubiquitin ligase signaling adaptor, Act1 (NF-κB activator 1), which is also known as CIKS (connection to IKK and SAPK/JNK), to propagate downstream signaling events.^[@R7]--[@R9]^ The binding of IL-17 to the heterodimeric IL-17 receptor (IL-17R) leads to Act1 recruitment through a homotypic SEFIR domain-dependent interaction. This results in incorporation of TNF receptor-associated factor 6 (TRAF6) into the signaling complex and the subsequent downstream activation of the NF-κB and MAPK pathways, leading to the induction of target genes in keratinocytes, epithelial cells, and fibroblasts stimulated with IL-17 family cytokines.^[@R9]--[@R11]^ The role of Act1 as a signal transducer is not confined to IL-17-responsive cells, as earlier studies found that Act1 is recruited to CD40 (TNF receptor superfamily member 5) and BAFFR (TNF receptor superfamily member 13C) upon stimulation of murine B-cells with CD40 ligand (CD40L, TNF superfamily member 5) and BAFF (tumor necrosis factor superfamily member 13b), respectively.^[@R12]^ One of the most robust genetic associations with psoriasis identified to date involves a variant in the TRAF3IP2 gene (rs33980500 C/T, which specifies a glutamic acid (D) to asparagine (N) change in Act1 (pD10N). This variant is unusual among complex disease susceptibility signals^[@R13]^ in that it specifies an amino acid change that appears to be functionally relevant. Moreover, this variant is associated with both cutaneous psoriasis and psoriatic arthritis^[@R14]--[@R17]^. Although this genetic association has been independently replicated, the manner in which the Act1 pD10N variant predisposes patients to psoriasis remains to be fully elucidated. An enigmatic feature of this association is that Act1 D10N appears to behave as a loss-of-function variant in acute signaling responses through IL-17R and CD40, whereas it appears to be pro-inflammatory in more broadly biological contexts. This enigma extends to mouse models in which Act1 has been silenced (see Discussion). In order to better understand the functional consequences of the Act1 D10N variant, we have undertaken an approach in which individuals of known Act1 genotype obtained through our previous GWAS studies of psoriasis^[@R14]--[@R17]^ were re-contacted and asked to provide biological samples for further functional analysis of Act1 genetic variation. Short-term responses, such as signal transduction responses, are desirable for such an analysis, because longer-term cellular responses involve the interaction of increasing numbers of proteins, whose functional variants cannot easily be controlled for due to random segregation of unlinked genes. To this end, we profiled peripheral blood mononuclear cells (PBMC) for Act1 expression and responsiveness to IL-17, making use of phospho-flow cytometry to measure short-term signaling responses. However, although multiple PBMC subsets expressed Act1, we were unable to identify any robust signaling responses to IL-17. Because Act1 has also been implicated in signaling events downstream of CD40^[@R12]^, we assessed the impact of the Act1 D10N variant on CD40L-stimulated B-cell signaling events in PBMCs from individuals homozygous, heterozygous, or nullizygous for the Act1 D10N allele. RESULTS {#S2} ======= To identify candidate assays for functional genetic testing of the Act1 D10N variant, we assessed Act1 protein levels in different PBMC subsets by flow cytometry. CD19^+^ B-cells, CD3^+^CD4^+^ ("helper") T-cells, CD3^+^CD8^+^ ("cytotoxic") T-cells and CD14^+^ monocytes were gated as shown in [Figure 1a](#F1){ref-type="fig"}. The percentages of Act1-positive cells were determined within each PBMC subset ([Figure 1b](#F1){ref-type="fig"}). We found the highest percentage of Act1-positive cells in monocytes (90.9% of CD14^+^ cells), with lower percentages in T-cells and B-cells (41.5% of CD4^+^ T-cells, 36.3% of CD8^+^ T-cells, and 54.9% of CD19^+^ B-cells, [Figure 1b](#F1){ref-type="fig"}). We also assessed Act1 expression levels by subtracting the median fluorescence intensity (MFI) of the isotype control mAb from that of the anti-Act1 Ab for each cell population ([Figure 1c](#F1){ref-type="fig"}). As assessed by MFI, all cell types expressed Act1, with CD14^+^ monocytes again expressing the highest levels, followed by CD4^+^ T-cells, CD8^+^ T-cells, and B-cells ([Figure 1c](#F1){ref-type="fig"}). The transcription factor NF-κB is a central mediator of immune and inflammatory responses.^[@R18]^ NF-κB is reported to be a downstream target of both IL-17R and CD40 signaling.^[@R12],\ [@R19],\ [@R20]^ To explore whether IL-17R and/or CD40 triggering leads to NF-κB activation in PBMC subsets, we combined surface marker phenotyping with intracellular assessment of phospho-NF-κB p65 (pp65) to assess IL-17A or CD40L-induced signaling by phospho-flow cytometry. As shown in the upper left panel of [Figure 2](#F2){ref-type="fig"}, pp65 was increased after CD40L stimulation for 15 minutes in B-cells but not in T-cells or monocytes. In contrast, no increase in pp65 was observed after 15 min of stimulation with 100 ng/ml or 200 ng/ml IL-17A in any of the IL-17A-treated PBMC subsets that we analyzed ([Figure 2](#F2){ref-type="fig"}, middle panels). In contrast, the levels of pp65 increased in all subsets when PBMCs were treated with phorbol myristate acetate (PMA) for 15 min as a positive control ([Figure 2](#F2){ref-type="fig"}, lower panels). Previous studies in Act1-deficient mice demonstrated a general increase in the numbers of peripheral B-cells.^[@R12],\ [@R21]^ However, Act1-deficient mice on a different background did not show this phenotype.^[@R22]^ To ask whether the same phenomenon might be observed in humans as a function of the Act1 D10N variant, we assessed the percentage of B-cells in individuals homozygous, heterozygous, or nullizygous for the D10N allele. The frequency of CD19^+^ B-cells did not differ between the Act1 D10N homozygotes and nullizygotes, although it was modestly higher in Act1 D10N heterozygotes ([Figure 3b](#F3){ref-type="fig"}). As shown in [Figure 3a](#F3){ref-type="fig"}, the frequency of CD3^+^ T-cells also did not differ significantly as a function of Act1 D10N genotype. We next investigated the functional consequences of the SNP-D10N on CD40 signaling in B-cells. The downstream pathways activated by CD40 triggering include not only NF-κB but also mitogen-activated protein kinases (MAPKs).^[@R23],\ [@R24]^ We thus sought to assay the phosphorylation state of the MAPKs p38 and Erk as potential functional measures of CD40 signaling. PBMCs from individuals homozygous, heterozygous, or nullizygous for the D10N allele were stimulated with CD40L for 15 minutes, and phosphorylation of p65, p38 and Erk was assayed by flow cytometry in the CD19^+^ B-cell subset. CD40L stimulation of PBMCs from all three genotype groups resulted in increased phosphorylation of p65, p38 and Erk in CD19^+^ B-cells ([Figure 4a](#F4){ref-type="fig"}). We found that the increase of pp65, pp38 and pErk (expressed as fold-change in MFI relative to unstimulated cells) determined across individuals were significantly correlated with each other ([Figure 4b](#F4){ref-type="fig"}). To determine the impact of the D10N variant on CD40 signaling in B-cells, we compared the increases of pp65, pp38 and pErk induced by CD40L among the three genotype groups. As shown in [Figure 4c](#F4){ref-type="fig"}, CD40L-mediated increases in pp65, pp38 and pErk were significantly attenuated in CD19^+^ B-cells from Act1 D10N homozygotes, relative to D10N nullizygotes. Moreover, the pp38 and pErk responses were significantly smaller in D10N homozygotes, relative to D10N heterozygotes. While responses did not differ significantly in heterozygotes relative to nullizygotes, there was a significant linear trend for decreasing phosphorylation of p65, p38 and Erk as a function of decreasing SNP-D10N variant counts across genotypes (pp65: P \< 0.01; pp38: P \< 0.01; pErk: P \< 0.01). DISCUSSION {#S3} ========== Genetic association studies of psoriasis have implicated a number of genes that function in NF-κB, IL-23R, and IL-17R signaling.^[@R14]--[@R16],\ [@R25]--[@R27]^ Act1, encoded by TRAF3IP2, is of particular interest due the fact that a disease-associated TRAF3IP2 SNP (rs33980500 C/T, encoding Act1 pD10N) appears to have functional impact^[@R15],\ [@R28]^, which may be due at least in part to its effect on the IL-17R signaling cascade.^[@R9]--[@R11]^ However, the direction of its effect seems paradoxical, in that the D10N mutation confers disease risk (i.e., is found at higher frequency in cases than in controls), yet it leads to loss of interaction with TRAF6^[@R15]^ and behaves as a hypofunctional allele in short-term signaling and gene expression assays.^[@R28]^ Reminiscent of the D10N mutation, Act1 knockout mice manifest many of the short-term signaling defects in IL-17 responses that are seen in Act1 D10N mice,^[@R7]^ and yet have been reported to manifest a skin inflammation phenotype characterized by epidermal hyperplasia and cell infiltration.^[@R12]^ Reports considering the effects of the Act1 D10N variant in the context of human cells are limited, with one previous report finding that fibroblasts from a D10N homozygous individual responded weakly to IL-17A.^[@R29]^ In the present study, we assayed Act1 expression and signaling responses to IL-17 treatment in different PBMC subsets by flow cytometry. As measured by both percentages of positive cells and by MFI, Act1 was expressed at the highest levels in monocytes. Lower levels of Act1 expression were observed in CD4^+^ T-cells, CD8^+^ T-cells and B-cells, though the rank order of expression depended upon the assay utilized ([Figure 1](#F1){ref-type="fig"}). However, despite the presence of Act1 in each cell type, PBMC-derived monocytes, as well as T-cells and B-cells, failed to respond to IL-17A stimulation of PBMC, as assessed by flow cytometric determination of NF-κB phospho-p65 ([Figure 2](#F2){ref-type="fig"}). This result was initially surprising, as a previous report indicated that human monocytes highly express both IL-17RA and IL-17RC, and that monocyte migration is stimulated by IL-17, in part due to PI3K and Erk-dependent induction of CCL2.^[@R30]^ However, that study involved longer-term assays using purified monocytes (a two-hour in vitro Boyden chamber chemotaxis assay, and a 24-hour in vivo sponge migration assay). While phosphorylation of p38 did increase after 15 and 30 minutes of IL-17 stimulation, three other IL-17-stimulated monocyte signaling events (pERK1/2, pJNK, and pAKT) were not significantly induced at 15 or 30 minutes, reaching maximal levels only after 60 to 180 minutes. In another paper, Krueger and colleagues^[@R31]^ reported gene expression profiling experiments in purified monocytes, which again involved longer IL-17 stimulation times (24 hours). In both of these studies, CD14^+^ monocytes were purified using magnetic beads, which might increase the sensitivity of monocytes to IL-17A treatment, whereas we utilized freshly-isolated PBMC. Future studies could utilize more sensitive detection methods, purified PBMC subsets, and/or assess longer-term responses to IL-17. Based on the aforementioned outcomes, we turned to B-cells, in which we reproducibly observed robust short-term responses to CD40L stimulation as assessed by p65, p38, and ERK phosphorylation ([Figures 2](#F2){ref-type="fig"} and [4](#F4){ref-type="fig"}). In addition to its key role in IL-17R signaling, Act1 was identified as an important adaptor molecule for CD40 signaling in B-cells.^[@R12]^ Ligation of CD40 by CD40L expressed on activated T-cells stimulates B-cell survival, proliferation, differentiation, isotype switching, and upregulation of surface molecules contributing to antigen presentation.^[@R32],\ [@R33]^ The CD40 receptor utilizes adapter molecules (TNF receptor associated factors, or TRAFs) that recognize binding sites on the CD40 cytoplasmic tail upon binding of CD40L. TRAFs then recruit TRAF-interacting kinases and together influence a number of well-characterized signal transduction pathways, including NF-κB and MAPKs.^[@R34]^ In the process, Act1 is activated by its recruitment to CD40 through its interaction with TRAF proteins.^[@R12],\ [@R20],\ [@R24]^ It would be worthwhile to further examine the impact of D10N mutation on the physical interaction of Act1 with CD40 and TRAFs. Engineering of the D10N variant into murine Act1 abolished its interaction with CD40 and BAFFR, suggesting that the D10N variant might phenocopy an Act1 knockout.^[@R28]^ However, previous studies in germ-line Act1 knockout mice revealed increased numbers of B-cells relative to WT littermates.^[@R12]^ These studies also found that activation of CD40 by treatment with a cross-linking anti-CD40 antibody increased survival of splenic B-cells, with stronger phosphorylation of IκBα and ERK.^[@R12]^ Taken together, these studies lead to the paradoxical conclusion that CD40 ligation activates short-term downstream signaling events in an Act1-dependent fashion, and yet in the long-term physiological context of transgenic animals, Act1 silencing increases overall B-cell numbers and function. A similar dichotomy has emerged from studies addressing the role of Act1 in IL-17 signaling in T-cells, as one strain of Act1 knockout mice develops inflammatory skin hyperplasia.^[@R28]^ It is possible that these paradoxical observations could be due to developmental influences and/or counter-regulatory feedback from other cell types, as the CD40L-dependent survival phenotype was much less evident in B-cells isolated from B-cell-targeted Act1−/− mice.^[@R12]^. Physiologically, CD40L is expressed primarily by activated T lymphocytes, therefore, B-cell signaling following CD40L-CD40 ligation represents a crucial component of the process known as T-cell "help" for antibody production. However, in addition to their role as antibody-producing cells, it has become apparent that B-cells markedly influence immunity through the secretion of cytokines.^[@R35]^ In particular, a subset of IL-10 producing B-cells has been found to have anti-inflammatory activity by diminishing both acquired and innate immune responses.^[@R36],\ [@R37]^ In humans, evidence is accumulating that B-cells can exert similar suppressive functions.^[@R38]^ Among many potential explanations for the "Act1 paradox" that remain to be tested, it is possible that the Act1 D10N variant might compromise the CD40-mediated function of B regulatory cells. In summary, we found that the Act1 D10N variant attenuated, rather than enhanced CD40L-dependent phosphorylation of NF-κB p65, p38, and Erk and MAPKs in B-cells ([Figure 4](#F4){ref-type="fig"}). However, we did not find a significant effect of the D10N variant on overall B-cell numbers ([Figure 3](#F3){ref-type="fig"}), as was previously observed in one strain of Act1 knockout mice.^[@R12]^ A limitation of our study is that our short-term signaling assays may not have probed more subtle levels of homeostatic regulation that are likely to be present in longer-term cellular responses and at the organismal level. In any case, our results establish a functional effect of the Act1 D10N variant in human cells for the first time, and as such, provide important albeit indirect evidence for the biological relevance of this variant to the immunopathogenesis of psoriasis. MATERIALS AND METHODS {#S4} ===================== Study sample and genotyping {#S5} --------------------------- All human subjects provided written informed consent and were enrolled according to the protocols approved by the institutional review board of the University of Michigan Medical School, in adherence with the Declaration of Helsinki principles. Genomic DNA was isolated from peripheral blood by phenol-chloroform extraction followed by ethanol precipitation. Sixty-three individuals of Caucasian descent (21 males, 42 females; 8 psoriasis cases, 55 controls, mean age 44 years, range 19 -- 83 years) were identified based on Act1 genotypes that were either measured directly or obtained from genome-wide or Immunochip-based association studies of psoriasis.^[@R17],\ [@R25],\ [@R27]^ and confirmed using a Taqman assay (Assay ID C_2473124_10, Applied Biosystems, Foster City, CA). The typing of these 63 subjects resulted in 13 Act1 D10N homozygotes, 20 Act D10N heterozygotes, and 30 Act1 D10N nullizygotes. Isolation of PBMCs {#S6} ------------------ For the isolation of PBMCs, anticoagulated blood was mixed with an equal volume of phosphate-buffered saline. The diluted blood was slowly layered over Histopaque^®^-1077 (Sigma-Aldrich, St. Louis, MO) solution by gently pipetting the diluted blood down the side of the tube, and was then centrifuged for 20 min at 800 g at 22°C, with no brake. The mononuclear cell layer was transferred into 10 ml phosphate-buffered saline and centrifuged for 10 min at 400 g at 4°C. After two washings with RPMI-1640, PBMCs were gently resuspended in freezing medium containing 90% FBS and 10% DMSO as described^[@R38]^ and cryopreserved with gradual cooling followed by storage in liquid nitrogen. Flow cytometric analysis of ACT1 expression {#S7} ------------------------------------------- PBMCs were thawed, washed by centrifugation, and 5 × 10^5^ cells were aliquotted to each tube. Dead cells were excluded from the analysis by staining with LIVE/DEAD Fixable Near-IR Dead Cell Stain (Catalog\#: L10119, Life Technologies, Waltham, MA). Antibodies to cell-surface markers were mixed and added to the cells, including PE-Cy5-coupled anti-CD3 (5 μl per 10^6^ cells; Catalog\#: 344808, BioLegend, San Diego, CA), PE-coupled anti-CD19 (Catalog\#: 12-0199-41, eBioscience, Santa Clara, CA), V450-coupled anti-CD14 (Catalog\#: 560349, BD Biosciences, San Jose, CA), PE-Cy7-coupled anti-CD4 (5 μl per 10^6^ cells; Catalog\#: 300512, BioLegend) and V500-coupled anti-CD8 (Catalog\#: 560774, BD Biosciences). After incubation for 30 min at 4°C;, the cells were washed twice with BD Pharmingen^™^ Stain Buffer (Catalog\#: 554656). The cells were then fixed and permeabilized for 60 min at 4°C; using the Fixation/Permeabilization diluent and concentrate (Catalog\#: 00-5123-43 and 00-5223-56, eBioscience). Cells were then stained with eFluor 660-coupled anti-Act1 (Catalog\#: 50-4040, eBioscience) or eFluor 660-coupled isotype control (eBioscience). The analysis was performed using flow cytometer (BD LSR II, BD Biosciences) and FlowJo 7.6.2 (TreeStar, Ashland, OR) software. Flow cytometric analysis of phosphorylated proteins {#S8} --------------------------------------------------- For determination of phosphorylated proteins after CD40 stimulation, PBMCs were thawed and rested for 2 h and then either left unstimulated or stimulated with trimeric CD40L (100 ng/ml) (Enzo Life Sciences, Farmingdale, NY) for 15 min, performed concurrently with decoration with V450-coupled anti-CD19 (Catalog\#: 560353, BD Biosciences). After incubation with BD Cytofix^™^ Fixation Buffer (Catalog\#: 554655) for 10 minutes at 37°C;, followed by permeabilization in BD Phosflow^™^ Perm Buffer III (Catalog\#: 558050) for 30 minutes on ice, cells were then stained with PE-Cy5-coupled anti-CD3 (BioLegend), FITC-coupled anti-pp38 (Catalog\#: 4551, Cell Signaling Technology, Danvers, MA), PE-coupled anti-pp65 (Catalog\#: 558423, BD Biosciences), or PE-coupled anti-pErk (Catalog\#: 612566, BD Biosciences). The analysis was performed using an LSR II flow cytometer (BD Biosciences) and FlowJo 7.6.2 software (TreeStar, Ashland, OR). Statistical analysis {#S9} -------------------- For flow cytometry experiments, CD3^+^ T-cell and CD19^+^ B-cell percentages and MFI values obtained for each donor belonging to the three genetic groups were assessed for normal Gaussian distribution using the Kolmogorov-Smirnov test and then analyzed by one-way analysis of variance, followed by post hoc least significant difference (LSD) test or Dunnett's test by using SPSS version 21.0 (IBM SPSS Statistics, New York). Pearson's coefficient was used to assess two-way correlations between pp65 and pp38, pp65 and pErk, and pErk and pp38. All results are shown as mean and the standard error of the mean (mean ± SEM). In this exploratory analysis, a P value \< 0.05 was considered to be significant. We thank the volunteers who provided blood samples for this study, and Dr. Andrew Johnston for advice on flow cytometry. This research was supported by National Institutes of Health (NIH) R01 grants AR042742, AR050511, AR054966, AR062382, and AR065183 to James T. Elder. We also acknowledge generous support from the Dawn and Dudley Holmes Memorial Fund and the Babcock Endowment Fund to the Department of Dermatology at the University of Michigan. James T. Elder is supported by the Ann Arbor Veterans Affairs Hospital. CONFLICT OF INTEREST The authors declare no conflict of interest. ![Act1 is expressed in different subsets of PBMCs. PBMCs were stained with LIVE/DEAD Fixable Near-IR Dead Cell Stain, PE-Cy5-coupled anti-CD3, PE-coupled anti-CD19, V450-coupled anti-CD14, PE-Cy7-coupled anti-CD4 and V500-coupled anti-CD8 and eFluor 660-coupled anti-Act1 or eFluor 660-coupled isotype control. CD19^+^ B-cells, CD4^+^ T-cells, CD8^+^ T-cells and CD14^+^ monocytes were identified. (a) Cell subset distributions from a representative flow cytometry experiment (one of four experiments). (b) Act1^+^ cells percentages within each PBMC population from a representative flow cytometry experiment (one of four experiments). Mean Act1^+^ percentages from all experiments (n = 4). Bars represent mean ± SEM. \*P \< 0.05, as assessed by one-way analysis of variance followed by least significant difference post hoc test. (c) Act1 expression levels for each PBMC population from a representative experiment (one of four experiments). Median fluorescence intensities (MFI) from all experiments (n = 4). Error bars represent SEM. \*P \< 0.05, \\P \< 0.01, as assessed by one-way analysis of variance followed by least significant difference post hoc test.](nihms902935f1){#F1} ![CD40L induces p65 phosphorylation in B-cells. Various PBMC subsets were assayed for phosphorylated p65 (pp65) after 15 minutes of stimulation with CD40L (100 ng/ml, upper panels), IL-17A (100 ng/ml, middle panels) or PMA (100 nM, lower panels). Results from one of four representative experiments are shown. Blue histograms demonstrate pp65 response after stimulation with CD40L, IL-17A, or PMA, as compared to unstimulated cells (red curves).](nihms902935f2){#F2} ![Percentages of T-cells and B-cells in PBMC, as a function of Act1 D10N status. PBMCs were collected from D10N homozygous (red), heterozygous (blue), and nullizygous (green) individuals. Percentages of CD3^+^ T-cells (a) and CD19^+^ B-cells (b) within the lymphocyte gate (defined by forward and side scatter) are shown. Each symbol represents three experiments averaged for one donor. Bars represent mean ± SEM. \*P \< 0.05, as assessed by one-way analysis of variance followed by least significant difference post hoc test.](nihms902935f3){#F3} ![D10N impairs CD40L-induced p65, p38 and Erk phosphorylation in PBMC-derived B-cells. PBMCs were cultured with (blue curves) or without (red shadow) 100 ng/ml CD40L for 15 minutes, followed by phospho-flow cytometric analysis of pp65, pp38 and pErk within the CD19^+^ gate. (a) Representative histograms (from one of two experiments performed for each genotype) are presented for PBMC-derived B-cells from Act1 D10N nullizygous (left panels), heterozygous (middle panels), and homozygous (right panels) individuals. (b) Correlations between MFI values for pp65, pp38 and pErk in stimulated CD19^+^ B-cells for all tested individuals (n = 63), Data are expressed as fold-change increases. Pearson correlations between pp65 and pp38, pp65 and pErk, pErk and pp38 were performed. Each symbol represents two experiments averaged for one donor (c) Phosphorylation of p65, p38 and Erk in B-cells from nullizygous, heterozygous, and homozygous groups in response to CD40L were expressed as fold MFI. Each symbol represents one individual donor. Summary data represented by horizontal lines represent mean ± SEM. \*P \< 0.05, \\P \< 0.01, as assessed by one-way analysis of variance followed by Dunnett's post hoc test.](nihms902935f4){#F4} [^1]: These authors contributed equally to this work	{ "pile_set_name": "PubMed Central" }
INTRODUCTION ============ Home bleaching is a popular and convenient treatment to improve the appearance of vital teeth.[@b1-dent5_p0157] During this procedure, whitening gel is deposited on moulded acrylic trays that allow the solution to act indistinctly over both enamel and pre-existing aesthetic restorations. Many investigations have shown that low-concentration bleaching substances may damage these structures[@b1-dent5_p0157]--[@b3-dent5_p0157] and thus increase their susceptibility to staining and bacterial adhesion.[@b4-dent5_p0157]--[@b6-dent5_p0157] Major colour improvements are sometimes enthusiastically desired by patients, whereas the adverse effects of whitening procedures may be unintentionally downplayed by professionals. Recently, manufacturers released a flood of high-concentration bleaching gels in the market. These gels are claimed to increase the effectiveness as well as longevity of at-home treatments. As expected, initial in vitro studies reported significant alterations to enamel[@b7-dent5_p0157],[@b8-dent5_p0157] or composite resins.[@b2-dent5_p0157],[@b9-dent5_p0157] However, these results were not confirmed by the scant evidence originating from clinical investigations.[@b10-dent5_p0157],[@b11-dent5_p0157] Since the effects of whitening substances have been evaluated separately and in a great diversity of regimens, it is difficult to extend those incompatible evidences to aesthetically restored teeth. Thus, this prospective clinical study aimed to assess simultaneously the effects of 16% carbamide peroxide on the external topography of enamel and composite resin. MATERIALS AND METHODS ===================== Preparation of specimens ------------------------ This study was initiated only after obtaining suitable approval from the Local Ethics Committee, which is in conformity with the Declaration of Helsinki (DoH). A total of five human impacted upper third molars were included in this study. All molars included fit the following criteria: 1) arising from different donors and 2) without structural damage, signs of hypomineralisation (white spots) or prior contact with saliva. After cleaning, teeth were individually positioned in a sectioning machine equipped with a water-cooled 300-μm thick diamond saw (Minitorm, Struers A/S, Copenhagen, Denmark). Roots were removed at the cemento-enamel junction. After that, the remaining crows were equidistantly sectioned in the mesiodistal and buccolingual directions to provide four enamel fragments (4 mm x 4 mm and 2 mm thickness) each. The pieces obtained were moulded into a light regular-set silicon (Virtual Light-Body, Ivoclair Vivadent, São Paulo, SP, Brazil), and the moulds were filled with increments of a micro-hybrid composite resin (EA2 Opallis, FGM Dentscare Ltda, Joinville, SC, Brazil). Moulds were sequentially polymerised over 20 seconds using a light-emitting diode unit set between 460 and 480 mW/cm^2^ (Ultra-Blue i5, DMC Vasconcellos, São Carlos, SP, Brazil) in order to produce enamel-like composite resin specimens. The resulting 40 specimens (20 of composite resin) were then individually placed at the end of polyvinyl chloride cylinders (3/4 inch diameter and 20 mm high), which were filled with chemically-activated acrylic resin (Technovit 4000, Heraeus Kulzer, Wehrheim, Germany). After 24 hours, the exposed side of each embedded fragment was flattened and polished by sequential use of wet 400-, 600- and 1200-grit silicon carbide papers (Norton®,Guarulhos, SP, Brazil) adapted in a slow-speed polishing machine set at 150 rpm (Ballerup, Struers A/S, Copenhagen, Denmark). The finished specimens were then removed from blocks and ultrasonically cleaned for 5 minutes before being sterilised and stored in distilled water (37±0.5ºC) until use. Ten volunteers (20--25 years old) were recruited to participate in this investigation. All volunteers presented the following inclusion criteria: absence of caries and gingivitis, no history of xerostomia, no previous bleaching therapy and willingness to understand and sign a consent form. Participants had their maxillary right teeth cleaned with a rotary toothbrush and dentifrice before isolation with a rubber dam and stainless steel clamps. The buccal sides of the right upper second premolar and first molar were etched with 37% phosphoric acid (Dentsply, Petrópolis, RJ, Brazil) for 30 seconds and immediately washed with running water for 60 seconds. To fix the enamel and composite resin randomly on these surfaces, a two-step adhesive (Adper™ Scotchbond, 3M, Sumaré, SP, Brazil) was brushed and light-polymerised for 15 seconds before an increment of low-viscosity composite resin (Natural Flow, DFL, Jacarepagua, RJ, Brazil) was used to glue the specimens on teeth using light-polymerisation for 40 seconds. These same procedures were performed on the adjacent lower right teeth to fix the enamel and composite resin on specimens not undergoing the bleaching process. These specimens thus served as paired controls (n=10). Bleaching process ----------------- An upper jaw impression was taken of each participant with alginate materials (Jeltrate, Dentsply, Petrópolis, RJ, Brazil), and impressions were poured with type II dental stone (Pasom, São Paulo, SP, Brazil) to fabricate a master cast. A 0.5 mm-thick wax layer was applied on the buccal face of each cast 1.5 mm away from the gingival margin to fabricate reservoirs for the bleaching gel. Soft trays (Cristal, Bio-art, São Carlos, SP, Brazil) were made using a heat/vacuum tray-forming machine (Plasvac P7, Bio-art, São Carlos, SP, Brazil). The trays were tested and trimmed to fit the gingival margin of each volunteer. The participants were instructed to use the tray at the upper arch and to keep the bleaching product (Whiteness Perfect 16%, FGM Dentscare Ltda, Joinville, SC, Brazil) in contact with teeth exactly 6 hours per day for 8 consecutive days. No restrictions on dietary or hygiene habits were imposed, but the necessity to remove the residual solution after each whitening session was emphasised. Every 48 hours, the volunteers were questioned about sensitivity of teeth or soft tissue. When hypersensitivity was confirmed, desensitising treatment was performed with 2% potassium nitrate (Desensibilize KF 2%, FGM Dentscare Ltda, Joinville, SC, Brazil). The four subjects were instructed to place the desensitizing gel in their tray and wear it for 20 minutes once a day, as recommended by the manufacturer. Surface analysis ---------------- After detachment from teeth, the specimens were left to dehydrate for 96 hours before being gold sputter-coated to permit analysis in a scanning electron microscope (DSM-940 A, Carl Zeiss, Oberkochen, Germany). Digital SEM photomicrographs were taken at four different areas over the surface of each fragment at 5000x magnification. Enamel and resin surface alterations were classified qualitatively by increasing order of scores ranging from 0 (no observable alterations) to 4 (heavy erosion with deep depressions). Prior to blind analysis of the 160 images, a single examiner was calibrated by viewing 30 additional SEM micrographs. A second assessment was repeated 4 weeks after the first evaluation, and the data obtained were considered as a whole for statistical comparison. Statistical analysis -------------------- Statistical differences between the experimental and appropriate control groups were executed with the Mann-Whitney Test adjusted to the 95% confidence interval. The intra-observer agreement was determined using Cohen's Kappa statistic. Statistical analyses were performed with SPSS software v.11.0 for Windows (SPSS Inc., Chicago, IL, USA). RESULTS ======= Representative SEM images are shown in [Figures 1](#f1-dent5_p0157){ref-type="fig"}--[4](#f4-dent5_p0157){ref-type="fig"}, and the results of the Mann-Whitney test are presented in [Table 1](#t1-dent5_p0157){ref-type="table"}. Following application of 16% carbamide peroxide, the enamel surface in the experimental group exhibited extensive mild erosion with shallow depressions and destruction of interprismatic matrix. This appearance differed significantly from that of non-bleached enamel fragments, which exhibited a smooth and amorphous aspect. The bleached composite resin specimens displayed a flat appearance, with slight erosion and some striation due to the grinding procedure. This appearance was statistically similar to that observed in the control group (P\<.05). Intra-observer agreement was 0.89. DISCUSSION ========== This study was the first to adhere human enamel and resin fragments to the buccal side of regular teeth to reproduce, as closely as possible, an actual home-bleaching situation. This manoeuvre permitted controlled and continuous exposure of specimens to saliva, beverages and oral hygiene habits, a procedure completely different from prior investigations in which removable appliances were adopted to carry the specimens.[@b1-dent5_p0157],[@b9-dent5_p0157]--[@b11-dent5_p0157] Therefore, such biases make previous studies difficult to interpret and compare. Scanning electron microscopy is a simple and effective method for identifying surface morphology alterations on enamel and composite resins submitted to home bleaching.[@b1-dent5_p0157],[@b12-dent5_p0157],[@b13-dent5_p0157] Nonetheless, dehydration and metal coating necessary to SEM analysis are able to change part of the specimens structure;[@b14-dent5_p0157] thus, the appearance of enamel observed may not correspond precisely to that of bleached enamel in its natural condition. Porosity and erosion resulting from demineralisation were demonstrated by SEM after in vitro and in vivo intermittent exposure of enamel to 10% carbamide peroxide.[@b1-dent5_p0157],[@b15-dent5_p0157],[@b16-dent5_p0157] In the present study, the time of contact between the enamel and whitening gel (48 hours in total) was shorter than those adopted in other investigations[@b7-dent5_p0157],[@b10-dent5_p0157],[@b15-dent5_p0157]--[@b17-dent5_p0157] and sufficient to produce an eroded appearance similar to that of acid-etched surfaces. This finding has already been observed in vitro by Adebayo et al[@b8-dent5_p0157] but not in vivo by Metz et al,[@b11-dent5_p0157] likely because the enamel specimens used in the latter research had prior contact with fluoride. The damage detected on enamel could be justified by the ability of urea, derived from the reaction of carbamide peroxide with water, to denature protein structures and thus cause structural and morphological changes through the degradation of organic molecules.[@b14-dent5_p0157] In addition, the slightly acidic pH of the gel tested (6.04)[@b18-dent5_p0157] may have contributed secondarily to these external modifications.[@b12-dent5_p0157],[@b17-dent5_p0157] However, a possible increased susceptibility of unerupted third molars' enamel to the action of chemical agents such as carbamide peroxide should be considered before this result can be generalized. Since these findings were detected in vivo, it may be suggested that the expected dissolving and buffering effects of saliva on the whitening gel[@b19-dent5_p0157] or the attenuation provided for a peroxide-consuming enzyme on the surface of teeth[@b20-dent5_p0157] were not able to minimise the deleterious impact of a highly concentrated bleaching gel on human enamel. Alterations to enamel topography caused by low concentrations of bleaching gels may undergo repair over time by precipitation of mineral phases derived from saliva into the existing porosities.[@b1-dent5_p0157],[@b16-dent5_p0157],[@b21-dent5_p0157] Tooth remineralisation is a slow process, and the high concentration of bleaching gel used in this study probably smothered the saliva recovery properties. Clinically, however, erosion reversal should be expected as soon as whitening gel application is discontinued. In recent laboratory investigations, the addition of fluoride to carbamide peroxide gel resulted in less demineralisation and shorter periods for enamel hardening recovery following bleaching.[@b22-dent5_p0157],[@b23-dent5_p0157] The absence of significant alterations on the surface of the composite resin bleached was already described for nano-hybrid and packable resins after extended periods of in situ and in vitro exposure to 15% carbamide peroxide.[@b6-dent5_p0157],[@b11-dent5_p0157] Since morphological alterations were noticed on enamel specimens subjected to the same home-whitening conditions, we inferred that salivary attenuation of the highly concentrated peroxide carbamide substance did not completely respond for resin resistance to bleaching detrimental effects.[@b19-dent5_p0157],[@b20-dent5_p0157] Enamel and composite are completely different materials, and an incompatibility between the solvents present in the whitening substances and the components in the polymer matrix of the resin-based material may justify the absence of deleterious effects on resin surfaces.[@b2-dent5_p0157],[@b11-dent5_p0157],[@b13-dent5_p0157],[@b24-dent5_p0157] Another possible explanation for the absence of statistical difference between groups relates to the subjective and qualitative evaluations necessary for the SEM micrographs and also to the low number of composite resin specimens used; it is possible that major changes in the surface of bleached resin have been underrated, despite the effort to calibrate the examiners, thus reducing the numerical difference between groups. Although not widespread as SEM, atomic force microscopy (AFM) enables quantitative and independent measurement of specimens' surface morphology in their natural condition.[@b17-dent5_p0157] Thus, the possibility of getting different results, if AFM had been used, cannot be disregarded. Colour improvements have been described separately for enamel[@b25-dent5_p0157] and composite resin.[@b26-dent5_p0157] Since these structures were bleached together in this investigation, we provide evidence that these substrates responded differently to 16% carbamide peroxide. This phenomenon may justify the common observation of poor colour matching of pre-existing restorations after home whitening.[@b11-dent5_p0157] Nonetheless, given the diversity of composite resins available, the susceptibility of their surfaces to this or other vital bleaching regimens and formulations may be different. To date, it is uncertain how the surface changes that occur during bleaching are related to colour improvement and/or increases in staining susceptibility.[@b5-dent5_p0157],[@b6-dent5_p0157] However, the complete repolishing of aesthetically restored teeth to recover smoothness and delay staining of composite resins[@b9-dent5_p0157],[@b27-dent5_p0157] may both be unnecessary, at least for the material studied, and thus avoid the irreversible loss of enamel structure. CONCLUSIONS =========== Within the limitations of this clinical study, we concluded that the occurrence of surface topography changes during home bleaching of aesthetically restored teeth with 16% carbamide peroxide was restricted to enamel. Dr. Elliot Watanabe Kitajima, from the Nucleus of Research Support, Electron Microscopy Center (NAP/MEPA-ESALQ/USP). We also thank all colleagues and volunteers who contributed to this research. ![Representative SEM micrograph of bleached enamel at original magnification 5000x.](DENT-5-160-g1){#f1-dent5_p0157} ![Representative SEM micrograph of unbleached enamel at original magnification 5000x.](DENT-5-160-g2){#f2-dent5_p0157} ![Representative SEM micrograph of bleached resin composite at original magnification 5000x.](DENT-5-160-g3){#f3-dent5_p0157} ![Representative SEM micrograph of unbleached resin composite at original magnification 5000x.](DENT-5-160-g4){#f4-dent5_p0157} ###### Median values and ranges of scores attributed to enamel and resin composite micrographs Substrate Treatment Median P value ----------------- ---------------- --------------------------------------------------- --------------------------------------------------- Enamel bleached 3.4 (3.0--4.0) 0.004[\](#tfn1-dent5_p0157){ref-type="table-fn"} unbleached 1.4 (1.0--2.0) 0.004[\](#tfn1-dent5_p0157){ref-type="table-fn"} Composite resin bleached 2.3 (1.0--3.0) 0.091 unbleached 1.6 (1.0--2.0) 0.091 statistically significant.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1} =============== A plasmacytoma is a tumor of plasma cells which often develops as a systemic spread of multiple myeloma, in bone as a solitary plasmacytoma of bone, or in soft tissue as extramedullary plasmacytoma (EMP). Extramedullary plasmacytomas are rare, comprising about 3% of all plasma cell neoplasms with bone marrow plasma cell infiltration \<5% of nucleated cells and demonstrating no evidence of myeloma \[[@B1]\]. While solitary bone plasmacytomas reveal single tumor inside the bone comprising of abnormal plasma cells. The majority (60%--80%) of extramedullary plasmacytomas are found in the head and neck region, particularly in the upper respiratory tract, and to a lesser extent in gastrointestinal tract. Cases of primary cutaneous plasmacytoma are comprised of 2--4% of extramedullary plasmacytomas. It is unusual to have a primary cutaneous plasmacytoma present as a lip lesion, as only 5 cases have been reported in the literature \[[@B2]\]. The first case reported was by Volk in 1936 \[[@B2]\]. Primary cutaneous plasmacytomas present as relatively slow growing papules or plaques or an erythematous-violaceous nodule either existing in solitary tumor (62%) or possibly involving multiple sites (38%). Patients usually present with a benign past medical history. Symptoms are dependent on location and size of lesion. Rarely, patients present with fever, malaise, night sweats, or weight loss. Other features that may be present are lymphadenopathy and hepatosplenomegaly. The lesion may be ulcerated and necrotic. The skin lesions range in diameter from 1 to 5 cm and the shape may vary. Histologically, it shows a nonepidermotropic dermal infiltrate of plasma cells at different stages of maturation. The plasma cells often show atypical, binucleation, and increased mitotic activity. The epidermis is generally spared. The mean age at diagnosis is 60 years with a male to female predominance of 3 : 1 and it is commonly reported in Asian men. The incidence rate increased with advancing age. Treatment for solitary or a few lesions can be treated with radiotherapy or surgical excision. Chemotherapy or intralesional corticosteroid therapy may be provided for multiple skin lesions. Due to this rarity, we feel the need to present this case in order to increase the awareness of the disease to better serve our patients. The clinicopathological features, prognostic factors, and treatment options with its unusual presentation of lip involvement is being discussed in this report. 2. Case Description {#sec2} =================== A 65-year-old male presented to the Ears, Nose, and Throat clinic in September 2015 with complaints of a nonhealing lower lip sore for the past 3 years. The slow growing lesion was described as a burning sensation with scabbing and occasional bleeding. It was slowly increasing in size. The patient had a greater than 50-year history of smoking tobacco (not chewing) which he quit about 7 months prior to his visit. He also complained of unintentionally losing 3 lbs in the last 2 weeks. He denied any fevers or chills. He had no other complaints. Physical examination was significant for a single lesion on the lower lip 1.5 × 1 cm. A biopsy was performed and sent to pathology. The patient was worked up thoroughly for evidence of multiple myeloma/light chain disease. Laboratory testing was significant for no anemia, normal serum calcium, normal serum albumin, normal total protein, and normal creatinine. There was no Bence Jones protein in urine. Bone survey was done which was negative. Bone marrow biopsy showed \<5% plasma cells, ruling out plasmacytoma of the bone. No monoclonal gammopathy was seen, which precluded immunofixation order. Kappa to lambda ratio was normal. Hematoxylin-eosin staining of the lip biopsy revealed a dense plasmacytic infiltration in the deep dermis (see [Figure 1(a)](#fig1){ref-type="fig"}). Immunohistochemical panel included CD3, CD20, and CD138 immunohistochemical stains as well as kappa and lambda light chain in situ hybridization. The superficial dermis and the epidermis were not seen to be involved. Immunohistochemistry was positive for CD138 (see [Figure 1(b)](#fig1){ref-type="fig"}) and revealed that most plasma cells were lambda positive and with rare kappa-positive plasma cells (see Figures [1(c)](#fig1){ref-type="fig"} and [1(d)](#fig1){ref-type="fig"}). Based on the ISCL/EORTC Proposed on TNM Classification of Cutaneous Lymphoma our patient was staged at T1aN0M0. The patient was treated in November 2015 with radiation therapy as follows: 3000 cGy in 15 fractions using 3D conformal technique to his lower lip and level I lymph nodes and he received 2000 cGy boost bringing his total dose to 5000 cGy in 25 fractions. He responded well to treatment. At his last follow-up, there was no evidence of multiple myeloma or light chain disease. 3. Discussion {#sec3} ============= Primary cutaneous plasmacytoma is a rare diagnosis that represents approximately 4% of extramedullary plasmacytoma cases. Of the 4% cases, only a few cases have been reported to be associated with lip lesions. In the latest comprehensive literature review by Tsang et al. in 2016, there are 5 case reports of primary cutaneous plasmacytomas of the lip \[[@B2]\]. Primary cutaneous plasmacytomas without underlying multiple myeloma is included in the 2005 WHO/EORTC classification among the primary cutaneous marginal zone B-cell lymphoma due to considerable overlapping features \[[@B3]\]. Primary cutaneous marginal zone B-cell lymphoma is considered as part of extranodal marginal zone lymphoma of MALT type \[[@B3]\]. Due to the infrequent and nonspecific presentation of cutaneous plasmacytomas, a clinical diagnosis is not sufficient. Diagnosis is based on clinical, histopathological, and immunohistochemical findings, and multiple myeloma must be ruled out by laboratory, radiologic, and bone marrow studies \[[@B4]\]. After ruling out multiple myeloma, the only findings will be a histopathologic dense monomorphic dermal plasmacytic infiltrate and a monoclonal immunohistochemistry. Consistent with published literatures, our male patient in this case, at the age of 65, presented with a slow growing lesion and insignificant systemic review. Tissue biopsy immunohistochemically shows cytoplasmic expression of immunoglobulins that are restricted to one immunoglobulin chain. Histopathologic finding of diffuse or nodular dermal and subcuticular infiltrate of plasma cells, displaying varying degrees of maturation and atypia but no epidermotropism, are findings that are consistent with our patient. The immunophenotype demonstrates plasma cells positive for CD138, CD38, and CD79a, but generally not CD20 and leukocyte common antigen, and shows monotypic cytoplasmic immunoglobulin light chain expression on paraffin sections \[[@B3]\]. On the other hand, cutaneous marginal zone B cells express CD 20. Our patient was positive for CD138 and negative for CD20. While 30% of all marrow-based myelomas are lysozyme positive, lysozyme is not expressed by cutaneous plasmacytomas. Laboratory, radiological, and bone marrow investigations must be done to exclude multiple myeloma. Laboratory studies include but are not limited to complete blood count, blood smear, blood sedimentation rate, renal function, electrolyte (calcium), serum and urine protein electrophoresis, quantitative Ig determination in serum, immunoelectrophoresis in serum and urine, and B-2 microglobulin determination in serum. All laboratory findings will be normal except for quantitative Ig elevation. Radiological studies include a full skeletal X-ray to rule out any osteolytic lesions that are consistent with multiple myeloma. A bone marrow biopsy will demonstrate plasma cell infiltration less than 5% of all nucleated cells. Plasma cells infiltration is common in reactive lesion and must be ruled out with staining for kappa or lambda light chain. In contrast to multiple myeloma, extramedullary plasmacytoma showed absence of cyclin D1 and infrequent expression of CD56. Other differential diagnoses included secondary cutaneous plasmacytoma in the setting of systemic disease, mucosal extramedullary plasmacytoma with secondary skin involvement, marginal zone B-cell lymphoma, plasma cell granuloma, non-Hodgkin lymphoma, high-grade malignant immunoblastic lymphoma, and poorly differentiated neoplasms, as well as infectious diseases that cause infiltration of plasma cells. Currently, there are approximately 50 cases of primary cutaneous plasmacytomas that have been reported in the literature \[[@B2]\]. Due to its rarity, little is known about its treatment. Treatment options include surgery, radiation, chemotherapy, and intralesional corticosteroid therapy. Type of treatment depends on the numbers (solitary versus multiple) and location of lesions. Most patients with solitary primary cutaneous plasmacytoma received radiotherapy as a component of treatment. Radiotherapy with curative intent (\>4000 cGy) is the standard of care and provides promising result since the tumor is radiosensitive. The optimal dose for local control is 40--50 Gy in 20 fractions delivered over 4--6 weeks. However, caution should be considered when using radiotherapy as local recurrence or lymph node metastases could develop thereafter \[[@B2]\]. In addition to high dose radiotherapy and fractions, adjuvant chemotherapy may also be considered for lesions with high-grade histology or tumors \>5 cm or for patients with multiple lesions. Palliative irradiation may also be used for the most symptomatic and large lesions. At the time when this case was written, the patient was in remission with no evidence of multiple myeloma or light chain disease. As reported in previous literatures, patient with solitary lesions received better outcomes than patient with multiple lesions at presentation. They tend to have less local regional or distant dissemination and remained confined to the skin. As a note, large solitary lesions might have a higher risk of progression to metastatic disease, although a size cut-off had not been determined \[[@B2]\]. The main prognostic factor seems to be the clinical presentation with more aggressive cases associated with multiple lesions, larger size, and IgA secretion by neoplastic cells. Multiple myeloma has been estimated to develop in one-third of patients \[[@B5]\]. It is found that progression to multiple myeloma is lower in soft-tissue plasmacytomas compared to osseous plasmacytomas \[[@B2]\]. Patients may also develop local recurrence and distant skin recurrence. Patients with multiple lesions had a higher likelihood of progression to multiple myeloma and subsequent death \[[@B2]\]. From previous case reported, recurrent-free survival increases from 1.4 years to 11 years and overall survival rate increases from 4.3 years to 17 years in patients with solitary primary cutaneous plasmacytomas compared to those with multiple primary cutaneous plasmacytomas lesions \[[@B2]\]. Currently, the pathophysiology is unknown; however there are a few developing hypothesis. Of all up to date cases of primary cutaneous plasmacytoma, five have been described in which local triggering stimulus from trauma may have contributed to the development of plasma cell proliferation. One of these cases reported a primary cutaneous plasmacytoma localized to the lower lip, which had been affected with recurrent herpes simplex virus-1 for 15 years. It was postulated that chronic stimulation of keratinocytes by herpes simplex virus-1, possibly through toll-like receptors, may have favored the release of interleukin-6 able to induce plasma cell proliferation, transformation, and survival \[[@B6]\]. Interleukin-6 may also have some oncogenic effects via enhancement of C-MYC and is antiapoptotic \[[@B7]\]. In the second hypothesis, it is thought that plasma cells expressing CXCR4 cytokine receptor can interact with the inflammatory cascade released due to tissue injury, facilitating the migration of neoplastic plasma cells into the skin \[[@B8]\]. Genetically, extramedullary plasmacytomas share many features of other plasma cell disorders, such as recurrent losses in chromosome 13, chromosome arm 1p, and chromosome arm 14q, as well as gain in chromosome arms 19p, 9q, and 1q \[[@B9]\]. According to the International Society for Cutaneous lymphoma (ISCL) and the cutaneous lymphoma task force of the European Organization of Research and Treatment of Cancer (EORTC), TNM classification system (see the list below) had been proposed to primarily document disease extension and to a lesser extent for prognostic guide \[[@B10]\]. ISCL/EORTC Proposed TNM Classification of Cutaneous Lymphoma Other Than Mycosis Fungoides and Sezary Syndrome (T) ⁡(T1) Solitary skin involvement(T1a) A solitary lesion \<5 cm diameter(T1b) A solitary \>5 cm diameter(T2) Regional skin involvement of multiple lesions limited to 1 body region or 2 contiguous body regions(T2a) All-disease encompassing in a \<15 cm diameter circular area(T2b) All-disease encompassing in \>15 and \<30 cm diameter circular area(T2c) All-disease encompassing in a 30 cm diameter circular area(T3) Generalized skin involvement(T3a) Multiple lesions involving 2 noncontiguous body regions(T3b) Multiple lesions involving equal to or greater than 3 body regions(N) ⁡(N0) No clinical or pathologic lymph node involvement(N1) Involvement of 1 peripheral lymph node region that drains an area of current or prior skin involvement(N2) Involvement of 2 or more peripheral lymph node regions or involvement of any lymph node region that does not drain an area of current or prior skin involvement(N3) Involvement of central lymph nodes(M) ⁡(M0) No evidence of extracutaneous non-lymph node disease(M1) Extracutaneous non-lymph node disease present In order to stage the disease, patients should undergo the following testing: a complete history and physical examination, complete blood count with differential count, chemistry panel with LDH, flow cytometry on peripheral blood, and imaging with computed tomography scan of the chest, abdomen, and pelvis with contrast or positron emission-tomography/CT as it can visualize metabolically active disease in relatively small lesions \[[@B10]\]. According to this classification, our patient was stage T1aN0M0 disease. As for treatment management, clinicians are recommended to follow the WHO-EORTC cutaneous lymphoma classification to understand the clinical behavior and prognostic variables in the heterogeneous group of cutaneous lymphomas in order to provide optimal care. Due to its shortcomings, the proposed TNM system will be reexamined as more data are obtained. 4. Conclusion {#sec4} ============= With few cases in the literature reported, this case is an example of a primary cutaneous lesion in an unusual location with no systemic disease. Follow-up care includes continual monitoring for progression to multiple myeloma. This consists of M-protein measurements and complete blood counts at six-week intervals for the first 6 months. This rarer and nonspecific presentation of a primary extramedullary plasmacytoma is important to keep in mind when evaluating lip lesions. Competing Interests =================== The authors declare that they have no competing interests. ![(a) Dense plasma cell infiltrate, H&E, 40x; (b) CD138 immunostain labeling plasma cells, IHC 40x; (c) Kappa in situ hybridization with only rare plasma cell labeling, ISH, 40x; (d) Lambda in situ hybridization with most plasma cells labeling, ISH, 40x.](CRIONM2017-4296802.001){#fig1} [^1]: Academic Editor: Ossama W. Tawfik	{ "pile_set_name": "PubMed Central" }
Related literature {#sec1} ================== For the biological and pharmacological properties of benzopyrans and their derivatives, see Brooks (1998[@bb2]); Hatakeyama et al. (1988[@bb7]); Hyana & Saimoto (1987[@bb8]); Tang et al. (2007[@bb13]). For a detailed account of the importance of 4H-chromenes, see Liu et al. (2007[@bb9]); Wang, Fang et al. (2003[@bb14]); Wang, Zhang et al. (2003[@bb15]). For hydrogen-bonding interactions and motifs, see: Bernstein et al. (1995[@bb1]); Desiraju (1989[@bb4]); Desiraju & Steiner (1999[@bb5]); Etter (1990[@bb6]). Experimental {#sec2} ============ {#sec2.1} ### Crystal data {#sec2.1.1} C~21~H~18~O~4~M ~r~ = 334.35Orthorhombic,a = 6.299 (2) Åb = 15.798 (6) Åc = 17.429 (6) ÅV = 1734.3 (11) Å^3^Z = 4Mo Kα radiationμ = 0.09 mm^−1^T = 293 (2) K0.35 × 0.32 × 0.29 mm ### Data collection {#sec2.1.2} Bruker SMART APEX CCD diffractometerAbsorption correction: multi-scan (SADABS; Sheldrick, 1998[@bb10]) T ~min~ = 0.969, T ~max~ = 0.97513979 measured reflections2055 independent reflections1793 reflections with I \> 2σ(I)R ~int~ = 0.062 ### Refinement {#sec2.1.3} R\[F ^2^ \> 2σ(F ^2^)\] = 0.046wR(F ^2^) = 0.098S = 1.012055 reflections226 parametersH-atom parameters constrainedΔρ~max~ = 0.10 e Å^−3^Δρ~min~ = −0.10 e Å^−3^ {#d5e448} Data collection: SMART (Bruker, 2007[@bb3]); cell refinement: SMART; data reduction: SAINT (Bruker, 2007[@bb3]); program(s) used to solve structure: SHELXS97 (Sheldrick, 2008[@bb11]); program(s) used to refine structure: SHELXL97 (Sheldrick, 2008[@bb11]); molecular graphics: PLATON (Spek, 2003[@bb12]); software used to prepare material for publication: SHELXL97. Supplementary Material ====================== Crystal structure: contains datablocks I, global. DOI: [10.1107/S1600536808039743/fb2126sup1.cif](http://dx.doi.org/10.1107/S1600536808039743/fb2126sup1.cif) Structure factors: contains datablocks I. DOI: [10.1107/S1600536808039743/fb2126Isup2.hkl](http://dx.doi.org/10.1107/S1600536808039743/fb2126Isup2.hkl) Additional supplementary materials: [crystallographic information](http://scripts.iucr.org/cgi-bin/sendsupfiles?fb2126&file=fb2126sup0.html&mime=text/html); [3D view](http://scripts.iucr.org/cgi-bin/sendcif?fb2126sup1&Qmime=cif); [checkCIF report](http://scripts.iucr.org/cgi-bin/paper?fb2126&checkcif=yes) Supplementary data and figures for this paper are available from the IUCr electronic archives (Reference: [FB2126](http://scripts.iucr.org/cgi-bin/sendsup?fb2126)). AN thanks Dr S. Kannan and Dr B. S. Krishnamurthy, School of Chemistry, Bharathidasan University, Tiruchirappalli, and Organica Aromatics Pvt Ltd Bangalore, India, for providing laboratory facilities. Comment ======= Chromenes (benzopyrans) and their derivatives have numerous biological and pharmacological properties (Tang et al., 2007) such as antisterility (Brooks, 1998) and anticancer activity (Hyana & Saimoto, 1987). In addition, polyfunctional chromene units are present in numerous natural products (Hatakeyama et al., 1988). 4H-chromenes are important synthons for some natural products (Liu et al., 2007). As a part of our structural investigations on 4H-chromene derivatives, the single-crystal X-ray diffraction study on the title compound was carried out. The chromene ring is almost planar: The puckering amplitude of the chromene ring is 0.097 (3) Å. In the related chromene derivatives (Wang, Zhang et al., 2003; Wang, Fang et al., 2003), the chromene ring is also planar. In the title structure, the interplanar angle between the chromene ring and the 2-phenyl ring is 6.1 (1)° thereby indicating their almost coplanar arrangement (Fig. 1). The propenyloxy substituents at both C5 and C7 are coplanar with the chromene ring with the respective interplanar angles 1.7 (2)° and 8.8 (2)°. The crystal structure is stabilized by the interplay of C--H···O and C--H···π interactions (Fig. 2, Table 1; Desiraju & Steiner, 1999; Desiraju, 1989). Each of C15--H15A···O12, C19--H19A···O16 and C25--H25···O1 interactions are involved in the S(5) motifs (Bernstein et al., 1995; Etter, 1990). Experimental {#experimental} ============ A suspension of chrysin (3.93 mmol, 1.00 g) and potassium carbonate (11.81 mmol, 1.64 g) in dimethyl formamide (10 ml) were added into a round bottom flask. The reaction mixture was heated to 383 K for 2--3 h. The reaction mixture was then cooled to 333 K and allyl bromide (15.74 mmol, 1.90 g) was slowly added to the reaction mixture with the help of a dropping funnel. The reaction mixture was maintained for 8--9 h at 333 K and monitored by high pressure liquid chromatography (HPLC). After completion of the reaction, the content was quenched with water and stirred for 30--45 min at 303 K. T he obtained crude solid was filtered and washed with plenty of water followed by methanol and dried under vacuum at 343 K. The compound was purified by column chromatography using ethyl acetate/n-hexane (1:1) as eluent. All highly pure column fractions were concentrated in a rota evaporator. The dried compound was dissolved in dichloromethane/hexane (1:1) mixture (10 ml). The clear solution was kept for a week and the resulting needle shaped crystals of average size 0.3 mm were washed with n-hexane. The crystals were dried over high vacuum at 343--348 K. Yield: 90% Refinement {#refinement} ========== In the absence of significant anomalous scattering effects, 1488 Friedel pairs have been merged. All the H-atoms were observed in the difference electron density map. However, they were situated into idealized positions with C--H = 0.93 and 0.97 Å for aryl and methylene H, respectively, and constrained to ride on their parent atoms with U~iso~(H) = 1.2U~eq~(C). Figures ======= ![The title molecule showing the displacement ellipsoids depicted at the 50% probability level for all non-H atoms. The hydrogen atoms are drawn as spheres of arbitrary radius.](e-65-000o1-fig1){#Fap1} ![The molecular packing viewed down the c-axis. Dashed lines represent weak C--H···O interactions.](e-65-000o1-fig2){#Fap2} Crystal data {#tablewrapcrystaldatalong} ============ ------------------------------- -------------------------------------- C~21~H~18~O~4~ D~x~ = 1.281 Mg m^−3^ M~r~ = 334.35 Melting point = 434--437 K Orthorhombic, P2~1~2~1~2~1~ Mo Kα radiation, λ = 0.71073 Å Hall symbol: P 2ac 2ab Cell parameters from 576 reflections a = 6.299 (2) Å θ = 1.8--26.0° b = 15.798 (6) Å µ = 0.09 mm^−1^ c = 17.429 (6) Å T = 293 K V = 1734.3 (11) Å^3^ Needle, colourless Z = 4 0.35 × 0.32 × 0.29 mm F(000) = 704 ------------------------------- -------------------------------------- Data collection {#tablewrapdatacollectionlong} =============== --------------------------------------------------------------- -------------------------------------- Bruker SMART APEX CCD diffractometer 3543 independent reflections Radiation source: fine-focus sealed tube 1793 reflections with I \> 2σ(I) graphite R~int~ = 0.062 Detector resolution: 0.3 pixels mm^-1^ θ~max~ = 26.4°, θ~min~ = 1.7° ω scans h = −7→7 Absorption correction: multi-scan (SADABS; Sheldrick, 1998) k = −19→19 T~min~ = 0.969, T~max~ = 0.975 l = −21→20 13979 measured reflections --------------------------------------------------------------- -------------------------------------- Refinement {#tablewraprefinementdatalong} ========== ------------------------------------- ------------------------------------------------------------------------------------------------- Refinement on F^2^ Primary atom site location: structure-invariant direct methods Least-squares matrix: full Secondary atom site location: difference Fourier map R\[F^2^ \> 2σ(F^2^)\] = 0.046 Hydrogen site location: difference Fourier map wR(F^2^) = 0.098 H-atom parameters constrained S = 1.01 w = 1/\[σ^2^(F~o~^2^) + (0.0343P)^2^ + 0.1512P\] where P = (F~o~^2^ + 2F~c~^2^)/3 2055 reflections (Δ/σ)~max~ \< 0.001 226 parameters Δρ~max~ = 0.10 e Å^−3^ 0 restraints Δρ~min~ = −0.10 e Å^−3^ 72 constraints ------------------------------------- ------------------------------------------------------------------------------------------------- Special details {#specialdetails} =============== ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Geometry. All e.s.d.\'s (except the e.s.d. in the dihedral angle between two l.s. planes) are estimated using the full covariance matrix. The cell e.s.d.\'s are taken into account individually in the estimation of e.s.d.\'s in distances, angles and torsion angles; correlations between e.s.d.\'s in cell parameters are only used when they are defined by crystal symmetry. An approximate (isotropic) treatment of cell e.s.d.\'s is used for estimating e.s.d.\'s involving l.s. planes. Refinement. Refinement of F^2^ against ALL reflections. The weighted R-factor wR and goodness of fit S are based on F^2^, conventional R-factors R are based on F, with F set to zero for negative F^2^. The threshold expression of F^2^ \> σ(F^2^) is used only for calculating R-factors(gt) etc. and is not relevant to the choice of reflections for refinement. R-factors based on F^2^ are statistically about twice as large as those based on F, and R- factors based on ALL data will be even larger. ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Fractional atomic coordinates and isotropic or equivalent isotropic displacement parameters (Å^2^) {#tablewrapcoords} ================================================================================================== ------ ------------- -------------- --------------- -------------------- -- x y z U~iso~\/U~eq~ O1 0.4396 (4) 1.06971 (12) 0.13245 (11) 0.0650 (6) C2 0.5860 (6) 1.0832 (2) 0.0765 (2) 0.0639 (8) C3 0.5782 (6) 1.0403 (2) 0.0102 (2) 0.0745 (9) H3 0.6761 1.0541 −0.0278 0.089\ C4 0.4277 (6) 0.9746 (2) −0.0056 (2) 0.0696 (9) C5 0.0891 (6) 0.90891 (19) 0.04852 (18) 0.0608 (8) C6 −0.0584 (6) 0.90372 (18) 0.10630 (18) 0.0622 (8) H6 −0.1728 0.8668 0.1020 0.075\* C7 −0.0358 (6) 0.95419 (19) 0.17160 (18) 0.0590 (8) C8 0.1332 (5) 1.00863 (19) 0.17997 (18) 0.0610 (9) H8 0.1495 1.0414 0.2239 0.073\* C9 0.2778 (5) 1.01254 (18) 0.12026 (18) 0.0557 (8) C10 0.2641 (5) 0.96435 (18) 0.05336 (17) 0.0559 (8) O11 0.4392 (5) 0.93093 (17) −0.06419 (14) 0.1049 (9) O12 0.0781 (4) 0.86278 (13) −0.01724 (12) 0.0748 (6) C13 −0.0923 (6) 0.8053 (2) −0.02818 (18) 0.0778 (10) H13A −0.2267 0.8353 −0.0279 0.093\* H13B −0.0942 0.7635 0.0125 0.093\* C14 −0.0582 (8) 0.7632 (2) −0.1044 (2) 0.0860 (11) H14 −0.1579 0.7232 −0.1196 0.103\* C15 0.0989 (7) 0.7778 (2) −0.15084 (19) 0.0906 (12) H15A 0.2021 0.8173 −0.1378 0.109\* H15B 0.1080 0.7487 −0.1971 0.109\* O16 −0.1951 (4) 0.94477 (13) 0.22389 (13) 0.0729 (7) C17 −0.1985 (6) 1.0013 (2) 0.28825 (18) 0.0773 (11) H17A −0.1929 1.0594 0.2704 0.093\* H17B −0.0753 0.9912 0.3204 0.093\* C18 −0.3953 (7) 0.9877 (2) 0.3335 (2) 0.0815 (11) H18 −0.4206 1.0255 0.3733 0.098\* C19 −0.5312 (6) 0.9305 (3) 0.3237 (2) 0.1011 (13) H19A −0.5136 0.8910 0.2846 0.121\* H19B −0.6496 0.9277 0.3555 0.121\* C20 0.7460 (6) 1.1464 (2) 0.0988 (2) 0.0670 (9) C21 0.9252 (7) 1.1595 (2) 0.0537 (2) 0.0818 (11) H21 0.9425 1.1292 0.0084 0.098\* C22 1.0767 (7) 1.2170 (3) 0.0757 (3) 0.0980 (13) H22 1.1961 1.2253 0.0452 0.118\* C23 1.0540 (8) 1.2625 (3) 0.1423 (3) 0.1075 (15) H23 1.1572 1.3015 0.1568 0.129\* C24 0.8787 (8) 1.2503 (3) 0.1873 (3) 0.1096 (14) H24 0.8634 1.2806 0.2327 0.131\* C25 0.7248 (6) 1.1932 (2) 0.1657 (2) 0.0887 (11) H25 0.6052 1.1859 0.1962 0.106\* ------ ------------- -------------- --------------- -------------------- -- Atomic displacement parameters (Å^2^) {#tablewrapadps} ===================================== ----- ------------- ------------- ------------- -------------- -------------- -------------- U^11^ U^22^ U^33^ U^12^ U^13^ U^23^ O1 0.0626 (14) 0.0716 (14) 0.0608 (14) −0.0033 (13) 0.0010 (14) 0.0035 (11) C2 0.059 (2) 0.066 (2) 0.067 (2) 0.008 (2) 0.003 (2) 0.0157 (18) C3 0.071 (2) 0.084 (2) 0.069 (2) −0.004 (2) 0.016 (2) 0.007 (2) C4 0.067 (2) 0.079 (2) 0.063 (2) 0.003 (2) 0.006 (2) 0.0007 (19) C5 0.072 (2) 0.0608 (19) 0.050 (2) 0.007 (2) −0.003 (2) 0.0017 (16) C6 0.066 (2) 0.0620 (18) 0.058 (2) −0.0021 (18) 0.003 (2) 0.0036 (17) C7 0.062 (2) 0.0673 (19) 0.048 (2) 0.0043 (19) 0.0072 (19) 0.0059 (17) C8 0.067 (2) 0.0646 (19) 0.0509 (19) 0.0001 (18) 0.001 (2) 0.0015 (16) C9 0.054 (2) 0.0572 (17) 0.056 (2) 0.0028 (18) −0.0023 (19) 0.0097 (16) C10 0.059 (2) 0.0577 (17) 0.051 (2) 0.0056 (19) 0.0027 (18) 0.0042 (15) O11 0.107 (2) 0.128 (2) 0.0792 (18) −0.029 (2) 0.0327 (17) −0.0308 (16) O12 0.0831 (16) 0.0798 (14) 0.0614 (15) −0.0099 (15) 0.0049 (14) −0.0105 (13) C13 0.086 (3) 0.076 (2) 0.071 (2) −0.013 (2) 0.002 (2) −0.0011 (19) C14 0.110 (3) 0.078 (2) 0.070 (3) −0.008 (3) −0.013 (3) −0.010 (2) C15 0.118 (3) 0.094 (3) 0.060 (2) 0.011 (3) −0.004 (3) −0.007 (2) O16 0.0753 (17) 0.0844 (14) 0.0591 (14) −0.0090 (13) 0.0117 (13) −0.0065 (12) C17 0.088 (3) 0.088 (2) 0.057 (2) −0.004 (2) 0.010 (2) −0.0027 (19) C18 0.092 (3) 0.092 (3) 0.060 (2) 0.012 (3) 0.000 (3) 0.003 (2) C19 0.082 (3) 0.125 (3) 0.096 (3) 0.003 (3) 0.010 (3) −0.010 (3) C20 0.061 (2) 0.0622 (19) 0.078 (3) 0.002 (2) −0.004 (2) 0.0161 (19) C21 0.070 (3) 0.084 (2) 0.092 (3) 0.000 (2) −0.002 (3) 0.028 (2) C22 0.072 (3) 0.095 (3) 0.127 (4) −0.012 (3) −0.002 (3) 0.044 (3) C23 0.092 (3) 0.083 (3) 0.147 (5) −0.021 (3) −0.015 (4) 0.012 (3) C24 0.096 (3) 0.100 (3) 0.132 (4) −0.022 (3) 0.000 (3) −0.019 (3) C25 0.077 (3) 0.082 (2) 0.107 (3) −0.010 (2) 0.005 (3) −0.010 (2) ----- ------------- ------------- ------------- -------------- -------------- -------------- Geometric parameters (Å, °) {#tablewrapgeomlong} =========================== --------------------- ------------ ----------------------- ------------ O1---C2 1.360 (4) C14---H14 0.9300 O1---C9 1.378 (3) C15---H15A 0.9300 C2---C3 1.340 (4) C15---H15B 0.9300 C2---C20 1.471 (5) O16---C17 1.434 (3) C3---C4 1.433 (5) C17---C18 1.485 (5) C3---H3 0.9300 C17---H17A 0.9700 C4---O11 1.233 (3) C17---H17B 0.9700 C4---C10 1.465 (4) C18---C19 1.256 (4) C5---O12 1.360 (3) C18---H18 0.9300 C5---C6 1.373 (4) C19---H19A 0.9300 C5---C10 1.410 (4) C19---H19B 0.9300 C6---C7 1.397 (4) C20---C25 1.387 (4) C6---H6 0.9300 C20---C21 1.391 (5) C7---O16 1.364 (4) C21---C22 1.373 (5) C7---C8 1.376 (4) C21---H21 0.9300 C8---C9 1.384 (4) C22---C23 1.373 (5) C8---H8 0.9300 C22---H22 0.9300 C9---C10 1.395 (4) C23---C24 1.368 (6) O12---C13 1.419 (4) C23---H23 0.9300 C13---C14 1.501 (4) C24---C25 1.377 (5) C13---H13A 0.9700 C24---H24 0.9300 C13---H13B 0.9700 C25---H25 0.9300 C14---C15 1.299 (5) C2---O1---C9 119.6 (3) C15---C14---H14 117.2 C3---C2---O1 121.0 (3) C13---C14---H14 117.2 C3---C2---C20 126.6 (4) C14---C15---H15A 120.0 O1---C2---C20 112.4 (3) C14---C15---H15B 120.0 C2---C3---C4 123.8 (3) H15A---C15---H15B 120.0 C2---C3---H3 118.1 C7---O16---C17 117.8 (3) C4---C3---H3 118.1 O16---C17---C18 109.7 (3) O11---C4---C3 121.7 (3) O16---C17---H17A 109.7 O11---C4---C10 124.1 (3) C18---C17---H17A 109.7 C3---C4---C10 114.2 (3) O16---C17---H17B 109.7 O12---C5---C6 123.5 (3) C18---C17---H17B 109.7 O12---C5---C10 115.0 (3) H17A---C17---H17B 108.2 C6---C5---C10 121.5 (3) C19---C18---C17 126.9 (4) C5---C6---C7 119.6 (3) C19---C18---H18 116.5 C5---C6---H6 120.2 C17---C18---H18 116.5 C7---C6---H6 120.2 C18---C19---H19A 120.0 O16---C7---C8 124.5 (3) C18---C19---H19B 120.0 O16---C7---C6 114.0 (3) H19A---C19---H19B 120.0 C8---C7---C6 121.5 (3) C25---C20---C21 118.3 (4) C7---C8---C9 117.2 (3) C25---C20---C2 121.2 (3) C7---C8---H8 121.4 C21---C20---C2 120.5 (3) C9---C8---H8 121.4 C22---C21---C20 120.3 (4) O1---C9---C8 113.6 (3) C22---C21---H21 119.8 O1---C9---C10 122.1 (3) C20---C21---H21 119.8 C8---C9---C10 124.3 (3) C21---C22---C23 120.7 (4) C9---C10---C5 115.9 (3) C21---C22---H22 119.7 C9---C10---C4 118.9 (3) C23---C22---H22 119.7 C5---C10---C4 125.2 (3) C24---C23---C22 119.7 (4) C5---O12---C13 119.6 (3) C24---C23---H23 120.2 O12---C13---C14 107.1 (3) C22---C23---H23 120.2 O12---C13---H13A 110.3 C23---C24---C25 120.2 (4) C14---C13---H13A 110.3 C23---C24---H24 119.9 O12---C13---H13B 110.3 C25---C24---H24 119.9 C14---C13---H13B 110.3 C24---C25---C20 120.8 (4) H13A---C13---H13B 108.5 C24---C25---H25 119.6 C15---C14---C13 125.6 (4) C20---C25---H25 119.6 C9---O1---C2---C3 −1.5 (4) O11---C4---C10---C9 174.5 (3) C9---O1---C2---C20 179.6 (2) C3---C4---C10---C9 −5.1 (4) O1---C2---C3---C4 −4.0 (5) O11---C4---C10---C5 −5.6 (5) C20---C2---C3---C4 174.8 (3) C3---C4---C10---C5 174.8 (3) C2---C3---C4---O11 −172.6 (3) C6---C5---O12---C13 −0.4 (4) C2---C3---C4---C10 7.1 (5) C10---C5---O12---C13 −179.7 (3) O12---C5---C6---C7 −179.3 (3) C5---O12---C13---C14 −178.8 (3) C10---C5---C6---C7 −0.1 (5) O12---C13---C14---C15 −0.3 (5) C5---C6---C7---O16 178.4 (3) C8---C7---O16---C17 6.5 (4) C5---C6---C7---C8 −0.8 (5) C6---C7---O16---C17 −172.7 (3) O16---C7---C8---C9 −177.9 (3) C7---O16---C17---C18 173.6 (3) C6---C7---C8---C9 1.2 (4) O16---C17---C18---C19 6.9 (5) C2---O1---C9---C8 −176.4 (2) C3---C2---C20---C25 173.0 (3) C2---O1---C9---C10 3.2 (4) O1---C2---C20---C25 −8.1 (4) C7---C8---C9---O1 178.8 (2) C3---C2---C20---C21 −8.2 (5) C7---C8---C9---C10 −0.9 (4) O1---C2---C20---C21 170.7 (3) O1---C9---C10---C5 −179.5 (2) C25---C20---C21---C22 0.4 (5) C8---C9---C10---C5 0.1 (4) C2---C20---C21---C22 −178.4 (3) O1---C9---C10---C4 0.3 (4) C20---C21---C22---C23 −0.1 (5) C8---C9---C10---C4 179.9 (3) C21---C22---C23---C24 0.2 (6) O12---C5---C10---C9 179.7 (3) C22---C23---C24---C25 −0.6 (7) C6---C5---C10---C9 0.4 (4) C23---C24---C25---C20 1.0 (6) O12---C5---C10---C4 −0.1 (4) C21---C20---C25---C24 −0.9 (5) C6---C5---C10---C4 −179.4 (3) C2---C20---C25---C24 178.0 (3) --------------------- ------------ ----------------------- ------------ Hydrogen-bond geometry (Å, °) {#tablewraphbondslong} ============================= --------------------- --------- --------- ----------- --------------- D---H···A D---H H···A D···A D---H···A C15---H15A···O12 0.93 2.35 2.691 (4) 101 C19---H19A···O16 0.93 2.42 2.749 (5) 101 C25---H25···O1 0.93 2.39 2.714 (4) 101 C17---H17B···O11^i^ 0.97 2.51 3.229 (4) 131 C14---H14···Cg1^ii^ 0.93 3.22 4.081 154 --------------------- --------- --------- ----------- --------------- Symmetry codes: (i) −x+1/2, −y+2, z+1/2; (ii) −x−1, y+3/2, −z+1/2. ###### Hydrogen-bond geometry (Å, °) D---H⋯A D---H H⋯A D⋯A D---H⋯A --------------------- --------- ------- ----------- ------------- C17---H17B⋯O11^i^ 0.97 2.51 3.229 (4) 131 Symmetry codes: (i) .	{ "pile_set_name": "PubMed Central" }
INTRODUCTION {#s1} ============ Regulation of reactive oxygen species (ROS) homeostasis plays key roles in living organisms \[[@R1], [@R2]\]. A moderate increase in ROS promote cell proliferation and differentiation, while the excessive amounts of ROS causes oxidative damage to lipids, proteins and DNA and induces cells undergoing apoptosis \[[@R3]\]. Cancer cells are characterized by persistently higher levels of ROS than non-transformed cells due to increased metabolic activity and the dysregulation of redox balance, which renders cancer cells more vulnerable to oxidative stress and a high dependency on antioxidant enzymes to detoxify from ROS \[[@R4]\]. As such, modulation of ROS homeostasis, via an increase in ROS levels or impairing antioxidant capacity, has been considered as an important strategy for cancer therapy \[[@R5]--[@R7]\]. Superoxide dismutase 1 (SOD1), which is involved in the conversion of toxic superoxide anions into molecular oxygen and hydrogen peroxide, is an important member in the intracellular ROS-scavenging system \[[@R8]\]. Active, mature SOD1 is a homodimeric protein containing two zinc (Zn^2+^) and two copper (Cu^2+^) ions for its stability and activity. The association with the copper chaperone for SOD (CCS) is essential for the activation of copper/zinc SOD, although an additional minor CCS-independent pathway has been reported in mammals \[[@R8]\]. CCS specifically delivers Cu to SOD1, which allows the formation of an intrasubunit disulfide bond between SOD1^Cys-57^ and SOD1^Cys-146^, and results in an enzymatically active homodimers of SOD \[[@R9], [@R10]\]. Thus far, CCS binding remains the most dominant mechanism for the regulation of the enzymatic activity of SOD1. Apart from CCS association, increasing evidence has indicated that diverse post-translational modifications, including nitration \[[@R11]\], phosphorylation \[[@R12]\], glutathionylaion \[[@R13]\] and glycation \[[@R14]\], are involved in the regulation of the dismutase activity of SOD1. Post-translational modifications have emerged as an important aspect in fine-tuning the signal process of SOD1 involved redox homeostasis. In the meanwhile, we have noticed that recent global proteomic profiling has identified lysine acetylation as a frequently occurred modification for cytoplasmic proteins, in particular metabolic enzymes including SOD1 \[[@R15]--[@R17]\], but the cellular functions of these modifications are still unknown. This study started from the validation of occurrence of SOD1 acetylation in cancer cells, and focused on the investigation of the biological significance of SOD1 acetylation. Our findings provided first evidence revealing the role of acetylation in modulating the SOD1 activity. The study highlighted a SOD acetylation mediated positive feedback loop in strengthening oxidative stress caused by genotoxic anticancer agents, and suggested the translational value of SOD1 acetylation for camptothecin-based chemotherapy. RESULTS {#s2} ======= SOD1 is acetylated at lysine 71 {#s2_1} ------------------------------- A number of mass spectrometry-based proteomic studies have suggested the occurrence of acetylation on SOD1 \[[@R15]--[@R17]\], but there lacks evidence to support acetylation of endogenous SOD1, and the biological significance of this modification remains unclear. We firstly validated the acetylation of SOD1 using a pan-specific anti-acetylated lysine antibody in cancer cells with ectopically expressed SOD1. Acetylation was detected on flag-tagged SOD1 enriched from HCT116 colon cancer cells. Treatment of protein deacetylase inhibitors, namely nicotinamide (NAM) and Trichostatin A (TSA), resulted in an increase in the acetylation of SOD1 (Figure [1A](#F1){ref-type="fig"}). ![SOD1 is acetylated on K71 residue\ A. Acetylation of exogenous SOD1. Ectopically expressed SOD1 was immunopreciptated and acetylation of SOD1 was examined using a pan-acetyl-lysine antibody. Cells were pretreated with TSA (500 nM) or/and NAM (10 mM) for 12 hr prior to immunoprecipitation; B., C. Impact of lysine mutant on SOD1 acetylation. B., wildtype (WT) SOD1 or indicated lysine mutants was transfected into cells and acetylation of each mutant was examined as described in A.. EV, empty vector; C., Acetylation of ectopically expressed SOD1WT, SOD1K71R, SOD1K71Q with or without TSA (500 nM) and NAM (10 mM) treatment was analyzed; D. Specificity of anti-acetyl-K71 antibody. Flag-tagged SOD1 WT or K71R mutant was transfected into HCT116 cells and acetylation of each purified protein was measured by immunoblotting using an anti-Ac SOD1 (K71) antibody; E. Endogenous SOD1 is acetylated at K71. Cells treated with NAM and TSA was analyzed by anti-Ac SOD1 (K71) antibody.](oncotarget-06-20578-g001){#F1} We next determined the main lysine sites where the acetylation occurred. SOD1 contains 11 lysine (K) residues, which are K4, K10, K24, K31, K37, K71, K76, K92, K123, K129 and K137. As lysine lysine (K)-arginine (R) replacement is widely used to generate acetylation-deﬁcient mutants \[[@R18]--[@R20]\], each of the lysine was individually mutated to a nonacetylatable arginine, and the impact on SOD1 acetylation was examined. Among the 11 mutants, only the K71R mutation largely abolished SOD1 acetylation (Figure [1B](#F1){ref-type="fig"}) and the treatment of deacetylase inhibitors failed to increase the detectable signaling of acetylation (Figure [1C](#F1){ref-type="fig"}), indicating the acetylation of SOD1 occurred at K71. Meanwhile, alignment of SOD1 protein sequence revealed that K71 was evolutionarily conserved across diverse species ([Supplemental Figure S1](#SD1){ref-type="supplementary-material"}). Further, we generated an antibody that specifically recognized SOD1 bearing acetylation at K71. The antibody was able to detect the acetylation of ectopically expressed wild-type SOD1 but not the K71R mutant (Figure [1D](#F1){ref-type="fig"}), demonstrating the selectivity of the antibody. This antibody enabled the first detection of the acetylation of endogenous SOD1, which was significantly increased by treatment of deacetylase inhibitors NAM plus TSA (Figure [1E](#F1){ref-type="fig"}). Meanwhile, SOD1 depletion by two independent siRNAs significantly decreased the SOD1 acetylation enriched by deacetylase inhibitors, further supporting the specificity of this antibody ([Supplemental Figure S2](#SD1){ref-type="supplementary-material"}). These results together demonstrated the acetylation of SOD1 at K71, which intrigued us to explore the biological significance of SOD1 acetylation. Acetylation inactivates the dismutase activity of SOD1 {#s2_2} ------------------------------------------------------ We asked whether acetylation of SOD1 affected its enzymatic activity. The dismutase enzymatic activity of SOD1 was measured using a specific in-gel enzymatic activity assay using the native polyacrylamide gel electrophoresis. Treatment with deacetylase inhibitors NAM or TSA, similar to SOD1 inhibitor DDTC, resulted in the reduction of SOD1 activity while the SOD1 protein level was not affected in parallel (Figure [2A](#F2){ref-type="fig"}), suggesting that acetylation of SOD1 negatively regulates the SOD1 activity. For further confirmation, we compared the enzymatic activity of wild type SOD1, K71R mutant and acetylation mimetic K71Q mutant. Flag-tagged wild type or mutant constructs was transfected into HCT-116 cells, and the enzymatic activity of endogenous and exogenous SOD1 was differentiated by their diverse migration in the native polyacrylamide gel electrophoresis. K71R mutant behaved similar to wildtype SOD1 in the activity assay, whereas the K71Q mutant showed a significant decrease in the catalytic activity (Figure [2B](#F2){ref-type="fig"}). These results suggested acetylated SOD1 as an inactive form of SOD1. ![SOD1 acetylation inactivates SOD1\ A. SOD1 activity is decreased by NAM/TSA treatment. Cells were treated by TSA (500 nM) and NAM (10 mM) for 12 hr and SOD1 activity was determined using an in-gel SOD1 activity assay; B. Mutations of the acetylation lysine residues affect SOD1 activity. Flag-tagged SOD1 WT, K71R or K71Q mutant was transfected into HCT116 cells and SOD1 activity was measured as in A.. endo, endogenous SOD1, exo, Flag-tagged SOD1; C. Sufficient supply of copper rescues the activity of K711Q mutant; D. The interaction between SOD1 and CCS is decreased by NAM/TSA treatment. HCT-116 cells transfected with Flag-tagged SOD1 were treated with NAM/TSA for 12 hr. The presence of CCS in the immunoprecipitated protein complexes was assessed by immunoblotting; E. Acetylation mimetic mutation at K71 decreased the interaction between SOD1 and CCS; F. The level of SOD1 homodimers is reduced under NAM/TSA treatment. HCT116 cells were treated with NAM/TSA for 12 hr. Cells lysis in the presence or absence of DMP (20 mM) were assessed by immunoblotting using SOD1 antibody; G. Acetylation mimetic mutation at K71 decreased the level of SOD1 homodimers. HCT116 cells stably knock-down SOD1 (shSOD1) were transfected with Flag-tagged SOD1 WT, SOD1K71R or SOD1K71Q. SOD1 homodimers were detected as described in C..](oncotarget-06-20578-g002){#F2} Acetylation of SOD1 disrupts its interaction with CCS {#s2_3} ----------------------------------------------------- We then asked how acetylation affected the SOD1 activity. To address this question, we inspected the multi-step process of SOD1 maturation, which involves zinc binding, copper loading by CCS, and homodimerization prior to turning into an active homodimeric enzyme. We firstly examined whether the impaired SOD1 activity was due to the impaired zinc or/and copper loading, which initiates the process of SOD1 maturation. To this end, the acetylation mimetic K71Q mutant was incubated with increasing amount of zinc or copper to examine whether the deficient SOD1 activity could be rescued by sufficient zinc/copper supplies. Indeed, we observed that copper incubation instead of zinc incubation was able to reverse the enzymatic activity of K71Q mutant to the similar level of wildtype SOD1 (Figure [2C](#F2){ref-type="fig"}). This data largely excluded the possibility of impaired zinc loading of the K71Q mutant, and led us to speculate that acetylation of SOD1 probably affected its interaction with CCS, a SOD1 binding partner specifically responsible for copper delivery. As such, flag-tagged SOD1 was transfected into HCT-116 cells and the interaction between SOD1 and CCS were assessed using co-immunoprecipitation assay. It was found that treatment with NAM and TSA, which effectively enriched cellular SOD1 acetylation, largely disrupted the interaction between SOD1 and CCS (Figure [2D](#F2){ref-type="fig"}). Further, we compared CCS binding ability between SOD1 wildtype and the two acetylation relevant mutants. The interaction between CCS and the acetylation defective K71R mutant was enhanced compared with that of wildtype SOD1, whereas K71Q mutant exhibited decreased association with CCS (Figure [2E](#F2){ref-type="fig"}). CCS binding and it mediated copper loading are required for the subsequent homodimerization and ultimate activation of SOD1. Therefore, we also examined the impact of acetylation on SOD1 homodimerization. The dimerized SOD1 was visualized by the treatment with cross-linking reagent dimethyl pimelimidate (DMP) prior to denatured polyacrylamide gel electrophoresis. SOD1 dimmers were recognized as a subset with shifted molecular weight. In agreement with our results above, deacetylase inhibition using NAM and TSA decreased the proportion of endogenous SOD1 homodimers (Figure [2F](#F2){ref-type="fig"}). Further, SOD1 wildtype, K71R or K71Q mutant was transfected into HCT-116 cells, where endogenous SOD1 was depleted using short hairpin (shRNA) to eliminate the interference of endogenous band for recognizing ectopically expressed proteins. Compared with the wildtype SOD1, acetylation mimetic mutant SOD1 K71Q showed a considerable decrease in the proportion of homodimers, whereas K71R barely effected SOD1 dimmers formation (Figure [2G](#F2){ref-type="fig"}). Together, we concluded that SOD1 acetylation at K71 disrupted the interaction between SOD1 and CCS, which impaired formation of SOD1 homodimers, and in turn attenuated the enzymatic activity of SOD1. SIRT1 deacetylates SOD1 acetylation {#s2_4} ----------------------------------- Protein acetylation is critically regulated by deacetylases, which are often dysregulated in cancer cells and result in aberrant acetylation status in cancer cells \[[@R21], [@R22]\]. To identify responsible deacetylases may help understand the physiological significance of SOD1 acetylation. HDACs and sirtuins represent the two major classes of protein deacetylases and can be respectively inhibited by the pan-inhibitor TSA and NAM. According to our results in Figure [1E](#F1){ref-type="fig"}, NAM treatment considerably increased the level of SOD1 K71 acetylation whereas TSA treatment only had a marginal effect, suggesting that sirtuin family members are primarily involved in SOD1 deacetylation. To further identify the involved sirtuin(s), endogenous sirtuin 1 (SIRT1) to 7 was individually depleted using specific siRNAs followed by the detection of SOD1 acetylation. Among the 7 sirtuins, only SIRT1 depletion dramatically increased endogenous level of SOD1 acetylation at K71 (Figure [3A](#F3){ref-type="fig"} and [Supplemental Figure S3](#SD1){ref-type="supplementary-material"}). SIRT1 depletion enriched the acetylation of the wildtype SOD1 but not that of the K71R mutant (Figure [3B](#F3){ref-type="fig"}), suggesting a dominant role of SIRT1 in the regulation of SOD1 acetylation at K71. Further, the enhanced SOD1 acetylation by SIRT1 knockdown was reversed by the ectopic expression of wildtype SIRT1 but not the catalytically inactive H363Y mutant. The alteration in the SOD1 acetylation was associated with the interaction change between SOD1 and CCS (Figure [3C](#F3){ref-type="fig"}). The impact of SIRT1 on SOD1 acetylation was also reflected by the enzymatic activity of SOD1. Reconstitution of wildtype SIRT1 in SIRT1 stably depleted cells partially rescued the suppressed SOD1 activity by SIRT1 downregulation, but the effect of H363Y mutant was indiscernible (Figure [3D](#F3){ref-type="fig"}). In support of all these observations, we detected the interaction between SOD1 and SIRT1, as shown by the detection of SIRT1 in the immunocomplex pull-downed by Flag-tagged SOD in HCT-116 cells (Figure [3E](#F3){ref-type="fig"}). We also tested whether the SOD1 and SIRT1 was able to interact endogenously. Immunoprecipitation of endogenous SOD1 using an anti-SOD1 antibody revealed the interaction of SOD1 with endogenous SIRT1 in HCT-116 cells (Figure [3F](#F3){ref-type="fig"}). ![SIRT1 deacetylates SOD1\ A. SIRT1 knockdown increased SOD1 acetylation. Cells were transfected with siRNA against SIRT1 to 7 and SOD1 acetylation was detected by anti-Ac SOD1 (K71) antibody; The Ac-SOD1 band was indicated by the solid triangle. B. SIRT1 deacetylated SOD1 acetylation on K71. HCT116 ShSIRT1 cells were transfected with flag-tagged SOD1 wildtype (WT) or K71R mutant. SOD1 acetylation was detected by using a pan-acetyl-lysine antibody following anti-Flag immunoprecipitation; C. Catalytic activity of SIRT1 is required for the deacetylation of SOD1. HA-tagged SOD1 was co-transfected with Flag-tagged SIRT1 wildtype (WT) or catalytically inactive mutant H363Y into SIRT1-depleted HCT-116 cells. Acetylation was detected as in B., D. Catalytic activity of SIRT1 is required for the regulation of SOD1 activity. Endogenous SOD1 activity was measured in shSIRT1 cells reconstituted with SIRT1 WT or H363Y mutant; E., F. SOD1 interacts with SIRT1. E, HA-tagged SOD1 was co-transfected with Flag-tagged SIRT1 into HCT-116 cells. Co-immunoprecipitation was performed using anti-Flag agarose. The presence of SOD1 in the immunoprecipitate was assessed by immunoblotting. F, interaction of endogenous SOD1 and SIRT1. Endogenous SOD1 was immunoprecipitated with a rabbit polyclonal antibody against SOD1. Rabbit Immunoglobulin G (IgG) was used as a negative control. The presence of SIRT1 in the immunoprecipitate was assessed by immunoblotting with an antibody against SIRT1.](oncotarget-06-20578-g003){#F3} These findings collectively indicated that SIRT1 deacetylates SOD1 at K71, which promotes its interaction with CCS, and enhances the enzymatic activity of SOD1. Genotoxic agents induce SOD1 acetylation via ROS generation {#s2_5} ----------------------------------------------------------- SIRT1 is essentially involved in coping with various stress including oxidative stress, while SOD1 plays a key role in scavenging cellular ROS \[[@R23]\], a natural byproduct of the normal oxygen metabolism but dramatically increased in environmental stress such as chemotherapy. Indeed, mounting evidence has suggested that cytotoxic anticancer agents induced oxidative stress contribute to the anticancer efficacy of these agents \[[@R24]--[@R29]\]. These information together implicate a possible involvement of SOD1 acetylation in cytotoxic agents caused oxidative stress. We then treated HCT-116 cells with various genotoxic anticancer agents including cisplatin (CDDP), camptothecin (CPT), etoposide (VP16) and bleomycin (BLM) to test the possible impact on SOD1 acetylation. Treatment with these agents all considerably increased SOD1 acetylation (Figure [4A](#F4){ref-type="fig"}). Interestingly, the impact on the increase of SOD1 acetylation was correlated with the level of ROS accumulation caused by these agents ([Supplemental Figure S4A](#SD1){ref-type="supplementary-material"} and [S4B](#SD1){ref-type="supplementary-material"}). Moreover, pretreatment with ROS scavenger N-acetylcysteine (NAC) or inhibition of NADPH oxidase using apocynin (APO), apparently reversed DNA damaging agents induced SOD1 acetylation (Figure [4B](#F4){ref-type="fig"}), implicating that genotoxic stress associated ROS generation accounted for the enhanced SOD1 acetylation. ![Genotoxic agents promote SOD1 acetylation and inactivate SOD1\ A., Genotoxic agent induces SOD1 acetylation. HCT-116 were treated with camptothecin (CPT, 100 nM), etoposide(VP-16, 10 μM), bleomycin (BLM, 10 μM) or cisplatin (CDDP, 20 μM) for 12 hr and SOD1 acetylation was examined using an anti-Ac SOD1 (K71) antibody; B. CPT-induced SOD1 acetylation is mediated by ROS. Cells were pretreated with or without NAC or APO for 2 hr before exposure to CPT (100 nM, 12 hr); C., D. The impact of CPT treatment on SOD1 acetylation and enzymatic activity. Time-dependent C. and dose-dependent D. induction of SOD1 acetylation and corresponding alteration in SOD1 activity was measured using immunoblotting or an in-gel enzymatic assay; E. The impact of CPT treatment on the interaction between SOD1 and CCS. Flag-tagged SOD1 WT, K71R, K71Q was transfected into HCT116 cells treated with or without 100 nM CPT for 12 hr. The presence of CCS in the immunoprecipitated protein complexes was assessed by immunoblotting. F. The impact of CPT on the formation of SOD1 homodimers. Flag-tagged SOD1 WT, K71R, K71Q was transfected into HCT116 cells treated with or without 100 nM CPT for 12 hr; G. The impact of CPT on the interaction between SOD1 and SIRT1. HA-tagged SOD1 were co-transfected with Flag-tagged SIRT1 into HCT-116 cells; H. The impact of SOD1 acetylation on cytosolic superoxide levels under CPT stimulation. HCT116 cells stably knock-down SOD1 (shSOD1) were transfected with Flag-tagged SOD1 (WT), SOD1K71R or SOD1K71Q, and cells were treated with CPT. The cytosolic superoxide level was measured by DHE staining and FACS analysis. Data represent mean ± SEM (n = 3), \\P \< 0.01. I. SOD1 Acetylation promotes cells sensitive to CPT treatment. The sensitivity of cells HCT116 shSOD1 cells expressing the wildtype, K71R or K71Q SOD1 were treated with CPT (1 nM) for 14 days and measured by a cologeneic assay.](oncotarget-06-20578-g004){#F4} We then used CPT as a representative to follow up the impact of DNA damage on SOD1 acetylation. CPT caused DNA damage, as reflected by p53 upregulation, induced SOD1 acetylation in a dose- and time-dependent manner. The induced increase of SOD1 acetylation was closely associated with a decline of SOD1 activity (Figure [4C](#F4){ref-type="fig"} and [4D](#F4){ref-type="fig"}), and the interaction with CCS ([Supplemental Figure S5A](#SD1){ref-type="supplementary-material"} and [S5B](#SD1){ref-type="supplementary-material"}). Consistently, CPT treatment disrupted the interaction between SOD1 and CCS but failed to affect the CCS binding to K71R mutant, suggesting an acetylation-dependent impact on CCS binding (Figure [4E](#F4){ref-type="fig"}). In line with these findings, CPT treatment suppressed dimerization of wildtype SOD1 but did not affect either K71R or K71Q mutant (Figure [4F](#F4){ref-type="fig"}, [Supplemental Figure S6A](#SD1){ref-type="supplementary-material"} and [S6AB](#SD1){ref-type="supplementary-material"}). Importantly, we also observed that treatment with CPT largely disrupted the interaction between SOD1 and SIRT1 (Figure [4G](#F4){ref-type="fig"}), which may explain the enhanced SOD1 acetylation upon DNA damage. Together, these data provided the first evidence showing that chemotherapies induce SOD1 acetylation and impair its enzymatic activity in cancer cells, which may result from disrupted interaction with SIRT1. SOD1 is essential for maintaining the redox homeostasis of cancer cells. CPT treatment increased ROS generation in cancer cells ([Supplemental Figure S4B](#SD1){ref-type="supplementary-material"}) whilst suppressed SOD1 activity (Figure [4C](#F4){ref-type="fig"} and [4D](#F4){ref-type="fig"}). This may suggest the further enhanced oxidative stress caused by SOD1 inactivation. We hence examined the impact of acetylation of SOD1 on cytosolic ROS level in HCT-116 cells, where the endogenous SOD1 was depleted using shRNA. shRNA-resistant SOD1 wild-type, K71R and K71Q SOD1 were transfected into HCT-116 cells respectively followed by the treatment of CPT. As expected, SOD1 depletion led to a dramatic increase of intracellular ROS level. Ectopic expression of wildtype SOD1 was able to rescue the ROS generation resulted from SOD1 depletion. In contrast, the acetylation-mimetic K71Q mutant, with impaired enzymatic activity, failed to reverse the intracellular ROS level (Figure [4H](#F4){ref-type="fig"}). However, it appeared puzzling that the nonacetylatable K71R mutant behaved similar to the wildtype SOD1 in ROS clearance upon CPT treatment. We speculated that this was likely due to the small proportion of acetylated form induced by CPT among total amount of exogenous SOD1. Indeed, cell sensitivity to CPT treatment measured by a cologeneic assay was consistent with results of the ROS clearance. Cells expressing acetylation-mimetic SOD1 K71Q mutant were more sensitive than either wild-type SOD1 and K71R mutant to CPT treatment (Figure [4I](#F4){ref-type="fig"}). SOD1 acetylation sensitizes cancer cells to DNA damaging agents {#s2_6} --------------------------------------------------------------- The substantial impact of K71Q mutant shown above suggests a possibility that the abundance of SOD1 acetylation may be a determinant of the sensitivity to the CPT-based chemotherapies, which are used in the clinical therapy of various types of human cancers including the first line treatment for colon cancer. We then probed the status of SOD1 acetylation at K71 in a panel of colon cancer cells, and found that the basal level of SOD1 acetylation varied largely across the cells (Figure [5A](#F5){ref-type="fig"}). Some cells lines, like HCT-8 and HCT-16, displayed massive abundance of intrinsic SOD1 acetylation. These data suggest that SOD1 acetylation status may confer a distinct antioxidant capacity across cancer cells, and those with low capacity may be more susceptible to CPT-induced oxidant stress. Indeed, we found that cells with high SOD1 acetylation were relatively more sensitive to CPT treatment. We also tested whether Ac-SOD1 level alteration in responses to CPT was correlated with the CPT sensitivity of these cells as well. Ac-SOD1 level was examined after 12 hr exposure to CPT treatment in the colon cancer cell lines ([Supplemental Figure S7](#SD1){ref-type="supplementary-material"}). It was found that cell with higher basal level of Ac-SOD1 showed more significant increase of Ac-SOD1 level upon CPT treatment, suggesting a correlation between Ac-SOD1 level change and the response to CPT treatment. ![SOD1 acetylation is associated with the response to CPT treatment\ A. SOD1 acetylation and cell sensitivity to CPT-11 in colon cancer cell lines. B., C. Intratumoral SOD1 acetylation level and response to CPT-11 treatment of patient-derived xenograft models. B. Growth curves of tumors from indicated PDX models treated with vehicle control or 10 mg/kg CPT-11 daily for 21 consecutive days. Data represent mean tumor volume ± SEM (n = 6), \\\*P \< 0.001; \*P \< 0.05; n.s., not significant. C. Correlation between SOD1 K71 acetylation and sensitivity to CPT-11. T/C%, relative tumor volume versus vehicle control on day 21.](oncotarget-06-20578-g005){#F5} Apart from colon cancer, we also tested whether basal Ac-SOD1 levels were correlated with the sensitivity to CPT treatment in lung cancer cells. The sensitivity of 13 lung cancer cells towards topotecan, a CPT analogue, was extracted from Cancer Cell Line Encyclopedia (CCLE) database. Immunoblotting detection of Ac-SOD1 level from those cells revealed that the basal level of Ac-SOD1 was correlated with the sensitivity to CPT treatment in lung cancer cells ([Supplemental Figure S8](#SD1){ref-type="supplementary-material"}). This data suggested that correlation of Ac-SOD1 and camptothecin-sensitivity could be a general mechanism beyond colon caner For further confirmation, we proceeded to validate this finding in patient-derived xenograft (PDX) models, which are believed to faithfully resemble the characteristics of human tumors in many aspects including heterogeneity, histology and genetic alterations \[[@R30]--[@R33]\]. CPT-11 efficacy was screened in a small panel of PDX models across different cancer types including lung cancer (LU0299, LU0743, LU0375, LU0350, LU0377), liver cancer (LI0941) and esophageal cancer (ES0204). We observed the diverse response of these models to CPT-11 treatment and subgrouped the models into CPT sensitive and resistant subset (Figure [5B](#F5){ref-type="fig"}). Meanwhile, we also measured the basal level of SOD1 acetylation on K71. The models with higher level of SOD1 acetylation were more responsive to the treatment (Figure [5C](#F5){ref-type="fig"}). These results suggest a potential value of SOD1 K71 acetylation in stratifying the responsive subset to CPT-11 based chemotherapy. DISCUSSION {#s3} ========== The increased generation of ROS and altered redox status in cancer cells offers an interesting therapeutic window that cancer cells are more sensitive than normal cells to agents causing further accumulation of ROS \[[@R4]\]. In fact, direct or indirect affects on ROS amount have been widely believed to contribute to the anticancer efficacy of cytotoxic anticancer agents, in particular genotoxic agents. Generation of high levels of ROS has been observed in patients receiving various chemotherapy treatment \[[@R24]--[@R29]\], though the mechanism of ROS generation may vary among the agents \[[@R34]\]. Apart from the widely studied ROS generation, the molecular insights into the ROS homeostasis changes by genotoxic agents have been very limited. In this study, we have provided the first evidence showing that genotoxic agents caused ROS accumulation was able to impair the antioxidant capacity of cancer cells via diminishing the activity of antioxidant enzyme SOD1. Our findings suggest the existence of a positive feedback mechanism in which ROS per se mediates the impairment of the antioxidative enzyme (defence) system of cancer cells (Figure [6](#F6){ref-type="fig"}). The feedback inhibition of SOD1 further raises the cytosolic ROS level, reinforces oxidative stress, and promotes the effectiveness of the anticancer agents. ![A schematic model showing the positive feedback of ROS induction resulted from SOD1 acetylation](oncotarget-06-20578-g006){#F6} It has long been noticed that the increase of ROS level and DNA damage, can be found one being caused by the other one; ROS induces DNA damage while DNA damage agents could also increase ROS generation. Cytotoxic anticancer agents, including cisplatin, mitomycin C, doxorubicin, CPT and ultraviolet radiation induced ROS are important for the induction of cell apoptosis and anticancer efficacy of these agents \[[@R24]--[@R29]\]. While in certain cancer cells, chemotherapeutic agents induced persistent ROS stress may induce adaptive stress responses including activation of redox-sensitive transcription factors, leading to an increase in the expression of ROS-scavenging enzymes, such as SOD and glutathione, to counteract with ROS stress. All these events enable cells to survive with the high level of ROS and render cancer cells more resistant to chemotherapeutic agents \[[@R6], [@R35]\]. Accordingly, modulating ROS-scavenging enzymes activity could enhance the anti-tumor activity of genotoxic agents via ROS mediated apoptosis induction. Intriguingly, our findings provided new insights by showing an apposing mechanism, in which the genotoxic agents, in parallel to ROS induction, are able to paralyze the antioxidant defence of cancer cells to facilitate their anticancer efficacy. Our findings are particularly interesting given the fact that cancer cells often maintain a high antioxidant capacity to cope with the massive ROS resulted from fast growth. This finding highlighted the role of antioxidant defence system in determining the efficacy the genotoxic anticancer agents, and may lead to a better understanding of the anticancer mechanism of genotoxic agents. The key molecular mechanism behind involves the acetylation of SOD1 on the lysine 71 residue. We have shown that acetylation decreases SOD1 activity by impairing the interaction between SOD1 and CCS, and hence decreasing the output of enzymatically active SOD1 homodimers in the maturation process of SOD1. We also noticed that the mutation of lysine 71 to arginine, which abolished the acetylation of SOD1, did not obviously promote the activity of SOD1. One possible explanation is that most cancer cells feature a high antioxidant capacity, in which SOD1, as a critical anti-oxidant enzyme, may maintain a high basal activity and can hardly to be further enhanced. Likewise, we have observed that the acetylation mimetic K71Q mutant exhibits striking impact on CPT-induced ROS generation and cell survival whereas the K71R mutant did not obviously promote scavenging of ROS or cell viability. We speculated that this might also partially result from the small proportion of acetylated form induced by CPT among total amount of exogenous SOD1. We also exploited how SOD1 K71 acetylation decreases CCS interaction with SOD1. As the human SOD1-CCS complex structure has not been solved yet, we took advantage of the solved crystal structure of yeast SOD1-CCS complex (ySOD1-yCCS) (PDBID: 1JK9) \[[@R36]\] and aligned the human SOD1 to yeast SOD1 in the ySOD1-yCCS crystal structure (PDBID:1JK9) using PyMol software. According to the structural alignment ([Supplemental Figure S9](#SD1){ref-type="supplementary-material"}), hSOD1 shares high structural similarity with ySOD1, and K71 (T71 in yeast SOD1) does not lie in the protein-protein interaction surface. Also, a previous crystallography study has suggested that the binding between SOD1 and CCS may involve conformational changes. We speculated that K71 acetylation may interfere with this conformational change through allosteric regulation and hinder its binding to CCS1. Previous studies have suggested two possible mechanisms by which CPT treatment may influence SOD1 acetylation. Genotoxic stress was reported to inactive SIRT1 via the ATM/ATR-DBC1 signaling \[[@R37], [@R38]\]. To test this possibility, we examined the SOD1 acetylation in ATM knockdown cells but did not observe the apparent reduction in either DBC1 phosphorylation or SOD1 acetylation, which largely ruled out the possible involvement of this pathway ([Supplemental Figure S10](#SD1){ref-type="supplementary-material"}). Alternatively, ROS induced by CPT is known to inhibit SIRT1 activity by evoking oxidative modifications on its cysteine residues \[[@R39]\]. In this case, we did not observe the alteration of SIRT1 activity upon CPT treatment ([Supplemental Figure S11](#SD1){ref-type="supplementary-material"}). Instead, our data suggested that CPT disrupted the interaction between SIRT1 and SOD1, which may possibly result from oxidative modification of SIRT1 by ROS. SIRT1 is known to play a critical role in coordinating the cellular response to stress. SIRT1 expression and activity are regulated by cellular stressors, including metabolic, genotoxic, oxidative and phototoxic stress. SIRT1 impacts cell survival by deacetylating substrate proteins to drive the cell towards a cytoprotective pathway \[[@R40]\]. Moderate overexpression of SIRT1 provides cells protection against oxidative stress by increasing the activity of catalase \[[@R41]\]. SIRT1 may also protect against oxidative stress through the modulation of family of forkhead transcription factors (FOXO) \[[@R42], [@R43]\] and the induction of manganese SOD (MnSOD) expression \[[@R44]\]. Our result supported and complemented the previous studies by showing that SIRT1 also regulates the oxidative metabolism via directly deacetylating and modulating the activity of SOD1. All these evidence suggests an important role of SIRT1 in determining the antioxidant capacity of cancer cells and hence the outcome of chemotherapy. Indeed, a previous study has observed the correlation between SIRT1 expression level and the CPT sensitivity in prostate cancer cells, although in very limited number of cell lines \[[@R45]\]. An apparent caveat is that SIRT1 expression level may not be necessarily associated with its activity. Indeed, we have observed that in colon cancer cell lines and the PDX tumors, SIRT1 protein level was not correlated with the CPT sensitivity (Figure [5A](#F5){ref-type="fig"} and [5C](#F5){ref-type="fig"}). According to our results, the basal level of SOD1 acetylation varies largely among either the cancer cell lines or patients tumor tissues; high level SOD1 acetylation is closely correlated with the increased response to CPT treatment. We speculate that while cancer cells often feature an increased antioxidant capacity, high level of SOD1 acetylation represents an intrinsic silencing of SOD1, and is also an indicator of low activity of SIRT1. Thus abundant basal level of SOD1 acetylation is able to stratify the subset with low capacity to copy with oxidative stress of cancer cells. The clinical value of SOD1 acetylation may deserve further investigation in clinical practice to increase the response rate of CPT-based chemotherapy regimen. MATERIALS AND METHODS {#s4} ===================== Plasmids {#s4_1} -------- The FLAG/HA-tagged form of SOD1 was generated by subcloning Xho I-Hind III an cassette of SOD1 into the Flag/HA-pcDNA3.1 mammalian expression vector. The plasmid pECE encoding SIRT1/SIRT1-H363Y with a FLAG tag was purchased from Addgene. The RNAi Consortium (TRC) Lentiviral shRNAs against SOD1 (Clone ID: TRCN0000039808, targeting the 3′UTR region of SOD1) and against SIRT1 (Clone ID: TRCN0000039808) were purchased from Thermo. Mutations in pcDNA3.1-SOD1-FLAG were introduced by the change site directed mutagenesis kit (Saibaisheng Gene Technolog, Shanghai, China). Sequences were verified by automated sequence analysis (Sangon Biotech, Shanghai, China). siRNA transfection {#s4_2} ------------------ For siRNA transfection, HCT116 cells were plated at 3×10^5^ cells/ml in OPTI-MEM serum-free medium and transfected with siRNA duplex using Lipofectamine RNAiMAX Reagent Agent (Life Technologies) according to the manufacturer\'s instructions. siRNAs were ordered from Sigma-Aldrich. The sequences were as follows: siSOD1 \#1: TTC GAG CAG AAG GAA AGT AAT GGA CCA dTdT; siSOD1 \#2: GGC CUG CAU GGA UUC CAU G dTdT. Cell culture {#s4_3} ------------ Human colon cancer HCT-116 cells purchased from American Type Culture Collection (ATCC) were cultured in McCOY\'s 5A medium (Life Technology) supplemented with 10% FBS. HCT116 cell lines stably transfected with short hairpin RNA targeting SOD1 or SIRT1 (ThermoFisher) (shSOD1/shSIRT1) were constructed according to manufacturer\'s instructions and maintained in McCOY\'s 5A medium supplemented with 1 μg/μl puromycin dihydrochloride (Sigma). Immunoprecipitation assay {#s4_4} ------------------------- Immunoprecipitation of Flag-tagged SOD1 was carried out using anti-FLAG M2 beads. Equal amounts of proteins in lysis Buffer were used for precipitation. Input samples represent \~1% of protein amounts used for immunoprecipitation. The following antibodies were used for immunoprecipitation and followup immunoblotting: monoclonal rabbit anti-SOD1 (Epitomics); monoclonal rabbit anti-Sir2/SIRT1 (Epitomics); monoclonal mouse anti-acetylated-Lysine(Cell Signaling Technology); monoclonal mouse anti-P53 (Santa Cruz Biotechnology); monoclonal mouse anti-CCS (Santa Cruz Biotechnology); polyclonal mouse anti-FLAG M2 affinity Gel (Sigma); monoclonal mouse anti-DYKDDDDK-Tag (Abmart, Shanghai, China); monoclonal mouse anti-HA-tag (Abmart); monoclonal rabbit anti-GAPDH (Epitomics). Antibodies specifically recognizing acetylation at lysine 71 were prepared by PTM BioLab, Inc. (Hangzhou, China). In-gel SOD1 activity assay {#s4_5} -------------------------- Cells were harvested and lysed in lysis buffer (25 mM HEPES; pH 7.9, 100 mM NaCl, 1 mM EDTA, 1% Triton X-100, 10% Glycerol, 1 mM Na~3~VO~4~, 1 mM PMSF and protease inhibitors). The supernatant was subjected to non-denaturing gel electrophoresis and nitro blue tetrazolium (NBT, Beyotime Institute of Biotechnology, Nanjing, China) staining to examine SOD1 activity as described previously with slight modifications \[[@R9]\]. Cell lysate was loaded directly without boiling to a nondenaturing 12% polyacrylamide gel. After electrophoresis, the gel was incubated for 45 min in the dark in 50 mM potassium phosphate buffer (pH 7.8) supplemented with 0.34 mM NBT and 14 mM riboflavin (Beyotime Institute of Biotechnology). The gel was subsequently exposed to light for 15 min after 1 μl/ml TEMED was added to the reaction mixture. Assessment of intracellular superoxide anions {#s4_6} --------------------------------------------- Cells were incubated with 3 μM dihydroethidium (hydroethidine, DHE; Invitrogen) in PBS for 30 min at 37°C following the treatment with CPT at 100 nM for 12 hr. Cells were washed once with PBS, trypsinized, and immediately analyzed using flow cytometry on a FACSCalibur flow cytometer (BD biosciences; software CellQuest Pro). Visualization of SOD1 monomers and dimers {#s4_7} ----------------------------------------- Cells were rinsed once in PBS prior to lysis in lysis buffer (25 mM HEPES (PH 7.5), 150 mM NaCl, 1% NP-40, 100 mM MgCl~2~ 6H~2~O, 1 mM Na~3~VO~4~, 1 mM EDTA, 1 mM PMSF, and protease inhibitor cocktail). 5 mg/ml protein lysis was incubated for 30 min at room temperature with or without 20 mM Dimethyl pimelate. The samples were incubated with 0.2 M Tris-HCl (PH 7.5) for 15 min at room temperature and analyzed using immunoblotting. Clonogenic assay {#s4_8} ---------------- Cells were plated in triplicate (500 cells/well in 6 well plates), treated with indicated doses of CPT and allowed to grow for 2 weeks. Plates were fixed in methanol and stained with 0.5% crystal violet solution. Colonies were counted and results were expressed as percentages of the untreated control. Animal studies {#s4_9} -------------- Animal studies using patient derived xenograft models were conducted by Crown Bioscience, Inc (Taicang, China) in strict accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. Prior to the initiation of treatment, mice were randomized into control and treated groups (n = 6 per group). Mice were treated intravenously with CPT-11 daily at 10 mg/kg or vehicle control for 21 consecutive days. Tumor growth was monitored by the measurement of tumor size using caliper every other day. Mice were sacrificed and tumor tissues were collected 2 hr after the last dosing. Statistical analysis {#s4_10} -------------------- A two-tailed independent Student\'s t test was used to compare the continuous variables between the two groups. The p value \< 0.05 was considered to be statistically signiﬁcant. SUPPLEMENTARY FIGURES AND TABLES {#s5} ================================ We thank Dr. Zhongyi Cheng (PTM BioLab, Inc, Hangzhou, China) for the assistance to generate Ac-SOD1-K71 antibody. This work was supported by grants from the National Science and Technology Major Project of the Ministry of Science and Technology of China (No. 2012ZX09301001-007, 2012ZX09301001-004, 2015ZX09101009), the National Marine "863" Project (No. 2013AA092902) and National Natural Science Foundation (No. 81102461). M.H. is a young scholar funded by the National Natural Science Foundation of China (NO. 81202549) and the Pujiang Scholar Program Grant by the Shanghai Metropolitan Government (No. 12PJ1410400). CONFLICTS OF INTEREST The authors declared no conflict of interest.	{ "pile_set_name": "PubMed Central" }
Background {#Sec1} ========== Betel quid chewing is one of the main causes leading to oral cancer in Taiwan \[[@CR1]\]. Arecoline, one of main effective components in betel quid, was reported to lead to DNA damage and apoptosis through the formation of reactive oxygen species (ROS) and contribute to oral carcinogenesis \[[@CR2]-[@CR5]\]. Therefore, the modulation of ROS level may be helpful for oral cancer prevention and therapy. Grape seed extract (GSE) is a common dietary health supplement due to its natural ROS modulating ability \[[@CR6]\]. Commercial preparations of GSE are marketed in the world as a dietary health supplements due to their natural free radical scavenging ability \[[@CR6]\]. The cancer chemoprevention and anticancer potential of GSE has been well reviewed previously \[[@CR7]\] including skin, colorectal, prostate, breast, lung, and gastric cancers. However, the GSE effects with respect to oral cancer cells are less studied as yet. ROS modulation effect has been well reviewed \[[@CR8],[@CR9]\]. For example, cellular ROS may regulate apoptosis through the mitochondrial pathway \[[@CR10]-[@CR13]\]. Pro-oxidants induce ROS specifically targeting cancer cells, thereby activating signal transduction pathways that are responsible for cell cycle arrest and/or apoptosis \[[@CR14]\]. Similarly, GSE was reported to generate a strong superoxide radical-associated oxidative stress and result in the apoptosis of non-small-cell lung cancer cells \[[@CR15]\] as well as in the induction of DNA damage \[[@CR16]\]. Different concentrations of GSE were reported to generate diverse biological effects in several cancer studies \[[@CR17]-[@CR20]\]. For example, high concentrations (25--100 μg/ml) of GSE showed cytotoxicity or antiproliferation of human bladder \[[@CR17]\], colorectal \[[@CR21]\], and breast \[[@CR18]\] cancer cell lines. In contrast, a low concentration (2.5 μg/ml) of GSE was reported to inhibit the micronuclei frequency and ROS generation in a lymphocyte culture, demonstrating that its antioxidant property has a protective effect during oxidative stress \[[@CR19]\]. However, more detailed mechanisms between cancer chemoprevention and anticancer effects of GSE in terms of concentration effects remain unclear. Since GSE is a natural ROS scavenger, we hypothesize that GSE modulates ROS to further regulate proliferation, apoptosis, mitochondrial function, and DNA damage. Since concentration responses of GSE for these regulations may be relevant, in this study we aim to define the critical concentrations that may or may not be able to induce apoptosis in oral cancer cells. Methods {#Sec2} ======= GSE source {#Sec3} ---------- The IH636 premium grade proanthocyanidin grape seed (Vitis vinifera) extract (GSE, commercially known as ActiVin®) was purchased from InterHealth Nutraceuticals Inc. (Benicia, CA, USA), which included 75--80% oligomeric proanthocyanidins and 3--5% monomeric proanthocyanidins as described previously \[[@CR22]\]. Cell cultures {#Sec4} ------------- Cell lines of human oral gingival cancer Ca9-22 \[[@CR23]\] and gingival fibroblast HGF-1 \[[@CR24]\] were routinely maintained in DMEM/F12 medium (Gibco, Grand Island, NY) containing 10% fetal bovine serum, 0.03% glutamine, 1 mM sodium pyruvate, and penicillin/streptomycin mixtures. Cells were kept at 37°C in a humidified incubator containing 5% CO~2~. Determination of cell viability {#Sec5} ------------------------------- Viability analysis was performed using Cell Titer 96™ Aqueous One solution cell proliferation (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4- sulfophenyl)-2H-tetrazolium) MTS) assay kit (Promega Madison, WI, USA) as described previously \[[@CR25]\] with minor modification. In brief, cells were treated with various concentrations of GSE in fresh media in triplicates. The non-toxic concentration of DMSO (less than 1% v/v) was used to prepare test solutions in all assays. The plates were then incubated for 24 h under standard growth conditions. Subsequently, MTS reagent was loaded to each well (5 mg/ml in PBS) and cells were again incubated for another 2 h. Then, absorbance of each well was recorded directly at 490 nm by ELISA multi-Plate Reader (MTX Lab Systems, Inc., Vienna, VA, USA). Determination of sub-G1 population {#Sec6} ---------------------------------- Measurement of DNA content for cell cycle analysis were carried out by flow cytometry, based on a previously described protocol \[[@CR26]\]. In brief, Ca9-22 cells were treated with either DMSO only or different GSE concentrations for 24 h. After incubation, cells were harvested for washing and fixing in 70% ethanol overnight. After harvest, cells were resuspended in 1 ml PBS containing 10 μg/ml PI (Sigma, St Louis, MO, USA) in the dark. Subsequently, cells were analyzed using a flow cytometer (FACScan; Becton-Dickinson, Mansfield, MA) at excitation and emission settings of 480 and 525 nm, respectively, and Win-MDI software (<http://facs.scripps.edu/software.html>). Determination of apoptosis by annexin V/PI {#Sec7} ------------------------------------------ The induction of apoptosis by GSE-treated Ca9-22 cells was analyzed by annexin V staining as previously described \[[@CR27]\]. Briefly, cells were treated with either vehicle or various GSE concentrations for 24 h. Subsequently, the cells were trypsinized, washed twice with PBS and stained with fluorescein isothiocyanate (FITC)-labelled annexin V. Then, the samples were measured with a flow cytometer (FACSCalibur; Becton-Dickinson) for the quantification of apoptotic cells at excitation and emission settings of 480 and 525 nm, respectively, and Win-MDI software. Determination of apoptosis by pan-caspase activity {#Sec8} -------------------------------------------------- The induction of apoptosis by GSE-treated Ca9-22 cells was analyzed by activation of caspases (caspase-1, 3, 4, 5, 6, 7, 8, 9) by the generic caspase activity assay kit (Abcam, Cambridge, UK) as previously described \[[@CR27]\]. Briefly, cells were treated with either vehicle or various GSE concentrations for 24 h. After harvest, the cells were suspended and stained with 1 X fluorescent TF2-Val-Ala-Asp (VAD)-FMK at the cell incubator for 1 h. Then, the samples were measured with a flow cytometer (BD Accuri C6; Becton-Dickinson, Mansfield, MA, USA) and a BD Accuri C6 Software (version 1.0.264) for the quantification of pan-caspase positive populations at excitation and emission settings of 480 and 525 nm, respectively. Determination of intracellular ROS {#Sec9} ---------------------------------- Intracellular redox state were determined by the ROS-sensitive dye 2′,7′-dichlorodihydrofluorescein diacetate (DCFH-DA) (Sigma Chemical Co., St. Louis, MO, USA) as previously described \[[@CR25],[@CR28]\]. Ca9-22 cells were treated with various concentrations of GSE for 24 h. Subsequently, cells were harvested, thoroughly washed, resuspended in 10 μM DCFH-DA in PBS and then incubated at 37°C for 30 min in darkness. After incubation, cells were washed, resuspended in PBS, and analyzed with a FACSCalibur flow cytometer at excitation and emission settings of 480 and 525 nm, respectively, and Win-MDI software. Determination of mitochondrial membrane potential {#Sec10} ------------------------------------------------- Mitochondrial membrane potential (MitoMP) was determined by flow cytometry using MitoProbe™ DiOC2(3) assay kit (Invitrogen, San Diego, CA, USA) as described previously \[[@CR25]\]. In brief, cells were incubated with various GSE concentrations at 37°C for 24 h. Subsequently, cells were incubated in culture medium (containing 50 μM of DiOC~2~(3)) at 37°C for 20 min in an incubator. After washing and resuspension in PBS, cells were subjected to flow cytometric analysis. The fluorescence intensity was measured using 488 and 525 nm filter settings for the excitation and emission wavelengths, respectively. The data were analyzed with Win-MDI software. Determination of DNA double strand breaks (DSBs) by γH2AX/PI double staining {#Sec11} ---------------------------------------------------------------------------- DSBs were measured by flow cytometry as described previously \[[@CR25]\]. Ca9-22 cells were incubated with various GSE concentrations for 24 h, followed by fixation with 70% ethanol overnight. After washing twice with BSA-T-PBS (1% bovine serum albumin and 0.2% Triton X-100 in PBS), cells were treated with 100 μl of BSA-T-PBS solution containing 0.2 μg monoclonal antibody against p-Histone H2A.X (Ser 139) (Santa Cruz Biotechnology, Santa Cruz, CA, USA) for overnight at 4°C. After washing, cells were resuspended in Alexa Fluor 488-tagged secondary antibody (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA) at a 1:100 dilution for 1 h at 4°C. After washing, cells were resuspended in 1 ml PBS containing 5 μg/ml PI and analyzed by a FACSCalibur flow cytometer and Win-MDI software. Statistical analysis {#Sec12} -------------------- Statistical analysis was performed with JMP 9 software. One-way ANOVA with Tukey's HSD Post Hoc Test was used to analyze significant differences between treatments. Unless otherwise indicated, all experiments were repeated in triplication. Results {#Sec13} ======= Cell viability {#Sec14} -------------- To access the potential harmful effects of GSE on Ca9-22 cells, cell viability was determined by MTS reagent. At low concentrations of GSE treatment (1--10 μg/ml), the cell viabilities maintained around 91% (Figure [1](#Fig1){ref-type="fig"}). In contrast, a significantly concentration-dependent decrease in cell viability (P \< 0.005--0.0001) was observed at high GSE concentrations (50, 100, 200 and 400 μg/ml). The IC~50~ of GSE for Ca9-22 cells was 150 μg/ml at 24 h incubation. However, both the low and high concentrations of GSE were not harmful to normal oral HGF-1 cells.Figure 1Cytotoxicity of GSE treatments on oral cancer Ca9-22 cells and normal oral HGF-1 cells. Cells were treated with either vehicle control (DMSO) or with 2, 4, 8, 10, 50, 100, 200 and 400 μg/ml of GSE for 24 h. Cell viability was detected by the MTS assay. The percent cell viability in the experimental groups was adjusted to the DMSO-treated group representing 100% viability. Data, mean ± SD (n = 10 and 5 for Ca9-22 and HGF-1 cells, respectively). Treatments with the same capital letter are nonsignificant. Cell cycle distribution by GSE treatments {#Sec15} ----------------------------------------- To investigate if GSE treatments cause change in cell cycle distribution in Ca9-22 cells, a standard PI-staining protocol was applied to GSE-treated Ca9-22 cells. In Figure [2](#Fig2){ref-type="fig"}A, the cell cycle distributions were stable at low concentrations (1--10 μg/ml) of GSE but the subG1 populations were gradually accumulated at high concentrations of GSE. In Figure [2](#Fig2){ref-type="fig"}B, the change in the sub-G1 populations (%) of Ca9-22 cells was not significant at low concentrations of GSE. However, the changes in the sub-G1 populations (%) significantly increased to 5.61, 16.73, 25.69 and 26.80 in a concentration-dependent manner (P \< 0.0001) when GSE concentrations were increased at 50, 100, 200 and 400 μg/ml, respectively. Additionally, the percentage changes in other cell cycle phases (i.e. the G1, S and G2/M phases) did not exhibit significant changes in all treatment groups compared to untreated samples.Figure 2Analysis on distribution of cell cycle in GSE-treated Ca9-22 cells. Ca9-22 cells were treated with indicated GSE concentrations (0--400 μg/ml) for 24 h before being harvested, fixed and stained with PI for cell cycle analysis. (A) Representative histograms for cell cycle phases in GSE-treated Ca9-22 cells. (B) Quantitative analysis on distribution of cell cycle phases. Data, mean ± SD (n = 3). Treatments with the same capital letter are nonsignificant. Apoptotic cell death: annexin V/PI {#Sec16} ---------------------------------- To determine the degree of apoptosis of GSE-induced cell death in Ca9-22 cells, the annexin V-FITC staining was determined by flow cytometry. In Figure [3](#Fig3){ref-type="fig"}A, the apoptosis signals were similar at low concentrations (1--10 μg/ml) of GSE but they gradually increased at high concentrations of GSE. In Figure [3](#Fig3){ref-type="fig"}B, the annexin V intensity of GSE-treated Ca9-22 cells was weak at low concentrations of GSE. However, the percentage changes in the annexin V intensity significantly increased to 14.19%, 20.03%, 44.53, and 72.86% in a concentration-dependent manner at high concentrations (50, 100, 200 and 400 μg/ml) of GSE (P \< 0.05--0.0001).Figure 3GSE induced annexin V/PI-based apoptosis in Ca9-22 cells. Cells were treated with indicated GSE concentrations (0--400 μg/ml) for 24 h. (A) Histograms of representative annexin V-FITC profile in GSE-treated Ca9-22 cells. (B) Quantitative analysis for the apoptotic cells (%). Apoptosis was counted at the intensity of right gated region in (A). Data, mean ± SD (n = 3). Treatments with the same capital letter are nonsignificant. Apoptotic cell death: caspase activity {#Sec17} -------------------------------------- To further determine the degree of apoptosis of GSE-induced cell death in Ca9-22 cells, the multiple caspase (pan-caspase) activity staining was determined by flow cytometry. In Figure [4](#Fig4){ref-type="fig"}A, the apoptosis signals based on pan-caspase intensities were similar at low concentrations (1--10 μg/ml) of GSE but they gradually increased at high concentrations of GSE. In Figure [4](#Fig4){ref-type="fig"}B, the pan-caspase positive intensity of GSE-treated Ca9-22 cells was weak at low concentrations of GSE. However, the percentage changes in the pan-caspase positive intensity concentration-responsively increased at high concentrations (50, 100, 200 and 400 μg/ml) of GSE, respectively (P \< 0.01--0.001).Figure 4GSE induced pan-caspase-based apoptosis in Ca9-22 cells. Cells were treated with indicated GSE concentrations (0--400 μg/ml) for 24 h. (A) Histograms of representative pan-caspase activity profile in GSE-treated Ca9-22 cells. Caspase positive % was indicated in the right gated region of each panel. (B) Quantitative analysis for the pan-caspase positive (%). Data, mean ± SD (n = 3). Treatments with the same capital letter are nonsignificant. Intracellular ROS {#Sec18} ----------------- ROS plays a pivotal role in regulating cellular apoptosis. To determine whether Ca9-22 cells exposed to GSE bore higher levels of ROS, we monitored the intracellular ROS levels using the fluorescent dye DCFH-DA as a specific ROS scavenger. In Figure [5](#Fig5){ref-type="fig"}A, the ROS signals were similar at low concentrations (1--10 μg/ml) of GSE but they were gradually increased at high concentrations of GSE. In Figure [5](#Fig5){ref-type="fig"}B, no significant elevation of ROS levels became apparent in Ca9-22 cells treated with GSE concentrations lower than 10 μg/ml. However, after exposing Ca9-22 cells to GSE at 50, 100, 200 and 400 μg/ml for 24 h, the ROS levels increased significantly in a concentration-dependent manner to 2.10%, 5.10%, 27.77%, and 63.17%, respectively (P \< 0.005--0.0001).Figure 5Concentration-dependent ROS generation in GSE-treated oral cancer Ca9-22 cells. Ca9-22 cells were incubated with indicated concentrations of GSE (0--400 μg/ml) for 24 h. (A) Representative histograms of flow cytometric analysis on ROS levels from GSE-treated cells. (B) Quantitative analysis of ROS intensity by means of DCF positivity percentage. ROS is counted at the intensity of right gated region in (A). Data, mean ± SD (n = 3). Treatments with the same capital letter are nonsignificant. MitoMP {#Sec19} ------ To examine the involvement of GSE-induced mitochondrial dysfunction in Ca9-22 cells, the flow cytometry-based Rh123 staining was performed. In Figure [6](#Fig6){ref-type="fig"}A, the mitoMP signals were similar in low concentrations (1--10 μg/ml) of GSE but they were gradually decreased in high concentrations of GSE after 100 μg/ml of GSE. In Figure [6](#Fig6){ref-type="fig"}B, there was not a significant elevation of mitoMP levels in Ca9-22 cells incubated with GSE concentrations lower than 10 μg/ml. In contrast, after exposing Ca9-22 cells to GSE at 50, 100, 200 and 400 μg/ml for 24 h, the mitoMP levels were significantly decreased in a concentration-dependent manner to 101.14%, 91.69%, 66.97%, and 15.01%, respectively (P \< 0.005--0.0001).Figure 6Reduction of MitoMP in GSE-treated Ca9-22 cells. Ca9-22 cells were treated with indicated GSE concentrations (0--400 μg/ml) for 24 h before the addition of fluorescent dye Rh123 (0.5 μg/ml) for 30 min. (A) Representative histograms on MitoMP levels for GSE-treated Ca9-22 cells. (B) Quantitative analysis on reduction of MitoMP in vehicle controls and GSE-treated cells. MitoMP is counted at the intensity of the left gated region in (A). Data, mean ± SD (n = 3). Treatments with the same capital letter are nonsignificant. DNA damages caused by GSE treatment {#Sec20} ----------------------------------- To detect whether GSE treatments cause DNA double strand break (DSB) in Ca9-22 cells, samples were analyzed using flow cytometry to quantify levels of the phosphorylated γH2AX protein. In Figure [7](#Fig7){ref-type="fig"}A, the γH2AX signals were similar in low concentrations (1--10 μg/ml) of GSE but they were gradually increased in high concentrations of GSE. In Figure [7](#Fig7){ref-type="fig"}B, there was not a significant elevation of γH2AX levels in GSE-treated Ca9-22 cells under low concentrations (lower than 10 μg/ml). In contrast, after exposing Ca9-22 cells to GSE at 50, 100, 200 and 400 μg/ml for 24 h, the γH2AX levels were significantly increased in a concentration-dependent manner to 3.38%, 5.88%, 19.02%, and 35.53%, respectively (P \< 0.001).Figure 7GSE treatments caused γH2AX-based DNA damage in Ca9-22 cells. Cells were treated with different GSE concentrations for 24 h before subjected to anti-phospho-γH2AX/PI double staining for measuring DNA double strand breaks by flow cytometry.(A) A representative γH2AX/PI profile for GSE-treated cells.(B) Quantitative analysis of γH2AX-based DNA damage in GSE-treated cells. The mean fluorescence intensity of phospho-γH2AX for gated regions in (A) was calculated. Data, mean ± SD (n = 3). Treatments with the same capital letter are non-significant. Discussion {#Sec21} ========== Accumulating evidence of the antiproliferative effect of GSE had been reported in several oral cancer cell lines. For example, high concentrations (50--600 μg/ml) of GSE of Vitis vinifera were found to inhibit cell proliferation and induce apoptosis of the KB cells but less harmful to non-cancerous human umbilical vein endothelial cells (HUVEC) by trypan blue assay at 24 h GSE treatment \[[@CR29]\]. Similarly, we found that the low and high concentrations of GSE to normal oral HGF-1 cells based on MTS analysis. The KB cells was used to be regarded as the oral cancer cell line, however, it was recently confirmed to be the contaminant cervical cancer HeLa cells \[[@CR30]\]. Moreover, the low concentrations of GSE were not investigated in this study. Recently, the differential concentration effect of GSE to differentially inhibit proliferation of oral cancer cells has been demonstrated. For example, low concentrations of GSE (10--20 μg/ml) did not displayed the antiproliferation of oral cancer CAL 27 cells but high concentrations of GSE (30--80 μg/ml) were able to inhibit its proliferation \[[@CR31]\]. Similarly, we found that low (1--10 μg/ml) and high (50--400 μg/ml) concentrations of GSE displayed the differential cytotoxic effects to cell viability of oral cancer Ca9-22 cells. Similar results also reported in other cancer cells. In the example of skin cancer HaCaT cells, high concentrations of GSE (IC~50~ = 76.44 μg GAE/ml) displayed the growth inhibitory effect, but low concentrations of GAE (10--20 μg GAE/ml) protected against UVB irradiation (50--100 mJ/cm^2^)-induced skin cancer \[[@CR20]\]. These findings suggested that different cancer cell lines may require different but high concentrations of GSE for antiproliferation purpose. ROS induction by GSE was reported in non-small-cell lung cancer H1299 and A549 cells but it only tested at high concentrations (20--100 μg/ml) without detecting the mitochondrial function \[[@CR15]\]. ROS generation of high GSE (40 μg/ml) also reported to induce apoptosis in head and neck cancer Detroit 562 and FaDu cells \[[@CR32]\]. In oral cancer CAL 27 cells, GSE also reported to induce mRNA overexpression of apoptosis-associated signaling such as caspase-2 and caspase-8 \[[@CR31]\]. In head and neck cancer cells, GSE also reported to induce DNA damage \[[@CR32]\]. Our results further validated that GSE at high concentrations (50--400 μg/ml) have high oxidative stress and apoptosis in terms of ROS generation, mitochondrial depolarization, annexin V/PI staining, and caspase activation but not for low concentrations (\<10 μg/ml) of GSE in oral cancer Ca9-22 cells. Moreover, this differential concentration effect of GSE was also found in cancer cell migration. For example, GSE was reported to inhibit migration and invasion of breast cancer MDA-MB231 cell \[[@CR18]\]. High concentrations (50--100 μg/ml) of GSE inhibited cell proliferation and induced apoptosis. Conversely, low GSE (25 μg/ml) concentrations decreased cell migration and invasion. Therefore, the differential concentration effect of GSE in oral cancer cell migration is warranted for further investigation. Conclusion {#Sec22} ========== We demonstrated that GSE shows differential concentration effects in the antiproliferation of oral cancer cells through differential expressions of apoptosis, oxidative stress, and DNA damage. We showed that the antiproliferative effect of high GSE concentrations is associated with an overproduction of ROS causing DNA damage and apoptosis of cancer cells. Competing interests The authors declare that they have no competing interests. Authors' contributions C-YY, M-FH, B-HC, and H-WC participated in the writing and the study design. J-YT and H-WC performed statistical analysis. Z-WY, K-TL, C-YH, and Y-HH performed survival assay and flow cytometer analysis. H-WH, S-YL, T-FF, B-HC, and H-WC analyzed and discussed the GSE data. B-HC and H-WC coordinated and oversaw the study. All authors read and approved the final manuscript. This work was partly supported by funds of the Ministry of Science and Technology (MOST 103-2320-B-037-008), the ChiMei-KMU Joint Project (103CM-KMU-09), the Kaohsiung Medical University "Aim for the Top Universities Grant, grant No. KMU-TP103A33", the Kaohsiung Municipal Ta-Tung Hospital (kmtth-102-011), the National Sun Yat-sen University-KMU Joint Research Project (\#NSYSU-KMU 104-p036), and the Health and welfare surcharge of tobacco products, the Ministry of Health and Welfare, Taiwan, Republic of China (MOHW104-TDU-B-212-124-003). We also thank for the help in English editing by Dr. Hans-Uwe Dahms and technical support with the flow cytometer by Mr. Yi-An Chung.	{ "pile_set_name": "PubMed Central" }
Introduction ============ Staphylococcus aureusis the most common cause of septic arthritis (SA) \[[@B1],[@B2]\]. SA has shown no change in incidence in spite of advances in antimicrobial therapy and is responsible for residual functional impairment and for a high mortality rate among debilitated patients. Risk factors include older age, diabetes mellitus, rheumatoid arthritis (RA), immunodeficiency, and a pre-existing joint disease. In SA, S. aureuscontributes to more than two thirds of identified organisms \[[@B3],[@B4]\]. In an epidemiological study of SA in an adult population of 116 patients by Abid and colleagues \[[@B5]\] between 1999 and 2004, S. aureuswas the most common organism isolated from blood as well as synovial fluid (18.8%). Cleeman and colleagues \[[@B6]\] studied 23 culture-positive cases of SA of the glenohumeral joint between 1986 and 2000, and 52% had a different primary site of infection identified, 70% of which were S. aureus-positive and 17% of which were methicillin-resistant. In a retrospective analysis by Moumile and colleagues \[[@B7]\] of the bacterial etiology of acute osteoarticular infections in 406 children with clinically suspected osteoarticular infections, 74 (18%) had a positive bacterial culture: 38 cases of SA and 36 cases of bone infections (osteitis and osteomyelitis), the most commonly recovered pathogen being S. aureus(44%). Goergens and colleagues \[[@B8]\] reviewed the clinical presentation, management, and organisms responsible for acute hematogenous osteomyelitis (AHO) and SA in Australia between 1998 to 2002, and S. aureuswas the most common identifiable causative organism, accounting for 76% of isolated organisms in AHO and 39% of isolated organisms in SA. S. aureusremains the most common organism causing AHO and SA, and multidrug-resistant S. aureus(MRSA) is on the increase as well. Progressive joint destruction despite appropriate antibiotic therapy and synovial fluid aspiration may indicate a potential role for host-derived proteases. Several matrix metalloproteinases (MMPs) are induced in host cells in response to infectious stimuli. Normally, MMPs assist in clearing infections, initiating immune responses, and in tissue remodeling \[[@B9]\]. Excessive MMPs cause matrix degradation and joint destruction as in various forms of arthritis \[[@B10]\]. Cytokines interleukin (IL)-1, IL-6, tumor necrosis factor (TNF)-α, and interferons (IFN-α and IFN-γ) are released from host cells in response to S. aureusinfection and these are potent inducers of MMPs \[[@B11]-[@B15]\]. Staphylococcal capsule polysaccharides, toxins, cell wall-attached adhesions, and possibly also the chromosomal DNA are virulence determinants in S. aureusarthritis. These bacterial components might affect the innate immune response and inflammation \[[@B1]\]. Alternatively, the bacterial products, secreted or intracellular, could directly affect the transcriptional machinery or signal transduction pathways related to MMP expression. Previous studies have shown the induction of proteolytic enzymes in chondrocytes in response to bacteria-free culture supernatants from S. aureus\[[@B16]\]. Also, peptidoglycan (PGN) from S. aureushas been shown to be capable of inducing arthritis \[[@B17]\]. A recent study showed that S. aureusPGN induces MMP-1, -3, and -13 in human synovial fibroblasts \[[@B18]\]. Purified PGN is chemically modified and may not really represent the native PGN. Also, there is a wide variety of bacterial components, including the superantigens, cell wall components, and extracellular toxins, which could stimulate the host cells. The full potential of synovial fibroblasts in terms of multiple MMP expression in response to S. aureuscomponents has not yet been addressed. To determine the global impact of S. aureuscomponents on primary human fibroblasts with respect to MMP expression, we exposed de-identified normal human dermal fibroblasts and synovial fibroblasts derived from de-identified patients with RA and osteoarthritis (OA) to whole cell lysate and culture supernatants (whole and fractionated) derived from S. aureuswild-type and mutant strains that induce less severe SA in murine models. Materials and methods ===================== Bacterial strains ----------------- S. aureusstrain isolated from a patient with SA was obtained from American Type Culture Collection (ATCC) (Manassas, VA, USA). A de-identified clinical isolate (U1) and mutants lacking staphylococcal accessory gene regulator (U155: Sar^-/-^) and accessory gene regulator (U929: Agr^-/-^) and a strain lacking both Sar Aand Agr(Sar-Agr^-/-^) derived from that clinical isolate were obtained from M. Smeltzer and were used in this study. The U155 strain was grown in the presence of tetracycline (5 μg/ml); U929 was grown in presence of kanamycin (50 μg/ml) and neomycin (50 μg/ml); and U930 was grown in the presence of tetracycline (5 μg/ml), 50 μg/ml kanamycin, and 50 μg/ml neomycin/ml for selection of the respective mutants. Strains grown in the presence of antibiotics were centrifuged, washed, and resuspended in Dulbecco\'s modified Eagle\'s medium (DMEM)/F-12 medium for inoculation in order to remove the antibiotics. Preparation of whole and fractionated bacterial culture supernatants and bacterial cell lysates ----------------------------------------------------------------------------------------------- To collect supernatants and bacterial cell pellets for experiments, bacterial strains (isogenic parent strain and mutants strains) were grown in DMEM/F-12 containing 2% fetal bovine serum (FBS) without any antibiotics. Supernatants and cell pellets were obtained by centrifugation from 12-hour bacterial cultures. The supernatants were sterilized through 0.22-μm filters to ensure that they were free of any bacteria. The cell pellets were treated with lysostaphin (20 U/ml) for 20 minutes at 37°C followed by repeated freezing and thawing. The lysates were clarified by centrifugation at 12,000 gfor 20 minutes and were filtered through 0.22-μm filters. The ATCC strain was also grown in the presence of 5 and 15 ng/ml recombinant human rhIL-1β (R&D Systems, Inc., Minneapolis, MN, USA). The cell lysates were prepared as described above. Total protein concentrations were measured by the calorimetric method (Bio-Rad, Hercules, CA, USA) in accordance with the manufacturer\'s instructions. The culture supernatants from the ATCC strain were fractionated into \<30, 30 to 50, and \>50 kDa molecular-weight fractions using respective Centricon filter centrifugation. Fibroblast cultures ------------------- Dermal fibroblasts from de-identified normal volunteers and synovial fibroblasts from de-identified RA patients and OA patients were maintained in DMEM/F-12 containing 10% FBS, 100 U/ml penicillin, and 100 μg of streptomycin. All the fibroblast cell lines were from a cell culture bank established by A. Postlethwaite in accordance with the full approval of the institutional review board of the University Of Tennessee Health Science Center (Memphis, TN, USA). Treatment of fibroblasts with S. aureussupernatants, lysates, and rhIL-1β/rhTNF-α ----------------------------------------------------------------------------------- For studies measuring MMP production, 10^5^fibroblasts harvested by trypsinization were added to each well of 24-well tissue-culture plates (Falcon; Falcon Plastics, Inc., Washington, PA, USA). Three days later, confluent monolayers of fibroblasts were treated with phosphate-buffered saline, 25 μg of total proteins from bacterial cell lysates, 25 μg of total proteins from culture supernatants, and 15 μg of protein from each fraction of culture supernatant per well (only from ATCC strain). Fibroblasts were cultured in an incubator containing a humidified atmosphere containing 5% CO~2~at 37°C. Fibroblasts were cultured for 8 hours for RNA analysis and 48 hours for protein analysis. Fibroblasts were also treated with a combination of 10 μg each of rhIL-1β/TNF-α for 8 hours and 48 hours. For mRNA analysis, cells were harvested after 8 hours of respective treatments, and total RNA was isolated using TRI-Reagent (Sigma-Aldrich, St. Louis, MO, USA) followed by isopropanol precipitation. The fibroblast culture supernatants were collected 48 hours after respective treatments and centrifuged to remove any cell debris. All samples were stored at -80°C until analyzed. Fractionation of S. aureusculture supernatants ------------------------------------------------ Culture supernatants from S. aureuswere purified using the Amicon Centricon filter device from Millipore Corporation (Billerica, MA, USA). Using this device, an approximately 2.0-ml volume was concentrated into an approximately 30-μl volume. Using the 10,000, 30,000, and 50,000 kDa cutoff filter devices, we fractionated the whole culture supernatants to \<30, 30 to 50, and \>50 kDa fractions. Assay for MMP and tissue inhibitor of metalloproteinases mRNAs -------------------------------------------------------------- Total RNA was reverse-transcribed into cDNAs by using avian myeloblastoid virus (AMV) RTase and Oligo dT primers. Qualitative profiling of multiple MMP mRNAs was performed using a Multi-MMP-mRNA kit from SuperArray Bioscience Corporation (Frederick, MD, USA) in accordance with the manufacturer\'s protocol. Relative quantification of MMP mRNAs was performed using SYBR green real-time polymerase chain reaction (PCR) on the cDNAs obtained. Assays for MMP-1, -2, -3, -7, -8, -9, -10, -11, -12, -13, and -14 were performed. Gene-specific oligonucleotide primers for tissue inhibitor of metalloproteinases (TIMP)-1, TIMP-2, and TIMP-3 were synthesized at Integrated DNA Technologies (Coralville, IA, USA). The messages of TIMP mRNAs were quantified using SYBR green real-time reverse transcription-PCR (RT-PCR). The message levels were expressed as ratios of the threshold cycle values of respective MMP or TIMP messages to those of the housekeeping gene glyceraldehyde phosphate dehydrogenase (GAPDH). Assays for MMP protein ---------------------- Multiple MMP proteins were determined using an MMP protein array kit from RayBiotech, Inc. (Norcross, GA, USA) strictly following the manufacturer\'s instructions. Briefly, the 100 μl of culture supernatants was applied to the membrane with arrayed antibodies. After blocking the free spaces on the membrane, a cocktail of biotin-labeled antibodies was added and the membranes were incubated for 2 hours at ambient temperature. After repeated washing, horseradish peroxidase-conjugated streptavidin was added onto the membrane and incubated for 2 hours at ambient temperature. After extensive washing, the detection buffer was added and the signals were detected by capturing the enhanced chemiluminiscence onto a Kodak x-omat AR film (Eastman Kodak, Rochester, NY, USA). The film was photographed and scanned for documentation. Semiquantitative profiling of mRNAs of MAPK family -------------------------------------------------- A human MAPK gene family multigene-12 RT-PCR profiling kit (Superarray Bioscience Corporation) was used for the qualitative assessment of extracellular signal regulated kinase (ERK) 2/MAPK2, ERK1, MAPK4, ERK3, ERK5, c-jun N-terminal kinase (JNK) 1, JNK2, JNK3, p38b MAPK, p38g MAPK, and p38delta MAPK mRNA in fibroblasts in response to S. aureusculture supernatant and cell lysate. Total RNA was reverse-transcribed into cDNAs using AMV RTase and Oligo dT primers, and the messages were amplified using the primer sets supplied by the manufacturer. The expression level of the housekeeping gene GAPDHin each sample was used to assess the qualitative differences in respective message levels between samples. The experiments were repeated three times and each time the assays were set up in duplicate. The PCR products were analyzed on a 2% agarose gel and were stained with SYBR green. The intensities of the bands were estimated by densitometric scanning software from Alpha Innotech Corporation (San Leandro, CA, USA). The values expressed are ratios of the densities of the MAPK genes to those of the housekeeping gene GAPDH. Cell-based enzyme-linked immunosorbent assay for tyrosine phosphorylation ------------------------------------------------------------------------- A cell-based phosphotyrosine enzyme-linked immunosorbent assay (ELISA) kit from RayBiotech, Inc., was used to quantitate tyrosine phosphorylation in human dermal fibroblasts in response to S. aureuscomponents and IL-1/TNF. Approximately 30,000 cells were seeded into each well in a 96-well plate. Cells were incubated at 37°C, 5% CO~2~overnight. The cells were then exposed to S. aureuscell lysate (25 μg/ml), S. aureus(ATCC) culture supernatant (25 μg/ml), or 10 ng/ml each of rhIL-1β and rhTNF-α for 30 minutes. The medium was removed from the wells, and the cells were treated with the fixing solution followed by quenching solution. The fixed, quenched cells were treated with blocking solution for 3 hours at ambient temperature, and after washing the cells were exposed to anti-phosphotyrosine-horseradish peroxidase for 1 hour followed by washing and the addition of one-step substrate solution. The plates were incubated in the substrate solution for 30 minutes, the color reaction was stopped, and the optical densities were read at 450 nm. The experiments were repeated three times and each time the experiments were run in triplicates. Statistical analysis -------------------- Each treated sample was compared with the untreated sample using Student\'s test. Sigma Stat (Systat Software, Inc., San Jose, CA, USA) program was used for statistical computation, and Sigma Plot (Systat Software, Inc.) was used to create graphs. A pvalue of less than 0.05 was considered significant. Results ======= Induction of multiple MMP proteins by S. aureusin human dermal fibroblasts ---------------------------------------------------------------------------- Culture supernatant and cell lysate from S. aureus(ATCC) induced the expression of immunoreactive proteins of MMP-1, MMP-2, MMP-10, and MMP-13 in dermal fibroblasts (Figure [1](#F1){ref-type="fig"}). Upregulation of TIMP-1 and TIMP-2 was also noted in S. aureusculture supernatant and cell lysate-treated fibroblasts. There were no notable changes in the expression levels of other MMP proteins (MMP-8 and MMP-9) in the cells in response to treatment. The expression pattern and level of expression were similar in S. aureuscomponents and IL-1β/TNF-α-treated fibroblasts. Induction of multiple MMP mRNAs by S. aureusin human dermal and synovial fibroblasts -------------------------------------------------------------------------------------- Multiple MMP mRNA profile in dermal and synovial fibroblasts in response to S. aureuscomponents was determined by SYBR green real-time PCR. Culture supernatants and cell lysate from S. aureussignificantly enhanced the expression of multiple MMP mRNAs (namely, MMP-1, -2, -3, -7, -10, and -11; p\< 0.05). As in the case of MMP protein expression pattern, the response of the fibroblasts in terms of MMP mRNA expression was similar in S. aureuscomponent-treated and rhTNF-α- and rhIL-1β-treated fibroblasts (Figure [2](#F2){ref-type="fig"}). Unlike untreated dermal fibroblasts (Figure [2](#F2){ref-type="fig"}), untreated synovial fibroblasts from patients with RA (Figure [3a](#F3){ref-type="fig"}) and OA (Figure [3b](#F3){ref-type="fig"}) had higher basal multiple MMP mRNA expression, indicating an activated status of the synovial fibroblasts from a pathological site. Similar to the dermal fibroblasts, most MMP mRNAs tested were elevated in RA and OA fibroblasts in response to S. aureuscomponents. Levels of significantly elevated MMPs (MMP-1, -2, -3, -7, -10, and -11; p\< 0.05) are shown in Figure [3a,b](#F3){ref-type="fig"}. The MMP mRNA expression pattern in response to IL-1β/TNF-α and S. aureuslysate was similar in RA and OA fibroblasts (Figure [3a,b](#F3){ref-type="fig"}) as well. Interestingly, no significant differences were noted in MMP-13 mRNA levels between the treated and untreated fibroblasts. All other MMPs tested were expressed at very low levels, could not be quantified, and hence were not included in the graph. Two more dermal fibroblast lines, RASF and OASF cell lines, were tested for multiple MMP mRNA expression profile upon exposure to S. aureusculture supernatants and bacterial cell lysates. Essentially the same profile as described above was obtained from the additional cell lines (data not shown). Because fibroblasts are heterogeneous in terms of their origin and some of their features, it is likely that fibroblasts from different sources may respond slightly differently in terms of MMP expression. Potentiation of MMP protein expression in human fibroblasts by S. aureusgrown in presence of rhIL-1β ------------------------------------------------------------------------------------------------------ We have observed significant changes in gene expression in S. aureusgrown in the presence of rhIL-1β (manuscript submitted). To test whether S. aureusgrown in the presence of rhIL-1β would have any impact on MMP expression, dermal fibroblasts were exposed to 25 μg/ml per well bacterial cell lysate obtained from S. aureusgrown in the presence of 5 or 15 ng/ml rhIL-1β. The supernatants were collected and expression of multiple MMP protein was assessed by multi-MMP-Array kit from RayBiotech, Inc., as described previously. The data presented in Figure [4](#F4){ref-type="fig"} show that production of MMP-2, -3, and -8 is greater in fibroblasts treated with cell lysate obtained from the S. aureusstrain grown in the presence of rhIL-1β. TIMP-4 expression was also slightly enhanced in fibroblasts treated with lysate obtained from the S. aureusgrown in the presence of 15 ng/ml IL-1β. Induction of MMP mRNA in human dermal fibroblasts by fractionated culture supernatants from S. aureus ------------------------------------------------------------------------------------------------------- The MMP-inducing active components in the culture supernatants were mostly in the 30 to 50 and \>50 kDa molecular-weight range as evidenced by significantly elevated expression of MMP-1 and MMP-3 by Centricon fractions 30 to 50 and \>50 kDa in dermal fibroblasts (Figure [5](#F5){ref-type="fig"}; p\< 0.05). Although the fractions are not identified beyond their molecular weight, this does rule out some of the already-characterized low-molecular-weight extracellular (secreted) products of S. aureus. MMP mRNA induction by Sar, Agr, and Sar/Agrmutants of S. aureus ----------------------------------------------------------------------- Blevins and colleagues \[[@B19]\] have shown that S. aureusstrains lacking the regulatory loci Saror Agrresult in less severe SA and osteomyelitis in murine models of these diseases. We therefore tested the ability of cell lysates and culture supernatants obtained from these mutants and their isogenic parent strain to induce MMP-1 and MMP-3 mRNAs in human dermal fibroblasts. The mutants and isogenic strains enhanced MMP-1 and MMP-3 production by fibroblasts to a similar degree (Figure [6](#F6){ref-type="fig"}). Induction of TIMP mRNA expression in human fibroblasts by S. aureuswild-type and Sar/Agrmutants --------------------------------------------------------------------------------------------------- TIMPs are members of the MMP gene family and play an important role in the overall availability of active MMPs. Hence, it is important to determine the TIMP expression profile of fibroblasts in response to S. aureusand S. aureuscomponents. In our current study, we used culture supernatants obtained from an S. aureusstrain isolated from synovial fluid of a patient with SA (ATCC), a clinical isolate (U1), and its Agr/Sar Adouble-loci-deleted mutant U930 (strains obtained from M. Smeltzer). The results presented in Figure [7a,b](#F7){ref-type="fig"} indicate a notably increased induction of TIMP-1, -2, and -3 mRNA by the Agr/Sar Adeletion mutant (U930) of the isogenic parent wild-type strain (U1) and the ATCC strain isolated from the synovium of a patient with arthritis. It may be speculated that the effective MMP available upon infection with Agr/Sardeletion mutant is likely to be less compared with the parent isogenic strain. However, further studies to examine expression of other MMPs as well as analysis to estimate enzymatically active MMPs by zymogram will be required to determine whether genes under the control of Saror Agrhave any effect on the expression of functional MMPs. MAPK gene expression in synovial fibroblasts from patients with RA and OA ------------------------------------------------------------------------- Members of the MAPK gene family (such as ERK1/2 and p38MAPK) are involved in the induction of MMPs through activation protein (AP-1) transcription factors. We therefore analyzed the mRNA expression levels of 12 members of the MAPK family using the MultiGene-12 RT-PCR profiling kit from Superarray Bioscience Corporation. Synovial fibroblasts obtained from patients with RA and OA were exposed to 25 μg of total proteins from bacterial culture supernatant or cell lysate, and total RNA was isolated 6 hours later, reverse-transcribed, and assayed for mRNA of 12 MAPK genes. Several of the MAPK family members were upregulated. The ratio between the intensities of each MAPK gene to that of GAPDHis depicted in Figure [7](#F7){ref-type="fig"}. Significant increases in ERK2, ERK1, MAPK4, JNK1, JNK2, p38b, and p38g were observed in dermal fibroblasts treated with S. aureusculture supernatant and cell lysate-treated compared with untreated fibroblasts (Figure [8a](#F8){ref-type="fig"}; p\< 0.05) and in synovial fibroblast-treated compared with untreated fibroblasts (Figure [8b](#F8){ref-type="fig"}; p\< 0.05). Similar increases in these MAPK gene family members were noted in IL-1β/TNF-α-treated fibroblasts (Figure [8a,b](#F8){ref-type="fig"}; p\< 0.05). Tyrosine phosphorylation in human dermal fibroblasts exposed to S. aureusculture supernatant ---------------------------------------------------------------------------------------------- Using a cell-based ELISA system (RayBiotech, Inc.), tyrosine phosphorylation was assessed in human dermal fibroblasts after 30-minute exposure to 25 μg of total protein from filtered culture supernatant of S. aureusand in fibroblasts treated with 10 ng/ml each of rhIL-1β and rhTNF-α. There was a significant increase in phosphotyrosine in S. aureusculture supernatant-treated cells, similar to that observed in IL-1β/TNF-α-treated cells (Figure [9](#F9){ref-type="fig"}; p\< 0.05). Overall, our data indicate that S. aureuscomponents induce multiple MMP expression in human dermal and synovial fibroblasts and that the response is similar to that induced by IL-1β/TNF-α. The expression pattern of MAPK gene expression also indicates the possibility of a signal transduction pathway akin to that induced by the inflammatory cytokine pathway. Our data also indicate that the virulence gene loci (namely, Sarand Agr) are not determinants of S. aureus-induced MMP mRNA expression. Discussion ========== We have shown that the culture supernatants and whole bacterial lysate from S. aureusinduce multiple MMPs from human dermal and synovial fibroblasts. Several genes of the MAPK pathways were upregulated in treated fibroblasts, and phosphotyrosine proteins were significantly elevated. Using fractionated S. aureusculture supernatants, we have shown that the best MMP induction was by components that fall within the molecular-weight range of 30 to 50 kDa. Interestingly, culture supernatants and bacterial cell lysates obtained from S. aureusgrown in the presence of rhIL-1β induced notably higher levels of MMPs compared with S. aureusgrown in the absence of rhIL-1β. The overall spectrum of MMP induction by S. aureuscomponents was similar to that elicited by a combination of IL-1β and TNF-α. Our in vitroMMP mRNA expression analysis showed that mutants lacking Sar Aand Agrloci and their parent isogenic strain induced comparable levels of MMP mRNAs; however, the mutant strains induced notably higher levels of TIMP-1, -2, and -3 mRNAs in human fibroblasts. To our knowledge, this is the first report on multiple MMP/TIMP induction by fractionated S. aureusculture supernatants and whole bacterial cell lysates in human dermal and synovial fibroblasts. SA is the most commonly reported bacterial complication of RA. The risk is highest in severe, longstanding, seropositive disease. The clinical presentation of joint infection is frequently atypical, and in 25% of cases, the infection is polyarticular. S. aureusis the most common causative organism \[[@B20]\]. Staphylococcal infections can be hard to eradicate from RA joints and often surgery is required \[[@B21]\]. TNF-α plays an important role in the host defense against infection. Inhibition of its activity could therefore be anticipated to augment the risk of infection in patients with RA. The reasons for the predominance of S. aureusin SA and the mechanisms of pathogenecity are not yet fully understood. The synovium of patients with RA is rich in IL-1β. We have previously shown that S. aureuscan bind to IL-1β and use it as a growth factor \[[@B22]\]. A recent report by McLaughlin and Hoogewerf \[[@B23]\] showed that the growth and replication of S. aureusin a biofilm are significantly increased by the addition of rhIL-1β. We have also observed that rhIL-1 can modulate the gene expression in S. aureusincluding the bicomponent leukotoxins and some of the surface adhesion molecules collectively called MSCRAMMs (microbial surface components recognizing adhesive matrix molecules) in addition to some of the genes in the pathogenecity island of S. aureus(Kanangat and colleagues, manuscript submitted). We speculate that the IL-1-rich synovial milieu might potentially contribute to the increased frequency of S. aureusin patients with RA-SA and that the host-derived MMPs induced by S. aureusmight accelerate the pathogenesis of SA. Our data on the induction of MMPs by S. aureusculture supernatants and cell lysates compares well with the previous report by Williams and colleagues \[[@B16]\], who demonstrated MMP-1 and -3 expression by articular cartilage upon exposure to purified culture supernatant from S. aureus. We have extended this observation by showing expression of a wide range of MMPs, including MMP-7 (with a proven in vivoassociation with S. aureus-induced SA), by human synovial as well as dermal fibroblasts in response to S. aureuscomponents. The profile was similar to that induced by a combination of IL-1/TNF, which might indicate the involvement of an inflammatory cytokine-mediated pathway in the observed induction of MMPs by S. aureus. S. aureusculture supernatants and cell lysates have a wide variety of proteins, and identification of the components that are actually responsible for inducing the MMP induction is essential to determine the mechanisms of induction as well as to rationally design intervening agents against bacterial products. Toward this we end, we have narrowed down the possible candidates to molecular-weight groups of the range of \>30 to 50 based on our experiments using Centricon filtration of the culture supernatants. Because the molecular weight of the chemically purified PGN used in previous studies is not known, we are not in a position to determine whether PGN is included in the stated molecular-weight range. At this time, we have not identified the components beyond the molecular level; nevertheless, this rules out the possibility of some of the recently described low-molecular-weight proteins such as the 19-kDa extracellular fibrinogen-binding protein that inhibits complement activation \[[@B24]\]. Certain complement components have been reported to activate MMPs \[[@B25]\]. The results of the fractionated supernatants also tentatively rule out the possibility of the exotoxin akin to the toxic shock syndrome protein described by Ren and colleagues \[[@B26]\] and the enterotoxin H described by Su and Wong \[[@B27]\]. Although speculative at this juncture, it is possible that the active components in the 30 to 50 kDa could potentially be the novel 38.5-kDa protein named extracellular matrix-binding protein described by Hussain and colleagues \[[@B28]\]. Joint destruction by S. aureusis very rapid if not treated appropriately. Although direct erosion of the joint architecture by S. aureusproteases/toxins cannot be completely ruled out, continued degradation of extracellular matrix component and the joint architecture even after clearing the infection and debris from the joint cavity indicates the possibility of host-derived proteases in causing joint pathology. Previous studies have shown the release of active MMP-1 and MMP-3 by human articular cartilage upon exposure to sterile purified S. aureusculture medium \[[@B16]\]. The enzymatic profile was similar to that induced by IL-1. The authors concluded that the collagenase and stromelysin released by articular cartilage could contribute to extensive destruction of human cartilage in SA. The exoproteases of S. aureushave been proposed as virulence factors during S. aureusinfections. Calander and colleagues \[[@B29]\], using wild-type S. aureusstrain 8325-4 and its mutants lacking aureolysin, serine protease, and cysteine protease, demonstrated in a murine SA model that inactivation of the exoprotease genes did not affect the frequency or the severity of joint pathology. Intra-articular injection of PGN into murine joints triggered arthritis in a dose-dependent manner \[[@B17]\]. A single injection of this compound caused massive infiltration of macrophages and polymorphonuclear cells with signs of cartilage and/or bone destruction, lasting for at least 14 days, indicating that PGN exerts a central role in joint inflammation triggered by S. aureus\[[@B17]\]. The significance of MMP-7 expression in SA was examined by Gjertsson and colleagues \[[@B30]\] using MMP-7-deficient mice and congeneic controls. These mice were inoculated (intravenously) with an arthritogenic dose of S. aureusLS-1, and the mice deficient for MMP-7 developed significantly less-severe arthritis both clinically and histologically despite significantly increased numbers of live bacteria within the internal organs. Interestingly, in vitroresponses to staphylococcal antigens and superantigens were not different between MMP-7^+/+^and MMP-7^-/-^mice in terms of cytokine production. MMP-7 facilitates migration of both macrophages and neutrophils, and the authors therefore conclude that modulation of SA by MMP-7 might be due to changes in peripheral leukocyte distribution. Also, studies by Wang and colleagues \[[@B31]\] have shown that addition of PGN to whole human blood resulted in enhanced levels of MMP-9 within 1 hour and significant enhancement of MMP-9 secretion from the neutrophils was obvious within 30 minutes of incubation with S. aureusPGN. Host-cell MMPs are necessary for efficient clearance of infection by accelerating the recruitment of effector cells to kill pathogen and to resolve inflammation and for subsequent tissue remodeling \[[@B9]\]. However, excessive MMPs after infection and inflammation may cause tissue damage leading to immunopathology. MMPs are secreted by both inflammatory and connective tissue cells such as fibroblasts, fibroblast-related cells, chondrocytes, neutrophils, and monocytes/macrophages in response to both infectious assaults and inflammatory cytokines such as TNF-α and IL-1β \[[@B32],[@B33]\]. The regulation of MMP secretion is dependent on the cell type and stimulus. Signal transduction pathways that involve the MAPK and prostaglandin E2/cyclic AMP pathways are considered to be crucial, although the involvement of other pathways cannot be ruled out \[[@B34],[@B35]\]. The transcription factors c-jun (cellular homologue of viral oncogene jun) and c-fos (cellular homologue of FBJ murine sarcoma virus oncogene), members of the AP-1 family, are involved in the transcription of MMP-1 and these AP-1 factors are induced through MAPK signal transduction. We determined the mRNA levels of some of the MAPK family members in synovial fibroblasts from patients with RA or OA treated with S. aureusand lysate culture supernatant or IL-1β/TNFα. mRNAs for several of the MAPK family members were upregulated by S. aureuslysates and culture supernatants similar to those treated with IL-1β/TNF-α. Also, the overall phosphotyrosine expression was enhanced in fibroblasts treated with S. aureuscomponents. The increase in phosphotyrosine in fibroblasts treated with S. aureuscomponents or IL-1β/TNFα was comparable. The importance of the proinflammatory cytokines TNF-α and lymphotoxin (LT)-α in an experimental model of S. aureussepsis and arthritis was examined by Hultgren and colleagues \[[@B36]\]. Using TNF-α/LT-α-double-deficient mice, they showed that in mice inoculated intravenously with a toxic shock syndrome toxin-1-producing S. aureusstrain LS-1, mortality in TNF-α/LT-α-deficient mice was 67%, with no mortality in the controls. Those results correlated with a significantly decreased phagocytosis in vitroand inefficient bacterial clearance in vivoin mice lacking the capacity to produce TNF-α/LT-α. Infection of mice with a lower dose of staphylococci resulted in an overall low mortality, but the frequency of arthritis was significantly higher in the wild-type group compared with the TNF-α/LT-α-deficient mice (40% versus 13%). Synovitis and erosivity were lower in TNF-α/LT-α-deficient mice compared with wild-type controls. This study demonstrates the detrimental role of TNF-α/LT-α as mediators of the inflammatory response in S. aureusarthritis. TNF-α is a potent inducer of several types of MMPs. Although IL-1 or TNF induced by S. aureuscould very well be contributing to the joint destruction (either through induction of MMPs or through other degradative pathways), studies by Kimura and colleagues \[[@B37]\] showed that blocking TNF and IL-1 does not significantly prevent the late-stage destruction of joint architecture in arthritis induced by S. aureus. In the murine heat-killed S. aureus-induced arthritis model, TNF-α and IL-1β peaked at 2 and 24 hours after the injection of heat-killed S. aureus, respectively. Simultaneous administration of anti-TNF-α monoclonal antibody (mAb) and IL-1 receptor antagonist (IL-1ra) with S. aureusresulted in significant inhibition (80%) of 12-hour leukocyte infiltration. However, leukocyte infiltration at 24 hours and beyond and the loss of proteoglycan in S. aureus-induced arthritis were not affected by anti-TNF-α mAb, IL-1ra, or their combination. These results suggest that TNF-α and IL-1β involvement in the pathogenesis of S. aureus-induced arthritis may be limited to the initial phases of inflammation. The authors suggested that suppressing TNF-α and IL-1 may not be effective in the clinical treatment of Gram-positive bacteria-induced arthritis. With respect to the molecular pathways involved in S. aureus-induced MMP expression in fibroblasts, our results suggest that S. aureuscomponents could use a pathway(s) similar to that of IL-1β/TNF-α given that the MMP expression pattern, MAPK gene expression, and phosphotyrosine levels were similar in fibroblasts treated with S. aureuscomponents or IL1β/TNF-α. It is also important to note that S. aureusis capable of inducing synthesis of inflammatory cytokines such as IL-1β and TNF-α from host cells \[[@B13]\]. Whether the MMP induction in fibroblasts by S. aureuscomponent(s) is due to the cytokine/chemokine induced by S. aureusis not known at present. Previous studies by Wang and colleagues \[[@B31]\] have shown that inhibitors of p38 MAPK (SB202190) and ERK1/2 (PD98059) and inhibitors of Src Tyrosine kinase (PP2) and PI3-K (LY294002) effectively blocked PGN-mediated MMP-9 upregulation in neutrophils. The potential involvement of the Toll-like receptor (TLR)-2 in S. aureusPGN-induced joint inflammation and destruction was postulated in a study by Kyburz and colleagues \[[@B18]\]. Cultured synovial fibroblasts obtained from patients with RA or OA were stimulated with PGN. The expression of various integrins was determined by fluorescence-activated cell sorting. TLR-2 and MMP mRNAs as measured by real-time PCR were upregulated in fibroblasts treated with staphylococcal PGN. The levels of IL-6 and IL-8 in the culture supernatants were also increased by treatment with PGN. We demonstrated that cultured synovial fibroblasts express low levels of TLR-2 and TLR-9 mRNA. Anti-TLR-2 mAbs significantly inhibited production of IL-6 and IL-8 induced by stimulation with PGN. The authors concluded that bacterial PGNs activate synovial fibroblasts, partially via TLR-2, to express integrins, MMPs, and proinflammatory cytokines. There was no mention regarding MMP expression after TLR blockade, and it remains unclear whether TLR is involved in MMP expression in a more direct way. Our preliminary results have shown that S. aureusculture supernatant and whole cell lysate induce the mRNA expression of several members of the TLR family, including TLR-2 (data not shown). To elucidate the MMP induction by S. aureus, we turned to two well-characterized mutant strains of S. aureuslacking Sar Aand Agr. Agrand Sarare the two best-characterized loci responsible for modulating the expression of S. aureusvirulence factors \[[@B38]\]. S. aureusstrains lacking either locus have been shown to result in attenuation of S. aureusin several models of staphylococcal diseases \[[@B39]-[@B41]\]. Recent investigations by Blevins and colleagues \[[@B19]\] have also shown that mutation of Sar Aand/or Agrcaused reduced capacity to induce both SA and osteomyelitis. The exact mechanisms of reduced effectiveness of Sar/Agrmutants to cause SA or osteomyelitis are not known. Studies by Nilsson and colleagues \[[@B41]\] showed that mice inoculated with the Sar A^+^staphylococcal strain exhibited a more pronounced T- and B-lymphocyte activation and higher levels of serum IL-6 and IFN-γ, compared with a Sar A^-/-^mutant, and infection with Sar A^+^staphylococci induced pronounced weight loss as well. These studies suggested that Sar Alocus might control molecules that are important virulence determinants in the induction and progression of SA. We therefore tested the MMP-1, -3, and -13 expression patterns in response to Sar, Agr, or Sar/Agrmutants in human dermal fibroblasts. The three MMPs were selected due to their known involvement in various models of arthritis and their respective degrading actions on collagen type I, II, and III and proteoglycan, which are important constituents of connective tissues and cartilage of the joints. Our results did not show any significant differences in MMP-1 and MMP-3 mRNA levels, and 13 mRNA levels were minimal and could not be quantified with reasonable accuracy in dermal fibroblasts upon exposure to culture supernatants or cell lysates obtained from the mutants and isogenic parent strain (Figure [6](#F6){ref-type="fig"}). However, interestingly, the expression of TIMPs was notably increased in fibroblasts treated with Sar/Agrmutants compared with isogenic parent strain (Figure [7a,b](#F7){ref-type="fig"}). This could mean that the effective biologically active MMPs are less abundant in cells treated with the Sar/Agrmutants compared with cells treated with isogenic parent strain. It will be necessary to estimate the levels of biologically active MMPs (by zymograms) to determine the net effect of Sar/Agrmutants on MMP expression. Temporal estimation of biologically active MMPs in the joints after infection with isogenic parent and mutants will help to clarify the issue of MMPs as a factor in the observed differences in severity of diseases caused by wild-type and mutant strains. Bacteria may secrete proteolytic enzymes such as the thermolysin family secreted by Pseudomonas aeruginosaand Vibrio cholerawhich activate pro-MMP-1, -8, and -9 \[[@B42]\]. Also, proteases from the oral pathogen Porphyromonas gingivalisactivate MMP-1, -3, and -9 \[[@B43]\]. If the bacterial-derived proteases are required for virulence, such proteases can be attractive therapeutic targets because their inhibition can be achieved without affecting the normal expression and function of MMPs. There are reports of other staphylococcal virulence factors associated with the pathogenesis and severity of SA \[[@B43]-[@B45]\]. Whether these virulence factors are associated with MMP/TIMP expression remains to be seen. In addition to the bone and joint infections, S. aureusis also the prime causative agent in many skin and soft tissue infections (SSTIs), which can be manifested as superficial to deep-seated and at times become life-threatening \[[@B46]-[@B48]\]. Due to lack of validated clinical evidence, it is often difficult to recommend general treatment options \[[@B49]\]. The pathogenesis of SSTI is not understood well, and the treatment is guided mostly by epidemiological pattern and microbiological information \[[@B50]\]. Due to the emergence of MRSA, it is important to understand the mechanisms of tissue destruction in soft tissue infections which could lead on the identification of novel therapeutic targets. Our current in vitrodata and the in vivodata reported previously by others implicate that host-derived metalloproteinases could be involved, at least in part, in tissue destruction. Excessive expression of these metalloproteinases induced by S. aureuscould lead to the destruction of the soft tissue connective tissue architecture. Conclusion ========== We have shown that S. aureusis a potent inducer of multiple MMPs in human dermal and synovial fibroblasts. Our studies also indicate that MAPK-mediated signal transduction pathway involving proteins that are phosphorylated at tyrosine residues might play a role in S. aureus-induced MMP expression. Enhanced expression of immunoreactive MMPs by cell lysate obtained from S. aureusgrown in the presence of rhIL-1 indicates that an inflamed milieu such as RA synovium might augment the MMP induction potential of S. aureus. More specific identification of the component(s) of S. aureusinvolved in the upregulation of MMP and associated signal transduction pathways may help in identifying novel targets for intervention. Based on our results, we propose that biologically active MMPs (reflected as a ratio of MMPs to TIMPs) induced by S. aureuscould potentially accelerate the joint destruction in SA. Abbreviations ============= Agr = accessory gene regulator; AHO = acute hematogenous osteomyelitis; AMV = avian myeloblastoid virus; AP-1 = activation protein; ATCC = American Type Culture Collection; DMEM = Dulbecco\'s modified Eagle\'s medium; ELISA = enzyme-linked immunosorbent assay; ERK = extracellular signal regulated kinase; FBS = fetal bovine serum; GAPDH = glyceraldehyde phosphate dehydrogenase; IFN = interferon; IL = interleukin; IL-1ra = interleukin-1 receptor antagonist; JNK = c-jun N-terminal kinase; LT = lymphotoxin; mAb = monoclonal antibody; MAPK = mitogen-activated protein kinase; MMP = matrix metalloproteinase; MRSA = multidrug-resistant Staphylococcus aureus; OA = osteoarthritis; PCR = polymerase chain reaction; PGN = peptidoglycan; RA = rheumatoid arthritis; RT-PCR = reverse transcription-polymerase chain reaction; SA = septic arthritis; Sar = staphylococcal accessory regulator; SSTI = skin and soft tissue infection; TIMP = tissue inhibitor of metalloproteinases; TLR = Toll-like receptor; TNF = tumor necrosis factor. Competing interests =================== The authors wish to state that they have no commercial or other association (such as pharmaceutical stock ownership, consultancy, advisory board membership, patents, or research funding) that might contribute to competing interests. Authors\' contributions ======================= SK conceived the idea, designed the study, and prepared the manuscript draft. AP, AK, KH, and DS all helped in the experimental design and data analysis and provided creative and critical suggestions after reviewing the manuscript. MS provided insight into the murine SA model, provided the mutant and isogenic strains of S. aureus, and critically reviewed the manuscript. WA assisted with the RT-PCR work, cell culture maintenance, MMP protein arrays, as well as in review of literature. All authors read and approved the final manuscript. Acknowledgements ================ This work was supported by a Public Health Service Grant (R03 AR47936-01) from the National Institute of Arthritis and Musculoskeletal and Skin Diseases (National Institutes of Heath); a University of Tennessee Health Science Center (UT-HSC) Connective Tissue Diseases\' Center of Excellence grant, a special fund from the Department of Medicine, UT-HSC, Memphis, TN, USA; and a pilot study grant from the Department of Veterans Affairs, Memphis, TN, USA. Figures and Tables ================== ![Induction of multiple matrix metalloproteinase (MMP) protein by Staphylococcus aureusin human dermal fibroblasts. Confluent monolayers of human primary dermal fibroblasts were exposed to (I)phosphate-buffered saline, (II)25 μg/well of total protein from clarified whole cell lysates of S. aureuscells, (III)25 μg of total protein from the filtered culture supernatant of S. aureus, and (IV)treated with 10 ng/m each of interleukin-1β/tumor necrosis factor-α (IL-1β/TNF-α) for 48 hours. The fibroblast culture supernatants were harvested and the MMPs and tissue inhibitors of metalloproteinases (TIMPs) were detected using the MMP-Array kit from RayBiotech, Inc. (Norcross, GA, USA). Culture supernatant and cell lysate from S. aureusinduced the expression of MMP-1, MMP-2, MMP-10, and MMP-13. A similar pattern was observed in IL-1β/TNF-α-treated fibroblasts, which were slightly more intense. There was upregulation of both TIMP-1 and TIMP-2 as well.](ar2086-1){#F1} ![Induction of multiple matrix metalloproteinase (MMP) mRNAs in human dermal fibroblasts (HDF) by Staphylococcus aureus. Confluent monolayers of fibroblasts were exposed to 25 μg/well of total protein from culture supernatants (SUP), 25 μg/well of total protein from bacterial cell lysate, or 10 ng/ml each of rhTNF-α and rhIL-1β for 8 hours. Total RNA was isolated and reverse-transcribed. The cDNA was amplified and quantified using SYBR green real-time polymerase chain reaction. All the MMPs tested were expressed in dermal fibroblasts exposed to S. aureuscomponents. Both culture supernatants and cell lysate from S. aureussignificantly enhanced the expression of multiple MMPs (namely, MMP-1, -2, -3, -7, -10, and -11; p\< 0.05). The dermal fibroblasts responded similarly to a combination of rhTNF-α and rhIL-1β in terms of multiple MMP mRNA expression. IL, interleukin; TNF, tumor necrosis factor.](ar2086-2){#F2} ![Induction of multiple matrix metalloproteinase (MMP) mRNAs in human synovial fibroblasts (SFs) by Staphylococcus aureus. MMP mRNA expression profile in confluent monolayers of SFs obtained from de-identified (a)rheumatoid arthritis (RA) and (b)osteoarthritis (OA) patients treated with culture supernatants (SUP) (25 μg of total protein/well) and cell lysates (25 μg of total protein/well) were determined using SYBR green real-time reverse transcription-polymerase chain reaction. Unlike untreated dermal fibroblasts, untreated SFs from RA and OA patients had higher basal multiple MMP mRNA expression. Similar to the dermal fibroblasts, most MMP mRNAs tested were elevated in RA and OA fibroblasts in response to S. aureuscell lysate and culture supernatants. Levels of significantly elevated MMPs (MMP-1, -2, -3, -7, -10, and -11; p\< 0.05) are shown in (a)and (b). The MMP mRNA expression pattern in response to IL-1β/TNF-α and S. aureuslysate was similar in RA and OA fibroblasts. IL, interleukin; TNF, tumor necrosis factor.](ar2086-3){#F3} ![Potentiation of matrix metalloproteinase (MMP) protein expression in human fibroblasts by Staphylococcus aureusgrown in the presence of recombinant human interleukin (rhIL)-1β. The cell pellets from S. aureusgrown in the presence or absence of rhIL-1β were washed repeatedly to prepare cell lysate. Enzyme-linked immunosorbent assay (R&D Systems, Inc., Minneapolis, MN, USA) was carried out to make sure that the bacterial cell lysate was devoid of any traces of rhIL-1β. Confluent monolayers of human dermal fibroblasts were exposed to 25 μg/well of IL-1β-free cell lysate obtained from S. aureusfor 48 hours. The supernatants were collected and expression of multiple MMP protein was assessed using multi-MMP-Array kit from RayBiotech, Inc. (Norcross, GA, USA) as described in the text. The data presented show that MMP-2, -3, and 8 are relatively more expressed in fibroblasts treated with S. aureuscell lysate obtained from strain grown in the presence of 5 and 15 ng/ml rhIL-1β. Tissue inhibitor of metalloproteinases (TIMP)-4 expression was also slightly enhanced in fibroblasts treated with S. aureuslysate obtained from strain treated with 15 ng/ml IL-1β.](ar2086-4){#F4} ![Induction of matrix metalloproteinase (MMP)-1 and MMP-3 mRNA in response to Staphylococcus aureuscomponents separated by molecular-weight difference. Fractionated S. aureusculture supernatant (15 μg/ml per well) obtained by centrifuging through Centricon filters with molecular-weight cutoff points \<30 kDa, 30 to 50 kDa, and \>50 kDa was added to confluent monolayers of dermal fibroblasts. Total RNA was isolated after 8 hours. The mRNA levels of MMP-1 and MMP-3 were measured by SYBR green real-time reverse transcription-polymerase chain reaction. The active components were mostly in the 30 to 50 and \>50 kDa molecular-weight range as evidenced by significantly elevated expression of MMP-1 and MMP-3 by Centricon fractions 30 to 50 and \>50 kDa (p\< 0.05; 30 to 50 kDa and \>50 kDa fractions were compared with \<30 kDa fractions).](ar2086-5){#F5} ![Matrix metalloproteinase (MMP) mRNA induction by staphylococcal accessory regulator (Sar), accessory gene regulator (Agr), and Sar/Agrmutants of Staphylococcus aureus. MMP-1 and MMP-3 mRNAs in confluent monolayers of human dermal fibroblasts exposed to 25 μg of total protein/well per milliliter from cell lysate and 25 μg/well per milliliter of total protein from culture supernatants (SUP) of parent (U1) mutant strains (U155 Sar^-/-^; U929 Agr^-/-^; U930 Sar/Agr^-/-^) were measured by SYBR green real-time reverse transcription-polymerase chain reaction 8 hours after exposure. No significant differences in mRNA levels of MMP-1 and MMP-3 induced by the mutants and their isogenic parent strain were observed.](ar2086-6){#F6} ![Enhanced expression of tissue inhibitor of metalloproteinases (TIMP) mRNAs in human dermal fibroblasts treated with culture supernatants (a)and whole bacterial cell lysates (b)obtained from Sar A^-/-^and Sar A^-/-^Agr^-/-^mutants of Staphylococcus aureusand isogenic parent strain. Culture supernatants and cell lysates obtained from an S. aureusstrain isolated from synovial fluid of a septic arthritis patient (obtained from American Type Culture Collection, Manassas, VA, USA), a clinical isolate (U1), its Sar Amutant (U929), and Agr/Sar Adouble-loci-deleted mutant (U930) (strains obtained from M. Smeltzer) were used. Equal amounts of total protein from all samples were added to confluent monolayers of fibroblast cultures. Total RNA was harvested from the cells 12 hours after exposure to the supernatants. The levels of TIMP-1, -2, and -3 mRNAs were estimated by real-time SYBR green reverse transcription-polymerase chain reaction. The values plotted represent the ratios of the threshold cycle values of TIMPs to that of the housekeeping gene GAPDH(glyceraldehyde phosphate dehydrogenase). Agr, accessory gene regulator; Sar, staphylococcal accessory regulator; SUP, culture supernatant.](ar2086-7){#F7} ![Mitogen-activated protein kinase (MAPK) family mRNA expression profile in human dermal fibroblasts (a)and human synovial fibroblasts (b)exposed to Staphylococcus aureuswhole cell lysate and filtered culture supernatant. Confluent monolayers of de-identified human dermal fibroblasts from normal volunteer and synovial fibroblasts from a patient with rheumatoid arthritis were treated with 25 μg/ml per well of S. aureuswhole lysate, filtered culture supernatant, and a combination of rhIL-1β and TNF-α (10 ng/ml each) for 8 hours. Cells were harvested and total cellular RNA was isolated and reverse-transcribed. The MAPK family gene expression was analyzed using the Human MAPK Gene Family I Multigene-12 reverse transcription-polymerase chain reaction (PCR) profiling kit. The semiquantitative values were generated by determining the ratios of the band intensities of respective PCR products to that of housekeeping gene GAPDH(glyceraldehyde phosphate dehydrogenase) in each sample. The experiments were repeated three times. The band densities were determined using three-dimensional densitometric scanning software from Alpha Innotech Corporation (San Leandro, CA, USA). Of the 11 tested Human MAPK Family I genes (ERK1, ERK2, MAPK2/4, ERK3, ERK5, JNK1, JNK2, JNK3, p38b MAPK, p38g MAPK, and p38delta), ERK1, ERK2, JNK1, JNK2, MAPK4, and p38b were elevated significantly in both dermal and synovial fibroblasts upon exposure to S. aureuscomponents (p\< 0.05 for individual MAPK gene family member when values from untreated fibroblasts and treated fibroblasts were compared). The response was similar in both cell lines in response to IL-1/TNF. ERK, extracellular signal regulated kinase; IL, interleukin; JNK, c-jun N-terminal kinase; SUP, filtered culture supernatant; TNF, tumor necrosis factor.](ar2086-8){#F8} ![Tyrosine phosphorylation in human dermal fibroblasts exposed to Staphylococcus aureusculture supernatant. Using a cell-based enzyme-linked immunosorbent assay system (RayBiotech, Inc., Norcross, GA, USA), tyrosine phosphorylation was assessed in human dermal fibroblasts exposed to 25 μg of total protein/ml from filtered culture supernatant of S. aureus, 25 μg/ml total protein from S. aureuscell pellet lysate, and in fibroblasts treated with 10 ng/ml each of rhIL-1β and rhTNF-α. These assays were carried out in 96-well plates supplied by the manufacturer. The cells were exposed to various agents for 30 minutes before assaying for tyrosine phosphorylation. The assays were carried out in triplicate. There was significant increase in tyrosine phosphorylation in S. aureusculture supernatant-treated cells similar to IL-1β/TNF-α-treated cells compared with untreated fibroblasts (p\< 0.05). IL, interleukin; OD, optical density; SUP, culture supernatant; TNF, tumor necrosis factor.](ar2086-9){#F9}	{ "pile_set_name": "PubMed Central" }
We thank Fox et al. ([@B2]) for their interest in our research and welcome this opportunity to respond to their commentary. They argue that participants in our studies enjoyed "just thinking" more than we claimed (Wilson et al., [@B5]). We found some irony in their position, because we began this line of research with a similar hypothesis. As the data came in we were surprised that participants did not enjoy deliberative thought very much, even when we went to some lengths to give them time to prepare and choose their topics (i.e., in our "prompted fantasy" conditions). We are thus in the rather amusing position of explaining why both Fox et al.\'s interpretation and our initial hypotheses are wrong. Our view can be summarized with a simple thought experiment. Imagine you are a venture capitalist, and a filmmaker comes to you asking for money to make a full-length feature film. You decide it would be wise to show the film maker\'s 15-min trailer to some focus groups. You find that (a) on average, the focus groups rate the trailer as much less enjoyable than a variety of everyday activities, such as listening to music and surfing the web; (b) between a third and a half of the people in the focus groups disobey the instructions to watch the trailer and instead prefer to do something else, such as playing with their cell phones; and (c) the majority of the men (and a quarter of the women) in the focus groups choose to give themselves an unpleasant electric shock while watching the trailer, even though they had said earlier that they would pay to avoid getting the shock. Would you invest? Probably not. These are, of course, precisely the results we found in our studies when we left people alone with their thoughts. As such, we do not think it is unreasonable to conclude that the participants found deliberative thought to be neither engaging nor enjoyable. Although our space is limited, we will respond to a few of Fox et al.\'s specific arguments. First, the results they report in Figures 1A--C are entirely consistent with the data we reported in our Table 1. Instead of quibbling over what it means for participants to rate something as somewhat enjoyable and somewhat boring, we believe it is more informative to compare the ratings of participants under different experimental conditions and to observe their actual behavior. In Study 8, participants rated everyday external activities as much more enjoyable than deliberate thinking (the distribution of responses in the two conditions was almost non-overlapping). Fox et al. suggest that we stacked the deck in favor of the external activities by letting people choose among activities that "were not banal or boring." But note that participants in the "thinking" condition were completely free to think about anything they wished, to choose topics that were "tailored to their personal interests," and to switch between topics. They could choose to think about past accomplishments, current interests, or future plans, or even, to build on Fox et al.\'s analogy, chocolate and sex. The fact that participants much preferred the everyday external activities is quite telling, we think, and does not speak well for the appeal of "just thinking." We also find it revealing that 32--54% of participants (depending on the study) disobeyed instructions to "just think" and performed external activities they were specifically asked to avoid, such as using their cell phones or doing school work. (Note that these figures include only those who admitted to doing the banned activities.) If deliberative thinking is as enjoyable as Fox et al. suggest, why did so many participants disobey explicit instructions to do it for relatively short periods of time? Fox et al. are unimpressed by the results of our study in which 67% of men and 25% of women administered at least one shock to themselves. Their argument that participants were simply curious is at odds with the fact that all participants had already experienced the shock and that we specifically selected only those participants who reported that they would pay to avoid another shock. Fox et al. argue that 14 of the 18 participants who shocked themselves reported that they did so out of "curiosity about (or interest in) the quality of the shock or its effects as their motivation." But note that what many of these participants were interested in was how unpleasant the shock would be, such as the participant who wrote, "I wanted to see if the shock felt as painful the second time as it did the first time." If thinking is so enjoyable, why would participants be willing to re-experience an unpleasant stimulus just to see if it is still unpleasant? (Not to mention the fact that participants\' reports about the reasons for their behavior are, in large part, post-hoc rationalizations that may not reflect their actual reasons; Nisbett and Wilson, [@B4]; Johansson et al., [@B3]). Lastly, we note an error in one of Fox et al.\'s statements. Participants did not rate their "enjoyment of 'just thinking,"' but rather their enjoyment of the thinking period, including any shocks they administered. Thus, if people shocked themselves because they were bored, and the shocks helped alleviate that boredom, it is not surprising that the shockers rated the overall experience similarly to the non-shockers. More fundamentally, Fox et al. seem to have misunderstood what our studies were about. They described our research as having investigated "spontaneous thought," which one of the authors defined as "the unintended, non-working, non-instrumental mental content that comes to mind unbidden and effortlessly" (Christoff, [@B1], p. 52). But in fact we examined directed conscious thought, because in virtually all of our studies we specifically instructed participants to think in enjoyable ways. As Christoff ([@B1]) noted, "It is impossible, after all, to instruct participants to have a spontaneous thought" (p. 57). No other studies, to our knowledge, have examined how successfully people can engage in deliberative thinking when the goal is enjoyment or positive affect. Fox et al.\'s comparison of our research to previous studies of spontaneous thought is thus moot. Clearly some kinds of thinking are enjoyable. As we said at the end of our article, "There is no doubt that people are sometimes absorbed by interesting ideas, exciting fantasies, and pleasant daydreams." But we also noted that "it may be particularly hard to steer our thoughts in pleasant directions and keep them there" (p. 77). The preponderance of the evidence supports this conclusion, we believe, and is contrary to both Fox et al.\'s and our own initial hypotheses. Conflict of interest statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The writing of this article was aided by National Science Foundation Grant BCS-1423747. [^1]: This article was submitted to Cognition, a section of the journal Frontiers in Psychology. [^2]: Edited by: Daniel Lakens, Eindhoven University of Technology, Netherlands [^3]: Reviewed by: Bernhard Hommel, Leiden University, Netherlands	{ "pile_set_name": "PubMed Central" }
Background {#Sec1} ========== End-stage renal disease (ESRD) is increasing worldwide; at the end of 2009 there were nearly 38,000 people living with end-stage renal disease in Canada \[[@CR1]\]. Most patients with ESRD do not receive a kidney transplant and require dialysis to survive. Quality of life for patients receiving dialysis has been rated similarly to patients with metastatic cancer \[[@CR2]\], in part because of the intrusiveness of ongoing dialysis (whether hemodialysis or peritoneal dialysis), and in part because of symptoms related to the disease, including depression which occurs in up to 30% of patients \[[@CR3], [@CR4]\]. Qualitative research provides insight into the views of respondents, including their attitudes, needs, beliefs and feelings. This provides some insight into patients' perspectives on their illness. Understanding patients' views can help health care providers tailor the type of treatments offered to patients, as well as understand when, how and what information to provide to patients about the treatment options. This is particularly relevant in patients with chronic diseases, such as kidney failure, since they deal with their illness and the consequences of its treatment on a daily basis. We build on prior work to determine the research priorities of patients on or nearing dialysis, their caregivers and the health care professionals who look after them. To develop the list of top research priorities, we initially surveyed patients, their caregivers and health care professionals across Canada, collating, and ranking their responses. A top ten list was subsequently developed at an in-person meeting \[[@CR5]\]. During this priority setting exercise, we noted that many of the items expressed by respondents represented not only unanswered research questions, but also expressed the needs, beliefs or feelings that were specifically related to their illness or its treatment. In this report, we use survey responses from patients and caregivers to conduct a thematic analysis of their views. This consideration of the views of patients and caregivers with kidney disease is meant to inform health care service providers about ways to support treatment decision-making, enhance communication, address psychosocial wellbeing and improve patient satisfaction. Methods {#Sec2} ======= We undertook a qualitative, descriptive study, based on a secondary analysis of survey data, using thematic content analysis to synthesize the views of patients on or nearing dialysis and their caregivers. Participants {#Sec3} ------------ Patients, caregivers and health care professionals who care for patients on or nearing dialysis were invited to complete a survey distributed online through partner organizations (Kidney Foundation of Canada), social media (Twitter, Facebook), or a paper-based version available in 10 Canadian hemodialysis centers. The survey was open for 3 months (October 2012 -- January 2013). For this analysis, we only analyzed the responses of patients and their caregivers. Survey {#Sec4} ------ The survey consisted of seven open-ended questions on topics relevant to patients with ESRD (overall management of kidney failure, treatment options, dialysis access options, prognosis, diet, symptoms and lifestyle). These open-ended questions were intended to elicit questions that were answerable by research, for example: "Are there questions about decisions regarding the way in which kidney failure can be treated that you would like to see answered by research?" In previous work, these research questions were prioritized to develop the top ten most important unanswered research questions from the perspective of patients and their caregivers (<http://www.cann-net.ca/patient-information/dialysis-research-priorities-survey#results>). While the goal of this previous work was to consider respondent statements in the context of developing a list of unanswered research questions, in the current work, we were interested in how the statements reflected respondents, needs, beliefs and feelings. In the survey responses, some patient statements only reflected unanswered research questions, and did not inform views. To eliminate these statements, which were less informative in the context of this work, two reviewers classified all responses independently, and separated these statements out. The remainder of the statements formed the dataset for the content analysis. Qualitative content analysis {#Sec5} ---------------------------- Statements were analyzed by thematic content analysis, a method that enables systematic analysis of the content in communication, such as answers to a survey, and reflects the content of the data set. Thematic content analysis is a method for identifying, analyzing, and reporting themes within data, where concepts or categories are derived from the data in an inductive or deductive manner \[[@CR6]\]. These concepts or categories serve to represent the data by providing knowledge and new insights, with the outcome a condensed and broad description of the phenomenon under study \[[@CR7]\]. Content analysis goes beyond the summary of the data and involves a qualitative interpretation by the researcher, providing a "vicarious experience" for the audience \[[@CR8]\]. A descriptive theme was defined as a subject that captured important data related to end-stage renal disease\[[@CR9]\]. Themes were not defined a priori but emerged after the initial readings of the data. Themes were identified as having captured an important element of the views of patients with chronic kidney disease and their caregivers and were not necessarily the most prevalent subjects within the data. Themes were not created to be mutually exclusive, but in order to represent the best conception of the data. Synthesis of findings {#Sec6} --------------------- We followed the process outlined by Braun et al. \[[@CR6]\] to synthesize the results. The data were first read by one author (LB) several times, to become familiar with the content and generate an initial idea of themes. The thematic content analysis involved three phases: the development of descriptive themes along with the creation of mutually exclusive rules for inclusion; statement by statement coding done independently by two authors (LB & LM); and the organization of the statements into meaningful categories by the same two authors (LB & LM). Any disagreements between the two authors were resolved by consensus, with the input of a third reviewer (BM) if necessary. We defined 4 descriptive themes, which were re-assessed to ensure that they worked with and represented the data. Lastly, a selection of powerful and compelling examples were selected, with context, to represent each theme. Results {#Sec7} ======= The survey elicited a total of 1820 statements from all respondents (patients, health care providers and caregivers) (Table [1](#Tab1){ref-type="table"}). Of these, 245 were removed because they were out of scope (not relevant to patients on or nearing dialysis) or were unclear. Of the remaining 1575, 1054 were from patients or caregivers. Statements that only expressed a research question, for example "can stem cell research improve my kidneys" or "what research is being done with regard to a mechanical replacement kidney, much like the artificial heart", were removed. There were a total of 544 remaining statements, from 189 individual respondents, included for the thematic content analysis.Table 1Patient and caregiver characteristicsType of respondentn (%)Total n= 189Patient receiving hemodialysis in a clinic91 (48.1)Patient on home hemodialysis32 (16.9)Patient on peritoneal dialysis22 (11.6)Patient on dialysis, no detail2 (1.1)Patient, within a year of starting dialysis6 (3.2)Care provider36 (19.0)Age18 -- 295 (2.6)30 -- 3915 (7.9)40 -- 4926 (13.8)50 -- 5935 (18.5)60 -- 6949 (25.9)70 -- 7923 (12.2)80 and over10 (5.3)Prefer not to answer26 (13.8)GenderMale73 (38.6)Female89 (47.1)Prefer not to answer27 (14.3)EthnicityAboriginal3 (1.6)Asian10 (5.3)Black9 (4.8)Mixed3 (1.6)Other7 (3.7)White123 (65.1)Prefer not to say34 (18.0)ProvinceAtlantic26 (13.8)British Columbia7 (3.7)Ontario59 (31.2)Prairies65 (34.4)Quebec4 (2.1)Territories1 (0.5)Prefer not to say27 (14.3) Four major descriptive themes were identified as being central to the views of patients with chronic kidney disease and their caregivers: gaining knowledge, maintaining quality of life, sustaining psychosocial wellbeing, and ensuring appropriate care. A thematic schema of the analytical framework is represented in Figure [1](#Fig1){ref-type="fig"}.Figure 1Components of each theme as an overview of the thematic analysis. Gaining knowledge {#Sec8} ================= The theme gaining knowledge encompassed seeking knowledge/lack of information, communication, timing of information, and included how patients and their caregivers learned about topics related to their chronic kidney disease. Seeking knowledge/lack of information {#Sec9} ------------------------------------- Respondents sought knowledge not only about the different treatment options available (either dialysis modality or access), but also wanted to determine which was the "best", and why, for both their overall health and quality of life. They specifically expressed the desire for a thorough explanation of all options. A caregiver stated: "Yes, I would like to see all of these options thoroughly explained at the very beginning. This is a whole new world for patients and care givers, making it very difficult to make good decisions." Respondents verbalized that they felt uninformed about many aspects of their treatment and care. A hemodialysis patient expressed: "Again left in dark till things happened, policy seems to be on a need to know. When you start dialysis why don\'t they warn you about the symptoms of crashing? Not nice not knowing what is happening". There was a recurring concern regarding symptom management, and respondents frequently voiced that they were unsure of why certain complications occurred (lack of energy, itchiness). They also felt that the lack of information interfered with their ability to cope with and manage these symptoms, as stated by a hemodialysis patient: "Why do we get restless legs, itchy skin and have difficulty sleeping? No one ever explained why nor what can be done". This lack of information extended to other areas of their disease, including fluid management and diet restrictions. A peritoneal dialysis patient wrote: "More information on why some foods are okay and others aren\'t. I had to find detailed lists of phosphate containing foods on the internet". Respondents also stated that they were unsure about their prognosis on dialysis, wanting to know how long they would need to stay on dialysis, the progression of their disease, as well as their life expectancy. A patient receiving hemodialysis in a clinic expressed: "Generally a summation of what will happen physically as the years pass on dialysis would be beneficial to me as a patient. What should I expect in terms of medical problems and deterioration due to the end stage renal disease". Lack of information extended to those seeking a transplant as well. A patient receiving hemodialysis in a clinic wanted to know: "At what age, and with what present medical issues does a person with kidney failure become ineligible for a transplant? Persons with kidney failure want to know all about the process of selection regarding a transplant". One dialysis respondent expressed that they wanted a holistic take on information, stating "We need to be better informed about the side effects of dialysis on your mind as well as your body, such as focus, cognitive ability and lack thereof. Also, how your whole life will change: how you have to live and interact with others and how totally frustrating it can be". Communication {#Sec10} ------------- In addition to not receiving the information they felt they needed, participants raised concerns about the communication between patient and healthcare provider, specifically in how the information was delivered. "An explanation in layman\'s terms would be useful". A hemodialysis patient felt information was not always communicated appropriately. "I started dialysis two weeks ago. There are a number of acronyms used, which I do not understand. How can we improve communication with patients?" This same respondent expressed that he found it "very frightening" when first discussing kidney disease. Respondents also expressed a desire for improved communication for managing their disease, and being informed. A home hemodialysis patient wondered: "Why can't the nurses share what's going on with our blood work and inform us as patients". In the particular case of diet restrictions, a long term dialysis patient stated: "Just because the patient was told this once at the beginning does not mean they remember. There should be some kind of annual review of the do's and don'ts as well as product updates". A respondent receiving hemodialysis in a clinic stated it as follows: "They need to go one step further than just give us a piece of paper saying what you can and cannot eat or drink and explain the consequences of not following a diet and watching what you drink". Many respondents verbalized their desire for a transplant, but felt that the selection process and position on the waiting list could be communicated in a clearer manner. Some felt this process was not transparent, perhaps due to a lack of communication on how the process works. "Why do I have to wait so long? Why haven\'t they made an effort to get me on the transplant list? Will I ever get my transplant?". Overall, some respondents voiced that they would like to feel part of the team and more involved in the decision making around their care. This was expressed by a patient receiving hemodialysis in a clinic through the following statement: "I've been through being diagnosed and I still see it happening: the nephrologist tells you what you have, the treatment and the diet, but never acts like you're a part of the team. Why is it that up until the moment you are told the above, you think you have a say in your treatment, but the specialist doesn't see it that way". Timing of information {#Sec11} --------------------- Respondents also questioned whether information could have been given earlier in the disease process. They felt that had they known the severity of their disease earlier on, they could have taken more action to prevent or delay the progression. Respondents also wanted to know early on about their treatment options for dialysis, given the impact of the decision. One patient wrote, "I feel that the patient should be told earlier about their choices. Also important knowing about how dialysis can affect your life and that of your family, friends, employer, etc. It is very traumatic and life changing". For those who had already made a decision about dialysis options, a respondent receiving hemodialysis in a clinic recounted feeling pressure to make a decision and needed more time. "Can you make the whole orientation process much clearer? Also, a strong focus was given to home hemodialysis, which I wasn\'t comfortable doing. Pushing us, especially those on the senior side, to use home dialysis with that machine is terrifying, even with training. Can you build layers of consulting to ease those of us into our new lifestyle? It may be repetitive but we\'d make better decisions in the long run". Another patient receiving hemodialysis in a clinic recounted his experience as follows: "When I first started dialysis, it would have been very helpful to know about the side effects of dialysis. I believe it would have helped prepare myself and my spouse for the onset of all this stuff. I received very little information when I first started". Source of information {#Sec12} --------------------- Peers appeared to be an important influence for both patients and caregivers on deciding their treatment options. Respondents valued the experience of their peers in the process of choosing their modality and were interested to know how they came to their decision. One caregiver stated: "Indeed it was a very difficult decision to make regarding the topic of which dialysis treatment would be best suited for us. We could have used more info on real people who have done both and what their experiences were and what the pros and cons are. Also the success rate of each and is there any difference or research on who does better on the different treatments". Respondents also looked to their peers in terms of their care, benchmarking their treatment against others, and wanting to know why there were differences. "How come some people have shorter times \[on dialysis\] than others, and more frequently during the week?". Maintaining quality of life {#Sec13} =========================== The theme maintaining quality of life encompassed the patient's general lifestyle, freedom to travel and sexual vitality. These were all identified as being important in maintaining a quality of life similar to pre-dialysis. General lifestyle {#Sec14} ----------------- Respondents most often expressed a strong desire to maintain their lifestyle, prior to the commencement of dialysis, what they commonly referred to as a "normal life". One respondent receiving hemodialysis in a clinic expressed: "What are the best options for a normal life?". Respondents also wanted to know how they could plan a life and maintain work while undergoing dialysis, as one respondent phrased it: "How to plan a life and work style that is compatible with staying fit while undergoing dialysis". One dialysis patient expressed a desire for normalcy and control in her ability to prepare the appropriate foods: "I see cookbooks for people of all kinds of conditions - heart problems, diabetes, gluten intolerance etc. I have never come across a cookbook for kidney patients. Can\'t some dietitian make one for us?". Respondents who experienced side effects or complications from dialysis or their illness often described the impact this had on their quality of life. "I\'ve lost my eyesight and leg because of diabetes-how can I have a normal life? My wife is fed up, tired of my problems, how can I make this better? I need her to look after me". Freedom to travel {#Sec15} ----------------- A large proportion of respondents verbalized their desire for freedom to travel and stated that traveling was extremely important in maintaining their quality of life. Respondents identified several barriers related to travel. Limited access to other dialysis centers was one barrier identified by a respondent on hemodialysis: "Can more access to hemodialysis machines be set up so that people who like to travel can get appointments in different parts of the country? It is very difficult to get time to dialyze where we boat and like to camp". Another respondent identified not knowing how to plan travel while on dialysis and resources available for travel (both personal resources required and those made available by the health system) as barriers. "Can persons on dialysis be provided with a comprehensive and detailed directory regarding travel anywhere throughout Canada, including any special perks or considerations available by public or private services in terms of costs, access, assistance, etc.". Respondents were also concerned with whether it was safe to travel while dialyzing and any possible consequences traveling had on their health. Sexual vitality {#Sec16} --------------- Respondents expressed the desire for more forthright communication around sexual vitality, emphasizing that sexuality was part of maintaining a normal life for them. A young woman with kidney disease expressed it as follows: "Sexuality is never spoken about. Young patients often have many questions about fertility, sex and how to deal with having a line or catheter and a sexual life". Respondents, and family members, were especially concerned about how they could maintain their sexuality, either by addressing erectile dysfunction, lack of sexual desire or other challenges to sexuality related to dialysis itself (e.g. femoral dialysis lines). "People are complaining about groin lines and trying to understand how they can still have sex. Sex is very important to any couple. Open communication regarding this topic would be much appreciated. All we need is to be on dialysis and divorced!". Sustaining psychosocial wellbeing {#Sec17} ================================= The theme sustaining psychosocial wellbeing included psychosocial support for both patients on or nearing dialysis and their caregivers, in addition to their overall wellbeing. Support for themselves & caregivers {#Sec18} ----------------------------------- Respondents were concerned about their access to support for themselves and their caregivers, whether at the start of treatment or after years on dialysis. "How about having a counseling session of encouragement before a patient goes on dialysis because I was terribly afraid and I wanted to run away". One caregiver expressed a need for psychosocial care. "What do you do about the mental care of a dialysis patient that suffers from depression due to or caused by lengthy treatment of dialysis?". Respondents were also concerned about their caregivers and the toll the disease took on them, illustrated by the following two quotes from patients on hemodialysis. "What types of things do family members have concerns about but are not voicing? What effect does CKD have on them?". "My wife left me because I was no longer that strong man she married, she wasn\'t getting the emotional love she needed and she thought I was going to die. Can more support/documents be made available to the spouses?". Some described the need for psychosocial support that extended beyond the dialysis unit. "Someone outside of the dialysis unit that is familiar with personal issues of chronic kidney disease other than my family doctor, especially dealing with sleeping and sexuality". Overall wellbeing {#Sec19} ----------------- Respondents emphasized that the lack of quality of life they experienced while on dialysis contributed to feelings of depression or inadequacy. "I cannot make myself breakfast/get dressed/put on socks. My quality of life is very bad but I feel all these symptoms are a part of my life. All I can do is smile and keep doing what I can, but it is very hard". Many comments reflected the overwhelming and deep-seated impact of dialysis and kidney disease on everyday life. "Without work you are living under the poverty line so are not only dealing with loss of health but now can barely afford to live. Family wants to distance themselves so they won\'t be reminded that death is imminent and won\'t feel the pain. Canada has made travel impossible. Most dialysis places don\'t want dialysis patients from another place, for various reasons; might infect their patients, too much trouble to accommodate. We cannot leave the country unless we are willing to pay for our own dialysis care. We are already destitute financially so that is impossible unless you have money. These kinds of restrictions, along with diet, and overall health add to feelings of depression". Ensuring appropriate care {#Sec20} ========================= The theme ensuring appropriate care included quality of care, care delivery and patient resources. Quality of care encompassed the satisfaction and standard of care patients received, while care delivery centered around the health system and its resources that impacted care of the patient. Patient resources focused on any perceived or real limitations to accessing health care due to constraints of personal financial resources of patients. Quality of care {#Sec21} --------------- Few respondents communicated concerns about their quality of care, with many expressing satisfaction with their nephrologists and other members of their health care team. "No questions but very satisfied with all treatment and information by all dialysis staff I have had contact with". Respondents did question whether their disease could have been diagnosed earlier, and suggested that education of their primary care physician may have led them to being treated earlier. "How do we educate family doctors to test for and refer patients at an early stage, so it is not such a shock when they are told they have kidney failure". Care delivery {#Sec22} ------------- Many of the views on care delivery expressed by respondents centered around perceived or real constraints on resources that affected their access to care, due to resources at the level of the health system. For example, some respondents perceived that home hemodialysis was a better treatment option, and that this should be offered more. "Why are more people not trained and guided towards self care, peritoneal dialysis or home hemodialysis". Another example was the lack of availability of all treatment options at all health centres, as stated by one patient receiving hemodialysis in a clinic. "I would like to try the button hole method but because of time and money and the clinic's part, I have not been able to try it. I'm trying to find a way to preserve the life of my vein". Patient resources {#Sec23} ----------------- As respondents are often limited in their ability to work and earn income, they were concerned about personal costs incurred because of their illness and the need for dialysis, as well as whether one dialysis therapy would incur more expenses to them: "What personal costs could I expect if I chose one \[dialysis therapy\] over the other?" Several dialysis patients expressed concerns about the impact that medication costs have on their finances, as highlighted by the following two comments. "We need certain medications to deal with kidney failure and dialysis, so why aren\'t they free?" "Why are some medications covered and others not, for example sensipar? Last year I spent over \$20,000 on medications. How does a 18-year dialysis patient survive his co-pay while maintaining their dignity and live with these costs?" Another hemodialysis patient expressed frustration about the lack of financial resources available to them. "Since dialysis patients can\'t work full-time because it takes time out of your life, why aren\'t we compensated financially? We need dialysis to survive!" Discussion {#Sec24} ========== We identified four themes that emerged from views of patients' on or nearing dialysis and their caregivers: gaining knowledge, maintaining quality of life, sustaining psychosocial wellbeing, and ensuring appropriate care. Respondents identified a lack of information and communication between patient and provider as a significant concern, particularly with respect to information about dialysis modality and access. Respondents also expressed significant concerns about their quality of life on dialysis, and expressed a desire to have as much of a "normal" life as possible. They identified a variety of issues that could be addressed by both the health system and health care providers to improve their quality of life. With respect to the health system, there is an urgent need for more psychosocial support, and resources to facilitate travel, particularly for patients on hemodialysis. For health care providers, better communication is key, along with more information on why symptoms happen, how they can best be managed, and information about their future prognosis. The majority of respondent's expressed a need for gaining knowledge, which was thought of either as seeking knowledge or lack of information. This need for information was expressed by both dialysis patients and caregivers, echoing previous studies that found that the primary need of families of chronic dialysis patients was information \[[@CR10], [@CR11]\]. The timing of when information was presented to patients and caregivers was also a common theme, as was the importance of re-enforcing and updating material previously communicated. The information needs of people as they pass through the various stages of their disease changes over time. A careful consideration of the type of information, including how much information, and at what stage of their illness it is provided, could enable a more efficient transfer of knowledge from health provider to patient. Further, education may alleviate concerns and stress, and stress has been found to affect dialysis modality selection \[[@CR12]\]. Providing predialysis education can not only enable patients to choose the modality best suited to them, but helped in their understanding of their disease\[[@CR13]\], and may extend their time prior to dialysis initiation \[[@CR14]\]. Indeed, previous research has noted that patients need information spread over an extended period of time, with increasing amount of detail and specificity as renal replacement therapy nears \[[@CR15]\]. Clinical practice guidelines in the UK specify that information should also be tailored to the stage of disease \[[@CR16]\]. Our results also highlighted that transferring knowledge is dependent on the quality of communication among people, echoing the results of a study that found that perceived knowledge of kidney disease was related to the quality of communication \[[@CR17]\]. Many respondent comments reflected the need for improved communication between patients/caregivers and health care providers, though increased communication is not merely repetition of the same information. Improving health literacy of patients with chronic kidney disease may be one step in not only ameliorating communication, but may also addressing the information needs of these patients, and how knowledge is retained \[[@CR18]\]. Health literacy encompasses communication between patients, their social networks, and providers. Limited literacy has been linked with reduced knowledge, less adherence, hospitalization and death \[[@CR19], [@CR20]\]. Finally, many respondents identified the importance of peer influence as a source of information, which may be particularly important since previous studies identified peer support as a means of providing practical information about kidney disease \[[@CR21]\]. Peer support may also give purpose to patients already on dialysis by valuing their experiences. As peers have "been there, done that", they have first hand comprehension of some of the difficulties of treatment, and may be a more influential source of information than clinicians for some important treatment decisions. We identified the need for more resources for patients and their caregivers, including improved access to travel, more information on how to cook meals within the limitations of the renal diet, more financial support, and more psychosocial care for themselves and their caregivers. Patients on hemodialysis consistently requested more information on how to arrange hemodialysis in other cities, and were concerned with the lack of hemodialysis spots available to enable travel. With respect to meal planning, although there is high-quality information available online to assist dialysis patients and caregivers (<http://www.kidneycommunitykitchen.ca/kkcookbook>), this information may not be sufficiently broad to address all cultures, tastes or financial means. Further, the number of comments we received requesting additional meal planning resources suggests patients are not aware of existing resources and are still facing barriers with respect to meal preparation. The difficulty of preparing meals to meet the dietary restrictions of those on dialysis is a need that has been met by pre-packaged frozen meals in the United States \[[@CR22]\]. Finally, with respect to the need for psychosocial care, the availability of trained professionals for psychosocial care in and out of the dialysis unit, along with support for caregivers, may help relieve the burden of living with this disease. Moreover, since depression is very common \[[@CR23]\], resources to help patients prevent and manage depression appears to be a critical need for patients. In addition to professional support, social support has also been found to improve outcomes, including compliance, in those with end-stage renal disease \[[@CR24], [@CR25]\] and resources should be made available to help foster important social relationships. In light of this information, kidney care and dialysis programs, in collaboration with patients and caregivers, should take inventory of what information they provide, the resources available, and how this information is communicated to patients. A careful examination of where a program's current resources are being directed and whether this aligns with the needs of patients and their caregivers is necessary. Further, detailed information on how to access resources should be made clear and reiterated often. Although the challenges that patients face may vary slightly across the country, there are likely more similarities than differences in many of the views of patients. As such, the development of certain educational resources, including materials and access to online resources, could be a collaborative process across programs, reducing the need for duplication of effort. While these educational materials will not replace face-to-face visits, they may provide helpful information for patients needing further information in between visits. Programs may also explore the idea of shared decision making as an innovative way of communicating and involving patients in critical decisions about their health \[[@CR26]\]. Our paper has limitations. We did a qualitative analysis of survey responses, which did not enable us to ask follow-up questions to further explore a theme. Analyzing written responses, without further clarification from respondents, could result in misclassification of certain statements. Though we sought to have a representative national response, our study was limited by having a low response rate from British Columbia and Quebec, and from some patient types (First Nations and the elderly). While we think that many of the views expressed in this report will be common across kidney care programs, the extent of the issue may vary based on geographic regions, health care systems and the availability of local resources. The information provided by this study could be supplemented by local questionnaires to best assess local needs, and inform local program changes. Conclusions {#Sec25} =========== We identified four themes from this analysis of the views of patients' on or nearing dialysis and their caregivers: gaining knowledge, maintaining quality of life, sustaining psychosocial wellbeing, and ensuring appropriate care. Importantly, respondents identified a variety of issues that could be addressed by both health systems and health care providers to improve their quality of life. These include the development of patient materials and resources, or sharing of existing resources across Canadian renal programs, along with adopting better communication strategies. Other concerns, such as the need for increased psychosocial and financial support, require consideration by health care funders. Competing interests The author(s) declare that they have no competing interests. Authors' contributions KK, BH, AL & BM have made substantial contributions to conception and design, or acquisition of data; LB, LM & BM have made substantial contributions to analysis and interpretation of data; all authors have been involved in drafting the manuscript or revising it critically for important intellectual content; all authors have given final approval of the version to be published; and all authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. Drs. Hemmelgarn and Manns are supported by Alberta Innovates - Health Solutions (AI-HS) (formerly Alberta Heritage Foundation for Medical Research (AHFMR) salary awards). Drs. Manns, and Hemmelgarn are supported by an alternative funding plan from the Government of Alberta and the Universities of Alberta and Calgary. This research was supported by two grants from CIHR (operating grant number 284262 team grant number 251048), an interdisciplinary team grant from Alberta Innovates -- Health Solutions, the Interdisciplinary Chronic Disease Collaboration, and from the Kidney Foundation of Canada. The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The authors are indebted to Erin Lillie for coordinating the primary study, which provided us with the responses for this study.	{ "pile_set_name": "PubMed Central" }
INTRODUCTION {#s1} ============ Nasopharyngeal carcinoma (NPC), derived from the epithelial lining of the nasopharrynx, is a commonly occurring cancer with the highest incidence of metastasis among head and neck cancers in southern China and Southeast Asia \[[@R1]--[@R4]\]. Because of its high radio-sensitivity, radiotherapy has been the main strategy for treatment of NPC. However, distant relapse remains the major cause of treatment failure in NPC \[[@R5]\], and the molecular mechanisms underlying NPC metastasis are poorly understood. In order to provide a basis for the development of novel therapeutics for NPC, it is crucial to obtain a better understanding of the molecular mechanisms used by cancer cells to facilitate their survival during metastasis. While there are many abnormal features of metastatic cancer cells, resistance to anoikis is particularly interesting as it enables cell survival with loss of attachment to the extracellular matrix (ECM) \[[@R6], [@R7]\]. Metastatic cancer cells develop anoikis resistance by triggering several signaling pathways \[[@R6], [@R8], [@R9]\], including the production of reactive oxygen species (ROS) and activation of mitochondrial metabolism \[[@R10], [@R11]\]. Under physiological conditions, ROS are constantly generated as by-products of aerobic metabolism in the mitochondria \[[@R12], [@R13]\]. Consistent with the pro-anoikis abilities of ROS, treatment with antioxidants suppresses anoikis in breast cancer \[[@R10]\]. Moreover, energy deficiency (diminished ATP levels) caused by ECM detachment can reduce the viability of breast cancer cells \[[@R11]\]. ROS is a collective term for the chemical species that are formed as a result of incomplete reduction of oxygen. ROS include superoxide anion radical (O~2~^·−^), peroxyradical (ROO−), hydrogen peroxide (H~2~O~2~), singlet oxygen (^1^O~2~), perhydroxyl radical (HO~2~.), and the extremely reactive hydroxyl radical (.OH). The mitochondrial enzyme manganese superoxide dismutase (MnSOD) efficiently converts O~2~^·−^ to H~2~O~2~ and thereby critically changes mitochondrial destructive effects. Because superoxide arises primarily from the mitochondria, MnSOD plays a pivotal role in its detoxification \[[@R14], [@R15]\]. However, the role of MnSOD as a critical feature of highly metastatic NPC has not been elucidated. In this study, we assess the role of MnSOD in mediating the survival of ECM-detached NPC cells. We provide evidence, both in vitro and in vivo, that up-regulation of MnSOD in a highly metastatic NPC cell line increases anoikis resistance after ECM detachment by decreasing mitochondrial O~2~ ^·−^ and accelerating hydrogen peroxide diminished. In addition, we show that the β-catenin pathway is robustly activated following a loss of ECM attachment. Activation of this pathway acts as a pro-survival signal, inducing a MnSOD-dependent cytoprotective response that is characterized by reduced mitochondrial O~2~ ^·−^ production and resistance to anoikis. Taken together, our findings suggest that β-catenin/MnSOD play central roles in mediating an antioxidant effect that enables cancer cells to survive and migrate to distant sites during metastasis. RESULTS {#s2} ======= Highly metastatic NPC cells have an advantage in anchorage-independent cell growth {#s2_1} ---------------------------------------------------------------------------------- Malignant cells can often survive and grow without adhesion to the ECM making this growth a hallmark of the malignant phenotype \[[@R16], [@R17]\]. There is a well-established model to evaluate the relationship between metastatic potential and anoikis resistance \[[@R17], [@R18]\]. In this model, a highly metastatic cellular clone, S18, is compared with the poorly metastatic clone, S26. We tested clonogenesis of the S18 and S26 cells using a soft agar assay, and assessed cell viability and apoptosis of cells by CCK8 and AnnexinV/PI staining, respectively. Even though there was no obvious difference in cell viability between S18 and S26 cells under attached culture conditions (Figure [1A](#F1){ref-type="fig"}), S18 cells survived better in anchorage-independent cell growth conditions (Figure [1B--1C](#F1){ref-type="fig"}). When the NPC cells were cultured in a three-dimensional model to form clones in Matrigel, the clones generated from S18 cells formed more fully-filled structures when compared with S26 cells (Figure [1D](#F1){ref-type="fig"}). Moreover, anoikis was induced in S26 cells after 12, 24, and 48 hours of suspension conditions (Figure [1E](#F1){ref-type="fig"}). These observations demonstrate that cell viability is elevated with the aggressiveness of human NPC cells in response to matrix detachment. ![S18 cells exhibit enhanced anoikis resistance\ The viability of S18 and S26 cells under adherence (A) and suspended (B) conditions. (C) Colony numbers of S18 and S26 cells in soft agar. (D) Acini were generated using the indicated cell lines cultured in Matrigel for 12 days, and cells were stained with DAPI. Representative images are shown (× 200). \*p \< 0.05. (E) Flow cytometric analyses of S18 and S26 cell apoptosis under suspended conditions for 12 hours, 24 hours, and 48 hours. Values were plotted as mean ± S.D. \*p \< 0.05 compared with control group.](oncotarget-07-32408-g001){#F1} ROS induced by matrix detachment plays a critical role in NPC anoikis {#s2_2} --------------------------------------------------------------------- Inadequate matrix attachment generates ROS and causes anoikis \[[@R10], [@R11]\]; therefore, we investigated the level of ROS in ECM-detached NPC cells. Since ROS are located in different subcellular compartments \[[@R19]\], we examined several different sub-cellular locations. There were no differences in DCF-DA fluorescence (for H~2~O~2~), DHE fluorescence (for Cytoplasmic O~2~^·−^) or Mito-SOX fluorescence (for mitochondrial O~2~^·−^) between S18 and S26 cells cultured on adherent plates. However, S18 cells had decreased H~2~O~2~ and mitochondrial O~2~^·−^ levels under detached conditions when compared with S26 cells (Figure [2A--2C](#F2){ref-type="fig"}). Since MnSOD and catalase are key antioxidant enzymes involved in scavenging ROS, we examined the protein levels of MnSOD and catalase in NPC cells grown in suspension. The MnSOD protein levels was up-regulated in S18 cells as compared to S26 cells. However, the protein levels of catalase was decreased in S26 cells when detached (Figure [2D](#F2){ref-type="fig"}). These results help understand why S18 cells have less O~2~^·−^ and H~2~O~2~. In addition, soft agar assays (Figure [2E](#F2){ref-type="fig"}) and Annexin/PI staining (Figure [2F](#F2){ref-type="fig"}) showed that treatment with 1 mM of an antioxidant compound N-acetyl-L-cysteine (NAC) stimulated the anchorage-independent growth in S26 cells, while treatment with 30 μM H~2~O~2~ decreased the anchorage-independent growth of S18 cells. These observations indicate that the level of ROS is inversely correlated with malign ant cell viability during the detachment process. ![Enhanced ROS is critical for anoikis in NPC cells\ (A--C) S18 and S26 cells were incubated for 30 min in the presence of DCH-DA (10 μM), DHE (10 μM) or Mito-SOX (5 μM). Representative flow cytometry plots show the separate analysis of H~2~O~2~ (A), cytoplasmic O~2~^·−^ (B) and mitochondrial O~2~^·−^ (C) content in S18 and S26 cells under attachment and detachment cell growth condition after 24 hours. (D) The S18 and S26 cells were cultured in suspension conditions for 24 h, total cell lysates were subjected to western blot analyses performed to detect MnSOD and Catalase levels.(E) Quantification of soft agar colony numbers of S18 and S26 cells treated with H~2~O~2~ (30 μM) or NAC (1 mM). (F) Flow cytometric analyses of S18 cells and S26 cells by AnnexinV/PI under suspended conditions in the present of H~2~O~2~ (30 μM) or NAC (1 mM) as described. Values were plotted as mean ± S.D. \*p \< 0.05 compared with control cells.](oncotarget-07-32408-g002){#F2} MnSOD expression is elevated in detached highly metastatic NPC cells {#s2_3} -------------------------------------------------------------------- We next investigated the molecular changes necessary for anoikis resistance and balancing the ROS concentrations. We examined Cu/ZnSOD activity as well as MnSOD activity in detached NPC cells. There was a difference in total SOD activity and MnSOD activity, but not Cu/ZnSOD activity, between S18 and S26 cells under detached conditions (Figure [3A--3C](#F3){ref-type="fig"}). When cells were cultured on adherent plates, there were no differences in Cu/Zn-SOD or MnSOD mRNA levels between S18 and S26 cells (Figure [3D](#F3){ref-type="fig"}). However, when cells were grown in suspension, both the mRNA and protein levels of MnSOD were upregulated in S18 cells as compared to S26 cells (Figure [3D--3E](#F3){ref-type="fig"}). These results suggest that MnSOD is necessary for the antioxidant defense in highly metastatic S18 cells and is critical for resistance to anoikis. ![Matrix detachment induces MnSOD in high-metastasis S18 cells\ (A--C) S18 and S26 cells were cultured under detachment condition for 12 hours. Then, total SOD (A), Cu/Zn-SOD (B) and MnSOD (C) enzyme activity was measured using an SOD Assay Kit-WST. (D) Transcript levels of Cu/Zn-SOD and MnSOD in S18 and S26 cells under attachment and detachment. (E) MnSOD levels in S18 and S26 cells by western blot analyses. Values were plotted as mean ± S.D. \*p \< 0.05 compared with control.](oncotarget-07-32408-g003){#F3} MnSOD reduces ROS levels and anoikis-mediated cell death {#s2_4} -------------------------------------------------------- To further confirm the role of MnSOD in providing resistance to anoikis, we utilized MnSOD siRNA and a MnSOD overexpression plasmid. siRNA knockdown of MnSOD in S18 cells (Figure [4A](#F4){ref-type="fig"}) resulted in higher levels of Mito-SOX fluorescence (for mitochondria O~2~^·−^) (Figure [4B](#F4){ref-type="fig"}) and apoptosis (Figure [4E](#F4){ref-type="fig"}), corresponding to over 20% reduction of survival after 24 hours in suspended conditions and rendering these cells sensitive to NAC (Figure [4E](#F4){ref-type="fig"}). Therefore, MnSOD-mediated elimination of mitochondrial O~2~^·−^ in response to matrix detachment protects against anoikis. Conversely, overexpression of MnSOD in S26 cells (Figure [4F](#F4){ref-type="fig"}) reduced Mito-SOX fluorescence (Figure [4G](#F4){ref-type="fig"}) and decreased the percentage of apoptotic cells (Figure [4J](#F4){ref-type="fig"}), further supporting an essential role of MnSOD in anoikis resistance. However, SOD converts superoxide radical into hydrogen peroxide thus may establish an elevation of H~2~O~2.~ Interestingly, subsequent analysis of the data demonstrated that MnSOD can promote protein expression level of catalase (Figure [4C, 4H](#F4){ref-type="fig"}) and H~2~O~2~ diminished (Figure [4D, 4I](#F4){ref-type="fig"}), which suggest that more MnSOD would be unlikely to yield more H~2~O~2~ in NPC cells. ![MnSOD confers resistance to anoikis\ S18 cells (A--E) were transfected with sicon or siMnSOD, S26 cells (F--J) were transfected with pcDNA3.1-vector or pcDNA3.1-MnSOD for 24 hours, after which they were cultured in poly-HEMA-coated plates under anchorage-independent conditions for 24 hours following which flow cytometry analysis for mitochondrial O~2~^·−^ (B, G), western blot analyses for catalase levels(C, H) were performed. H~2~O~2~ concentration (D, I) and apoptosis (E, J) were detected with Amplex^®^ Red Hydrogen Peroxide Kit and AnnexinV/PI, respectively. The data are presented as mean ± S.D. of triplicate determinations. For differences between two groups, the Students t-test was employed. \*p \< 0.05.](oncotarget-07-32408-g004){#F4} β-catenin signaling upregulated MnSOD and critical for anoikis resistance {#s2_5} ------------------------------------------------------------------------- Recently, several studies have suggested involvement of the Wnt/β-catenin pathway in ROS scavenging, and Wnt signaling is also known to function in tumor metastasis \[[@R20]\]. β-catenin transcriptionally regulates several antioxidant genes in response to oxidative stress \[[@R21]\]. Hence, we studied the putative connection between MnSOD and the canonical Wnt/β-catenin pathway. Interestingly, loss of β-catenin expression disrupted cell-cell adhesion in S26 cells and increased the nuclear accumulation of β-catenin in S18 cells after 6 hours of suspension (Figure [5A--5B](#F5){ref-type="fig"}). Knockdown of β-catenin in S18 cells reduced MnSOD protein levels in suspension (Figure [5C](#F5){ref-type="fig"}). Flow cytometric analysis confirmed that mitochondrial O~2~ ^·−^ levels were elevated in siβ-catenin cells compared with control cells following 12 hours of suspension (Figure [5D](#F5){ref-type="fig"}) and accompanied by increased anoikis (Figure [5E](#F5){ref-type="fig"}). Moreover, S26 cells overexpressing β-catenin had increased MnSOD protein levels (Figure [5F](#F5){ref-type="fig"}) and suppressed anoikis when in suspension (Figure [5H](#F5){ref-type="fig"}). On the contrary, Mito-SOX fluorescence intensity was reduced in S26 cells overexpressing β-catenin (Figure [5G](#F5){ref-type="fig"}). Together, these findings provide evidence for the involvement of β-catenin working upstream of MnSOD to promote anoikis resistance in metastatic NPC cells. ![β-catenin signaling promotes anoikis resistance via MnSOD up-regulation\ (A) S18 cells were under adherent or suspended condition for 6 hours and 12 hours as indicated, total cell lysates, MnSOD, and β-catenin were subjected to western blotting. (B) Immunofluorescence staining of β-catenin. β-catenin in the cellular nucleus of S26 had a low intensity compared with that in the S18 after grown in suspension for 24 hours while there had no obvious difference of that when attachment cultured. (C) S18 cells were transfected with siRNA against β-catenin (siβ-catenin) or sicon. Immunoblots for indicated proteins were measured. (D) Flow cytometric analysis of mitochondrial O~2~ ^·−^ in S18 cells after cultured in suspension condition for 24 hours. (E) Flow cytometric analysis of cells by AnnexinV/PI under attachment or suspended conditions for 6 h,12 h, 24 h and 48 h as indicated. (F--H) S26 cells were transfected with empty vector or β-catenin for 24 hours and grown in attached or suspension condition for another 24 hours. Immunoblots of β-catenin (F), Flow cytometric analysis of mitochondrial O~2~^·−^ (G) and apoptosis rate (H) were shown. Values are plotted as mean ± S.D. \*p \< 0.05 compared with control.](oncotarget-07-32408-g005){#F5} Knockdown of MnSOD inhibits metastasis by promoting anoikis {#s2_6} ----------------------------------------------------------- Additionally, we utilized the shRNA (shMnSOD) to develop a stably transfected S18-shMnSOD cell line (Figure [6A](#F6){ref-type="fig"}). Under detached conditions, there was a difference in H~2~O~2~ and mitochondrial O~2~^·−^ (but not cytoplasmic O~2~^·−^) between S18-shcon and S18-shMnSOD cells (Figure [6B](#F6){ref-type="fig"}). S18-shMnSOD cells also displayed increased sensitivity to anoikis and decreased cell viability following 24 hours of suspension, as compared to S18-shcon cells (Figure [6C--6E](#F6){ref-type="fig"}). Activation of caspase3, and 9 were involved in mitochondrial ROS-induced apoptosis (Figure [6F](#F6){ref-type="fig"}). To test the consequences of MnSOD loss in vivo, we used a nude mouse tumor model administering S18-shcon or S18-shMnSOD cells via intravenous injection to force them to grow in suspension. After six weeks, mice injected with S18-shMnSOD cells had amassed less tumor burden in the lungs compared with the S18-shcon group (Figure [6G](#F6){ref-type="fig"}). In addition, there were discernible differences in tumor histology when comparing lung samples from S18-shcon to S18-shMnSOD injected animals (Figure [6H](#F6){ref-type="fig"}). These data suggest that MnSOD is critical for the NPC colonization of lung tissue in vivo. ![Diminished MnSOD expression promotes anoikis in vitro and reduces tumor formation in vivo\ (A) Using lentiviral transduction of shRNA, MnSOD expression was reduced in S18 cells. Western blot analysis for MnSOD and β-actin confirms the success of the established stable cell lines. (B) Cells were incubated for 30 min in the present of DCH-DA (10 μM), DHE (10 μM) or Mito-SOX (5 μM). Representative flow cytometry plots show the separate analysis of H~2~O~2~, cytoplasmic O~2~^·−^ and mitochondrial O~2~^·−^ content in S18-shcon and S18-shMnSOD cells under detachment cell growth condition for 24 hours. (C) Cell viability by CCK8 assays under suspension conditions. (D) Colony numbers of S18-shcon and S18-shMnSOD cells grown in soft agar. (E) The percent of apoptotic cells shown in response to cell culture under suspension as indicated. (F) Western blot analysis for markers of apoptosis (caspase 3 and caspase 9) in S18-shcon and S18-shMnSOD cells grown in suspension cultures for 24 hours was performed. (G) S18-shcon and S18-shMnSOD cells (2 × 10^6^ cells per injection) were injected into nude mice via tail vein (n = 8). Six weeks after injection, mice were killed. Representative images of shcon and shMnSOD tumors on the lungs fixed in picric acid and the average numbers of tumor nodules were shown. (H) Immunohistochemistry staining with hematoxylin and eosin was performed on shcon or shMnSOD lung tumor tissues, lack of tumor nodules in shMnSOD lung tissues. Scale bars represent 500 μm or 200 μm. The data are presented as mean ± S.D. For differences between two groups, the Students t-test was employed. \*p \< 0.05.](oncotarget-07-32408-g006){#F6} MnSOD expression is elevated with progressing NPC tumor stage {#s2_7} ------------------------------------------------------------- We next used immunohistochemistry to analyze MnSOD expression in tumors from NPC cancer patients (tissue microarrays, Shanghai Outdo Biotech Co., Ltd.). In NPC samples stratified by stage, MnSOD expression increased with histologic tumor grade, being highest at histologic grade IV and lowest in grade I (Figure [7A--7B](#F7){ref-type="fig"}). We also examined MnSOD expression in NPC samples with known clinical outcomes. In a Kaplan-Meier model, MnSOD protein expression was a strong predictor of poor survival rate, P = 0.0015 (Figure [7C](#F7){ref-type="fig"}). Correlations between MnSOD expression and various clinic pathological features are summarized in Table [1](#T1){ref-type="table"}. MnSOD expression did not correlate with sex; however, it was closely associated with tumor metastasis. These results show that MnSOD expression is elevated with progressing tumor stage and correlated with poor survival, suggesting an association between high MnSOD expression and tumor aggressiveness in human NPC. ![Upregulation of MnSOD in nasopharyngeal carcinoma\ (A) Representative immunostaining of MnSOD in tissue microarrays across different pathologic stages. (B) Quantification of MnSOD in tissue microarrays (Shanghai Outdo Biotech Co., Ltd., n = 118). The IHC intensity of MnSOD was analyzed using ImageJ and statistical analysis was performed using one-way analysis of variance. Error bars represent SD, \*p \< 0.05. (C) Kaplan-Meier survival curves based on MnSOD staining for patients with stage I--IV (log-rank test).](oncotarget-07-32408-g007){#F7} ###### Correlation between MnSOD expression and various clinical pathological features Variables No.of cases (n = 118) Expression of MnSOD P ----------------------- ------------------------- --------------------- ----- ---- ------- Gender Male 87 8 32 47 0.974 Female 31 3 12 16 Pathology grading I 5 3 1 1 0.000 II 42 6 29 7 III 40 1 9 30 IV 31 1 5 25 Primary tumor T1 9 3 3 3 0.004 T2 43 4 24 15 T3 30 2 9 19 T4 36 2 8 26 Lymph node metastasis N0 40 5 17 18 0.002 N1 45 2 23 20 N2 28 2 3 23 N3 5 2 1 2 Distant metastasis M0 105 10 43 52 0.046 M1 13 1 1 11 Correlations between MnSOD staining index scores and other categorical factors were analyzed using the Pearson chi-square test of independence. DISCUSSION {#s3} ========== Anoikis is apoptosis induced by cell detachment from the extracellular matrix is a phenomenon \[[@R22]\]. Tumor cells need anoikis resistance to survive before metastasis, making it an essential trait of malignancy \[[@R8]\]. However, the mechanisms underlying anoikis resistance remain poorly understood in nasopharyngeal carcinoma (NPC) cells, which have a high rate of metastasis. In the present study, we found that the highly metastatic NPC cells, S18, have an advantage in anchorage-independent cell growth and exhibit strong anoikis resistance when compared with the poorly metastatic S26 NPC cell line (Figure [1](#F1){ref-type="fig"}). These results suggest that anoikis resistance is a critical feature of highly aggressive NPC cancer cells. When adherent cells detach from the extracellular matrix (ECM) it causes metabolic defects, including marked increases in the amount of reactive oxygen species (ROS) \[[@R10]\]. Although low levels of ROS regulate cellular signaling and play a role in normal cell proliferation, recent studies have shown that excessive amounts or persistent elevation of ROS might lead to increased anoikis. Neutralization of ROS by antioxidant enzymes or NAC enhances survival of breast cancer cells when deprived of ECM \[[@R10], [@R11]\]. Considering the importance of ROS and their implications in anoikis, as anticipated, we found that the highly metastatic S18 cells have lower intracellular ROS production. Specifically, S18 cells have low mitochondrial O~2~^·−^and H~2~O~2~ but not cytoplasmic O~2~^·−^, when compared with S26 cells (Figure [2](#F2){ref-type="fig"}). These results suggest that the level of ROS is inversely correlated with malignant cell viability during the detachment process. MnSOD is an antioxidant mitochondrial matrix protein that catalyzes O~2~^·−^ into H~2~O~2~ \[[@R23], [@R24]\]. Our data show that the highly metastatic S18 cell line display higher MnSOD protein and activity levels compared to the S26 cell line, when cells are grown in suspended conditions. These results are consistent with other reports, which found that induction of MnSOD by detachment allows cells to survive longer in suspension and depletion of MnSOD sensitizes cells to anoikis \[[@R15]\]. However, SOD converts superoxide radical into hydrogen peroxide may establish a steady flow of H~2~O~2~ originating from mitochondria. Interestingly, MnSOD can promote protein expression level of catalase thus would be unlikely to yield more H~2~O~2~ (Figure [4](#F4){ref-type="fig"}). Futhermore, though we observed Cu/Zn-SOD (a cytosolic superoxide dismutase \[[@R25]\]) was highly expressed in both S18 and S26 cells, it was not induced by detachment from the ECM in the first 24 hours (Figure [3](#F3){ref-type="fig"}). These results indicate that MnSOD, but not Cu/Zn-SOD contributes to maintaining mitochondrial bioenergetics and promoting anoikis resistance in NPC. To further confirm the role of MnSOD in anoikis resistance, we found that the anchorage-independent growth ability of S18 cells is inhibited upon MnSOD knockdown, while growth is enhanced with MnSOD overexpression in S26 cells (Figure [4](#F4){ref-type="fig"}). These findings suggest that increased MnSOD activity contributes to anoikis resistance, which is a critical feature of highly aggressive NPC cancer cells. These results are strongly supported by our in vivo studies, which showed that the lung tumor burden in mice injected with MnSOD-deficient cells is reduced (Figure [6G](#F6){ref-type="fig"}). Clinically significant elevations in MnSOD expression are associated with increased tumor invasion and metastasis in certain cancer types including gastric and esophageal cancer, breast cancer, lung carcinoma \[[@R26]--[@R29]\]. However, the role of MnSOD as a primary participant in the malignant transformation process in NPC cancer cells has been not investigated. Our clinical studies demonstrated that MnSOD was expressed at increased levels with progressing tumor stage, indicating an association between high MnSOD expression and tumor aggressiveness. Moreover, high expression of MnSOD correlates with extremely poor survival rates (Figure [7](#F7){ref-type="fig"} and Table [1](#T1){ref-type="table"}). Another implication of the data presented here is that activation of β-catenin plays a role in the anoikis resistant of NPC. We found that the β-catenin levels were substantially decreased following matrix detachment in poorly metastatic NPC cells. Upon cell detachment, β-catenin accumulates and translocates to the nucleus, as shown with immunofluorescent staining of highly metastatic S18 cells (Figure [5B](#F5){ref-type="fig"}). Knockdown of β-catenin blocks MnSOD expression and induces S18 cells to become sensitive to apoptosis in suspension culture. Furthermore, overexpression of β-catenin increases MnSOD expression and represses anoikis in S26 cells. These results support an emerging role for β-catenin in regulating MnSOD in NPC, thereby making it an attractive therapeutic target to prevent tumor metastasis. Even though we document the properties of β-catenin/MnSOD in anoikis resistance, the precise mechanism by which β-catenin is activated in suspended cells and how β-catenin is involved in the regulation of the MnSOD warrants further investigation. In summary, we established the critical role of MnSOD in conferring mitochondrial O~2~^·−^ and H~2~O~2~ degradation during anoikis resistance. Activation of β-catenin up-regulates MnSOD and might be the underlying molecular mechanism that defines the aggressive form of NPC. Thus, β-catenin could potentially be exploited as a therapeutic target to induce anoikis in circulating tumor cells and prevent metastatic growth. MATERIALS AND METHODS {#s4} ===================== Cells culture and generation of stable cell lines {#s4_1} ------------------------------------------------- S18 and S26 were graciously provided by Professor Qian (Department of Nasopharyngeal Carcinoma, Sun Yat-sen University Cancer Center, Guangzhou, China). Cells were maintained in RPMI 1640 supplemented with 10% FBS. Penicillin and streptomycin were added to all cultures. Puromycin (0.5 ug/mL) was used for the generation of stable cell lines. Cell viability assays {#s4_2} --------------------- Cells were plated at a density of 20,000 cells per well in 48-well or 150,000 cells per well in 6-well poly-HEMA-coated (or normal) plates. Cell viability was measured using the CCK8 assay (Dojindo, Japan) according to the manufacturer\'s protocol. For cell counting, detached cells were stained with trypan blue and placed in a hemocytometer. Soft agar colony assays {#s4_3} ----------------------- A total of 20,000 cells were suspended in RPMI 1640 (10% FBS) containing 0.7% agarose and layered on solidified medium containing 1.2% agarose in six-well plates. After solidification, the top layer was covered with the medium. When necessary, 1 mM NAC and 30 μM H~2~O~2~ was added to each layer and the cover medium. The medium was replaced every week. Images were taken after 3--4 weeks and analyzed with ImageJ software. Acini formation assays {#s4_4} ---------------------- 12-well plates were initially coated with 30 ul 100% Matrigel and allowed to solidify, forming a gelled bed of basement membrane measuring approximately 1 mm in thickness. Cells (1 × 10^4^/ml) were seeded onto this bed as a single-cell suspension in RPMI 1640 (10% FBS) containing 2% Matrigel. The medium was replaced every 4 days. Images were obtained using a confocal laser scanning microscope after 2 weeks. Anoikis assays {#s4_5} -------------- Apoptosis was assessed by the AnnexinV/PI detection. Cells (4 × 10^5^) were plated onto poly-HEMA-coated six-well plates in growth medium to prohibit attachment. After 12, 24, and 48 h in suspension, cells were harvested, clumps were separated by trypsin (0.05%) and a mesh screen process utilized to prepare monoplast suspensions which were stained per the manufacturer\'s recommendation. Experiments were repeated three times. ROS assays {#s4_6} ---------- Cells were plated at a density of 200,000 cells per well in 6-well poly-HEMA-coated (or normal) plates. After 12 h and 24 h, H~2~DCF-DA (sigma), dihydroethidium (Life Technologies) or MitoSOX (Invitrogen) was added to each well at a concentration of 10 μM, 10 μM or 5 μM, respectively. After 30 min, the samples were run on the flow cytometer to detect fluorescence. Each sample was collected using 20,000 events. Amplex^®^ Red Hydrogen Peroxide/Peroxidase Assay Kit (Invitrogen, A22188) was also used to detect H~2~O~2~ according to the manual operation. Superoxide dismutase (SOD) activity {#s4_7} ----------------------------------- SOD activity was measured using a Cu-Zn/Mn-SOD assay kit (WST) (Beyotime Institute of Biotechnology, Jiangsu, China). Briefly, total SOD activity was measured by reduction rate inhibitions of 2-(4-iodophenyl)-3-(4-nitrophenyl)-5-(2,4-disulfophenyl)-2H-tetrazolium and monosodium salt (WST-1). MnSOD activity was measured by adding 10 mM potassium cyanide to inactivate Cu-Zn/SOD activity. The difference between total SOD and MnSOD activity was considered as the Cu-Zn/SOD activity. SOD activity was expressed as units/mg of protein (one unit was defined as the amount of enzyme that inhibited WST-1 reduction by 50%). RNA isolation and quantitative RT-PCR {#s4_8} ------------------------------------- Total RNA isolation from cells grown under attached and suspended conditions was carried out with TRIzol (Invitrogen, Grand Island, NY, USA), following manufacturer\'s protocol. 500 ng total RNA was used for reverse transcription and quantitative real-time PCR analysis (qRT-PCR). Relative mRNA quantities were determined using the comparative cycle threshold (ΔΔCt) method. β-actin was used for normalization. The primers used for MnSOD, Cu, ZN-SOD, and ACTB are as follows: MnSOD forward, 5′-CTG TTG GTG TCC AAG GCT CA-3′, MnSOD reverse, 5′-GTA GTA AGC GTG CTC CCA CA-3′. Cu, Zn-SOD forward, 5′-TGG TTT GCG TCG TAG TCT CC-3′, Cu, Z-SOD reverse, 5′-CTT CGT CGC CAT AAC TCG CT-3′, ACTB forward, 5′-GCA CTC TTC CAG CTT CCT T-3′, ACTB reverse, 5′-GTT GGC GTA CAG GTC TTT GC-3′. The experiments were performed in triplicate. siRNA transfection and short hairpin RNA {#s4_9} ---------------------------------------- To establish stable cells expressing either MnSOD short hairpin RNA(shRNA), S18 cells were transfected with Lentiviral RNAi system (The lentivirus skeleton plasmid were gifts from Prof. Peng Xiang (Sun Yat-sen University, China) and selected by 0.5 ug/mL puromycin (Invitrogen, Carlsbad, CA, USA) for 15 days. For siRNA knockdown, MnSOD siRNA and a control siRNA were purchased from RiboBio. According to the manufacturer\'s instructions, transfections were performed at approximately 60% confluency using HiPerFect Transfection Reagent (Qiagen). Western blotting {#s4_10} ---------------- Cells were harvested and lysed for total protein extraction. Protein concentration was determined using Bio-Rad DC protein assay kit (Bio-Rad Laboratories) according to manufacturer\'s protocol. Aliquots of equal amounts of protein from the cell lysates were subjected to Western blot analysis. Antibodies used include those specific for MnSOD(BD Biosciences), catalase (proteintech), caspase3 (Cell Signaling Technology), caspase9 (Cell Signaling Technology), β-catenin (Cell Signaling Technology) and β-actin (Sigma, MO, USA). Animal experiments {#s4_11} ------------------ Male athymic mice between 5 and 6 weeks of age were obtained from Shanghai Institutes for Biological Sciences (Shanghai, China). 2 × 10^6^ cells were injected into the lateral tail vein of 7-week-old nude mice. Six weeks after injection, the mice were sacrificed, and lungs were removed. Lungs were fixed in Bouin\'s solution for 24 hours and stored in 70% ethanol before analysis. All the animal procedures were outlined in the guidelines of Institutional Animal Care and Use Committee at SYSU. Immunofluorescence {#s4_12} ------------------ S18 and S26 cells plated on culture slides or smear after the suspension culture were fixed in ice-cold 4% paraformaldehyde. Then, samples were permeabilized for 20 min with 0.01% Triton X-100 in PBS and blocked with normal goat serum at 37°C for 1 hour. After washing 3 times with PBS, they were incubated with β-catenin (Cell Signaling Technology) overnight at 4°C, then were followed by fluorescent secondary antibodies (1:200 dilution in blocking buffer, 1 h, room temperature). After immunolabeling, cells were washed, stained with 4, 6-di-amino-2-phenylindole (DAPI) (Sigma, St. Louis, MO, USA) for 5 min at room temperature. Cells were visualized under a confocal microscope (Axio Observer Z1, ZEISS, Jena, Germany). Immunohistochemistry {#s4_13} -------------------- Tissue microarray (Shanghai Outdo Biotech Co., Ltd., n = 118) were deparaffinized and dehydrated with graded alcohol. The samples were pretreated with 0.01 M citrate buffer (pH 6.0) for 2 min at 100°C in an autoclave; then the slides were allowed to cool to room temperature. Endogenous peroxidase activity was quenched by incubation in methanol containing 3% H~2~O~2~ for 10 min at room temperature. After several washes in Phosphate-Buffered Saline Tween-20 (PBST) (pH 7.2), the sections were blocked with goat serum for 60 min at room temperature and then incubated with MnSOD antibody (BD Biosciences) overnight at 4°C in a humidified chamber. After a brief rinse in PBST, sections were incubated for 40 min at 37°C with a biotin-conjugated secondary antibody (mouse) followed by incubation with DAB for 15 s. After rinsing with distilled water, sections were counterstained with hematoxylin. As a negative control, slides were incubated in PBS in place of primary antibody. The IHC intensity of MnSOD was analyzed using ImageJ. Statistical analyses {#s4_14} -------------------- Data are presented as means ± SD. The difference between experimental groups was assessed by ANOVA or two-tailed paired Student\'s t-test. The log-rank test was used for survival analysis. A value of p \< 0.05 was considered statistically significant. We thank the members of the laboratory for their helpful comments on the manuscript. CONFLICTS OF INTEREST No potential conflicts of interest was disclosed. GRANT SUPPORT This study was supported by National Nature Science Foundation of China, Grant Number: 81172163, 81272338, 81272515, 81200706, 8137 0945, 81471033, 81572342, 81570871, 81570764; National Key Sci-Tech Special Project of China, Grant Number: 2013ZX09102--053, 2015GKS-355; Program for Doctoral Station in University, Grant Number: 20120171110 053,20130171110053 Key Project of Nature Science Foundation of Guangdong Province, China, Grant Number: 2015A030311043; Guandong Natural Science Fund, Grant Number: 2014A020212023, 2014A0303 13073, 2015A030313103, 2015A030313029; Key Sci-tech Research Project of Guangzhou Municipality, China, Grant Number: 2014J4100162, 201508020033; Chang jiang Scholars and Innovative Research Team in University, number: 985 project PCSIRT 0947; Fundamental Research Funds for the Central Universities of China (Youth Program 13ykpy06, 31610046). The funding agencies had no role in study design, data collection, and analysis, decision to publish, or preparation of the manuscript.	{ "pile_set_name": "PubMed Central" }
CASE REPORT {#s1} =========== A 16-year-old girl developed multiple painful and tender soft tissue swellings over her neck, trunk and extremities since 2 years of age. These swellings appeared spontaneously or after trivial injury and either used to resolve spontaneously or they ossified into mature bone. These flares of heterotopic ossification resulted in gradual restriction of movements of neck, trunk, shoulder, elbow, hip and knee joints. There was associated congenital hallux valgus bilaterally (Fig. [1](#OMV022F1){ref-type="fig"}a). There was no history of similar illness in her family. The radiographs of the girl showed certain characteristic features of fibrodysplasia ossificans progressiva (FOP). The cervical skiagram showed characteristic anomalies of the cervical spine in form of large posterior elements and fusion of the facet joints between C2 and C7 (Fig. [1](#OMV022F1){ref-type="fig"}b) \[[@OMV022C1]\]. A CT scan of dorso-lumbar spine showed a peculiar 'Y-shaped' soft tissue ossification extending from the inferior angle of bilateral scapulae to the midline soft tissue of posterior back (lattisimus dorsi muscles and thoracolumbar fascia). It also demonstrated dorso-lumbar scoliosis and fusion of posterior elements of thoracic and lumbar vertebrae (Fig. [1](#OMV022F1){ref-type="fig"}c). Figure 1:(a) Characteristic malformed great toe and hallux valgus. (b) The cervical skiagram shows large posterior elements and fusion of the facet joints between C2 and C7. (c) A CT scan of dorso-lumbar spine shows Y-shaped soft tissue ossification extending from the inferior angle of bilateral scapula to the midline soft tissue of posterior back (lattisimus dorsi muscles and thoracolumbar fascia), dorso-lumbar scoliosis with convexity to left, squaring of thoracic and vertebral bodies and fusion of posterior elements of thoracic and lumbar vertebrae. DISCUSSION {#s2} ========== FOP is a rare genetic disorder of connective tissue characterized by congenital malformed great toes and progressive heterotopic ossification leading to severe disability. Classic FOP is caused by a recurrent activating mutation (617G\>A; R206H) in the gene ACVR1/ALK2 encoding Activin A receptor type I/Activin-like kinase 2, a bone morphogenetic protein type I receptor \[[@OMV022C2]\]. This genetic defect which leads to the formation of a heterotopic skeleton involves normal skeletal morphogenesis at heterotopic sites \[[@OMV022C3]\]. During the first decade of life, sporadic episodes of painful soft tissue swellings (flare-ups) occur which are often precipitated by trivial injury. These flare-ups transform skeletal muscles, tendons, ligaments, fascia and aponeuroses into heterotopic bone causing restriction of movements of various joints. The flares may partially respond to a short course of steroids, but overall prognosis is reduced with median lifespan of 40 years. Most patients die of complications of thoracic insufficiency syndrome \[[@OMV022C4]\]. CONFLICT OF INTEREST STATEMENT {#s3} ============================== None declared.	{ "pile_set_name": "PubMed Central" }
All relevant data are within the paper. Data can also be found online: doi:[10.5061/dryad.mb622](https://doi.org/10.5061/dryad.mb622) Introduction {#sec001} ============ Temperature influences the rates of virtually all developmental processes in bird embryos, including metabolism, development, and growth \[[@pone.0184670.ref001]--[@pone.0184670.ref003]\]. Thus, it is not surprising that, within a zone of thermal tolerance, higher mean temperatures result in higher avian embryonic growth rates than lower mean temperatures \[[@pone.0184670.ref004]--[@pone.0184670.ref013]\], albeit with some potentially negative effects on hatchability. Many birds---especially domesticated species---maintain the incubation temperature of their eggs relatively constant and close to optimal temperature (typically 35--38°C) through behavioral and physiological manipulations \[[@pone.0184670.ref014]\]. As a result, most studies evaluating temperature effects on avian development use a constant incubation temperature equal to mean daily-temperatures of a particular species' natural environment \[[@pone.0184670.ref014]\]. For economically important species such as the domestic chicken (Gallus gallus) or turkey (Meleagris gallopavo), for example, constant temperatures during incubation represent the natural incubation conditions and certainly the commercial industry standard. Consequently, relatively few studies have used cyclic thermal protocols to evaluate hatchability or post-natal growth---e.g., \[[@pone.0184670.ref015]\]. However, constant (non-cyclic) incubation temperatures do not fully represent the naturally encountered daily thermal fluctuations that eggs of many precocial, non-domesticated birds. Some species may lay multiple eggs in a clutch (e.g., ducks, gallinaceous birds, and ratites), depositing eggs in their nest at a rate of one egg per day. In such species, true incubation by one or more parents does not begin until the penultimate or ultimate egg is laid (see \[[@pone.0184670.ref002]\] for review, \[[@pone.0184670.ref016]\]). Thus, for many precocial birds there is an extended pre-incubation period---the focal period of this study---where eggs that are laid first are unattended for the entire pre-incubation period and thus lack the thermal protection and stabilization provided by an incubating parent. Pre-incubation egg temperatures closely follow what can be large diurnal thermal fluctuations of the environment \[[@pone.0184670.ref017]\]. The magnitude of the diurnal thermal fluctuations varies locally, seasonally, and from year to year, being for example typically greater during drought years than non-drought years. Pre-incubation egg temperatures can paradoxically reach actual incubation temperatures, which could begin development of the fertilized egg. In fact, potentially lethal temperatures have been measured in pre-incubation eggs found in ground-nests of precocial birds. For example, internal egg temperatures of pre-incubated northern bobwhite eggs (Colinus virginianus) in their nest but not yet being tended, followed ambient thermal fluctuations in non-drought years and often exceeded 40°C during peak thermal intensity \[[@pone.0184670.ref012], [@pone.0184670.ref017], [@pone.0184670.ref018]\]. In fact, bobwhite nest temperatures in drought conditions regularly peaked at 45°C and were recorded as high as 60°C \[[@pone.0184670.ref019]\]. Bobwhite eggs in ground nests could be severely impacted by further increases in temperature variability and magnitude \[[@pone.0184670.ref007]\]. Indeed, the impacts of climate change and local habitat warming, especially on the thermal biology of species including birds and their development, only continue to grow---for an entry into the voluminous and expanding literature see \[[@pone.0184670.ref020]--[@pone.0184670.ref024]\]. The objective of this study was thus to determine if pre-incubation is actually an unrecognized critical period for subsequent avian morphological and physiological development that can be revealed by experimental thermal protocols (e.g., \[[@pone.0184670.ref023]\]). Specifically, we investigated if exposure to diurnally oscillating temperatures or a constant temperature of equal heating degree-hours (the product of temperature (°C) X time (hours)) during pre-incubation affects subsequent morphological and physiological development of northern bobwhite embryos during their actual incubation period, the duration of incubation, the hatchability of eggs, and the hatching synchrony of eggs. We hypothesized that bobwhite embryos exposed to oscillating temperature treatments would exhibit a differential rate of development, metabolism, and hatching compared to those exposed to a constant temperature of the same mean value, and that an increased heat load would negatively affect development, metabolism and hatching success. Materials and methods {#sec002} ===================== Fertilized northern bobwhite eggs were collected from captive-reared breeding pairs of flight ready birds at Lake Cumberland Game Bird Farm (Mill Springs, KY, USA). Eggs were packaged and shipped to the University of North Texas (Denton, USA) on the day of collection. Eggs arrived on-site within 2 business days with a written record of the date and time of egg collection. Upon arrival, bobwhite eggs were randomly divided into 5 groups of 15 eggs, placed on plastic egg trays blunt end up, and given a unique identifying number with an indelible marker. Each egg was then weighed to the nearest 0.01 g with a digital scale (Ohaus Explorer Pro, Pinebrook, NJ, USA) and placed into the assigned pre-incubation treatment. Lake Cumberland Game Bird Farm was approved for egg production by the United States Department of Agriculture (USDA) and certified by the USDA National Poultry Improvement Plan. This research was approved by the University of North Texas Institutional Animal Use Care Committee, protocol \# 0808. Pre-incubation protocol {#sec003} ----------------------- Each of the 5 egg groups (N = 15 eggs per group) were assigned to a thermal chamber (G.Q.F. 1583 Hova-Bator with circulated air) with a unique experimental thermal protocol of oscillating or constant temperatures for a 12-d pre-incubation period. The groups consisted of a low and high oscillating temperature regime (LT~Osc~, HT~Osc~), a low and high constant temperature regime (LT~Const~, HT~Const~), and a commercial regime (T~Comm~) that was standard in the commercial poultry production industry \[[@pone.0184670.ref025]\]. The specifics of these incubation regimes are given in [Table 1](#pone.0184670.t001){ref-type="table"}. 10.1371/journal.pone.0184670.t001 ###### Diel thermal regimes for oscillating and constant treatments for eggs of northern bobwhites during a 12-d pre-incubation period. ![](pone.0184670.t001){#pone.0184670.t001g} Time of Day Oscillating Thermal Regime Constant Thermal Regime ------------- ---------------------------- ------------------------- --------- --------- --------- 0000--0759 25.00°C 30.00°C 28.85°C 33.85°C 20.00°C 0800--1059 30.00°C 35.00°C 28.85°C 33.85°C 20.00°C 1100--1359 35.00°C 40.00°C 28.85°C 33.85°C 20.00°C 1400--1659 40.00°C 45.00°C 28.85°C 33.85°C 20.00°C 1700--2359 25.00°C 30.00°C 28.85°C 33.85°C 20.00°C Peak temperatures of the low and high groups were selected based on thermal studies which described temperatures peaking at ≥40°C in non-drought years and ≥45°C in simulated drought years \[[@pone.0184670.ref017], [@pone.0184670.ref019]\], in Texas. The experimental temperature regime for each group (low and high) ensured that the temperature of the constant treatment was equivalent to the mean of the oscillating treatment, and heating degree-hours were equivalent within the low and high groups. A heating degree-hour was defined as 1°C above physiological zero (25°C for Colinus virginianus) \[[@pone.0184670.ref026]\] for 1 hour, such that 26°C for 1 h = 1 heating degree-hour, and 27°C for 1 h = 2 heating degree-hours. Thus, the mean temperature for the LT~Osc~ group was equal to that of the LT~Const~ group (28.85°C) and mean temperature for the HT~Osc~ group was equal to that of the H~Const~ group (33.85°C). Further, each low group (LT~Osc~ and LT~Const~) received 92.4 heating degree-hours per day, and each high group (HT~Osc~ and H~Const~) received 212.4 heating degree-hours per day, for the entire pre-incubation. The commercial regime (T~Comm~; 20°C) received no heating degree-hours per day, since pre-incubation temperatures did not reach physiological zero. Oscillating chambers were kept at base temperatures of 25°C (LT~Osc~) and 30°C (HT~Osc~) from midnight (0000h) to 0800h and were increased by 5°C at 0800h, 1100h, and 1400h. At 1700h, chamber temperatures were returned to their base temperature. These diurnal fluctuations were carried out each day of the 12-d pre-incubation period. The constant temperature chambers were kept at the intended temperatures throughout the entire 12-day pre-incubation period ([Table 1](#pone.0184670.t001){ref-type="table"}). All thermal chambers maintained a relative humidity (RH) of 60% to minimize compounding factors with temperature treatments. Eggs were not turned during this pre-incubation period. Assessment of development during pre-incubation {#sec004} ----------------------------------------------- To establish a baseline of pre-incubation development for egg groups subjected to each pre-incubation thermal regime, and to investigate whether various pre-incubation thermal groups affected subsequent embryonic development, eggs from each group were randomly selected for analysis (23 eggs per thermal regime group were selected over the duration of the study). Following the pre-incubation protocol, on the morning of the 13^th^ day of incubation, selected eggs were weighed to the nearest 0.01 g to determine the amount of water loss during pre-incubation. Further, eggs were opened and if embryos were present, they were separated from the egg (yolk-free), weighed to the nearest 0.01 g, aged, and staged according to morphological indicators of development \[[@pone.0184670.ref027], [@pone.0184670.ref028]\]. Incubation protocols {#sec005} -------------------- After the pre-incubation period, on the morning of the 13^th^ day, eggs from each group were removed from the thermal chamber, weighed to the nearest 0.01 g, and immediately placed into the incubator (G.Q.F. 1502 Sportsman incubator, G.Q.F. Manufacturing Co., Savannah, GA, USA). Incubator temperature was maintained at a constant 37.5±0.5°C with a RH of 60%. Eggs were turned automatically every 3 h for the first 19 days of the 23 day incubation \[[@pone.0184670.ref001]\]. On day 20 (D20) of incubation, eggs were weighed to the nearest 0.01 g and placed in the hatching chamber of the same incubator and egg turning was stopped \[[@pone.0184670.ref001]\]. Hatching was determined when eggs were star-pipped, defined as an externally pipped egg where the embryo created a small hole in the shell (approximately 3 mm^2^) to initiate hatching \[[@pone.0184670.ref024]\]. This definition of hatching was used to compensate for any artificial hatching difficulties. Upon star-pipping, the duration of incubation, degree of hatching synchrony, and percentage of eggs hatched was recorded. The stage at the time of death was recorded for eggs that did not hatch. A subset of incubating eggs was allowed to go to hatching, and the hatchlings were subsequently maintained for one day at 37.5±0.5°C with a RH of 60%, for oxygen consumption measurements (see below). Oxygen consumption {#sec006} ------------------ During incubation, on incubation days 10, 12, 14, 16, 18, 20, 22, and on 1-d post hatch, VO~2~ of the embryos in ovo was recorded via flow-through respirometry as an indicator of development and timing of physiological processes (e.g., internal pipping). On these selected days, 6 eggs from an individual group were removed and immediately placed into individual metabolic chambers located in a modified incubator (37.5°±0.5°C) as part of the flow-through respirometry system. The respirometry system was operated by pumping compressed normoxic air (159.22 mmHg) into an airflow manifold (MF-8 Sable Systems International, Las Vegas, NV, USA) which split the airflow into 7 channels and set the rate air flow through the system at 100±1 ml·min^-1^ \[[@pone.0184670.ref028]\]. The seven Nalgene air tubes then went into the incubator in a sealed opening and connected to copper coils of the same diameter, to warm the air to incubation temperature. Air then flowed through Nalgene tubes connected the copper coils, then to the 7 metabolic chambers (modified 2.5 oz sterilized Gerber baby food jar) simultaneously. Chamber 1 was kept empty (the "blank") and used as a reference for normoxic air, while chambers 2--7 contained eggs. From each chamber, air flowed to a multiplexor (Sable Systems International, Las Vegas, NV, USA) that regulated the sequence of airflow from the chambers to a column of Drierite and soda lime, which removed water vapor and carbon dioxide, respectively. A sub-sample pump then pulled a low-pressure air sample of each chamber, sequentially through the oxygen analyzer (FC-1B, Sable Systems International, Las Vegas, NV, USA). Each chamber was sampled at 5 sec consecutive intervals for 15 min. Oxygen measurements from the analyzer were sent to the computer via a Sable Systems Universal Interface II and processed by system software (Data acquisition system 2.0 and Datacan V, Sable Systems International, Las Vegas, NV, USA). VO~2~ was measured as the difference in oxygen concentration between the normoxic air flowing into the blank chamber (chamber 1) and the expired air flowing out of each treatment chamber (chambers 2--7). Statistical analyses {#sec007} -------------------- All data was tested with a Shapiro--Wilks normality test \[[@pone.0184670.ref029]\] and Hartley's Fmax test \[[@pone.0184670.ref029]\] before specific statistical analyses were performed. An independent t-test \[[@pone.0184670.ref029]\] was used to compare oscillating to constant temperature egg groups. An ANOVA \[[@pone.0184670.ref029]\] was used to identify the relationship between groups for all experiments. A Kaplan-Meier survival analysis \[[@pone.0184670.ref030]\] was used to compare survival rates and time to mortality among groups. Significance between groups was determined with a Student--Newman--Keuls (SNK) multiple range post hoc test \[[@pone.0184670.ref029]\]. All statistical tests were conducted using SigmaStat 3.5 software (Systat Software Inc. San Jose, CA), or manually \[[@pone.0184670.ref029], [@pone.0184670.ref031]\]. Statistical decisions were made with a 0.05 level of probability. All data are reported as mean±S.E. unless otherwise indicated. Results {#sec008} ======= Pre-incubation development {#sec009} -------------------------- For the subsample of embryos extracted directly after the 12-d pre-incubation period, mean stages of development were different among pre-incubation treatments (ANOVA, P\<0.001). Oscillating and constant groups significantly differed from each other ([Fig 1](#pone.0184670.g001){ref-type="fig"}) and developmental stage of all groups were statically unique (P\<0.05). During pre-incubation, eggs of the high-constant temperature (H~Const~) group were the most developed, advancing to a mean developmental stage of 21.4±0.2, equivalent to 5.3 incubation days at 37.5°C \[[@pone.0184670.ref027], [@pone.0184670.ref028]\]. Eggs exposed to high-oscillating temperatures (HT~Osc~) developed to a mean stage of 13.7±0.2, equivalent to 2.2 incubation days at 37.5°C. Low-oscillating (LT~Osc~) embryos developed to a mean stage of only 2.5±0.1, equivalent to 0.5 incubation days at 37.5°C, while pre-incubation development was lowest in low-constant (LT~Const~) eggs, which progressed to a mean stage of just 1.1±0.1, or \<0.5 incubation days at 37.5°C ([Fig 1](#pone.0184670.g001){ref-type="fig"}). Commercial pre-incubation eggs (T~Comm~) had no pre-incubation development. ![Mean developmental stages of northern bobwhite embryos after exposure to specific thermal groups during a 12-d pre-incubation period.\ Numbers above bars show days of incubation equivalent to the indicated developmental stages. Letters indicate statistical groupings. n = 23 eggs per group. Commercially incubated eggs did not have any pre-incubation development.](pone.0184670.g001){#pone.0184670.g001} Survival curves hatching success {#sec010} -------------------------------- Mortality occurred prior to hatching in all groups including the commercial group, as anticipated. The survival rate and stage of development at the time of death was significantly different between groups (Log rank survival analysis; P\<0.001). A Holm-Sidak Pairwise MCP statistically grouped the treatments receiving the lowest amounts of heating degree-hours (LT~Osc~, LT~Const~, and T~Comm~) into one group. These groups had a statistically higher survival rate compared to the high heating degree-hour groups (HT~Osc~ and H~Const~; P\<0.001, [Fig 2](#pone.0184670.g002){ref-type="fig"}). Mean survival time for LT~Const~ eggs was 22.2±1.4 days. The low oscillating group had a mean survival time of 22.1±0.7 days, and this was not statistically different from the LT~Const~ group (P = 0.25). Mean survival for HT~Osc~ eggs was 12.2±1.8 days which was significantly different from H~Const~ eggs (4.47±0.3 days), none of which hatched during any trials ([Fig 2](#pone.0184670.g002){ref-type="fig"}). ![Kaplan-Meier survival curves showing survival rate and time to mortality of northern bobwhite embryos exposed to different thermal protocols during a 12-day pre-incubation period.\ Stair-step declines show time of mortality. n = 45 eggs per treatment group. Letters indicate statistical groupings with a P\<0.05 level of significance.](pone.0184670.g002){#pone.0184670.g002} Commercial eggs hatched at a significantly higher rate (80.8±3.6%) than all treatment groups ([Fig 3](#pone.0184670.g003){ref-type="fig"}; ANOVA, P\<0.001). Hatching success of the LT~Const~ treatment (63.3±6.7%) was not significantly different from the LT~Osc~ treatment (53.5±6.4%; P = 0.24). HT~Osc~ hatch rates (6.0±2.1%) did not statistically differ (P = 0.53) from H~Const~ hatch rates (0.0%). Hatch rates for LT~Osc~ was significantly higher than for the HT~Osc~ groups (P\<0.001). ![Hatching success of northern bobwhite embryos exposed to different thermal protocols during a 12 day pre-incubation period.\ n = 45 eggs per group. Letters indicate statistical groupings.](pone.0184670.g003){#pone.0184670.g003} Pre-incubation thermal treatment not only affected survival and hatching success, but also the timing of specific hatching events---internal pipping, external pipping, and hatching. All of these developmental landmarks were significantly different between commercial and treatment groups ([Fig 4](#pone.0184670.g004){ref-type="fig"}; ANOVA, P\<0.001). HT~Osc~ eggs internally pipped earlier (D18.6±0.2) than LT~Const~ eggs (D19.7±0.2), LT~Osc~ eggs (D20.5±0.2), and T~Comm~ eggs (D20.6±0.2). Similarly, HT~Osc~ eggs externally pipped earlier (D20.0±0.0) than LT~Const~ eggs (D21.5±0.3), LT~Osc~ eggs (D21.5±0.2), and T~Comm~ eggs (D21.7±0.2). HT~Osc~ eggs also hatched earlier (D21.0±0.0) than T~Comm~ eggs (D23.0±0.0) and LT~Osc~ eggs (D23.5±0.2), while LT~Const~ eggs took the longest to hatch of treatment groups (D23.8±0.1). ![Time to internal and external pipping and hatching for northern bobwhite eggs exposed to T~Comm~, LT~Const~, LT~Osc~, and HT~Osc~ thermal treatments during a 12 day pre-incubation period.\ Sample size in parentheses. Events are statistically grouped (0.05 level of significance) with lower-case, italicized, and capital lettering.](pone.0184670.g004){#pone.0184670.g004} Additionally, the duration from internal pip to hatching was significantly different between treatment groups (P\<0.001) ([Fig 4](#pone.0184670.g004){ref-type="fig"}). The hatching duration was not significantly different between LT~Const~ (3.7±0.4 days) and LT~Osc~ (3.2±0.1 days) treatments (P = 0.16), but HT~Osc~ had a shorter duration of hatching events than LT~Osc~ (P\<0.001). The duration between internal pip and hatch was the shortest for HT~Osc~ (2.4±0.2 days) and T~Comm~ eggs (2.4±0.2 days. All eggs in a given group hatched essentially synchronously--i.e., within 24 h of each other. Oxygen consumption {#sec011} ------------------ Oxygen consumption (VO~2~) increased significantly as a function of development ([Fig 5](#pone.0184670.g005){ref-type="fig"}). VO~2~ in LT~Const~ and LT~Osc~ groups was similar in overall pattern throughout incubation, but absolute values differed significantly (P\<0.05) on D10, D15, D18, D20, and D21 of incubation ([Fig 5A](#pone.0184670.g005){ref-type="fig"}). HT~Osc~ embryos had a significantly higher VO~2~ than H~Const~ embryos on D10 but not on D15, however H~Const~ embryos did not survive past day 15 of incubation, so no further comparisons were possible ([Fig 5B](#pone.0184670.g005){ref-type="fig"}). Hatchling VO~2~ was higher than at any point in embryonic development in all groups, but there were no significant differences between treatments in hatchlings. ![Oxygen consumption of northern bobwhite embryos as a function of development and pre-incubation thermal groups.\ Statistically similar means are grouped within boxes. \* No H~Const~ eggs survived past day 15.](pone.0184670.g005){#pone.0184670.g005} To further aid in VO~2~ comparison, semi-log plots were created ([Fig 6](#pone.0184670.g006){ref-type="fig"}). Slopes relating VO~2~ to development were not significantly different between LT~Const~ and LT~Osc~ pre-incubation treatments ([Fig 6A](#pone.0184670.g006){ref-type="fig"}; t test for slopes, P\>0.50). Slopes relating VO~2~ to development were significantly different between LT~Osc~ and HT~Osc~ pre-incubation treatments ([Fig 6B](#pone.0184670.g006){ref-type="fig"}; t-test for slopes, 0.02\>P\>0.001). As noted above, H~Const~ eggs did not survive past day 15 of incubation, so no oxygen consumption comparisons were made with this treatment group. ![Regressions relating oxygen consumption to development.\ A) No significant differences existed between the slope of the lines for LT~Const~ and LT~Osc~ (ANOVA; P\>0.50). B) The slope of the line for relating LT~Osc~ was significantly higher than for HT~Osc~ (0.02\>P\>0.001). Coefficient of determination (r^2^) and n values are shown.](pone.0184670.g006){#pone.0184670.g006} Egg mass and water loss {#sec012} ----------------------- Mean egg weight loss among thermal groups during the pre-incubation period was higher in LT~Osc~ and HT~Osc~ egg groups than in all other pre-incubation groups (P\<0.001), with a percentage loss of 4.9±0.2% and 3.9±0.5% respectively. There was no difference between HT~Const~ eggs (2.8±0.5%) and LT~Const~ eggs (2.6±0.1%) and between LT~Const~ eggs and T~Comm~ eggs (2.14±0.04%; P\>0.05) ([Fig 7](#pone.0184670.g007){ref-type="fig"}). ![Mean percentage of fresh egg weight loss for northern bobwhite eggs measured after exposure to oscillating and constant thermal treatment for a 12 day pre-incubation period.\ n = 45 eggs per group. Letters (lower-case, italicized, and capital) indicate statistical groupings within each treatment.](pone.0184670.g007){#pone.0184670.g007} Unlike pre-incubation, no significant difference (P = 0.88) in the percentage of fresh egg weight loss during incubation was detected between HT~Osc~ and H~Const~ (P = 0.05) or LT~Osc~ and LT~Const~ (P = 0.88) groups. However, there was a significant (P\<0.05).difference in egg weight loss between T~Comm~ eggs (11.5±0.3%) and H~Const~ eggs (14.9±1.5%) during incubation ([Fig 7](#pone.0184670.g007){ref-type="fig"}). Combining the pre-incubation and incubation periods, all treatment groups lost a greater percentage of fresh egg mass than T~Comm~ eggs (ANOVA, P\<0.05). The mean percentage of fresh egg mass lost by T~Comm~ eggs eggs was 11.5±0.3%, lower than H~Const~ (17.7±1.9%), HT~Osc~ (14.9±1.5%), LT~Const~ (15.9±0.6%), and LT~Osc~ eggs (18.3±0.7%). No difference was detected in percentage of fresh egg mass lost between HT~Osc~ and H~Const~ (P = 0.28), or LT~Osc~ and LT~Const~ (P = 0.12) ([Fig 7](#pone.0184670.g007){ref-type="fig"}). Discussion {#sec013} ========== Pre-incubation modification of subsequent embryonic development {#sec014} --------------------------------------------------------------- This study shows that quail embryos actually begin to develop during pre-incubation when eggs are exposed to temperatures above "physiological zero". As expected, more development was obtained with more heating degree-hours and less development occurred with fewer heating degree-hours, even when slightly above physiological zero. Eggs exposed to high heat loads (HT~Osc~ and H~Const~) and low heat loads (LT~Osc~ and LT~Const~) had equal quantities of heating degree-hours within groups, but they all exhibited differential pre-incubation development and growth rates. Groups exposed to a constant temperature had the largest variation between high and low groups, with a difference of \~5 days of development between H~Const~ and LT~Const~, even though each group received the same amount of heating degree-hours ([Fig 1](#pone.0184670.g001){ref-type="fig"}). The difference in groups with oscillating temperature was much less, with about 1.7 days of development difference between HT~Osc~ and LT~Osc~. Thus, at least for quail development, the quantity of heat (e.g., number of heating degree-hours) is not the only driver of development and that the nature of the heat (oscillating or constant) is also a major factor. Recently, intragenerational epigenetic modification of embryonic development has been induced by the length of pre-incubation storage in chicken embryos \[[@pone.0184670.ref032]\]. These changes were not the result of high temperatures, since the eggs were stored well below physiological zero. Whether an oscillating temperature at cool temperatures with no likelihood of beginning embryonic development would actually alter embryonic development remains to be investigated. Survival and hatching assessment {#sec015} -------------------------------- The thermal protocol used during the pre-incubation period in Colinus virginianus had a profound effect on subsequent embryonic survival. While high pre-incubation temperatures severely affected subsequent embryonic survival, it is interesting to note that an equivalently high oscillating temperature, while leading to only poor survival, nonetheless was more successful in producing viable embryos than a constant high temperature. Observations of the effect of pre-incubation temperature and other conditions on subsequent embryonic survival have mostly derived from interest in poultry production or basic experimentation \[[@pone.0184670.ref032]--[@pone.0184670.ref038]\]. Such studies on the pre-incubation period have almost invariably considered steady-state storage conditions, rather than oscillations in environmental parameters, such as the oscillations in temperature employed in the current study. Few studies on oscillating temperatures exist for any aspect of avian egg storage---the few such studies have been directed towards inducing temperature oscillations once formal incubation has begun \[[@pone.0184670.ref015]\]. Embryonic survival and hatching success was not statistically different between constant and oscillating groups. However, H~Const~ embryos did not hatch, dying at \~4 to 15 days of incubation ([Fig 2](#pone.0184670.g002){ref-type="fig"}). The death of H~Const~ embryos is noteworthy, since eggs were kept for 12 days at 33.85°C, well within the natural incubating range of 30--35°C measured in nature for this species \[[@pone.0184670.ref017]\]. While the objective of this study was not to determine the cause of death in eggs at constant high temperature, we speculate that constant temperatures near incubation temperature require eggs to be turned daily for survival. Indeed, studies have shown that, while turning of quail eggs is not required when eggs are stored at 12.8°C, turning is required during incubation at temperatures between 35--38°C \[[@pone.0184670.ref001], [@pone.0184670.ref039], [@pone.0184670.ref040]\]. High mortality was not experienced in LT~Const~ groups (28.85°C) presumably because the mass of the embryo was that at \<0.5 day of age (i.e. a much smaller embryo), and thus was not affected by the absence of turning. Bobwhite quail eggs exposed to high oscillating temperatures internally pipped, externally pipped, and hatched two days earlier than all other groups, ultimately showing a reduced incubation period ([Fig 4](#pone.0184670.g004){ref-type="fig"}). This should not be surprising since the high oscillating temperature protocol (HT~Osc~) during pre-incubation enabled approximately 2 additional days of development than in LT~Osc~ and LT~Const~ groups. Both LT~Osc~ and LT~Const~ eggs, which developed approximately 0.5 days during the pre-incubation period, exhibited an extended incubation time of 0.5 d ([Fig 4](#pone.0184670.g004){ref-type="fig"}). HT~Osc~ eggs not only started development earlier, but also showed overall reduced incubation duration of HT~Osc~ eggs. This may be an adaptation to the thermally stressful environments inhabited by northern bobwhite quail, where both high temperature and temperature fluctuation are common and would affect the eggs during their pre-incubation period. Collectively, our results suggest that oscillating temperatures act differently on developing organisms than constant temperatures of equal heating degree-hours. The impact of these differences is increased as heat loads are increased, i.e., higher heating degree-hours result in more variation in development in embryos exposed to constant and oscillating temperatures. Further, within oscillating groups, higher oscillating heat loads caused an increase in the rate of development during pre-incubation, a decrease in the rate of development during incubation, and had more of a negative effect on hatching. Hatching synchrony {#sec016} ------------------ While the current study did not focus on hatching synchronization, it is interesting to note that, irrespective of thermal regime, all quail eggs hatched within 24h of each other--i.e. "synchronously" \[[@pone.0184670.ref041]\]. Northern bobwhites do not begin incubation until the last egg is laid \[[@pone.0184670.ref042]\]. Consequently, all eggs receive equal amounts of incubation, resulting in synchronous hatching. This hatching adaptation seems beneficial by allowing the incubating parent to quickly transition behavior from incubating sessile eggs to brooding motile hatchlings \[[@pone.0184670.ref041], [@pone.0184670.ref043]\]. Importantly, limited studies have investigated hatching synchrony in warmer climates where nest temperatures fluctuate during pre-incubation and can often exceed physiological zero (the temperature above which embryonic development begins). In warmer climates, those eggs laid early could actually begin to develop prior to incubation, resulting in asynchronous development at the onset of incubation and asynchronous hatching at the other end of embryonic development. In this case, eggs might hatch asynchronously unless a synchronizing mechanism takes place \[[@pone.0184670.ref006], [@pone.0184670.ref044]--[@pone.0184670.ref046]\]. Hatching synchrony has been observed in the laboratory among quail eggs that differed in developmental stages by 2 days \[[@pone.0184670.ref046]\], but not by up to 5 days, which is the probable result of northern bobwhites in warm climates given our experimental findings on stage of development in high temperature groups at the onset of incubation. Interestingly, eggs must remain in physical contact with each other for synchronous hatching to occur, especially in the last 20% of incubation when pipping occurs \[[@pone.0184670.ref046], [@pone.0184670.ref047]\]. Regardless of relatedness or order of laying, viable bobwhite eggs of the same age hatch synchronously if touching \[[@pone.0184670.ref048]\]. When isolated, their time of hatch is spread over a longer period \[[@pone.0184670.ref045]\]. Bobwhite eggs accelerated hatching when placed in contact with more-advanced eggs (24 h advanced in development) and delayed hatching when placed in contact with more-retarded eggs (24 h retarded in development) \[[@pone.0184670.ref045]\], suggesting that both an acceleration and delay of hatching time occur between earlier laid and later laid eggs in the wild. Oxygen consumption and development {#sec017} ---------------------------------- Oxygen consumption during avian embryonic development has been very well documented, and the reader is directed to recent discussions of that literature \[[@pone.0184670.ref016], [@pone.0184670.ref049]--[@pone.0184670.ref051]\]. Many factors influence avian embryonic VO~2~, most notably body mass, oxygen availability, and environmental temperature. Yet, in precocial birds, there is a representative pattern of VO~2~ change during development in which VO~2~ typically increases until approximately 80% of incubation time. Thereafter, rate of VO~2~ change plateaus until the embryo internally pips into the air cell. VO~2~ increases once more, followed by final increase at the time the chick pips the shell and hatches \[[@pone.0184670.ref047]\]. The oxygen consumption pattern of change in the current study ([Fig 5](#pone.0184670.g005){ref-type="fig"}) was generally that described for precocial birds \[[@pone.0184670.ref047]\], and the pattern is consistent with reports of developing northern bobwhite quail \[[@pone.0184670.ref052]\] as well as the Coturnix quail, Coturnix coturnix \[[@pone.0184670.ref041]\], a precocial bird with similar sized eggs. Importantly, in our study the thermal protocol during pre-incubation significantly affected VO~2~, which was different between LT~Const~ and LT~Osc~ thermal groups. The slopes of the normalized VO~2~ data were not different between LT~Const~ and LT~Osc~ treatments ([Fig 6A](#pone.0184670.g006){ref-type="fig"}), indicating no overall differences in rate of development to match the elevation of development induced by higher temperature. There were significant differences in VO~2~ between HT~Osc~ and LT~Osc~ groups logically due to HT~Osc~ embryos having developed more during pre-incubation ([Fig 1](#pone.0184670.g001){ref-type="fig"}), thus requiring more oxygen on any given day during incubation. However, the lower slope of VO~2~ for HT~Osc~ eggs during incubation showed that HT~Osc~ eggs developed at a slower rate during incubation than LT~Osc~ eggs after a higher rate of development during pre-incubation ([Fig 6B](#pone.0184670.g006){ref-type="fig"}). Additionally, HT~Osc~ eggs did not exhibit the same oscillating VO~2~ patterns in the last 20% of incubation as LT~Osc~ and LT~Const~ eggs. Rather, VO~2~ of HT~Osc~ eggs stayed relatively constant during the pipping process ([Fig 5B](#pone.0184670.g005){ref-type="fig"}). Why and how VO~2~ during incubation is affected by pre-incubation thermal protocols awaits further investigation. Implications for natural bobwhite quail populations {#sec018} --------------------------------------------------- The intention of comparing LT~Osc~ (simulated non-drought temperatures) and HT~Osc~ (simulated drought temperatures) groups was to better understand the differences between drought and non-drought conditions as well as the potential impacts of global warming on avian development generally, and quail populations specifically. In this context, one of the most interesting findings is the significant difference in hatch rates between high-oscillating (HT~Osc~, 6.0±2.1%) and low-oscillating (LT~Osc~, 53.5±6.4%) groups ([Fig 2](#pone.0184670.g002){ref-type="fig"}). The HT~Osc~ thermal group was pre-incubated at a temperature 5.0°C higher than the LT~Osc~ group, a temperature increase coincident with the predicted increase in global average temperature of 5±1°C in the United States by the period 2071--2100 relative to the period 1961--1990 \[[@pone.0184670.ref053]\]. Our results show that a 5°C increase in pre-incubation temperature of just 12 days reduces hatching rate of northern bobwhites by approximately one half. Reductions in the percentage of juvenile bobwhite quail at different lattitudes \[[@pone.0184670.ref054]\] or during drought \[[@pone.0184670.ref055], [@pone.0184670.ref056]\] could, according to our findings, be caused by increased heat loads during the 12 day pre-incubation period. Eggs in ground nests could be severely impacted by further increases in temperature variability and magnitude as global warming continues. Thus, our study indicates that the dynamic nature and magnitude of diel temperature should be considered when evaluating the biological response to predicted temperatures of global warming. We thank D. Reyna, G. Bowling, and H. Valls for providing logistical support with egg incubation, and E. Dzialowski for providing training on measuring oxygen consumption. [^1]: Competing Interests:The authors have declared that no competing interests exist.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1} =============== Cancer is a complex disease that is very heterogenic and variable at cellular level and also differs from one patient to the other in its behaviour, development, and outcome \[[@B1]\]. Physical, metabolic, and behavioural variations of cancer cells from normal ones arise through the accumulation of genetic modifications and help them to proliferate rapidly, escape from host immune surveillance, and ultimately invade distant tissues \[[@B2]\]. Histopathological, genetic, and epigenetic and clinical outcome variations between and within different types of cancers have been the greatest challenge to understand the disease and develop novel therapies \[[@B3]\]. Surgery and radiation therapy were the most preferred means of treatment to control cancer before 1950 and after 1960, respectively \[[@B4]\]. Chemotherapy can be done before surgery to shrink the tumor or after surgery to kill the remaining cancer cells \[[@B5]\]. However, most of the chemotherapeutic drugs lack specificity and tend to rapidly damage normal dividing tissues, causing side effects such as immunosuppression, neurotoxicity, and hair loss \[[@B6]\]. Moreover, resistance has also reduced therapeutic efficacy of some anticancer chemotherapeutic drugs \[[@B7]\]. In order to address these limitations, tapping nature as a major source of chemically diverse novel anticancer compounds is a consistently proven track \[[@B8]\]. Screening natural products yield more hit with more "drug-like" characteristics (absorption and metabolism) as compared to screening of rationally designed compounds \[[@B9]\]. Furthermore, screening medicinal plants based on traditional use provides a higher chance of finding active plants relative to the random approach \[[@B10]\]. Ethiopia has a rich and diverse heritage of traditional medical practices, known for using plants to prepare more than 90% of the remedies \[[@B11]\]. In addition, the country has more than 6,500 higher plant species of which, around 12% are endemic \[[@B12]\]. Reports indicate that up to 80% of the population relies on traditional remedies as a primary source of health care \[[@B13]\]. Only few ethnobotanical reports from different agroecological zones of Ethiopia are available in the literature regarding medicinal plants used for cancer treatment. These include Bersama abyssinica, Buddleja polystachya, Clerodendrum myricoides, Dovyalis abyssinica, Ekebergia capensis, Myrsine melanophloeos, Olea capensis, Pentas lanceolata, Sideroxylon oxyacanthum, and Zingiber officinale \[[@B14]\]; Bidens macroptera, Clematis simensis, Ferula communis, and Punica granatum \[[@B15]\]; Rumex abyssinicus \[[@B16]\]; Zanthoxylum chalybeum \[[@B17]\]; Phytolacca dodecandra and Vinca rosea \[[@B18]\]; Kalanchoe lanceolata, Stephania abyssinica, and Vernonia hymenolepis \[[@B19]\]; Plumbago zeylanica \[[@B20]--[@B22]\]; Acalypha acrogyna, Carissa spinarum, Maytenus ovatus, and Salvia nilotica \[[@B23]\]; Croton macrostachyus \[[@B24]\]; and Dorstenia barnimiana \[[@B25], [@B26]\]. In view of this fact and considering the weak traditional recording and knowledge transfer system and an alarming rate of environmental degradation, finding anticancer plants and documenting their ethnobotanical information constitute an urgent and indispensable task. Therefore, the main aim of this study was to establish an inventory of medicinal plants traditionally used to treat cancer in eleven districts of Ethiopia. 2. Materials and Methods {#sec2} ======================== 2.1. Description of the Study Areas {#sec2.1} ----------------------------------- This ethnobotanical study was conducted in four national regional states of Ethiopia: Oromia, Amhara, Afar, and Southern Nations, Nationalities, and People. The survey included different districts from each region, namely, Bale Robe and Goba from Oromia, Bahir Dar Zuria and Filiklik from Amhara, Gewane from Afar, and Wondo Genet, Sodo Zuria, Doyo Gena, North Bench, Mizan Aman, and Shako from Southern Nations, Nationalities, and People Regional State ([Figure 1](#fig1){ref-type="fig"}). These geographically, culturally, and agroecologically different study areas ([Table 1](#tab1){ref-type="table"}) were selected mainly based on the availability of traditional healers and recommendations from health workers. 2.2. Data Collection {#sec2.2} -------------------- A team comprising a botanist and researchers from Addis Ababa University was set up, and health authorities were contacted for permission and identification of traditional herbalists living in each study area. Altogether, 117 traditional healers were approached using the snowball technique and 74 traditional healers who used herbs to manage cancer-like symptoms were selected. Ethnobotanical data were collected between January and August 2016, mainly through individual interviews with the selected traditional herbalists using a semistructured interview questionnaire. The questionnaire was prepared in Amharic language and translated to different local languages for traditional healers who do not speak Amharic. This questionnaire was designed to obtain information in the following areas: (i) general data on the informant, (ii) school attendance, (iii) use of plants for cancer treatment, (iv) source of the plant material, (v) part of the plant used, (vi) method of medicinal preparation, (vii) route of administration, and (viii) side effects. A traditional healer for the purpose of this study is "a person who is recognized by the community in which s/he lives as competent to provide healthcare by using plants and plant products." Each traditional healer was approached, briefed about the purpose of the research, and asked for his/her verbal consent in talking about cancer and its treatment. They were assured of the confidentiality of the information they provided. If plants were mentioned for their anticancer purposes, a botanical sample was collected. These specimens were pressed and preserved for later identification at the National Herbarium, Addis Ababa University, Addis Ababa, and a voucher specimen of each plant was deposited in the institute. All botanical names have been transcribed according to the nomenclature system used by the Plant List ([http://www.theplantlist.org](http://www.theplantlist.org/)). 2.3. Data Analysis {#sec2.3} ------------------ The relative importance of medicinal plants used in the management of cancer-like symptoms in study areas was assessed using the relative frequency of citation (RFC), use value (UV), informants consensus factor (ICF), and cultural importance index (CI). ### 2.3.1. Relative Frequency of Citation (RFC) {#sec2.3.1} The RFC was calculated by dividing the number of informants that cite a particular plant species (FC) by the total number of informants in the survey (N) \[[@B29]\]:$$\begin{matrix} {\text{RFC} = \frac{\text{FC}}{N}.} \\ \end{matrix}$$ ### 2.3.2. Use Value (UV) {#sec2.3.2} The use value demonstrates the relative importance of plant species to treat particular ailment, and it is determined by the following formula \[[@B30]\]:$$\begin{matrix} {\text{UV} = {\sum\frac{U_{i}}{N_{i}}},} \\ \end{matrix}$$where "UV" stands for the use value of a species, "U~i~" stands for the number of use reports cited by informants for that plant species, and "N~i~" is the total number of informers who reported the particular plant species i. ### 2.3.3. Informant Consensus Factor (ICF) {#sec2.3.3} Informant consensus factor (ICF) was calculated to determine the homogeneity of the information collected about particular plant species to treat specific ailment. It was estimated using the following formula \[[@B31]\]:$$\begin{matrix} {\text{ICF} = \frac{\text{Nur} - \text{Nt}}{\text{Nur} - 1},} \\ \end{matrix}$$where Nur is the number of use reports of informants for particular ailment category and Nt refers to the number of species used for the ailment category by all informants. ### 2.3.4. Cultural Importance Index (CI) {#sec2.3.4} Cultural importance index (CI) is calculated by the sum of the use reports (UR) of informants mentioning each species use (from i~1~ to i~N~) in each use category and adding all the UR of each category (from u~1~ to u~NC~) divided by the total number of informants N. This index is determined by the following formula \[[@B29]\]:$$\begin{matrix} {\text{CI}_{i}{\sum\limits_{u = u_{1}}^{u_{\text{NC}}}{\sum\limits_{i = i_{1}}^{i_{N}}\frac{\text{UR}u_{i}}{N}}},} \\ \end{matrix}$$where CI is an ethnobotanical index that indicates the spread of the use along with the diversity of uses of each species. 3. Results {#sec3} ========== The informants consisted of 66 male and 8 female traditional healers and they were divided into three age groups: 20--40, 41--60, and ≥ 61 years. Out of 74 interviewed traditional healers, most of them (N%) were adults aged between 41 and 60 years. Majority of the respondents (70.2%) gained their knowledge from family members and 82% of all interviewed respondents practiced ethnomedicine for more than 25 years. More than 70% of the respondents were either only at their primary level of education or did not have a formal education at all ([Figure 2](#fig2){ref-type="fig"}). Traditional healers usually used their intuition and relied on the chronicity and growth of external mass, as a means to diagnose cancer. Lumpy growth was the most commonly cited criteria used to diagnose cancer, followed by ulcerative wounds and bleeding ([Table 2](#tab2){ref-type="table"}). However, there were instances where some of the healers claimed to have treated patients already diagnosed with cancer at modern health institutions. Traditional healers identified cancer as "Nekersa" in Bahir Dar Zuria and Filiklik, "Naqarsa" in Bale Robe and Goba, "Sissac" in Gewane, "Xoka or Toka" in Doyo Gena, "Balamo" in Wondo Genet, "Kums or niamt" in North Bench, and "Kanser" in Sheko and Sodo Zuria district. Out of the 6 specific cancer types (skin, breast, lung, cervical, throat, and intestinal) claimed to be treated by the respondents, skin cancer was a dominant one followed by breast cancer. A total of 53 plant species belonging to 30 families were reported for their anticancer use ([Table 3](#tab3){ref-type="table"}). The result of this study showed that shrubs (49.1%), herbs (33.9%), trees (13.2%), and climbers (3.8%) were the main sources of potential anticancer medicinal plants. This study also indicated that leaves (56.7%) were the most commonly used plant parts followed by roots (21.7%), bark (6.7%), stem (1.7%), seeds (1.7%), whole plant (1.7%), leaves and roots (5%), leaves or stem (1.7%), and leaves or seeds (1.7%) ([Figure 3](#fig3){ref-type="fig"}). Most of the reported plants occurred naturally in wild (96.2%); however, cultivation was also a source (3.8%). Reported medicinal plants have been traditionally claimed to be used to treat different types of ailments including cancer. However, only few have been scientifically investigated for their antiproliferative or cytotoxic activity ([Table 4](#tab4){ref-type="table"}). While comparing the amount and distribution of anticancer plants in the past ten years, regardless of the study areas, all respondents believed that the amount and distribution of these plants are reduced. In the current study, the highest UVs were recorded for Aloe spp. (6), Albizia schimperiana (4), Sida schimperiana (4), Achyranthes aspera (4), Brucea antidysenterica (4), Cleome brachycarpa (3), Leonotis ocymifolia (3), and Prunus africana (3). The lowest UVs were obtained for Acokanthera schimperi, Acmella caulirhiza, Cineraria abyssinica, and Gnidia involucrata ([Table 3](#tab3){ref-type="table"}). A total of 228 use reports have been documented and categorized into seven categories ([Table 5](#tab5){ref-type="table"}). Among these, other ailments (46.3%) and skin cancer (26.5%) had the highest use reports. Furthermore, ICF values were also calculated and ranged from 0 to 0.42. The highest ICF values were recorded in other ailments (0.42) and breast cancer (0.32) followed by skin cancer (0.23) category ([Table 5](#tab5){ref-type="table"}). The other ailments category comprises of diseases such as stomach ache, malaria, wart, swelling, wounds, evil eye, toothache, bleeding, gastrointestinal disorder, headache, bone fracture, cough, snake bite, herpes simplex, tonsillitis, hypertension, dandruff, fever, and hemorrhoid. The ICF value of the remaining four categories (lung cancer, colon cancer, cervical cancer, and throat cancer) was zero. Quantitative ethnobotanical indexes such as RFC and CI were calculated in this study to analyze the ethnobotanical information. According to RFC values, Croton macrostachyus (0.1), Vernonia auriculifera (0.04), Clematis simensis (0.04), and Acmella caulirhiza (0.04) are the most frequently cited among all reported plants. Croton macrostachyus (0.16), Dorstenia barnimiana (0.12), and Aloe spp. (0.08) rank 1^st^, 2^nd^, and 3^rd^ in position, respectively, according to the CI reference. Our result also shows that the Pearson correlation coefficient of RFC was positively and negatively correlated to CI and UV, respectively ([Table 6](#tab6){ref-type="table"}). Most of the reported remedies, prepared from these plants, were either applied topically (50%) or taken orally (41.7%). The remaining remedies were prepared to be administered either topically or orally (3.3%), both topically and orally (1.7%), and intranasally (1.7%). Usually, fresh plants were finely chopped, dried, and pounded to powder form. Then, the powder of either one or the combination of more than one plant was either mixed with drinking water or pasted and applied topically. In other cases, fresh plant parts were decocted and taken orally or crushed and applied topically. Water was the main medium in preparation of most remedies and additives like honey, milk, and butter were also used. To determine the amount of plant parts used to prepare remedies, traditional healers used spoon, fingertip, and number (in case of fresh leaves). Adverse effects reported by respondents include vomiting, diarrhea, and skin ulcers. 4. Discussion {#sec4} ============= Despite the rich biodiversity of the study areas, broad acceptability, and centuries-old tradition of using traditional medicines, the number of anticancer plants reported in this study is far less than expected. As it was reported by different ethnobotanical studies conducted in different parts of Ethiopia, this could be attributed to the attitude of many traditional healers to guard their indigenous medical knowledge as a family secret and hence hesitant to share with the researchers \[[@B13], [@B32], [@B73]\]. Justifying the lower number of female traditional healers (8, 11%) participated in this study, these studies also inferred that traditional healers usually pass their knowledge to the first son of the family. In this study, in agreement with the studies conducted in Fiche district \[[@B35]\], Ghimbi district \[[@B20]\], and Hawassa city \[[@B17]\] of Ethiopia, the predominant botanical families recorded, listing over 5 plant species each, were Asteraceae, Fabaceae, and Lamiaceae. This could be due to the fact that these families are the largest in the flora of Ethiopia and Eritrea \[[@B15], [@B21], [@B143]\]. Moreover, cytotoxicity studies conducted on different Mexican plants reported that the highest number of plant species with both in vitro and in vivo antineoplasic activities was from these families \[[@B20]\]. The highest UVs recorded in this study include Aloe spp. (6), Achyranthes aspera L. (4), Albizia schimperiana (4), Sida schimperiana (4), and Brucea antidysenterica (4). The highest ICF value (0.42) recorded for "other ailments" category, in this study, suggests that informants are in agreement with the use of particular plant species to treat ailments in this category. The lowest ICF value (0) obtained was for lung, colon, cervical, and throat cancer categories. This might be due to the cultural and ecological differences of the study sites and the difficulty to pinpoint the physical symptoms of lung, colon, cervical, and throat cancer as compared to the breast and skin cancer. The present study also revealed that RFC and CI values of some reported species are similar. However, there is a distinct difference in species ranking using each index. C. macrostachyus is placed in the first position according to both RFC and CI index. This could be due to the fact that this species is mentioned by many informants and is the most recognized plant in most study areas. Furthermore, CI value of C. macrostachyus is also high, suggesting the diversified use of the plant. V. auriculifera and C. simensis ranked next to C. macrostachyus, according to RFC index. On the other hand, D. barnimiana and Aloe spp. ranked 2^nd^ and 3^rd^ by CI index. It has been suggested that UV value is a good measure of use diversity, than the number of citations \[[@B144]\]. In agreement with this, UV value in our study is driven by species with greatest number of use rather than those cited by more informants. The Pearson correlation coefficient of −0.36, between RFC and UV, shows significant negative association between the local importance of each medicinal plant and relative importance of use of plants. This result is in contrast to previous studies that reported a significant positive correlation between RFC and UV \[[@B145], [@B146]\]. On the other hand, there is a significant positive correlation between RFC and CI (r^2^ = 0.74, p \< 0.001) implying that their pattern matches across species. The species with larger RFC value usually have higher CI, such as Croton macrostachyus and Vernonia auriculifera. Leaves and roots are the most commonly used plant parts in the preparation of remedies in the study districts. Similarly, other ethnobotanical studies conducted in different parts of Ethiopia also reported that leaves are the dominant plant part followed by root \[[@B16]--[@B18], [@B20]\]. The preference towards leaves may be because leaves are the main photosynthetic organs in plants and the primary reservoirs for secondary metabolites with medicinal values \[[@B36]\]. In contrary to other ethnobotanical studies \[[@B17], [@B18]\], where the common use of concoctions and oral route were reported, in the current study majority of the reported remedies are prepared from a single plant species and applied topically. Comparative analysis of this study with other ethnobotanical surveys of plants used traditionally in treating and managing cancer in Ethiopia \[[@B18]\], Kenya \[[@B147]\], Cameroon \[[@B37]\], Nigeria \[[@B19], [@B38]\], South Africa \[[@B39]\], and Bangladesh \[[@B148]\] revealed some similarities in the plants cited in these surveys. Of the 30 plant species cited to be used in Ethiopia \[[@B18]\], 7 species are identified in our study: Bersama abyssinica Fresen., Brucea antidysenterica JF. Mill., Calpurnia aurea (Ait.) Benth. Dodonaea angustifolia L.f., Dorstenia barnimiana Schweinf, Kalanchoe petitiana A. Rich., and Prunus africana (Hook. f) Kalkm. Although herbal remedies are believed by the general public to be safe \[[@B46]\], some research findings suggested otherwise. For instance, traditionally used Thai anticancer plants Ganoderma lucidum (Fr.) Karst., Houttuynia cordata Thunb., and Saussurea involucrata Matsum. & Koidz. were reported to cause side effects such as headache, insomnia, constipation, and diarrhea \[[@B62]\]. Similarly, side effects such as vomiting, diarrhea, and skin necrosis, associated with the use of traditional herbal remedies, were reported in this and other ethnobotanical studies conducted in Ethiopia \[[@B149], [@B150]\]. Few side effects reported in this study, as compared to other ethnobotanical studies conducted in Ethiopia, could be attributed to the frequent use of the topical route of administration. Nevertheless, considering the probability of underreporting adverse effects, extensive toxicological investigations should be conducted to protect the public. In vitro cytotoxicity and antioxidant properties of some of the plants reported in our study have also been studied. Among these plants, potent cytotoxic activity was reported for knipholone anthrone isolated from Kniphofia foliosa, with IC~50~ value that ranges between 0.9 ± 0.1 and 3.3 ± 0.4 μg/mL \[[@B89]\]. Similarly, Nibret and Wink reported the cytotoxic activity of the crude extract of Acokanthera schimperi with IC~50~ value of 7.1 μg/mL \[[@B73]\]. Studies conducted on the leaves of Cineraria abyssinica \[[@B100]\], bark of Senna singueana \[[@B116]\], and bark and leaves of Rumex nepalensis \[[@B79]\] also revealed potent radical scavenging activity of these plants. 5. Conclusion {#sec5} ============= The present study showed that traditional healers in eleven districts of Ethiopia use different medicinal plants to manage cancer-like symptoms. Frequency of citation value ranked Croton macrostachyus Del., Clematis simensis Fresen., Dorstenia barnimiana Schweinf, Vernonia auriculifera Hiern, and Acmella caulirhiza Del. as most cited plant species in study areas. Hence, based on these findings, we are currently evaluating the in vitro antiproliferative activities of reported medicinal plant species with a higher frequency of citation against human breast adenocarcinoma (MCF-7), human uterine cervical adenocarcinoma (SiSo), human lung carcinoma (A-427), and human bladder cancer (RT-4) cell lines using crystal violate assay. However, considering the rapid disappearance of the traditional knowledge of medicinal plants and an urgent need for new anticancer agents, additional studies have to be conducted to document and scientifically validate traditionally used Ethiopian anticancer plants. The authors wish to thank all traditional healers, individuals, data collectors, and local administrative authorities in all study districts that made this survey possible. This research was supported by the thematic research grant from the Addis Ababa University (grant number: TR/35/2015). Data Availability ================= The authors declare that all data supporting the finding of this study are included in this article and its supplementary information files. Ethical Approval ================ Ethical approval was obtained from Addis Ababa University, College of Health Sciences Ethics Review Board (Ref no. ERB/SOP/126/12/2015). Consent ======= Each participant consented before the interview. Conflicts of Interest ===================== The authors declare that they have no conflicts of interest. Authors\' Contributions ======================= A.B., E.E., and K.A. jointly conceived the study. S.T. conducted the ethnobotanical study and taxonomical identification. S.T., A.B., E.E., T.G., and K.A. enriched the draft manuscript for its intellectual content. All authors read and approved the final manuscript. ![Map of Ethiopia showing the location of study districts.](ECAM2020-7683450.001){#fig1} ![Demographic details of the interviewed informants.](ECAM2020-7683450.002){#fig2} ![Frequency of plant parts used for the preparation of medicinal remedy.](ECAM2020-7683450.003){#fig3} ###### Vegetation type, climatic condition, and demographic data of the study areas \[[@B27], [@B28]\] (source: National Meteorological Service Agency of Ethiopia). District Distance from capital city (km) Approximate population (2015) Number of interviewed healers Area size (km^2^) Geographical location Average elevation above sea level (m.a.s.l) Vegetation type Climatic condition (2014) ----------------- --------------------------------- ------------------------------- ------------------------------- ------------------- ------------------------------- --------------------------------------------- ----------------------- --------------------------- ------------ Bale Robe 432 65,284 2 8.87 7°07′11.65″ N 40°00′24.82″ E 2480 DAF 745.6 9.2--23.2 Goba 444 47,135 7 20.15 7°00′41.66″ N 39°58′33.96″ E 2614 DAF 736.3 9.5--23.8 Bahir Dar Zuria 578 206,708 16 1443.37 11°34′27.15″ N 37°21′40.87″ E 1800 CTW, DAF, and FLV/MFS 1547.1 12.7--27.6 Filiklik 188 142,722 7 806.98 10°02′12.74″ N 38°14′27.65″ E 1853 CTW and DAF 880.2 12.9--22.0 Gewane 344 39,186 6 967.85 10°29′59.99″ N 40°44′59.99″ E 568 ACB 586.7 19.5--36.7 Wondo Genet 270 196,277 12 226.45 7°05′3.01″ N 38°37′8.02″ E 1742 DAF 928.7 15--29.6 Sodo Zuria 383 145,092 2 25.62 6°51′10.11″ N 37°45′39.49″ E 1854 CTW and DAF 1569.2 14.8--25.2 Doyo Gena 258 95,393 14 130.57 7°21′20.22″ N 37°47′07.15″ E 2300 DAF 1334.5 11--22.8 North Bench 587 126,308 4 392.65 6°37′53.43″ N 35°33′56.83″ E 2367 CTW 1671.8 16--33.3 Mizan Aman 565 64,996 3 24.45 6°59′37.13″ N 35°34′55.92″ E 1441 CTW and MAF 1963.7 14.8--28.8 Shako 617 51,195 1 48,089.63 ha 7°33′42.37″ N 35°39′11.83″ E 1800 CTW and MAF 1906.9 11.4--22.4 Note. Vegetation type: DAF: dry evergreen Afromontane forest and grassland complex; CTW: Combretum-Terminalia woodland and wooded grassland; FLV/MFS: freshwater marshes and swamps, floodplains, and lake shore vegetation; ACB: Acacia-Commiphora woodland and bushland proper; MAF: moist evergreen Afromontane forest. m.a.s.l: meter above sea level; mm: millimeter; °C: degree Celsius; km^2^: kilometer square. ###### Symptoms that are used by traditional healers to diagnose cancer. Cancer types Reported symptoms Number of traditional healers ------------------------------------------------------------------------------ ----------------------------------------------------------------------------------------- ------------------------------- Skin Lumpy growth 32 Spreading pea-sized growth 1 Ulcerative growth and oozing blood 1 Breast Lumpy growth 17 Lumpy growth on one breast and progressive weight loss 1 Ulcerative wounds on breast 5 Ulcerative wounds on breast and swelling on armpit and neck 1 The patient was receiving anticancer treatment for breast cancer in hospital 12 Cervical Foul-smelling bloody vaginal discharge, pain during sexual intercourse, and weight loss 1 Colon Chronic rectal bleeding and weight loss 1 Lung Coughing up blood 1 Throat Coughing and swelling on the neck 1 ###### List of candidate medicinal plants traditionally used for cancer treatment in the study areas. Voucher number Botanical name (family) Vernacular name Districts Growth form Habitat Parts used Preparation Type of cancer treated Application UV RFC CI ---------------- ---------------------------------------------------------- ------------------- ----------------- ------------- ------------ ------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------ ------------------ ------ ------- ------- Acanthaceae Bele-047 Justicia schimperiana (Hochst. ex Nees) T. Anderson Kitkit North Bench Shrub Wild Roots Fresh roots are crashed and boiled, and the cool decoction is drunk before meal Lung Oral 1 0.027 0.067 Bele-057 Justicia schimperiana (Hochst. ex Nees) T. Anderson Gulbana Doyo Gena Shrub Wild Leaves Fresh leaves are pounded, and the juice is applied on the affected area Skin Topical Aloaceae Bele-060 Aloe sp. Gurta waqota Doyo Gena Shrub Wild Leaves Fresh roots are crashed, and the sap is applied on the affected area Skin Topical 6 0.014 0.081 Amaranthaceae Bele-044 Achyranthes aspera L. Koch ashite Mizan Aman Herb Wild Leaves Leaves are roasted on metal plate, pounded into powder, mixed with animal butter, and smeared on the affected part Skin Topical 4 0.014 0.054 Apiaceae Bel-046 Centella asiatica (L.) Urb. Gorongoch Sheko Herb Wild Leaves Young leaves are crashed, and the sap sniffed Throat Intranasal 2 0.014 0.027 Bel-002 Hydrocotyle mannii Hook.f Ye\'ti medhanit North Bench Herb Wild Leaves Young leaves are crashed and applied on the affected area Skin Topical 1 0.014 0.014 Apocynaceae Bel-003 Acokanthera schimperi (A.DC.) Schweinf. Merenz Bahir Dar Zuria Shrub Wild Leaves Young leaves are crashed and applied Skin Topical 0.5 0.027 0.027 Bel-009 Carissa spinarum L. Agam Bahir Dar Zuria Shrub Wild Leaves Leaves are crashed and infused in cold water overnight and drunk before meal and applied on the affected area Skin Oral 1 0.027 0.027 Asclepiadaceae Bel-040 Calotropis procera (Aiton) Dryand. Qumbo Gewane Shrub Wild Roots Fresh roots are crashed, and the sap is applied on the affected area Breast Topical 3 0.014 0.027 Bel-036 Pentarrhinum insipidum E. Mey. Barohula Gewane Shrub Wild Roots Fresh roots are crashed, and the sap is applied on the affected area Breast and skin Topical 1 0.014 0.014 Bel-037 Echidnopsis dammanniana Sprenger Mureli Gewane Herb Wild Stem Stems are cut, and the sap is applied Skin Topical 2 0.014 0.027 Asphodelaceae Bel-020 Kniphofia foliosa Hochst. Shushube Bale Goba Shrub Wild Roots Dry roots are pounded, and the powder is mixed with honey Cervical and breast Oral 1 0.027 0.027 Asteraceae Bel-045 Acmella caulirhiza Delile Kust asht Mizan Aman Shrub Wild Leaves Young leaves are chewed by the healer and spit on Breast Topical 0.67 0.04 0.054 Bel-049 Acmella caulirhiza Delile Bitisa Wondo Genet Shrub Wild Leaves Fresh leaves are crashed and infused in cold water Breast Oral Bel-030 Artemisia absinthium L. Natrara Sodo Zuria Herb Wild Leaves Dried leaves are ground and macerated in coffee or tea Breast Oral 2 0.014 0.027 Bel-029 Artemisia afra Jacq. ex Willd. Agufa Doyo Gena Herb Wild Leaves Dried leaves are ground and macerated in coffee or tea Breast Oral 1 0.014 0.014 Bel-031 Artemisia annua L. Artemisia Sodo Zuria Tree Cultivated Leaves Dried leaves will be ground and decocted in hot water Breast Oral 1 0.014 0.014 Bel-021 Cineraria abyssinica Sch.Bip. ex A.Rich. Unknown Bale Robe Herb Wild Leaves Fresh leaves are pounded, and the sap is applied on the affected area Skin Topical 1.5 0.027 0.054 Bel-058 Guizotia scabra (Vis.) Chiov. Sheshota Doyo Gena Shrub Wild Leaves Fresh leaves are pounded, and the sap is applied on the affected area Skin Topical 1 0.014 0.014 Bel-034 Solanecio gigas (Vatke) C. Jeffrey Arbaba Doyo Gena Shrub Wild Leaves Fresh leaves are pounded and the sap is applied on the affected area Skin Topical 2 0.014 0.027 Bel-025 Vernonia auriculifera Hiern Barawa Doyo Gena Shrub Wild Leaves Fresh leaves are pounded, and the sap is applied on the affected area Skin Topical 1.33 0.041 0.081 Bel-056 Vernonia auriculifera Hiern Reji Wondo Genet Shrub Wild Leaves Fresh leaves are chewed by the healer and spit on Skin Topical Capparidaceae Bel-039 Cleome brachycarpa (Forssk.) Vahl ex DC. Berbere Gewane Herb Wild Leaves Fresh leaves are pounded, and the sap is applied on the affected area Breast and skin Topical 3 0.014 0.014 Commelinaceae Bel-026 Commelina benghalensis L. Laluncha Doyo Gena Herb Cultivated Roots Fresh roots are pounded, and the sap is applied on the affected area Skin Topical 2 0.014 0.027 Crassulaceae Bel-019 Kalanchoe petitiana A. Rich. Anchura Bale Goba Shrub Wild Leaves Fresh leaves are roasted for 2 minutes and applied on the affected area Breast and skin Topical 1.5 0.027 0.041 Euphorbiaceae Bel-012 Croton macrostachyus Hochst. ex Delile Bisana Filiklik Tree Wild Leaves or stem Fresh leaves or succulent stems are crashed, and the sap is applied on the affected area Breast and skin Topical 0.75 0.1 0.16 Bel-035 Croton macrostachyus Hochst. ex Delile Besena Doyo Gena Tree Wild Bark Dry bark is pounded, and the powder is applied on the affected area Skin Topical Bel-048 Croton macrostachyus Hochst. ex Delile Masichoo Wondo Genet Tree Wild Leaves Fresh leaves are crashed, macerated in cold water, and drunk Breast and skin Oral Bel-032 Euphorbia schimperiana Scheele Gendalelata Doyo Gena Shrub Wild Roots Fresh roots are pounded, and the sap is applied on the affected area Skin Topical 1 0.014 0.014 Fabaceae Bel-014 Albizia schimperiana Oliv. Sessa Filiklik Tree Wild Leaves The mixture of fresh leaves of Albizia schimperiana and Carissa spinarum is macerated in cold water for 2 days, and the macerated liquid is drunk Breast, intestinal, and skin Oral 4 0.014 0.014 Bel-004 Calpurnia aurea (Aiton) Benth. Digita Bahir Dar Zuria Shrub Wild Leaves or seeds Dry leaves or seeds are ground, macerated in cold water, and drunk Breast Oral 2 0.014 0.027 Bel-023 Crotalaria agatiflora Schweinf. Unknown Bale Goba Shrub Wild Seeds Dry seeds are ground, mixed with honey, and applied Skin Topical 1 0.014 0.014 Bel-028 Crotalaria incana L. Chelke Doyo Gena Shrub Wild Leaves Fresh leaves are crashed, and the sap is applied on the affected area Skin Topical 1 0.014 0.014 Bel-007 Senna singueana (Delile) Lock Gefa Bahir Dar Zuria Shrub Wild Leaves Fresh leaves are crashed, macerated, and drunk Skin Oral 2 0.014 0.027 Lamiaceae Bel-043 Ajuga leucantha Lukhoba Tiks asht North Bench Herb Wild Leaves Fresh leaves are crushed, and the sap is applied on the affected area Breast Topical 1 0.014 0.014 Bel-024 Leonotis ocymifolia (Burm.f.) Iwarsson Armagusa Bale Goba Herb Wild Leaves Fresh leaves are crashed, macerated overnight, and drunk Breast and skin Oral 3 0.014 0.014 Bel-054 Ocimum gratissimum L. Mekedesisa Wondo Genet Herb Wild Roots Fresh roots are crushed, boiled, and drunk Skin Oral 2 0.014 0.027 Bel-059 Pycnostachys abyssinica Fresen. Tontona Doyo Gena Herb Wild Leaves Fresh leaves are crushed, and the sap is applied on the affected area Skin Topical 2 0.014 0.027 Bel-042 Salvia nilotica Juss. ex Jacq. Barnbanch North Bench Shrub Wild Whole plant Dry plant parts are ground, mixed with honey, and applied Breast Topical 2 0.014 0.027 Bel-022 Thymus schimperi Ronniger Tosigne Bale Goba Herb Wild Leaves Dry leaves are decocted and drunk Breast Oral 2 0.014 0.027 Malvaceae Bel-051 Sida schimperiana Hochst. ex A. Rich. Kotijebessa Wondo Genet Shrub Wild Roots and leaves Fresh leaves and roots are crashed, macerated, and drunk Breast and skin Oral 4 0.014 0.027 Melianthaceae Bel-001 Bersama abyssinica Fresen. Azamir Bahir Dar Zuria Shrub Wild Bark Dry bark is ground, macerated, and drunk before meal Breast Oral 1 0.014 0.014 Moraceae Bel-008 Dorstenia barnimiana Schweinf. Work Bemeda Bahir Dar Zuria Herb Wild Roots Dry roots are ground, mixed with water and honey, and drunk, or dry roots are ground, mixed with honey, and applied on the affected area Breast Oral or topical 0.6 0.068 0.12 Myrtaceae Bel-006 Syzygium guineense (Willd.) DC. Dokima Bahir Dar Zuria Tree Wild Leaves and roots Dry leaves and roots of Syzygium guineense and dry leaves of Osyris quadripartita are ground, mixed, decocted, and drunk Skin Oral 2 0.014 0.027 Oxalidaceae Bel-052 Oxalis corniculata L. Qinta Wondo Genet Herb Wild Leaves and roots Fresh leaves and roots are crashed and applied with a bandage Breast Topical 2 0.014 0.027 Polygonaceae Bel-018 Rumex nervosus Vahl Emboacho Filiklik Shrub Wild Roots Dry roots are ground, macerated, and drunk Skin Oral 3 0.014 0.041 Bel-033 Rumex nepalensis Spreng. Goecho Doyo Gena Herb Wild Roots Dry roots are ground and taken with food Colon Oral 1.5 0.027 0.041 Bel-053 Rumex nepalensis Spreng. Sharibicho Wondo Genet Herb Wild Bark Fresh bark is crashed and squeezed, and the sap is applied Skin Topical Ranunculaceae Bel-010 Clematis simensis Fresen. Yeazo Hareg Bahir Dar Zuria Climber Wild Leaves Fresh roots of Dorstenia barnimiana mixed with fresh leaves of Clematis simensis, pounded, and applied Breast Topical 0.67 0.041 0.054 Rosaceae Bel-011 Prunus africana (Hook.f.) Kalkman Tikur enchet Bahir Dar Zuria Tree Wild Bark Dry bark is ground, decocted, and drunk Breast and skin Oral 3 0.014 0.014 Rutaceae Bel-016 Clausena anisata (Willd.) Hook.f. ex Benth. Limich Filiklik Shrub Wild Leaves Dry leaves are ground, mixed with honey, and eaten Breast Oral 2 0.014 0.027 Santalaceae Bel-013 Osyris quadripartita Salzm. ex Decne. Keret Filiklik Shrub Wild Leaves Dry leaves are ground, decocted, and drunk Breast Oral 2 0.027 0.027 Sapindaceae Bel-005 Dodonaea viscosa subsp. angustifolia (L.f.) J.G.West Kitkita Bahir Dar Zuria Tree Wild Roots Dry roots are ground, mixed with honey, and applied or dry roots are ground, decocted, and drunk Breast, skin and cervical Topical or oral 1 0.014 0.041 Simaroubaceae Bel-017 Brucea antidysenterica J.F.Mill. Abalo Filiklik Tree Wild Leaves Dry leaves are ground, pasted with cold water, and applied Skin Topical 4 0.014 0.054 Solanaceae Bel-027 Discopodium penninervium Hochst. Chechanga Doyo Gena Shrub Wild Leaves Fresh leaves are crashed and applied Skin Topical 1 0.014 0.014 Thymelaeaceae Bel-055 Gnidia involucrata Steud. ex A.Rich. Bito Bahir Dar Zuria Herb Wild Roots Dry roots are ground, mixed with honey, and eaten Breast Oral 0.5 0.027 0.027 Verbenaceae Bel-050 Lantana trifolia L. Hanshebello Wondo Genet Shrub Wild Leaves Fresh leaves are ground, macerated in cold spring water, and drunk Breast and skin Oral 2 0.014 0.014 Bel-015 Lippia adoensis Hochst. Kessie Filiklik Shrub Wild Leaves Dry leaves are ground, macerated in cold water, and drunk Skin Oral 2 0.014 0.027 Vitaceae Bel-038 Cyphostemma serpens (Hochst. ex A.Rich.) Desc. Eiriti Gewane Climber Wild Roots Dry roots are ground, pasted with honey and eaten, and applied Skin Oral and topical 1 0.014 0.014 UV = use value; RFC = relative frequency of citation; CI = cultural importance index. ###### Cross-reference of cancer treatment candidate plant species collected from the study areas with the published literature. ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Botanical name (family) Biological activity/chemical constituents Illnesses/symptoms claimed to be treated traditionally ------------------------------------------------------------------------ ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Justicia schimperiana (Hochst. ex Nees) T. Anderson (Acanthaceae) Saponins, alkaloids, terpenoids and flavonoids \[[@B32]\]\ Wound \[[@B15], [@B21]\]; rabies \[[@B15], [@B18]--[@B20], [@B36]--[@B39]\]; jaundice \[[@B15], [@B16], [@B21], [@B23], [@B38], [@B40]\]; gonorrhea \[[@B17], [@B36], [@B39]\]; liver cirrhosis \[[@B18], [@B26]\]; seizure \[[@B19], [@B41]\]; stomach ache \[[@B15], [@B25], [@B38]\]; helminths \[[@B15], [@B42], [@B43]\]; skin burn/lesion \[[@B23], [@B44]\]; arthritis \[[@B21], [@B23]\]; hepatitis \[[@B45], [@B46]\]; evil eye \[[@B15], [@B46]\]; dysentery \[[@B15], [@B21]\]; malaria \[[@B36], [@B39]\]; common cold, asthma, and headache \[[@B36], [@B39], [@B47]\]; otitis \[[@B48]\]; toothache \[[@B49]\]; and rheumatism \[[@B50]\] In vitro cytotoxicity \[[@B33]\]; in vitro antioxidant activity on DPPH assay \[[@B34]\]; in vivo suppression of parasitaemia on Plasmodium berghei-infected mice in the 4-day suppressive test \[[@B32]\]; and in vivo hepatoprotective activity in mice intoxicated with CCL~4~ \[[@B35]\] Aloe sp. (Aloaceae) Anthrones and chromones \[[@B51]\], pyrones, coumarins, alkaloids, glycoproteins, naphthalenes, and flavonoids \[[@B52]\]\ Wound \[[@B21], [@B55]\]; eye disease \[[@B21], [@B46], [@B48], [@B56]\]; snake bite \[[@B21], [@B48], [@B56]\]; malaria \[[@B20], [@B21], [@B44], [@B48], [@B54]\]; easing labour \[[@B44]\]; tropical ulcer, colon cleaner, and gallstone \[[@B48]\]; amoeba, abdominal pain, impotence, and urine retention \[[@B21]\]; dandruff \[[@B46], [@B56]\], hemorrhoids and hepatitis B \[[@B46]\]; ascariasis \[[@B21]\]; diabetes \[[@B54]\]; asthma \[[@B55]\]; foot strain \[[@B57], [@B58]\]; wart and anthrax \[[@B20]\]; external injury \[[@B59]\]; and liver swelling, splenomegaly, and skin inflammation \[[@B56]\] 7‐O‐methylaloeresin showed in vitro antioxidant activity in DPPH assay \[[@B51]\], and methanol and ethanol extract showed in vivo parasitaemia suppression on Plasmodium berghei-infected mice in the 4-day suppressive test \[[@B53], [@B54]\] Achyranthes aspera L. (Amaranthaceae) Phytosteroids, polyphenols, and saponins \[[@B60]\]\ Bleeding \[[@B21], [@B24], [@B26], [@B62]--[@B64]\]; retained placenta \[[@B21], [@B62]\]; stomach ache and external swelling \[[@B17]\]; rhesus factor incompatibility in pregnancy \[[@B40], [@B55]\]; epistaxis \[[@B19]\]; hepatitis and evil eye \[[@B24]\]; tonsillitis \[[@B21], [@B57]\]; snake bite and paralysis \[[@B21]\]; dysentery \[[@B59]\]; herpes zoster \[[@B26]\]; anthrax \[[@B21], [@B49]\]; nasal infection and ophthalmic infection \[[@B64]\]; excessive menstruation and tape worm infection \[[@B15]\]; and gonorrhea \[[@B65]\] Methanol extracts have showed in vivo wound healing activity \[[@B61]\] Centella asiatica (L.) Urb. (Apiaceae) Terpenoids (triterpenes, asiaticoside, centelloside, madecassoside, brahmoside, brahminoside (saponin glycosides), asiaticentoic acid, centellic acid, centoic acid, madecassic acid, terminolic acid, betulic acid, β-caryophyllene, trans-β-farnesene and germacrene D (sesquiterpenes), α-pinene, and β-pinene \[[@B66], [@B67]\]\ Genital infection and lymphadenitis \[[@B63]\]; topical swelling \[[@B26], [@B70]\]; gastritis, headache, and evil eye \[[@B70]\]; bleeding \[[@B40]\]; wound \[[@B24]\]; abdominal ache \[[@B71]\]; meningitis \[[@B72]\]; and tinea corporis \[[@B47]\] Methanol extract inhibited the proliferation of human gastric adenocarcinoma (MK-1), human uterine carcinoma (HeLa), and murine melanoma (B16F10) cells in vitro \[[@B68]\]; aqueous extracts induced apoptosis in colonic crypts and exerted chemopreventive effect on colon tumorigenesis in male F344 rats \[[@B69]\] Hydrocotyle mannii Hook.f (Apiaceae) No previous reports Eye infection \[[@B63]\] and cataract \[[@B72]\] Acokanthera schimperi (A.DC.) Schweinf. (Apocynaceae) In vitro cytotoxicity \[[@B73]\]; in vitro antiviral activity against coxsackie B3, influenza A, and herpes simplex type1 virus \[[@B74]\]; in vitro antimicrobial activity against Staphylococcus aureus, Pseudomonas aeruginosa, Trichophyton mentagrophytes \[[@B75]\]; and in vivo parasitaemia suppression in Plasmodium berghei-infected mice \[[@B76]\] Wound \[[@B16], [@B44], [@B77], [@B78]\]; hepatitis \[[@B15], [@B16], [@B22], [@B44]\]; gonorrhea \[[@B19], [@B25]\]; evil eye \[[@B62]\]; bone fracture \[[@B24]\]; hemorrhoids \[[@B44]\]; scabies \[[@B21]\]; malaria and tonsillitis \[[@B48], [@B56]\]; psychiatric disease \[[@B55]\]; and skin diseases \[[@B65]\] Carissa spinarum L. (Apocynaceae) In vitro antioxidant activity on DPPH assay and in antiproliferative activity \[[@B79]\] Throat cancer \[[@B23], [@B80]\]; evil eye \[[@B16], [@B21], [@B24], [@B49], [@B62], [@B70], [@B72], [@B81]\]; snake bite \[[@B23], [@B80]\]; gonorrhea \[[@B20], [@B65]\]; stomach ache \[[@B20], [@B70]\]; impotence and headache \[[@B20]\]; tonsillitis \[[@B17], [@B56], [@B70]\]; wound and febrile illness \[[@B16]\]; bleeding after delivery \[[@B44]\]; muscle cramps \[[@B49]\]; toothache \[[@B47]\]; and premature ejaculation \[[@B56]\] Calotropis procera (Aiton) Dryand. (Asclepiadaceae) Latex contains phytochemicals such as alkaloids, sterols, fatty acids, starches, sugars, oils, tannins, resins, and gums, and enzymatic proteins such as proteases, chitenases, lipases, peptidases, esterase, peroxidases, papain, hevein, and lectins \[[@B82]\]\ Wound \[[@B16], [@B21], [@B81]\]; hemorrhoids \[[@B16], [@B19], [@B44]\]; wart \[[@B16], [@B57]\]; snake bite \[[@B23], [@B87]\]; kidney stone, tuberculosis, and scabies \[[@B16]\]; swelling \[[@B58]\]; skin rash \[[@B21], [@B49]\]; tinea capitis \[[@B21]\] In vivo hepatoprotective \[[@B83]\]; hypoglycemic effect \[[@B84]\]; strong anti-implantation (antifertility) \[[@B85]\]; crude latex showed antioxidant and antiapoptotic activities against the toxicity of 4-nonylphenol \[[@B86]\] Pentarrhinum insipidum E. Mey. (Asclepiadaceae) No previous reports Echidnopsis dammanniana Sprenger (Asclepiadaceae) No previous reports Snake bite \[[@B56]\] Kniphofia foliosa Hochst. (Asphodelaceae) 2-Acetyl-1-hydroxy-8-methoxy-3-methylnaphthalene, 10-(chrysophanol-7′-yl)-10-(ξ)-hydroxychrysophanol-9-anthrone, chryslandicin, knipholone, and chrysophanol \[[@B88]\]\ No previous reports 10-(Chrysophanol-7′-yl)-10-(ξ)- hydroxychrysophanol-9-anthrone showed in vitro antiplasmodial activity against chloroquine-sensitive 3D7 strain of Plasmodium falciparum and knipholone selectively inhibited leukotriene metabolism in in vitro a human blood assay \[[@B88]\]; knipholone anthrone showed in vitro cytotoxicity \[[@B89]\] and antioxidant activity on DPPH assay \[[@B90]\] Acmella caulirhiza Delile (Asteraceae) Unsaturated alkylamides like spilanthol and N-isobutylnona-2E,4E-dien-8ynamide \[[@B91]\]\ Swelling \[[@B15]\]; tonsillitis \[[@B20], [@B63]\]; and toothache \[[@B40], [@B87]\] In vitro antiplasmodial activity \[[@B92]\] Artemisia absinthium L. (Asteraceae) Camphor, davanone, ethyl (E)-cinnamate, (E)-nerolidol, and chamazulene \[[@B93]\]\ Hypertension, stomach ache, severe abdominal cramp \[[@B18]\] and sour throat \[[@B40]\] Essential oils showed in vitro antiparasitic effects against promastigote and axenic amastigote forms of Leishmania donovani and Leishmania aethiopica and in vitro cytotoxicity on THP-1 (human leukaemia) cell lines \[[@B93]\]; and in vitro cytotoxicity on human leukaemia cell lines \[[@B94]\] Artemisia afra Jacq. ex Willd. (Asteraceae) Epoxylinalol and dihydrocostunolide \[[@B94]\]; camphor, davanone, bornyl acetate, 4-terpineol, and chamazulene \[[@B95]\]\ Stomach ache \[[@B18], [@B42]\]; evil eye \[[@B16], [@B17], [@B62]\]; headache \[[@B42], [@B77]\]; eye disease, tinea capitis infection, hematuria, and stabbing pain \[[@B77]\]; antifertility agent \[[@B33]\]; malaria \[[@B42], [@B62]\]; ascariasis \[[@B18]\]; epilepsy and febrile illness \[[@B46], [@B65]\] In vitro cytotoxicity on human leukaemia cell lines \[[@B73]\]; and in vitro antioxidant effect on DPPH assay \[[@B95]\] Artemisia annua L. (Asteraceae) In vitro inhibition of immune mediators of angiogenesis \[[@B96]\]; the sesquiterpene (Z)-7-acetoxy-methyl-11-methyl-3-methylene-dodeca-1,6,10-triene showed moderate cytotoxic activities against the human tumor cell lines of HO8910 (ovary), 95-D (lung), QGY (liver), and HeLa (cervix) by MTT assay and induced apoptosis on 95-D tumor cells \[[@B97]\]; artemisinin and quercetagetin 6,7,3′,4′-tetramethyl ether showed significant cytotoxicity against P-388, A-549, HT-29, MCF-7, and KB tumor cells \[[@B98]\] No previous reports Cineraria abyssinica Sch.Bip. ex A.Rich. (Asteraceae) In vitro radical scavenging activity on DPPH assay \[[@B99]\]; flavonoidal glycoside (rutin) showed in vitro antibacterial activity \[[@B100]\] No previous reports Guizotia scabra (Vis.) Chiov. (Asteraceae) In vitro cytotoxicity on human leukaemia cell lines \[[@B73]\], and in vitro antiviral activity \[[@B101]\] Wound \[[@B20]\]; epilepsy \[[@B40]\]; and ectoparasite infestation \[[@B47]\] Solanecio gigas (Vatke) C. Jeffrey (Asteraceae) In vitro antiviral activity against human immunodeficiency virus type 1 and type 2 cytotoxicity on human T-lymphocytic MT-4 cell lines \[[@B102]\] Skin diseases \[[@B62]\]; retained placenta \[[@B40]\]; hepatitis \[[@B64]\]; evil eye \[[@B15]\] Vernonia auriculifera Hiern (Asteraceae) Tannins, flavonoids, terpenoids, and saponins \[[@B103]\] Toothache \[[@B72]\]; snake bite \[[@B42]\]; skin cut \[[@B47]\] Cleome brachycarpa (Forssk.) Vahl ex DC. (Capparidaceae) No previous reports Commelina benghalensis L. (Commelinaceae) Phlobatannins, carbohydrates, tannins, glycosides, volatile oils, resins, balsams, flavonoids, and saponins \[[@B104]\]\ Helminths \[[@B65]\]; skin infection \[[@B72]\] Ethanol extract showed in vivo sedative and anxiolytic activity \[[@B105]\] Kalanchoe petitiana A. Rich. (Crassulaceae) Polyphenols, alkaloids, flavonoids, tannins, saponins, and steroids \[[@B106]\]\ Breast and skin cancer \[[@B107]\]; swelling \[[@B40], [@B77]\]; tapeworm infection, trachoma, and syphilis \[[@B77]\]; lymphadenopathy and evil eye \[[@B22]\]; sore muscles \[[@B108]\]; itching skin \[[@B63]\]; and bone fracture \[[@B23]\] In vitro antimicrobial activity against Escherichia coli, Pseudomonas aeruginosa, and Staphylococcus aureus \[[@B75]\]; and in vivo wound healing activity \[[@B106]\] Croton macrostachyus Hochst. ex Delile (Euphorbiaceae) Ethanol extract showed in vitro antioxidant activity on DPPH assay \[[@B79]\] Tumor, rabies, and wart \[[@B24]\]; skin cancer and wound \[[@B17]\]; gonorrhea \[[@B20], [@B23], [@B62]\]; headache \[[@B18], [@B109]\]; snake bite \[[@B18], [@B72]\]; malaria \[[@B16], [@B18]--[@B20], [@B110]\]; helminths \[[@B18], [@B111]\]; tinea nigra \[[@B40]\]; ringworm \[[@B17], [@B62]\]; tinea versicolor \[[@B16], [@B25]\]; heart failure \[[@B62]\]; bleeding \[[@B18], [@B24]\]; hepatitis \[[@B16], [@B18], [@B24]\]; stomach ache \[[@B16], [@B18], [@B23]\]; diarrhea \[[@B16], [@B18]\]; lymph adenitis and rheumatism \[[@B18]\]; bloat, scabies, and urine retention \[[@B16]\]; retained placenta and leprosy \[[@B19]\] Euphorbia schimperiana Scheele (Euphorbiaceae) In vitro cytotoxic effect against breast cancer (MCF7), hepatocellular carcinoma (HEPG2), and cervix cancer (HELA) cells \[[@B112]\] Syphilis \[[@B108]\] Albizia schimperiana Oliv. (Fabaceae) In vitro cytotoxicity on human leukaemia cells \[[@B73]\] Evil eye \[[@B20]\]; kidney infection and liver cirrhosis \[[@B18]\] Calpurnia aurea (Aiton) Benth. (Fabaceae) 3β,4α,13α-Trihydroxylupanine, calpaurine, lupinine, and epilupinine calpurmenine and calpurmenine pyrrolecarboxylic acid ester, 13-hydroxylupanine, its tiglate and pyrrolecarboxylic acid esters (calpumine), virgiline and virgiline pyrrolecarboxylic acid ester \[[@B113]\]; 4β-hydroxy-13α-O-(2′-pyrrolylcarbonyl)-lupanine (digittine) and 4β,13α-dihydroxylupanine \[[@B114]\]; alkaloids, tannins, flavonoids, and saponins \[[@B35]\]\ Tumor \[[@B22], [@B26], [@B80]\]; stomach ache \[[@B21], [@B62], [@B70], [@B81]\]; wound and skin infection \[[@B62]\]; Gonorrhoea and syphilis \[[@B16]\], amoebiasis \[[@B16], [@B80]\]; ascariasis and gastric ulcer \[[@B23]\]; diarrhea \[[@B21], [@B38], [@B70]\]; scabies and pubic hair louse \[[@B40]\]; diabetes mellitus and hypertension \[[@B19]\]; herpes zoster, hemorrhoids and tinea capitis \[[@B21]\]; and swelling and tuberculosis \[[@B58]\] Methanol extract showed in vitro antimicrobial activity against Staphylococcus aureus, Escherichia coli, and Pseudomonas aeruginosa \[[@B75]\] and type 1 and type 2 human immunodeficiency virus and showed cytotoxicity on human T-lymphocytic MT-4 cell lines \[[@B102]\]; methanol and dichloromethane crude extracts showed in vitro cytotoxicity on human leukaemia cells \[[@B73]\]; and ethanol extracts showed in vitro antioxidant activity on DPPH assay \[[@B79]\] Crotalaria agatiflora Schweinf. (Fabaceae) Methanol and dichloromethane crude extracts showed in vitro cytotoxicity on human leukaemia cells \[[@B73]\] No previous reports Crotalaria incana L. (Fabaceae) Dihydrosenecionine isomer, nemorensine isomer, integerrimine and anacrotine \[[@B115]\]\ Methanol and dichloromethane crude extracts showed in vitro cytotoxicity on human leukaemia cell lines \[[@B73]\] Senna singueana (Delile) Lock (Fabaceae) Methanol extracts showed in vitro antioxidant activity on DPPH assay \[[@B116]\] Stomach ache \[[@B58], [@B62], [@B70]\]; wound and swellings \[[@B62]\]; teeth infection and sprain \[[@B58]\] Ajuga leucantha Lukhoba (Lamiaceae) No previous reports Diarrhea \[[@B70]\] Leonotis ocymifolia (Burm.f.) Iwarsson (Lamiaceae) Methanol and dichloromethane crude extracts showed in vitro cytotoxicity on human leukaemia cells \[[@B73]\] Ascariasis \[[@B62]\], febrile illness \[[@B16], [@B62]\]; eye disease \[[@B16]\]; headache and neck ulcer \[[@B55]\]; and snake bite \[[@B15]\] Ocimum gratissimum L. (Lamiaceae) Essential oil contains constitutes γ-terpinene, β-phellandrene, limonene, and thymol and showed in vivo antiplasmodial activity against Plasmodium berghei infection \[[@B117]\] Allergy reaction \[[@B18], [@B20]\]; rheumatism, headache and eye disease \[[@B18]\]; febrile illness and general malaise \[[@B40]\]; sun stroke \[[@B24]\]; malaria \[[@B44]\] Pycnostachys abyssinica Fresen. (Lamiaceae) No previous reports Eye disease \[[@B18], [@B47]\]; ascariasis and wound \[[@B18]\]; diarrhea, stomach ache, amoebiasis, stomach bloating, and food poisoning \[[@B70]\]; headache \[[@B63]\] Salvia nilotica Juss. ex Jacq. (Lamiaceae) Essential oil contains germacrene D, guaiol, and trans-caryophyllene as major constituents and showed activity against both Gram-positive and Gram-negative pathogenic bacteria; the oil also showed in vitro antioxidant activity on DPPH assay \[[@B118]\] Tonsillitis and constipation \[[@B62]\]; herpes simplex \[[@B18], [@B38]\]; wound \[[@B40]\]; lymphadenitis \[[@B63]\]; and hemorrhoids and diarrhea \[[@B65]\] Thymus schimperi Ronniger (Lamiaceae) Phenol and flavonoid compounds, and aqueous methanol extract showed in vitro radical scavenging ability, iron reducing power, and total antioxidant capacity \[[@B119]\] Diabetes \[[@B62]\]; hypertension \[[@B18], [@B40]\]; tonsillitis \[[@B18]\]; toothache \[[@B18], [@B21]\]; abdominal pain \[[@B21]\]; and cough \[[@B38], [@B55]\] Sida schimperiana Hochst. ex A. Rich. (Malvaceae) No previous reports "Shotelaye" (hydrops fetalis) \[[@B21], [@B22]\]; cough and fever \[[@B62]\]; diarrhea \[[@B18]\]; wound \[[@B25], [@B62]\]; bleeding and evil eye \[[@B24]\]; glandular disease and rabies \[[@B40]\]; amoebic dysentery, and liver disease \[[@B65]\]; paralysis \[[@B21]\]; epilepsy \[[@B43]\] Bersama abyssinica Fresen. (Melianthaceae) Flavonol glycosides isoquercetrin, hyperoside, quercetin-3-O-arabinopyranoside, kaempferol-3-O-arabinopyranoside, xanthone glycoside, mangiferin \[[@B115]\]\ Tumor, dysentery and roundworms \[[@B107], [@B109]\]; ascariasis \[[@B15], [@B38], [@B81], [@B109]\]; wound \[[@B20]\]; stomach ache \[[@B17]\]; snake bite and liver diseases \[[@B70]\]; tonsillitis \[[@B72]\]; bronchitis and febrile illness \[[@B42], [@B43]\] Ethanol water extracts showed in vitro antioxidant activity on DPPH assay and antiproliferative activity on human liver carcinoma cell line and normal human fetal lung cells \[[@B79]\]; methanol extract showed in vitro antioxidant activity on DPPH assay \[[@B115]\], and antiviral activity against type 1 human immunodeficiency virus \[[@B102]\] Dorstenia barnimiana Schweinf. (Moraceae) Phytochemical screening showed the presence of coumarins \[[@B34]\] Cancer \[[@B26]\]; hepatitis, syphilis and rabies \[[@B25], [@B26]\]; skin cancer, dysentery, wart and fever \[[@B25]\]; pulmonary tuberculosis, leprosy, and stomach illness \[[@B22]\] Syzygium guineense (Willd.) DC. (Myrtaceae) Methanol and dichloromethane crude extracts showed in vitro cytotoxicity on human leukaemia cells \[[@B73]\] and antimicrobial activity \[[@B120]\] Stomach ache \[[@B17]--[@B19], [@B23]\]; diarrhea \[[@B15], [@B18], [@B19], [@B24]\], kidney infection, liver cirrhosis, and tonsillitis \[[@B18]\]; syphilis \[[@B23], [@B80]\]; malaria, hemorrhoid, internal worms, snake bite, and gonorrhea \[[@B65]\] Oxalis corniculata L. (Oxalidaceae) In vivo antitumor activity against Ehrlich ascites carcinoma on mice \[[@B121]\] Wound \[[@B17]\]; arthritis \[[@B63]\]; tape worm infection \[[@B21]\] Rumex nervosus Vahl (Polygonaceae) Alkaloids, flavonoids, terpenoids, tannins, glycosides, and volatile oils \[[@B122]\] Breast cancer, gastritis, and snake bite \[[@B16]\]; wart \[[@B15], [@B22]\]; hepatitis \[[@B49], [@B55]\]; skin rash \[[@B16], [@B21]\]; bleeding \[[@B15], [@B40], [@B81], [@B109]\]; wound \[[@B40], [@B49], [@B55], [@B62], [@B109], [@B110]\]; scabies and acne vulgaris \[[@B62]\]; ascariasis and herpes simplex \[[@B21]\]; stomach ache and dysentery \[[@B22]\]; diarrhea \[[@B49]\]; eye problems and round worm \[[@B55]\] Rumex nepalensis Spreng. (Polygonaceae) Anthraquinones, naphthalenes, tannins, stilbenoids \[[@B123]\]\ Wound, ascariasis, abdominal bleeding, gastric ulcer, and hemorrhage \[[@B23], [@B80]\]; gastritis \[[@B18]\]; stomach problems \[[@B108]\]; leishmaniasis \[[@B25]\]; abdominal cramp and ear infection \[[@B63]\]; tonsillitis \[[@B18], [@B25]\] Ethanol water extracts showed in vitro antiproliferative activity on human liver carcinoma cell line and on normal human fetal lung cells and antioxidant activity on DPPH assay \[[@B79]\], and methanol and dichloromethane crude extracts showed in vitro cytotoxicity on human leukaemia cells \[[@B73]\] Clematis simensis Fresen. (Ranunculaceae) Triterpenoids, saponins, alkaloids, polyphenols, and unsaturated sterols \[[@B120]\]\ Cancer and hemorrhoid \[[@B15]\]; wart and evil eye \[[@B24], [@B40]\]; wound \[[@B15], [@B24], [@B40], [@B63], [@B81]\]; tonsillitis \[[@B62]\]; eye infection \[[@B63]\]; leg swelling, malaria, and mental illness \[[@B49]\]; stomach ache \[[@B47]\] In vivo anti-inflammatory and antinociceptive activities \[[@B124]\] Prunus africana (Hook.f.) Kalkman (Rosaceae) No previous reports Benign prostatic hyperplasia and prostate gland hypertrophy \[[@B20]\]; cancer, respiratory disorders, bad breathe, diarrhea, gonorrhea, tuberculosis, and ear problems \[[@B22]\]; swelling \[[@B40]\]; wounds \[[@B19], [@B22]\]; tonsillitis \[[@B23], [@B80]\] Clausena anisata (Willd.) Hook.f. ex Benth. (Rutaceae) Carbazole alkaloids, peptide derivatives, sitosterol, and stigmasterol \[[@B125]\]\ Skin irritation \[[@B20]\]; toothache \[[@B40]\]; ascariasis \[[@B19]\]; evil eye \[[@B24], [@B25], [@B63]\] Methanol and dichloromethane crude extracts showed in vitro cytotoxicity on human leukaemia cells \[[@B73]\] Osyris quadripartita Salzm. ex Decne. (Santalaceae) Alkaloids, phenols, terpenoids, tannins, saponins, and flavonoids \[[@B126]\]\ Cancer \[[@B62]\]; anaphylactic shock, evil eye, and epilepsy \[[@B18]\]; eczema \[[@B40]\]; toothache \[[@B46]\] Methanol extracts showed in vitro antimicrobial activity against Escherichia coli, Pseudomonas aeruginosa, Staphylococcus aureus, Candida albicans, and Trichophyton mentagrophytes \[[@B11]\]; in vitro inhibition of NO production and cytotoxicity against MCF-7 and NCI-H460 cell lines \[[@B127]\] Dodonaea viscosa subsp. angustifolia (L.f.) J.G.West (Sapindaceae) Alkaloids, terpenoids, saponins, tannins, sugars, phenolics, and flavonoids \[[@B128]\]\ Malaria \[[@B57]\] Methanol extracts showed in vivo nonsensitizer effect in mice using the mouse ear swelling test method \[[@B129]\], in vitro antiviral effect against type 1 human immunodeficiency virus \[[@B102]\], and in vitro free radical scavenging activity on DPPH assay \[[@B128]\] Brucea antidysenterica J.F.Mill. (Simaroubaceae) Flavonoids, amino acids, and vitamin C \[[@B130]\]\ Cancer/tumor \[[@B107]\]; wart \[[@B24]\]; rabies \[[@B18], [@B62]\]; leprosy \[[@B62]\] In vitro antiplasmodial activity against Plasmodium berghei infection \[[@B131]\] Discopodium penninervium Hochst. (Solanaceae) 5α,17β-Dihydroxy-6α,7α-epoxy-1-oxowitha-2,24-dienolide, withanone, and withanolide A \[[@B132]\], 5,6-epoxy-16-oxygenated withanolides, jaborosalactone-L, and 17-epiacnistin-A \[[@B133], [@B134]\]; 6α,7α-epoxy-1-oxo-5α,12α,17α-trihydroxywitha-2,24-dienolide and a coloratane sesquiterpene, 7α,11α-dihydroxy-4(13),8-coloratadien-12,11-olide, withanone, 5α,17β-dihydroxy-6α,7α-epoxy-1-oxowitha-2,24-dienolide, 7α,11α-dihydroxy-8-drimen-12,11-olide, withasomnine, and (E,Z)-9-hydroxyoctadeca-10,12-dienoic acid \[[@B135]\]\ Skin detoxification \[[@B62]\]; and liver disease \[[@B70]\] Jaborosalactone-L showed cytotoxicity only to the murine macrophage cell line, RAW 264.7, but the 16α-oxygenated withanolides exhibited cytotoxicity to both human (COR-L23 and ECV 304) and murine (L929 and RAW 264.7) carcinoma cell lines with IC~50~ values ranging from 1.2 to 150 μM \[[@B136]\]. 6α,7α-Epoxy-1-oxo-5α,12α,17α-trihydroxy-witha-2,24-dienolide inhibited COX-2 and LTB4 formation; 7α,11α-dihydroxy-4(13),8-coloratadien-12,11-olide and withasomnine inhibited LTB~4~ biosynthesis but showed minor inhibition of COX-1 and COX-2 \[[@B135]\] Gnidia involucrata Steud. ex A.Rich. (Thymelaeaceae) Flavonoids and glycosides \[[@B137]\] Ascariasis, evil eye, anthrax, intestinal helminths, and gland swelling \[[@B18]\] Lantana trifolia L. (Verbenaceae) Flavone glycosides (scutellarein-7-O-β-D-apiofuranoside and apigenin-7-O-β-D-apiofuranosyl-(1⟶2)-β-D-apiofuranoside) and the flavone celtidifoline (5,6,40,50-tetrahydroxy-7,30-dimethoxyflavone) \[[@B138], [@B139]\] Headache \[[@B70]\]; malaria \[[@B71]\] Lippia adoensis Hochst. (Verbenaceae) Limonene, perillaldehyde, piperitenone, and 2-methyl-6-methylene-2,7-octadien-4-one \[[@B140]\], sesquiterpene hydrocarbon (germacrene D) \[[@B141]\]\ Eczema, fungal infections, common cold, and cough \[[@B62]\]; intestine swelling \[[@B18]\]; gastrointestinal disorder \[[@B40]\]; abdominal irritation and acute stomach illness \[[@B46]\] Methanol extract showed in vitro cytotoxicity on human leukaemia cell lines \[[@B73]\], and antimicrobial activity against Staphylococcus aureus, Escherichia coli, and Pseudomonas aeruginosa \[[@B75]\]; water extracts showed in vivo protection/relieve against acetic acid induced writhing in mice model \[[@B142]\] Cyphostemma serpens (Hochst. ex A.Rich.) Desc. (Vitaceae) No previous reports ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ###### Informants consensus factor for different ailment categories. No. Category No. of species \% of all species No. of use reports \% of all use reports ICF ----- --------------- ---------------- ------------------- -------------------- ----------------------- ------ 1 Skin 25 30.5 32 26.5 0.23 2 Breast 20 24.4 29 23.9 0.32 3 Cervical 1 1.22 1 0.83 0 4 Colon 1 1.22 1 0.83 0 5 Lung 1 1.22 1 0.83 0 6 Throat 1 1.22 1 0.83 0 7 Other disease 33 40.2 56 46.3 0.42 Total 82 ^∗^ 121 ^∗^Each plant species may be listed in several categories. ###### Summary of stats for relative frequency of citation (RFC) and cultural importance index (CI). -------------------------------------------------------------------- ------------ -------------------- --------- --------- Mean Standard deviation Minimum Maximum UV 1.8 1.1 0.5 6 RFC 0.02 0.015 0.014 0.1 CI 0.034 0.027 0.014 0.16 Association between RFC and CI by using Pearson correlation method UV RFC CI UV 1 RFC −0.36^∗^ 1 CI 0.003 0.858^∗∗^ 1 -------------------------------------------------------------------- ------------ -------------------- --------- --------- ^∗^Correlation is significant at 0.05 level. ^∗∗^Correlation is significant at 0.001 level. [^1]: Academic Editor: Sebastian Granica	{ "pile_set_name": "PubMed Central" }
Introduction {#s1} ============ Insomnia is both a symptom of depression and a separate disorder that may precede depression [@pone.0094674-Baglioni1]. Women sleep poorly during pregnancy and sleep even less in the postpartum period [@pone.0094674-Dorheim1], [@pone.0094674-Ko1]. At the same time, the perinatal period is a time of increased risk for depressive disorders [@pone.0094674-MunkOlsen1]. Insomnia during pregnancy may, therefore, influence the risk of postpartum depression. Insomnia is defined as repeated difficulty with sleep initiation, duration, consolidation, or quality that occurs despite adequate time and opportunity for sleep, and results in some form of daytime impairment [@pone.0094674-AASM1]. Among postpartum women, there may not be adequate time and opportunity for sleep; the newborn baby requires attention and will disturb maternal sleep in the first months after delivery [@pone.0094674-Insana1]. However, postnatal women may be exposed to similar practical challenges with sleep and infant care postpartum, but some may develop larger sleep problems than others or be more vulnerable to these changes. A review of 21 longitudinal studies found that individuals with insomnia had a twofold risk of developing depression and that residual insomnia after recovery from depression was associated with an increased risk of relapse [@pone.0094674-Baglioni2]. A large, population-based study from Norway found that not only did depression precede insomnia but also that insomnia preceded depression by many years [@pone.0094674-Sivertsen1]. Similar results have been found in Finland [@pone.0094674-Salo1]. Furthermore, treating insomnia in depressed patients leads to a better outcome of depression than treating depression on its own [@pone.0094674-Manber1]. Several studies suggest that insomnia may be co-morbid with depression, with a shared genetic component and a common final pathway [@pone.0094674-Zhang1]. Cross-sectional studies have found that insomnia and depressive symptoms are conditions associated with each other both before [@pone.0094674-Dorheim1], [@pone.0094674-Kizilirmak1] and after delivery [@pone.0094674-Dorheim2], [@pone.0094674-Marques1]. Depression during pregnancy is a risk factor for preterm birth [@pone.0094674-Dayan1] and may affect the physiology of the newborn\'s brain [@pone.0094674-Diego1]. Furthermore, maternal depression both during pregnancy and after delivery may affect the infant\'s cognitive and emotional development [@pone.0094674-Evans1], [@pone.0094674-Gerardin1], and these effects may persist through the childhood and teenage years [@pone.0094674-Murray1], [@pone.0094674-Murray2]. Prenatal depression is associated with sleep disturbances in the newborn (less deep sleep and more disorganized sleep) [@pone.0094674-Field1], and both prenatal and postnatal depression is associated with infant sleep problems one year after delivery [@pone.0094674-Gerardin1], [@pone.0094674-Pinheiro1]. Poor maternal sleep has been linked to perceived sadness of the infant (perceived by the mother) [@pone.0094674-Tikotzky1]. Poor sleep may also influence the relationship between the parents of the baby [@pone.0094674-Insana1], [@pone.0094674-Donnellan1], and a poor relationship with the partner is a risk factor for postpartum depression [@pone.0094674-OHara1]. Longitudinal studies of change in sleep and depression across childbirth are few. One study of 44 low-risk women suggested that change in subjective sleep, more than objective sleep, predicts postpartum mood in the first week postpartum [@pone.0094674-Bei1]. Marques et al., in a study of 382 women, found that insomnia in pregnancy was a predictor for postpartum depressive symptoms along with negative affect, but insomnia did not remain a risk factor when controlling for lifetime depression [@pone.0094674-Marques1]. Conversely, Okun et al., in a study of 51 perinatal women with a history of postpartum depression, found that women with fewer sleep complaints in late pregnancy had a more rapid recurrence (less than 4 weeks) of postpartum depression, whereas women with sleep problems during pregnancy developed postpartum depression later on [@pone.0094674-Okun1]. Change in sleep in the postpartum period may predict the development of depression more than changes in hormones do [@pone.0094674-Okun2]. More information is needed regarding the longitudinal relationship between insomnia and depressive symptoms in the perinatal period. Aims of the study {#s1a} ----------------- The aims of the study were to describe changes in sleep patterns, insomnia, and depressive symptoms before and after delivery, and to evaluate whether insomnia during pregnancy may predict postpartum depression. The study was done in a large, population-based cohort of pregnant women followed through childbirth until eight weeks after delivery, controlling for a range of known risk factors for perinatal depression. Material and Methods {#s2} ==================== Study population and design {#s2a} --------------------------- The Akershus Birth Cohort is a longitudinal questionnaire study targeted at all women giving birth at Akershus University Hospital, Norway. The hospital is located near the capital city of Oslo and serves a population of 350,000 from urban and rural areas. All women scheduled to give birth at the hospital were approached in gestational week 17, when they underwent routine fetal ultrasound. Women were included if they gave consent to participate and were able to complete a questionnaire in Norwegian. Recruitment took place from November 2008 until April 2010. Consenting women were handed a questionnaire at gestational week 17 and, thereafter, received a questionnaire by mail at week 32 of pregnancy and eight weeks after delivery. Information was also retrieved from the birth records at the maternity ward. [Figure 1](#pone-0094674-g001){ref-type="fig"} displays a flowchart of the recruitment and retention of study participants. The current paper examined insomnia and depressive symptoms as measured in week 32 of pregnancy and week eight postpartum. In total, 2,386 women returned the third questionnaire, making the participation rate 51.2% (n = 2386) of the 4,662 women who originally consented to participate, and 81.1% of the women who returned the second questionnaire. The final sample consisted of 2,088 women, as 298 women had missing data for the main outcome variables in one or both questionnaires (insomnia or depressive symptoms). ![Study flow chart.\ Legend: EPDS -- Edinburgh Postnatal Depression Scale. BIS -- Bergen Insomnia Scale.](pone.0094674.g001){#pone-0094674-g001} Variables {#s2b} --------- ### Measure of sleep {#s2b1} The Bergen Insomnia Scale (BIS) was used to assess for insomnia (Appendix A) [@pone.0094674-Pallesen1]. This questionnaire consists of six items, of which the first four pertain to night factors such as sleep onset delayed more than 30 minutes, waking up for more than 30 minutes during the night, waking up more than 30 minutes earlier than planned, and not feeling adequately rested after sleep. These items correspond to the DSM-IV-TR criterion A for insomnia [@pone.0094674-APA1]. The last two items assess level of daytime impairment (affecting work/studies or private life) due to sleepiness and dissatisfaction with sleep, corresponding to the DSM-IV-TR criterion B [@pone.0094674-APA1]. Each item is rated on average occurrence from 0 to 7 days per week, giving a possible total sum score from 0 to 42. Women were scored as having insomnia if they had experienced at least one A criterion and one B criterion for three days or more per week during the last month. The BIS has been validated against other self-report scales, as well as polysomnographic data [@pone.0094674-Pallesen1]. Cronbach alpha for the BIS was 0.85 in week 32 of pregnancy and 0.74 in week eight postpartum. Three questions from the Pittsburgh Sleep Quality Index (PSQI) [@pone.0094674-Buysse1], [@pone.0094674-Pallesen2] were included concerning the average time at which participants went to bed, average wake-up time in the morning, and average total sleep duration per night during the previous month. From these data, habitual sleep efficiency (time asleep divided by total time spent in bed) was calculated. When calculations resulted in sleep efficiencies exceeding 100% (due to women\'s self-reports), sleep efficiency was recorded as 100%. The women also reported the estimated time their baby was awake during the night. The use of sleep medication was reported for the final 10 weeks of pregnancy. ### Measure of depressive symptoms {#s2b2} The Edinburgh Postnatal Depression Scale (EPDS) [@pone.0094674-Cox1], [@pone.0094674-EberhardGran1] was used to measure depressive symptoms. The EPDS is a 10-item, self-rating questionnaire developed to screen for depression in the postpartum period; it addresses symptoms present during the last seven days. The scale also has good psychometric properties for use during pregnancy [@pone.0094674-Bunevicius1]. Each question has four possible responses, related to scores from 0 to 3, for a maximum score of 30. Cronbach alpha for the EPDS was 0.85 in late pregnancy and 0.86 in week eight postpartum. A cutoff of 10 or above was found to have good psychometric properties for a diagnosis of depression among Norwegian postpartum women, and this cutoff has also been used in previous studies of pregnant women [@pone.0094674-EberhardGran2]. Information on previous depression was measured by the Lifetime Major Depression Scale [@pone.0094674-Kendler1]. This scale consists of five questions (concerning sadness, appetite changes, lack of energy, self-blame, and concentration) constructed to measure lifetime history of major depression based on the DSM-IV criteria. Prior depression was defined as having had at least three symptoms at the same time with duration of at least two weeks. ### Other study factors {#s2b3} The demographic information collected included maternal age, marital status (married or cohabitating versus single/widowed/divorced), number of previous children, and level of education (elementary school, completed high school, or higher education). In pregnancy week 32, we asked about the experience of 10 specific, stressful life events during the last year, as used by previous studies in Norway [@pone.0094674-Dorheim2], [@pone.0094674-EberhardGran3]. Fear of childbirth was measured by the Wijma Delivery Expectancy/Experience Questionnaire version A (W-DEQ) [@pone.0094674-Wijma1], [@pone.0094674-GarthusNiegel1]. This is a 33-item, self-assessment scale where each item ranges from 0 to 5, with a total sum ranging from 0 to 165. Serious fear of childbirth was defined as a W-DEQ sum score ≥85. This cutoff is commonly used to distinguish women with a fear of childbirth from women without a fear of childbirth [@pone.0094674-Ryding1]. Symptoms of anxiety were measured by the first 10 items in the Hopkins Symptom Checklist (SCL-25) that comprise the anxiety score (HSCL-A) [@pone.0094674-Derogatis1]. Each item ranges from "not at all" (score 1) to "extremely" (score 4). The Norwegian version of SCL-25 has been validated against the ICD-10 criteria for anxiety and depression [@pone.0094674-Sandanger1]. Statistical methods {#s2c} ------------------- The data distributions were checked for normality using p-p charts. For numerical data, means and standard deviations were calculated. Differences in means between groups were tested by student\'s independent t-test, and differences between mean scores before and after delivery for the same women were tested by paired samples t-tests. For categorical data, statistical differences in proportions were examined with chi square. Differences in prevalence of insomnia before and after delivery were tested by McNemar\'s test. Pearson\'s correlation was used to test the association between BIS and EPDS. Multivariate linear regressions were used for testing associations with postpartum BIS and EPDS values simultaneously. Factors individually associated with either of these two variables were entered into the model. Parameters with non-normal distribution (EPDS and HSCL) were log transformed before the analyses, and validity of the model was tested by Box\'s test of equality of covariance matrices and Levene\'s test for equality of variances. Wilks\' Lambda was calculated to test the significance of the multivariate tests, and non-significant variables, except age, were excluded. All analyses were performed in SPSS 18.0 for Windows. The significance level was set to a probability (p) of less than 0.05. Ethical considerations {#s2d} ---------------------- All women asked to participate were given written information explaining the purpose of the study and informed that participation was voluntary. Written informed consent was obtained from all participants, by asking them to sign with full name, address and their date of birth at a separate form. These forms were sent to the project leader, and kept separate from the other data. At the field for date of birth, the following information was given "If you are less than 18 years old, you cannot participate in the study". In spite of this, three girls aged 17 returned questionnaires, and one had returned all three questionnaires and was included in the present study. This was not discovered until a later phase of the study. The age of majority for health in Norway is 16 years, therefore guardians were not asked for consent, as this would break confidentiality. Since the girl had taken time to complete the questionnaires, and the study did not involve any active interventions, we decided to keep her data in the study. The Regional Committee for Ethics in Medical Research in Norway approved the inclusion of the women aged 17 in the study, provided she now, as adult, gave her consent. Authors did not have any access to identifying information. Anonymization was done before transferring the data from the birth records to the researchers, by personnel at the birth department having this as their job. The forms were sent out to the women\'s home addresses containing only an anonymous form number, by people not having access to the completed forms returned, and who did not participate in the analyses. Therefore, the data was anonymized upon collection. The study was approved by the Regional Committee for Ethics in Medical Research in Norway, approval number S-08013a. Results {#s3} ======= The mean age of the postpartum women was 31.5 years (SD 4.7; range 17.4--45.7), and the number of children each woman had (including the new-born) ranged from 1 to 6, median one child, interquartile range 1--2; 11.8% had three or more children. Primipara comprised 50.8% (n = 1,058) of the sample. Two-thirds of the women (n = 1,345; 67.5%) had an education above 12 years of school. Nearly all the women were in stable relationships; only 48 (2.3%) were single. A majority of the women, 70.6% (n = 1,464), were exclusively breastfeeding; 17.8% (n = 370) combined breastfeeding with supplemental formulas, and 241 women (11.6%) did not breastfeed. The response rate among women scoring high for depressive symptoms (EPDS ≥10) in week 32 was 68.0%, significantly lower than among low scorers (74.1%; p = 0.008). [Table 1](#pone-0094674-t001){ref-type="table"} displays changes in sleep from late pregnancy to postpartum week eight. Mean sleep duration was significantly reduced after delivery (mean change 49 minutes; p\<0.001). The amounts of time women spent in bed after delivery were similar to the time spent in bed in week 32 of pregnancy, resulting in the mean sleep efficiency being reduced from 84.0% to 74.9%, p\<0.001. The babies were awake, on average, for one hour and 53 minutes during the night, 25 minutes less than their mothers. Women reported significantly more frequent night-time awakenings in the postpartum period compared to late pregnancy, but the total sum score of insomnia (BIS) improved from 17.2 to 15.4 (p\<0.001). The prevalence of insomnia according to the DSM-IV criteria also decreased from 61.5% in late pregnancy to 53.9% after delivery (p\<0.001). We did not ask about the present use of sleep medication, but only 12 women (0.6%) reported having used sleep medication during the final months of pregnancy. 10.1371/journal.pone.0094674.t001 ###### Sleep before and after delivery among 2088 women. ![](pone.0094674.t001){#pone-0094674-t001-1} Sleep variable Week 32 of pregnancy Week 8 postpartum Mean change 95% CI of change p value ------------------------- ---------------------- ------------------- ------------- ------------------ --------------- Sleep duration (h:mm) 7:19 (1:28) 6:30 (1:18) −0:49 −0:53; −0:44 \<.001 Sleep efficiency (%) 84.0 (14.0) 74.9 (13.9) −9.19 −9.94; −8.42 \<.001 Time in bed (h:mm) 8:44 (1:16) 8:47 (1:26) 0:03 −0:01; 0:07 .21 Time baby awake (h:mm) \- 1:53 (1:10) \- BIS. sum score 17.19 (10.3) 15.47 (8.9) −1.72 −2.17; −1.27 \<.001 Sleep initiation 2.73 (2.4) 1.49 (2.0) −1.24 −1.34; −1.13 \<.001 Sleep maintenance 2.89 (2.3) 4.24 (2.8) 1.35 1.21; 1.49 \<.001 Early awakening 2.30 (2.3) 1.29 (2.0) −1.01 −1.12; −0.88 \<.001 Non-restorative sleep 3.54 (2.2) 3.69 (2.3) .15 −.26: −.03 .013 Daytime impairment 2.28 (2.2) 1.79 (2.0) −.49 .38; .58 \<.001 Satisfaction with sleep 3.46 (2.3) 2.98 (2.3) −.48 .37; .60 \<.001 n (%) n (%) McNemars test Insomnia DSM-IV 1284 (61.6) 1126 (53.8) −7.8% −10.4; −5.2 \<.001 Criterion A Nighttime 1687 (80.8) 1786 (85.4) 4.6% 2.6; 6.6 \<.001 Criterion B Daytime 1350 (64.8) 1150 (55.0) −9.8% −12.3; −7.2 \<.001 BIS - Bergen Insomnia Scale - sub scores represent number of days per week with the insomnia symptom. h:mm - hours:minutes. Primiparous women reported significantly lower sleep efficiencies (3.8% difference; p\<.001), and reported that their babies were awake significantly longer (26 minutes more; p\<.001) during the night compared to babies of multiparous women ([Table 2](#pone-0094674-t002){ref-type="table"}). They had significantly higher BIS scores (15.9 versus 14.9; p\<0.001), mainly due to more problems with sleep-onset latency and night-time awakenings. This did not, however, result in a higher proportion of insomnia. Women who breastfed did not have significantly different BIS scores or prevalence of insomnia compared to women who were not breastfeeding. However, they reported fewer nights with sleep onset insomnia (BIS 1 sub score 1.4 versus 2.1; p\<0.001) and more nights with wakefulness after sleep onset insomnia (BIS 2 sub score 4.3 versus 3.7; p = 0.003). 10.1371/journal.pone.0094674.t002 ###### Sleep after delivery, according to parity among 2088 women. ![](pone.0094674.t002){#pone-0094674-t002-2} Sleep variable Primipara Multipara Mean difference 95% CI p-value ------------------------- ----------------- ----------------- ----------------- ----------------- ------------ mean (SD) mean (SD) t-test Sleep duration (h:mm) 6:33 (1:24) 6:26 (01.11) −0:07 −0:14; 0:00 .031 Sleep efficiency (%) 73.1 (14.6) 76.8 (12.8) 3.77 2.57; 4.96 \<.001 Time in bed (h:mm) 9:07 (1:33) 8:26 (1:13) −0:40 −0:47; −0:33 \<.001 Time baby awake (h:mm) 2:07 (1:12) 1:41 (1:05) −0:26 −0:32; −0:20 \<.001 mean (SD) mean (SD) BIS, sum 15.97 (9.1) 14.94 (8.8) −1.03 −1.79; −.26 \<.001 Sleep initiation 1.74 (2.1) 1.23 (1.9) −.51 −.67; −.32 \<.001 Sleep maintenance 4.55 (2.7) 3.92 (2.8) −.63 −.87; −.40 \<.001 Early awakening 1.35 (2.1) 1.22 (2.0) −.13 −.31; .04 .13 Non-restorative sleep 3.64 (2.3) 3.74 (2.3) .10 −.10; .29 .34 Daytime impairment 1.74 (2.0) 1.84 (2.0) .10 −.08; .27 .27 Satisfaction with sleep 2.95 (2.3) 3.00 (2.4) .05 −.15; .25 .78 n (%) n (%) Chi X^2^ Insomnia DSM-IV 567 (53.3%) 556 (53.9%) 0.6% −3.4; 4.7 .79 Criterion A Nighttime 916 (86.1%) 866 (84.6%) −1.5% −4.4; 1.3 .17 Criterion B Daytime 579 (54.6%) 568 (55.1%) 0.5% −3.5; 4.6 .79 BIS - Bergen Insomnia Scale - sub scores represent number of days per week with the insomnia symptom. h:mm - hours:minutes. Women who recovered from being depressed during pregnancy (EPDS ≥10 in pregnancy only) still had significantly higher insomnia (BIS) scores postpartum compared to women who scored low on the EPDS at both times ([Table 3](#pone-0094674-t003){ref-type="table"}). Furthermore, [table 3](#pone-0094674-t003){ref-type="table"} shows that women who developed new depression (EPDS ≥10 postpartum only) had higher insomnia (BIS) scores in late pregnancy compared to women who remained well. Estimated sleep duration and sleep efficiency did not predict recovery from depression or later development of depression. Women who developed new depression postpartum had the largest negative change in mean sleep duration (−90 minutes) and sleep efficiency (−13.6%). The longer periods of wakefulness at night for postpartum women scoring 10 or more on the EPDS were mainly accounted for by the baby being awake more, as the differences between maternal and infant wake times were similar across the groups. 10.1371/journal.pone.0094674.t003 ###### Differences in insomnia scores and other sleep parameters according to depressive status[+](#nt105){ref-type="table-fn"} before and after delivery among 2088 women. ![](pone.0094674.t003){#pone-0094674-t003-3} Not depressed[+](#nt105){ref-type="table-fn"} pre or postpartum New onset postpartum Recovered postpartum Depressed[+](#nt105){ref-type="table-fn"} both pre and post --------------------------- ----------------------------------------------------------------- -------------------------------------------------------------------------------------- --------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------- BIS, sum score 3rd trimester 15.8 (9.9) 18.9 (9.9)[\\](#nt108){ref-type="table-fn"} [a](#nt110){ref-type="table-fn"} 22.7 (10.1)[\\](#nt108){ref-type="table-fn"} [a](#nt110){ref-type="table-fn"} 26.2 (10.0)[\\](#nt108){ref-type="table-fn"} ^b,^ [\](#nt109){ref-type="table-fn"} ^c^ Postpartum 14.0 (8.3) 22.1 (9.2)[\\](#nt108){ref-type="table-fn"} [a](#nt110){ref-type="table-fn"} 17.8 (8.7)[\\](#nt108){ref-type="table-fn"} [a](#nt110){ref-type="table-fn"} 23.8 (8.2)[\\](#nt108){ref-type="table-fn"} ^c^ Sleep duration (h:mm)* 3rd trimester 7:22 (1:24) 7:22 (1:18) 7:11 (1:54) 6:50 (1:36)[\](#nt109){ref-type="table-fn"} ^b^ Postpartum 6:36 (1:18) 5:52 (1:18)[\\](#nt108){ref-type="table-fn"} ^a^ 6:32 (1:30) 5:56 (1:30)[\](#nt109){ref-type="table-fn"} ^c^ Baby wake time 1:50 (1:08) 2:23 (1:15)[\\](#nt108){ref-type="table-fn"} ^a,^ [c](#nt112){ref-type="table-fn"} 1:50 (1:04) 2:15 (1:23)[\](#nt109){ref-type="table-fn"} ^a,^ [c](#nt112){ref-type="table-fn"} Difference mum/baby 0:26 0:19 0:24 0:25 Sleep efficiency* 3rd trimester 84.8% (13.3) 84.2% (13.6) 79.3% (17.5)[\\](#nt108){ref-type="table-fn"} ^a^ 78.6% (15.9)[\](#nt109){ref-type="table-fn"} ^b^ Postpartum 75.5% (13.3) 70.6% (15.5)[\\](#nt108){ref-type="table-fn"} ^a^ 75.3% (14.1) 70.8% (17.4)[\](#nt109){ref-type="table-fn"} ^c^ Depression was defined by EPDS (Edinburgh Postnatal Depression Scale) score ≥10. BIS - Bergen insomnia Scale SD: Standard Deviation. h:mm - hour: minutes. \\p\<.001. \p\<.01. a- difference from women with no depression before or after delivery. b- difference from women with new onset depression postpartum. c- difference from women who recovered from depression in pregnancy. [Table 4](#pone-0094674-t004){ref-type="table"} shows the results of a multivariate linear regression analysis of factors associated with higher insomnia (BIS) scores after delivery and factors associated with postpartum depression (EPDS scores). Insomnia scores during pregnancy were not a risk factor for depressive symptoms after delivery when adjusted for previous depression. 10.1371/journal.pone.0094674.t004 ###### Risk factors for insomnia and for depressive symptoms eight weeks after delivery, results from a multivariate linear regression analysis among 1914 women. ![](pone.0094674.t004){#pone-0094674-t004-4} Variable B Adj. B P value 95% CI p value ------------------------------------------------- ------------------------- ------- ----------- ---------- ------------ ---------- ----------------------------------- Insomnia* [a](#nt113){ref-type="table-fn"} Intercept −4.30 .213 −11.08 2.48 .037 (BIS score) Primipara .98 1.51 \<.001 .74 2.29 \<.001 after delivery Previous depression 3.89 1.00 .018 .17 1.82 \<.001 Fear of delivery 5.72 2.23 .002 .82 3.63 .008 Stressfactors last year .67 .19 .11 −.04 .42 .001 BIS week 32 .35 .30 \<.001 .26 .33 \<.001 Log EPDS week 32 7.27 1.91 .008 .51 3.30 \<.001 Log HSCL-A week 32 12.27 3.09 .012 .67 5.51 \<.001 Education level .33 .81 .023 .11 1.50 .020 Maternal age .01 .08 .058 −.003 .17 .16 Depression [b](#nt114){ref-type="table-fn"} Intercept −.39 .001 −.63 −.15 [\](#nt115){ref-type="table-fn"} (log EPDS) Primipara .03 .04 .002 .02 .07 after delivery Previous depression .22 .09 \<.001 .06 .11 Fear of delivery .23 .02 .35 −.03 .07 Stressfactors last year .05 .02 \<.001 .01 .02 BIS week 32 .008 .00 .66 −.001 .002 Log EPDS week 32 .58 .43 \<.001 .38 .48 Log HSCL-A week 32 .80 .23 \<.001 .14 .31 Education level −.01 .03 .025 .004 .05 Maternal age −.004 .00 .76 −.003 .004 R^2^ = 0.209. R^2^ = 0.356. \p values for Wilk\'s Lambda for the significance of the multivariate tests (the p values are common for insomnia and depression). CI - Confidence Interval. BIS - Bergen Insomnia Scale. EPDS - Edinburgh Postnatal Depression Scale. HSCL A - Hopkins Symptoms Checklist, Anxiety module. Discussion {#s4} ========== We found that women slept fewer hours at night after delivery compared to during late pregnancy, and that they reported more nights with night-time awakenings. Hence, sleep efficiency decreased from 84% to 75%. In spite of this, the self-reported insomnia scores improved, and the prevalence of insomnia according to the DSM-IV criteria decreased. Primiparous women reported lower sleep efficiency and higher insomnia scores than multiparous. Primiparous women also had increased risk of depressive symptoms after delivery. Insomnia scores in pregnancy, but not sleep duration or sleep efficiency, were associated with depressive symptoms postpartum. However, when adjusted for previous depression, insomnia in pregnancy did not predict depressive symptoms after delivery. Change in sleep from pregnancy to postpartum {#s4a} -------------------------------------------- We found that women slept fewer hours and less efficiently after delivery when compared to sleep during pregnancy. Montgomery-Downs et al. found longer sleep duration after delivery than in our study (7.2 hours compared to 6.5 hours) but similar low sleep efficiency [@pone.0094674-MontgomeryDowns1]. In spite of reduced sleep time and sleep efficiency, insomnia scores and prevalence of insomnia were reduced in our study. Although they reported more nights of being awake for longer than 30 minutes, the postpartum women had fewer evenings with delayed sleep onset and fewer mornings with awakenings more than 30 minutes earlier than desired. Women reported less daytime impairment and were more satisfied with their sleep after delivery compared to during pregnancy. New mothers may have expected sleep disturbances after delivery to be more severe than they experienced, and therefore they were more satisfied with the (shorter amount of) sleep they got. In addition, waking up because of an infant may be more tolerable and may produce less worry compared to being unable to sleep for other reasons. Worry about one\'s sleep may contribute to insomnia, as individuals who tend to focus cognitively on their insomnia and ruminate about their poor sleep are less likely to get a good night\'s sleep [@pone.0094674-Baglioni1]. This worry may consequently lead to hyper-arousal, an important factor in maintaining insomnia [@pone.0094674-Riemann1]. None of the women were working outside the home at the time of the postpartum follow-up, due to the right of paid maternity leave in Norway for the first 45 postnatal weeks. This may explain some of the lower insomnia rate, as sleep problems are more common during periods of active employment [@pone.0094674-Lampio1]. However, the majority of the women in our study, 63%, were on sick leave in week 32, and these women had a higher rate of insomnia compared to working pregnant women [@pone.0094674-Dorheim3]. Taking care of a new-born baby the first weeks after delivery may be hard work and not considered leisure time. However, compared to work outside the home, there may be more flexibility in terms of completing tasks and more opportunities for daytime napping. Our study did not measure daytime napping, as this is not part of the Bergen Insomnia Scale or the Pittsburgh Sleep Quality Index, but a study of pregnant women in Taiwan found decreased sleep among women with longer work hours and women who had longer daytime naps [@pone.0094674-Tsai1]. Future studies should include questions regarding daytime naps. Primiparous women reported more days with delayed sleep-onset time and were awake during more nights compared to multiparous women, possibly due to less experience and confidence in child care and more worry and hyper-arousal. Coo Calcagni et al. found that first-time mothers experienced more stress and had poorer subjective sleep than multiparous women [@pone.0094674-CooCalcagni1]. Other studies have also found poorer sleep among first-time mothers [@pone.0094674-Dorheim2], [@pone.0094674-Signal1]. Salo et al. found that difficulties initiating or maintaining sleep increased the long-term risk of depression in the general population [@pone.0094674-Salo1]. These sleep factors, more prevalent among primipara, may be mediators that also increase the risk of postpartum depression. However, primiparous women also had an increased risk of depression when adjusting for level of insomnia, so there may be factors other than insomnia contributing to this increased risk for depression among first-time mothers. Previous studies from Norway have found conflicting results regarding parity and postpartum depression [@pone.0094674-Dorheim2], [@pone.0094674-EberhardGran3], [@pone.0094674-Berle1]. Parity does not seem to be among the strongest risk factors for postpartum depression [@pone.0094674-OHara1]. We found no associations between breastfeeding and total insomnia scores, which is in line with other studies [@pone.0094674-Dorheim2], [@pone.0094674-MontgomeryDowns2]. From our data, it seems that although breastfeeding may increase the number of nights with awakenings, it also decreases nights with prolonged sleep-onset latency. Insomnia and depression {#s4b} ----------------------- We found that insomnia, but not sleep efficiency or sleep duration, was a risk factor for postnatal depression in the univariate analyses. This is similar to findings by Park et al., using actigraphy and subjective sleep reports. They found that subjective assessments of sleep may be more accurate predictors of postpartum depression than total amount of sleep measured by actigraphy [@pone.0094674-Park1]. In our study, women who recovered from antepartum depression still had residual insomnia symptoms compared to women who scored low for depression at both time points. Similarly, women who developed depression after delivery but were not depressed in pregnancy had higher insomnia scores in pregnancy compared to women who remained well. This may reflect the co-morbid and bidirectional relationship between insomnia and depression [@pone.0094674-Manber1], [@pone.0094674-JanssonFrojmark1]. It may also reflect a pre-clinical state where women have insomnia as a precursor for or an early symptom of depression. The residual insomnia found among women who recovered from depression in pregnancy has previously been reported in non-pregnant populations and is associated with an increased risk of relapse of depression [@pone.0094674-Manber1], [@pone.0094674-Li1]. This may explain why insomnia in pregnancy did not predict postpartum depression when adjusted for previous depression. Antenatal depression and a prior history of depression are among the strongest risk factors for postnatal depression [@pone.0094674-Milgrom1]. Marques et al. found similar results to our study: Insomnia in pregnancy seemed to predict depression after delivery; however, when adjusted for lifetime history of depression and negative affect in pregnancy, the association disappeared [@pone.0094674-Marques1]. In a study of women with previous postpartum depression, Okun et al. found that changes in postpartum sleep during the first 17 weeks predicted recurrence of depression in these at-risk women [@pone.0094674-Okun2]. Insomnia in pregnancy may thus be seen as a mediator between previous depression and a new depressive episode postpartum for women at risk. As insomnia is an easy complaint to look for during pregnancy, insomnia may thus warn the clinician of an increased risk of postpartum relapse of depression. This may in turn provide opportunities for early interventions to prevent clinical depression. Treating insomnia during pregnancy and the early postpartum period for women with previous depression may be a way of preventing postpartum depression. One pilot study of insomnia treatment among 12 women with postpartum depression and insomnia showed promising results on sleep efficiency, total wake time, subjective mood, insomnia severity, sleep quality, and fatigue [@pone.0094674-Swanson1]; however, more studies on treatment of postpartum women are needed. An encouraging finding of our study is that women with insomnia in pregnancy who have never been depressed before may be reassured that insomnia in itself does not seem to increase the risk for postpartum depression. For many, insomnia is a chronic condition. The present study found that a woman who slept poorly during pregnancy was more likely to sleep poorly after delivery as well, regardless of any depression. Morin and colleagues found that 74% of people with insomnia had the complaint for more than one year, and more than 25% of those who recovered experienced relapse [@pone.0094674-Morin1]. The perinatal period may therefore be no exception to this. Hyper-arousal may be one factor explaining that both anxiety, as measured by the HCSL-Anxiety scale, and fear of delivery were factors associated with postpartum insomnia [@pone.0094674-Riemann1]. There may be an interaction between cognitive and autonomic hyper-arousal that contributes to the maintenance of both insomnia and continued emotional disturbance [@pone.0094674-Baglioni3], [@pone.0094674-LeBlanc1]. Strengths and limitations {#s4c} ------------------------- This is a large study from a general population of pregnant women. However, as women who did not understand the Norwegian language were excluded, these results may not be representative for the newer immigrant population. The longitudinal design makes it possible to examine the direction of effects and to evaluate whether depressive symptoms precede insomnia or whether insomnia is a precursor for depression in this period with decreased sleep among all women. For longitudinal population-based studies, a retain rate of more than 50% after three questionnaires is acceptable, and a large number of participants remained in the study. According to Galea and Tracey, declines in participation rates are not likely to have substantial influence on exposure-disease associations [@pone.0094674-Galea1]. Women with depressive symptoms during pregnancy were less likely to remain in the study, and this may have led to selection bias. We found, however, a strong correlation between depression and insomnia in our sample with possibly fewer depressed women than in the population, and the real relationship may be stronger than the results presented here. The questionnaires included several known risk factors for postpartum depression, making it possible to adjust for confounders. We controlled for stressful life events the previous year, but did not assess for past life events, such as childhood adversities, which may also be of importance for emotional dysfunction during pregnancy [@pone.0094674-Dayan2]. The tools used are all validated and have been used in several other studies in Norway. Depression and insomnia were measured by self-report and not verified by clinical interviews or sleep diaries. As the diagnosis of insomnia relies on the subjective experiences of the individual experiencing sleep problems [@pone.0094674-AASM1], self-report may be an adequate way of measuring insomnia. The BIS is based on recall of insomnia symptoms during the past month, and the prospective use of sleep diaries might have given a more accurate estimate of current insomnia. However, BIS scores correspond well with polysomnographic sleep registration, except for item 4 (feeling adequately rested after sleep) and item 5 (daytime functioning) [@pone.0094674-Pallesen1]. The EPDS is not diagnostic of depression, and a cut-off at 9/10 at the Norwegian version is relatively low compared to validations in other languages [@pone.0094674-EberhardGran4]. There may therefore be a risk of some false positives in [Table 3](#pone-0094674-t003){ref-type="table"}. However, the continuous scale of the EPDS has been recommended for use in population research to find factors associated with depression [@pone.0094674-Green1]; hence the cut-off on the EPDS did not influence the risk factors identified in [Table 4](#pone-0094674-t004){ref-type="table"}. Suggestions for clinical intervention and further research {#s4d} ---------------------------------------------------------- Insomnia during pregnancy may be a valid clinical marker for recurrence of depression postpartum among women at risk. Further studies should evaluate whether treatment of insomnia during pregnancy may prevent postpartum depression among women with other risk factors for the condition. Cognitive behavioral therapy for insomnia may have a positive effect on anxiety [@pone.0094674-Belleville1], and treatment of depression with mindfulness meditation improves subjective sleep quality with corresponding changes in polysomnography [@pone.0094674-Britton1]. Lee and Gay found that treatment of sleep problems among new parents using a modified sleep-hygiene intervention improved sleep, especially among less socioeconomically advantaged women [@pone.0094674-Lee1]. Low socioeconomic status is a risk factor for postpartum depression [@pone.0094674-OHara1], and improving sleep for these mothers may therefore be one way of preventing or relieving depressive symptoms in this group. The authors thank the women who volunteered their time to participate in this study. We also thank Tone Breines Simonsen, Wenche Leithe, and Ishtiaq Khushi for assistance in the data collection, and Ingvild Dalen, Stavanger University Hospital, and Geir Egil Eide, Western Norway Regional Health Authority, for assistance in the statistical analyses. [^1]: Competing Interests:Signe Dørheim has had paid speaking engagements for AstraZeneca and Pfizer. Bjørn Bjorvatn has had paid speaking engagements for Glaxo and Confex. Malin Eberhard-Gran has indicated no financial conflicts of interest. This does not alter the authors\' adherence to PLOS ONE policies on sharing data and materials. [^2]: Conceived and designed the experiments: SD BB MEG. Performed the experiments: MEG. Analyzed the data: SD BB MEG. Contributed reagents/materials/analysis tools: SD. Wrote the paper: SD BB MEG.	{ "pile_set_name": "PubMed Central" }
Today, most patients who require renal replacement therapy are treated using dialysis membranes. To improve the outcomes of chronic dialysis patients, unique developments have been made in the past several decades. Cellulose membranes have been mostly replaced by synthetic polymeric membranes with improved biocompatibility. The development of high-flux membranes and more efficient treatment modes, such as hemodiafiltration, have resulted in improved removal rates of uremic toxins, particularly those in the middle molecular-weight range (\>500 Da). Beyond the membranes used for conventional hemodialysis, new membranes with increased pore size have been developed for specific treatments. These specialized membranes allow for the removal of higher molecular-weight molecules, such as mediators of sepsis/inflammation, or the removal of nephrotoxic light chains of immunoglobulins. However, these membranes allow the passage of plasma proteins, such as albumin ([@bibr1-ijao.5000603]), whose loss is undesirable. The newest generation of highly selective and permeable medium cut-off (MCO) membranes enables the removal of large molecules, as do high cut-off (HCO) membranes, while simultaneously maintaining low passage of albumin. For uremic solutes in the 15,000 to 45,000 Da- size range, the MCO membranes offer improved clearance in comparison with that of high-flux membranes used in HD mode and equivalent clearance to that of high-flux membranes used in high-volume hemodiafiltration (HDF) mode ([@bibr2-ijao.5000603]). Therefore, the use of MCO membranes simplifies the delivery of high-removal treatments for end-stage renal disease (ESRD) patients, with a removal spectrum that extends the current possibilities of the best available therapies. This improvement should allow clinicians to surpass the benefits of HDF while using regular HD equipment, i.e., not requiring large amounts of high-quality fluid and a more complex setup. These special MCO membranes should aid in raising the standard of treatment available for all chronic HD patients and may potentially decrease inflammation and improve general patient outcomes. Future studies are being planned to demonstrate the clinical benefits of this new membrane type. Development of Dialysis Membranes with Different Permeabilities {#section1-ijao.5000603} =============================================================== Renal failure (RF) is characterized by the loss of the ability of the kidneys to excrete wastes, concentrate urine, conserve electrolytes, and maintain fluid balance. Acute kidney injury can be a life-threatening illness, with a mortality of between 50% and 80% ([@bibr3-ijao.5000603]). ESRD occurs in the late stages of chronic kidney disease and is associated with an irreversible loss of kidney function. Renal replacement therapy (RRT) replaces kidney function in patients with both types of renal failure. There are 2 possible alternatives to treat chronic RF: organ transplantation or, more commonly, dialysis. However, the natural secretion of kidney hormones, which influences blood pressure, cannot be achieved with dialysis treatment modalities. Over 2 million people worldwide currently receive treatment with dialysis or a kidney transplant to stay alive, yet this number may represent only 10% of people who actually need treatment to live ([@bibr4-ijao.5000603]). Of the 2 million people who receive treatment for kidney failure, the majority are treated in only 5 countries: the United States, Japan, Germany, Brazil, and Italy. These 5 countries represent only 12% of the world population. Only 20% are treated in approximately 100 developing countries that make up over 50% of the world population ([@bibr4-ijao.5000603]). In medicine, hemodialysis (HD) is the process of removing blood from a patient, purifying the blood through an artificial kidney (dialyzer), and then returning it to the patient\'s bloodstream. On the basis of the developments of Willem Kolff and Nils Alwall in the 1940s, the dialyzer membrane systems have undergone multiple development cycles and are now the basis of an effective, reliable, and cost-effective treatment approach. In the early days of dialysis, large and unwieldy plate dialyzers made of cellulose membranes were used. Until the 1970s, cellulosic membranes were used exclusively and represented the majority of membranes used in hemodialysis worldwide ([@bibr5-ijao.5000603]). However, cellulosic membranes may activate the complement system ([@bibr6-ijao.5000603]) and induce other adverse biological reactions, e.g., leucopenia, owing to the accumulation of granulocytes in lung capillaries ([@bibr7-ijao.5000603]), inhibition of granulocyte metabolism, and release of enzymes from granulocytes and monocytes ([@bibr8-ijao.5000603]). The drawbacks associated with the decreased hemocompatibility of cellulosic membranes can be mitigated by the partial substitution of the hydroxyl groups by (i) acetylation or the (ii) introduction of diethylamino groups or (iii) benzyl groups ([@bibr9-ijao.5000603]). Cellulose acetate is a cellulose-derived material that has been modified by esterification. Depending on the degree of substitution, such a membrane can be categorized as cellulose acetate, cellulose diacetate, or cellulose triacetate. Another biocompatibility improvement of cellulosic membranes is surface-coating them with polyethylene glycol (PEG), polyacrylonitrile (PAN-RC®) ([@bibr10-ijao.5000603]), or vitamin E (Excebrane®). The use of vitamin-E-coated membranes, compared with uncoated cellulose membranes, appears to significantly decrease the activation and migration of monocytes and granulocytes ([@bibr11-ijao.5000603]). Cellulosic membranes belong to the class of "low-flux" membranes, which have a low permeability: substances of up to 5,000 Da can be eliminated, whereas nearly 100% of the (ß2-microglobulin is rejected. The area of synthetic polymeric membrane development began with the discovery by Geyjo in 1985 that (ß2-microglobulin is the amyloidogenic precursor of AB-amyloidosis ([@bibr12-ijao.5000603]), together with the introduction of the hemofiltration treatment mode. Because the mass transfer of low-flux membranes is mainly diffusive and is not suitable for use in therapies that require high convective transport, such as hemofiltration (HF) and hemodiafiltration (HDF), membranes with higher permeability were required. Synthetic polymeric membranes close this gap, showing higher permeability and better biocompatibility. The hydrophobic base material is mainly polysulfone or polyethersulfone (polyarylethersulfone). These high-flux membranes allow for the removal of so-called middle molecular-weight molecules. Membranes prepared from hydrophobic/hydrophilic polymer blends are the predominant type of synthetic polymeric membrane ([@bibr9-ijao.5000603]). High-flux membranes are highly permeable to β2-microglobulin. The pore structure of high-flux membranes allows for the passage of so-called middle molecules of up to 20,000 Da. In addition, significant protein loss, for example, in albumin (68,000 Da), is prevented. Beyond the immunoglobulins, albumin is the most important plasma protein that must be retained in chronic dialysis patients. High-flux membranes are suitable for hemodialysis, hemofiltration, and hemodiafiltration. High-flux synthetic membranes show good transport properties and exhibit only slight complement activation ([@bibr13-ijao.5000603]), a minimal decrease in leukocytes ([@bibr14-ijao.5000603]), and a low release of leukocyte elastase ([@bibr15-ijao.5000603]). In a recent study, no significant survival benefit with either high-flux or low-flux membranes has been found in the overall population, but the use of high-flux membranes confers a significant survival benefit in patients with serum albumin ≤4 g/dL ([@bibr16-ijao.5000603]). Beyond membranes used for conventional hemodialysis, high cut-off (HCO) membranes with increased pore size have been developed for specific treatments. In high cut-off membranes, the pore sizes are shifted toward a larger pore diameter than that of conventional high-flux membranes showing a higher permeability for substances above 15,000 Da ([@bibr17-ijao.5000603]). Examples of these substances include inflammatory cytokines, which are over-expressed in acute inflammatory diseases and contribute to the pathogenesis of septic acute kidney injury or free light-chain proteins. Monoclonal light chains of immunoglobulin are excessively produced by malignant plasma cells in patients with multiple myeloma (MM) and can cause renal lesions such as cast nephropathy. In this condition, serum concentrations may be elevated by more than 100-fold compared with normal serum concentrations. The increase in the pore sizes of HCO membranes translates into an increase in both the convective and diffusive permeability, thus allowing for the removal of molecules, such as mediators of sepsis/inflammation ([@bibr18-ijao.5000603]), or the removal of nephro-toxic light chains of immunoglobulins ([@bibr19-ijao.5000603]). However, the albumin loss with this membrane is substantial, from approximately 9 to 23 g/treatment ([@bibr20-ijao.5000603], [@bibr21-ijao.5000603]), and it is not indicated for use in chronic dialysis but instead is recommended mainly for acute applications ([@bibr22-ijao.5000603]). Over the past decade, various membranes, such as the cellulose triacetate-based FH 70/150 (Sureflu; Nipro ([@bibr23-ijao.5000603])), the polysulfone-based APS-1050 (Asahi ([@bibr24-ijao.5000603])), the PMMA-based BK-F/BG 2.1(Toray ([@bibr25-ijao.5000603])), and the Helixone polysulfone based FX-E (Fresenius Medical Care ([@bibr26-ijao.5000603])), have been manufactured with the purpose of increasing membrane permeability, given the need for the increased clearance of low molecular weight proteins and protein-bound solutes. Because the loss of albumin is associated with the removal of protein-bound uremic toxins, protein-leaking membranes have been developed for hemodialysis. Dialysis treatments with such membranes, compared with conventional high-flux dialysis membranes, provide greater clearance of those molecules, albeit at the cost of some albumin loss into the dialysate ([@bibr27-ijao.5000603]). The newest generation of highly selective and permeable medium cut-off (MCO) membranes meets both requirements for high-quality and effective dialysis treatment: the removal of large middle-molecules up to a molecular weight of 45,000 Da, similar to a high cut-off membrane, and a low loss of albumin (high-flux membrane). For uremic solutes in the 15,000 to 45,000-Da size range, the MCO membranes offer improved clearance in comparison with those of high-flux membranes used in HD mode and equivalent clearance to high-flux membranes in HDF mode ([@bibr2-ijao.5000603]). Uremic Toxins and Removal Mechanisms {#section2-ijao.5000603} ==================================== Because 'artificial kidneys' remove uremic toxins primarily via size exclusion, the classification of these solutes on the basis of their molecular weight and removal behavior is key for membrane development. In maintenance hemodialysis patients, the insufficient removal of mid-sized and protein-bound uremic toxins is associated with endothelial injury and chronic inflammation, thus contributing to subsequent cardiovascular disease. Studies demonstrating the role of these uremic toxins in inducing cardiovascular effects or other negative effects, such as coordination disturbances or polyneuritis in dialysis patients, have attracted increasing attention over the past decade ([@bibr28-ijao.5000603]). The molecular weights of these uremic toxins vary widely from very small compounds, such as urea (MW = 60 Da), to globulins (e.g., \[ß2-microglobulin MW = 11,800 Da). Generally, uremic toxins are split into 3 groups: small water-soluble compounds, middle molecules and protein-bound solutes ([@bibr29-ijao.5000603]). In the latest literature research performed by the European Uremic Toxin (EUTox) Work group, 56 newly reported solutes have been found ([@bibr30-ijao.5000603]). Examples of solutes classified into the 3 aforementioned groups are shown in [Table I](#table1-ijao.5000603){ref-type="table"}. ###### Examples of Uremic Solutes and their Molecular Weights, Classified into 3 Groups: Small Water-soluble Compounds, Middle Molecules and Protein-bound Solutes ([@bibr28-ijao.5000603], [@bibr33-ijao.5000603]) ![](10.5301_ijao.5000603-table1) Molecule Molecular weight Example ------------------------------- ------------------ ---------------------------------------------------------------------------------------------------------------------------- Small water-soluble compounds \<500 Da asymmetric dimethylarginine, guanidine, uric acid, oxalate, ethylamine, methylguanidine, neopterin, phenylacetic acid Middle molecules \>500 Da β2-microglobulin, adiponectin,α1-acid glycoprotein, cystatin C, prolactin, osteocalcin, vascular endothelial growth factor Protein-bound solutes Variable p-cresylsulfate, indoxyl sulfate, phenol, indol-3-acetic acid, hippuric acid, homocysteine, carboxymethyllysine, acrolein Small water-soluble compounds are also referred to as low molecular weight molecules, which have molecular weights \<500 Da and are soluble in water. The second group, consisting of middle molecular weight toxins, includes compounds with molecular weights above 500 Da. Protein-bound solutes constitute the third group of uremic toxins, which have molecular weights less than 500 Da but demonstrate strong binding to albumin. Small water-soluble compounds are easily removed, diffusively, across the semipermeable membrane, owing to the concentration gradient between the blood and dialysis fluid. Conventional hemodialysis is an effective treatment mode for the removal of those compounds, whereas it is poorly suited for middle molecules such as (ß2-microglobulin, even with the use of high-flux membranes. This phenomenon is attributable to a decrease in the diffusion coefficient with increasing molecular size. Alternative therapies such as hemofiltration or hemodiafiltration include convective transport mechanisms and the improved removal of both low- and middle-molecular-weight toxins such as (ß2-microglobulin ([@bibr31-ijao.5000603]). The driving force for convective transport is a pressure gradient across the membrane, thus leading to ultrafiltration (UF), in which the molecules are pulled from the blood into the dialysate. An increase in the ultrafiltration rate increases the solute removal of low- and middle-molecular-weight solutes. In hemodialysis, UF is limited to the removal of only excess body fluid from the blood, whereas in convective therapies UF is increased beyond the target weight loss, and fluid balance is preserved via the infusion of a physiological solution into the blood ([@bibr32-ijao.5000603]). In clinical practice, the UF rate is limited by the blood flow rate because only a portion of the volume, usually 25% to 30%, is filtered before the blood cell concentration increases to a level at which cellular damage and dialyzer clotting occur ([@bibr32-ijao.5000603]). Given the ability to obtain ultrapure water and dialysis fluids for use in water treatment systems and online HDF machines, the use of online hemodiafiltration is increasing steadily in Europe ([@bibr34-ijao.5000603]). Online HDF post-dilution is considered the most effective and safest dialysis treatment, owing to its superior blood purification of all uremic toxins and its decreased association with the incidence of cardiovascular events ([@bibr35-ijao.5000603]). However, this therapy is more complex and requires high exchange volume rates of up to 24 L per treatment. The automated machine settings aim to optimize the filtration fraction with the lowest possible rate of machine alarms but do not regulate the 2 primary determinants of the convection volume: the treatment time and the blood flow rate ([@bibr36-ijao.5000603]). Furthermore, in comparison with other dialysis modalities, HDF also provides superior removal of certain protein-bound uremic solutes ([@bibr37-ijao.5000603]). However, the removal of protein-bound solutes via dialysis strategies is still less efficient than the removal of non-protein-bound solutes of similar molecular weights, owing to the resistance induced by the protein binding ([@bibr28-ijao.5000603]). The total removal depends on how rapidly the solute unbinds from its carrier protein as the free concentration decreases ([@bibr32-ijao.5000603]). Investigations into protein-bound uremic toxin removal are often restricted to some representatives, such as indoxyl sulfate, conjugates of p-cresol, p-cresylsulfate, and p-cresylglucoronide, and phenylacetic acid ([@bibr28-ijao.5000603], [@bibr38-ijao.5000603]). With the deployment of adsorptive therapies that were first implemented for the treatment of severe liver failure, the removal of these molecules may be enhanced ([@bibr38-ijao.5000603][@bibr39-ijao.5000603]--[@bibr40-ijao.5000603]). Because the loss of albumin is associated with the removal of protein-bound uremic toxins, protein-leaking membranes have been developed for hemodialysis. Dialysis treatments with such membranes, compared with conventional high-flux dialysis, provide greater clearance of these molecules, but this comes at the cost of a certain degree of albumin loss into the dialysate ([@bibr28-ijao.5000603]). Albumin Removal in Different Treatment Modalities {#section3-ijao.5000603} ================================================= Hypoalbuminemia is associated with mortality in patients with ESRD. Albumin removal during dialysis treatment leads to a decrease in the serum albumin concentration, because it is influenced by the albumin synthesis rate, catabolism, distribution between the intra- and extravascular compartments and external loss under pathological conditions ([@bibr41-ijao.5000603]). The limits of albumin removal across a dialyzer in terms of patient tolerance have been addressed in 2 reviews published in 2003 and 2005, but this issue has not been resolved ([@bibr27-ijao.5000603], [@bibr42-ijao.5000603]). Nevertheless, albumin removal during different treatment modalities has been elucidated: a small but non-significant decrease in the serum albumin concentration has been observed over a period of 24 months in a group of 17 stable peritoneal dialysis patients, despite protein losses in the range of 5 to 7 g per day ([@bibr43-ijao.5000603]). The albumin removal during hemodialysis with conventional high-flux membranes is generally reported to be in the range of 0 to 2 g per 4-hour treatment, depending on the membrane material and the surface area ([@bibr27-ijao.5000603]). The reported albumin removal levels in online-HDF treatments vary widely and depend on the dilution mode, degree of flux across the membrane, type of membrane, and other treatment parameters. The albumin loss per HDF treatment typically ranges from 1 to 4 g and can sometimes reach values of greater than 5 g ([@bibr44-ijao.5000603]). The protein removal levels with the use of reused dialyzers after bleaching procedures are in the range of 10 to 12 g per treatment; these levels are associated with a significant decrease in the serum albumin concentration ([@bibr45-ijao.5000603]). However, comparable average serum albumin concentrations have been reported in another study in which bleach reprocessing was limited, equivalent to an average dialyzer albumin removal rate of 4.3 g per hemodialysis session ([@bibr46-ijao.5000603]). More recent data published as an abstract at the World Congress of Nephrology 2009 have indicated dialysate albumin losses in the range of 0.48 g for Evodial, with 2.2 to 15.5 g per 4-hour session for the FDY 210 dialyzer (Nikkiso) used in hemodiafiltration mode ([@bibr47-ijao.5000603]). The most recent studies examining widely used membranes have reported albumin losses of 3.1 ± 2.4 g ([@bibr48-ijao.5000603]) and 3.0 ± 2.4 to 4.3 ± 3.5 g ([@bibr49-ijao.5000603]). Currently, it is unclear how much albumin loss per treatment session is tolerated in ESRD patients. There has also been discussion on the utility of a certain degree of loss during albumin removal without triggering antioxidant effects versus facilitating the synthesis of new albumin with antioxidant effects ([@bibr50-ijao.5000603]). Medium Cut-off (MCO) Membranes -- a New Generation of Hemodialysis Membranes {#section4-ijao.5000603} ============================================================================ In the years since the development of hemodialysis as a treatment for ESRD, a number of improvements have been made to dialyzer membranes. Despite these advances, the overall clinical outcomes for patients still present challenges. Many observational studies have supported the hypothesis that higher-molecular-weight toxins are responsible for a number of dialysis comorbidities, such as chronic inflammation and related cardiovascular diseases ([@bibr51-ijao.5000603]), immune dysfunctions ([@bibr52-ijao.5000603]), anemia and EPO hyper responsiveness ([@bibr53-ijao.5000603]), thus in turn influencing the mortality risk. However, when high cut-off membranes are used to filter higher-molecular-weight toxins, the patients lose unacceptable amounts of albumin and other essential proteins. In an open, randomized, cross-over, 2-center, controlled, prospective clinical study, dialysis patients have been treated with high cut-off HCO1100 dialyzers (Baxter-Gambro, Hechingen, Germany) in series with the low-flux dialyzer PF14L (HCO/LF-HD) or the high-flux dialyzer PF210H (HF-HD) ([@bibr1-ijao.5000603]). The patients treated with HCO/LF-HD showed significantly greater decrease in multiple immune mediators such as slL-2R, sTNF-R1, sTNT-R2, and FLCs but also a significant decrease in albumin from 36.2 ± 3.5 to 31.0 ± 4.7 g/L after 3 weeks. Thus, hemodialysis treatment with high cut-off membranes achieves a more effective removal of larger uremic toxins, thus decreasing inflammatory activity, at the cost of higher albumin removal. The existing gap in the high selective membrane permeability has now been closed with the development of medium cut-off membranes (MCO). Today, advances in polymer recipes, spinning technologies, and the whole dialyzer-manufacturing concept, including the increased use of high-tech equipment, have resulted in the production of improved, safer and higher-quality dialyzer products. Different approaches in terms of the membrane and dialyzer design, such as fiber undulation, high package density, and improved flow distribution of dialysate fluid as well as decreased fiber diameter to increase internal filtration, have led to improved dialyzer performance but remain insufficient with regard to the removal of middle-molecular-weight products and simultaneous retention of proteins such as albumin. These goals require an extremely narrow pore-size distribution of the dialysis membrane, which has not previously been obtained through the phase inversion method but has recently been achieved through a tailored and well-controlled spinning technology that creates larger, uniformly sized, and densely distributed pores. The newly developed medium cut-off membrane exhibits high selective membrane permeability. After the in vitro characterization of the membrane properties in the context of dextran filtration, Boschetti-de-Fierro et al ([@bibr54-ijao.5000603]) have classified MCO membranes as having permeabilities closest to that of the natural kidney, as compared with other conventional dialysis membranes, including all permeability classes (low- and high-flux membranes, high cut-off, and protein-leaking membranes). This classification has introduced a new term, the molecular weight retention onset (MWRO). According to the authors' definition, the MWRO is the molecular weight at which the sieving coefficient is 0.9. Together with the general known molecular weight cut-off (MWCO), at which the sieving coefficient is 0.1, it is possible to map different types of blood purification membranes. Different dialyzer families can be distinguished on the basis of their membranes, such as low-flux, high-flux, protein-leaking, high cut-off and medium cut-off membranes. Medium cut-off membranes expand the known limits of high-flux membranes. This expansion in membrane permeability and selectivity represents a large step toward the realization of ideal dialysis membrane separation properties. Boschetti-de-Fierro et al have described 4 dialysis membrane groups in their general classification and their typical performances ([@bibr17-ijao.5000603]). The data, including the water permeabilities and sieving coefficients of (ß2-microglobulin and albumin, of these dialysis membrane groups, extended to include the class of MCO membranes, are given in [Table II](#table2-ijao.5000603){ref-type="table"}. ###### General classifications and typical performance of membranes used in dialysis ![](10.5301_ijao.5000603-table2) Dialyzer type Water permeability^[a](#table-fn2-1-ijao.5000603){ref-type="table-fn"}^ Sieving coefficient^[b](#table-fn2-2-ijao.5000603){ref-type="table-fn"}^ ----------------- ------------------------------------------------------------------------- -------------------------------------------------------------------------- ------------ Low-flux 10--20 -- \<0.01 High-flux 200--400 0.7--0.8 \<0.01 Protein-leaking 50--500 0.9--1.0 0.02--0.03 High cut-off 1100 1.0 0.2 Medium cut-off 600--850 1.0 0.008 with 0.9% wt.-% sodium chloride at 37°C ± 1 ° C and QB 100--500 mL/min. according to EN1283. In vitro data suggest expanded toxin removal similar to that observed with high cut-off membranes, with simultaneous retention of albumin, such that medium cut-off membranes are appropriate for regular use in conventional treatment schedules and treatment modes, e.g.,4-hour treatments, 3 times weekly, in Europe. The first 2 studies, which were designed to compare MCO dialyzers with the last generation of high-flux dialyzers, have confirmed this assumption ([@bibr2-ijao.5000603]). In the first randomized hemodialysis study, 3 different MCO prototype versions MCO AA, BB, and CC (Gambro Dialysatoreny, a subsidiary of Baxter International) with different membrane pore sizes/permeabilities AA \<BB \<CC54 were compared with a high-flux FX CorDiax 80 dialyzer (Fresenius Medical Care) ([@bibr55-ijao.5000603]). The study was conducted in the LKH University hospital in Graz and at the LKH in Bruck, both in Austria, and 19 ESRD patients were included. The primary endpoint was the overall clearance of λ-free light chains (λ-FLCs) with a molecular weight of \~45,000 Da. The secondary outcomes were the removal of medium-sized solutes such as κ--Ig (MW \~22,500 Da) free light chains,α1-microglobulin, complement factor D, myoglobin,(ß2-microglobulin and small solutes, as well as the safety of these prototypes. The studied MCO dialyzer prototypes achieved significantly higher overall clearances of κ-FLC and λ-FLC (MCO AA, BB and CC vs. FX CorDiax 80: 8.5 ± 0.54,11.3 ± 0.51,15.0 ± 0.53 vs. 3.6 ± 0.51 mL/min), and the removal of other medium-sized solutes was significantly greater, whereas the total mass of the albumin removal was moderate (medians \[range\] of MCO AA, BB and CC vs. FX CorDiax 80: 2.9 g \[1.5 -- 3.9\], 4.8 g \[2.2 -- 6.7\] and 7.3 g \[1.9 -- 9.7\] vs. \<0.3g\[\<0.3-\<0.3\]). In the second study, the performance of 2 MCO prototype dialyzers (MCO AA and MCO BB) in hemodialysis mode was compared with the performance of 2 high-flux dialyzers, FX CorDiax 80 and FX CorDiax 800 (Fresenius Medical Care) in hemodialysis and high-volume hemodiafiltration modes (post-dilution volume-controlled mode with a target total convective ultrafiltration volume of ≥23 L). The study was conducted at the dialysis center in Elsenfeld am Main, Germany. The primary outcome of this study was the λ-FLC overall clearance of MCO prototypes in hemodialysis mode in comparison to the high-flux dialyzers used in hemodialysis and hemodiafiltration mode. The results of this study indicated greater overall λ-FLC clearance by the MCO dialyzers in hemodialysis mode compared with both the HD and HDF treatments with the last-generation high-flux dialyzers (\[least squares mean (standard error)\]): MCO AA 10.0 (0.57), MCO BB 12.5 (0.57) vs. high-flux HD 4.4 (0.57) and HDF 6.2 (0.58) mL/min). The clearances of α1-microglobulin, complement factor D, κ-FLC, and myoglobin were generally greater for MCO than for high-flux HD and similar or greater than in HDF treatments, whereas the albumin removal was moderate with MCO but greater than that of high-flux HD and HDF (medians \[range\] of MCO AA and BB vs. FX CorDiax 80 and FX CorDiax 800; 3.2 g \[1.9 -- 3.9\] and 4.9 g \[1.1 -- 7.2\] vs. 0.2 g \[0.2 -- 0.9\] and 0.4 g \[0.3 -- 0.8\]). On the basis of both studies, the MCO membranes in hemodialysis mode demonstrate the effective removal of a wide range of middle molecules and exhibit substantially better performance than that of standard high-flux hemodialysis treatment, even exceeding the performance of high-volume postdilution HDF for large solutes, particularly λ-FLC. Because MCO membranes are associated with albumin removal in the range reported in the literature for HDF, the effects of albumin removal with MCO dialyzers on serum albumin levels would be expected to be similarly low, thus providing safe treatment in routine hemodialysis mode. Conclusions {#section5-ijao.5000603} =========== The newest generation of highly selective and permeable MCO membranes meets both requirements for high quality and good performance for dialysis treatment, featuring the removal of large middle-molecules up to a molecular weight of 45,000 Da typical of a high cut-off membrane, and the low removal of albumin as in state-of-the-art high-flux membranes. For uremic solutes in the 15,000- to 45,000-Da size range, MCO membranes offer improved clearance compared with that of high-flux membranes used in HD mode and equivalent clearance to that of high-flux membranes in high-volume HDF mode. The benefit is performance equivalent to that of high-volume HDF without a need for the online production of substitution fluid or for vascular access required for high blood flow rates. Therefore, the use of MCO membranes simplifies the delivery of high-removal treatment for ESRD patients, providing a removal spectrum that extends the current capabilities of the best therapy available. This improvement should allow clinicians to surpass the benefits provided by hemodiafiltration, which requires large amounts of high-quality fluid and a more complex setup, while utilizing regular HD equipment. These special MCO membranes should raise the standard of treatment available for all chronic HD patients, potentially decrease inflammatory responses, and generally improve patient outcomes. Future studies are planned to demonstrate the clinical benefits of this innovative product. Financial support: This work has been financed by Baxter International. Conflict of interest: All authors are employees of the Gambro Dialysatoren GmbH, Hechingen (Germany) or Gambro Lundia AB. Gambro AB (including all direct and indirect subsidiaries) is now part of Baxter International Inc. Baxter is a manufacturer of dialysis devices. None of the authors has proprietary interest. All experimental information is given in great detail to exclude any bias on the results.	{ "pile_set_name": "PubMed Central" }
Introduction {#Sec1} ============ Anthocyanin pigments are important flavonoid compounds that exhibit a wide range of biological functions in plants^[@CR1]--[@CR3]^, including as attractants for pollinators and seed dispersers and in protecting plants against abiotic and biotic stresses^[@CR4],[@CR5]^. More importantly, anthocyanins show beneficial effects to human health and exhibit potential protective functions against cancer and heart disease^[@CR6],[@CR7]^. These properties are partially attributed to their strong antioxidant capacity^[@CR8]^. According to studies in the model plant Arabidopsis thaliana, regulatory anthocyanin genes can be mainly divided into positive and negative regulatory genes based on whether they promote or inhibit the expression of structural genes for anthocyanin biosynthesis, respectively^[@CR9]^. Positive regulatory factors primarily include three types of genes: R2R3-MYB^[@CR10]^, basic helix-loop-helix (bHLH), and WD40 transcription factors^[@CR11]^. These genes promote the biosynthesis of anthocyanins. Additionally, there are two main types of negative regulatory genes, including the R3-MYB transcription factor^[@CR10],[@CR12]^ and the nitrogen-induced LBD transcription factor^[@CR13]^. The bioavailability and activity of anthocyanins and their regulatory genes vary widely across plants^[@CR5],[@CR14]^. "Zicaitai" (Brassica rapa L. ssp. chinensis var. utilis), a representative purple variety of B. rapa, exhibits purple petioles and flower stalks^[@CR15]^. The purple pigment of Zicaitai has been confirmed as anthocyanins^[@CR4]^. Although several studies have characterized anthocyanins in Brassica crops^[@CR16]--[@CR19]^, there is limited information on the genes involved in anthocyanin biosynthesis in Zicaitai. Quantitative trait locus (QTL)-Seq is a method that combines bulked segregant analysis (BSA) and high-throughput whole-genome re-sequencing to detect the major locus of a certain quantitative trait in a segregating population. BSA was proposed in 1991^[@CR20],[@CR21]^. It selects parents that show a contrasting phenotype on a trait of interest to build a segregating population---either F~2~ recombinant inbred lines, double haploid, or backcross populations---and then selects two groups of individual plants, each showing segregation of the trait to one of the parents, as two mixed pools to perform genotype analysis. The genomic position of the polymorphic molecular markers that shows significant segregation of genotypes is the region that harbors the major QTL. Currently, BSA has been updated to QTL-Seq through the replacement of traditional markers such as RAPD (random amplification polymorphic DNA) or RFLP (restriction fragment length polymorphism) to SNP (single nucleotide polymorphism) markers, accompanied by high-throughput re-sequencing and SNP-index analysis^[@CR22]^. QTL-Seq shows much higher efficiency than traditional QTL mapping, which is typically time-consuming and involves labor-intensive genotyping and maintenance of the mapping populations. However, QTL-Seq analysis always locates one locus for a trait at one time, while traditional QTL mapping may locate several loci for one trait in one experiment. QTL-Seq has been widely and successfully used in many crop populations, such as in the mapping of flowering traits in cucumber^[@CR23]^, tomato fruit weight^[@CR14]^, and 100-grain weight and root traits of chickpeas^[@CR24],[@CR25]^. In the present study, we performed QTL-Seq analysis and located different major loci to that obtained by linkage mapping using the same F~2~ population of Zicaitai and "Caixin". A simulated population was generated and analyzed, and the results supported that QTL-Seq and linkage mapping are able to detect different major loci under the combination of different modes of inheritance. We further assembled the draft-genome sequences of Zicaitai and Caixin. By combining the assembled sequences with the locus located by QTL-Seq, we determined the causal genes and their functional mutations responsible for anthocyanin enrichment in the Zicaitai accession. Materials and Methods {#Sec2} ===================== Plant materials and whole-genome re-sequencing {#Sec3} ---------------------------------------------- Zicaitai DH line ZCT095 was used as the receptor parent and Caixin DH line L58DH was used as the donor parent to construct an F~2~ population of 200 plants. The picture of the two parental lines have been shown in previous study^[@CR3]^. 30 plants exhibiting an extreme purple phenotype were selected from the population as the purple group, while 30 non-purple plants were selected as the green group. The frequency distribution on the anthocyanin accumulation in the F~2~ population were also shown in previous report, plants that have more than 0.4 mg/g dry weight of total anthocyanin content were selected and considered as the purple group. DNA from both parents and two groups of 60 samples was extracted from fresh leaves at the six-leaf stage. The DNA samples from the two groups were then combined into two pools for library construction. Following this, 100-bp pair-end reads were generated by the Illumina Solexa sequencing platform from BerryGenomics Biotech Co., Ltd. (Beijing, China). Variant calling with re-sequencing data {#Sec4} --------------------------------------- The paired-end reads of Zicaitai, Caixin, and the two sample pools were aligned to the B. rapa Chiifu reference genome version 3.0 using BWA^[@CR26]^ software with the method "mem"^[@CR27]^. Samtools^[@CR28]^ was then used to call the SNP and InDel variants from the aligned reads of Zicaitai and Caixin. Low quality (Q \< 10, DP \< 5), multi-allelic, or heterozygous variants in the VCF files of Zicaitai and Caixin were filtered out using Bcftools^[@CR29]^. Finally, SNP datasets were searched for polymorphic loci between Zicaitai and Caixin. Following this, the genotypes of these polymorphic loci were called out from the aligned reads of the two sample pools. Draft-genome assembly of the two parental lines {#Sec5} ----------------------------------------------- The adaptors, duplicates, and low-quality reads were filtered from the raw Illumina Sollexa sequencing reads produced from three DNA libraries with insert sizes of 180 bp, 350 bp, and 500 bp. The filtered datasets were then submitted for assembly into contig sequences using SOAPdenovo (Version r240)^[@CR30]^, implementing the default parameters. And then gaps in the assembly were closed using the tool GapCloser. Finally, the two parental genomes were assembled into scaffolds with N50 lengths of 4.5 kb and 6.5 kb, with a total length of 403.9 Mb and 350.8 Mb for Zicaitai and Caixin, respectively. The sequences of the draft assembly were deposited in BRAD database (<http://brassicadb.org/brad/datasets/pub/Zct_Cx/assemble/>). QTL-Seq analysis {#Sec6} ---------------- The SNP-index was calculated for all SNPs in the two pools of mixed samples, with the genotype of Zicaitai as the reference. During the calculation, we filtered SNPs with a SNP-index \<0.3 or \>0.7, which denotes a co-segregation of certain genotypes in the two pools. A 200 kb sliding window with a 20 kb increment was applied to slide across the genome, and the average value of the SNP-index was calculated in each window. The Δ(SNP-index) of Zicaitai (range from −1 to +1) was calculated by using the SNP-index of the purple pool minus the SNP-index of the green pool. We further repeated these calculations with the genotype of Caixin as the reference to obtain the Δ(SNP-index) of Caixin (range from −1 to +1). Finally, the Δ(SNP-index) of the population (range from −2 to +2) was calculated from the sum of Δ(SNP-index) of Zicaitai and Δ(SNP-index) of Caixin and plotted along the 10 chromosomes of B. rapa to show the signals detected by the QTL-Seq analysis. Linkage mapping {#Sec7} --------------- The software MapQTL (<https://www.kyazma.nl/index.php/MapQTL/>) version 6 was used to perform linkage analysis with genotype datasets from both populations of Zicaitai/Caixin and the simulation. The algorithm "MQM mapping" was selected to perform the calculation. The output logarithm of odds (LOD) scores were plotted along the genetic distances of the markers analyzed. Population simulation and analysis {#Sec8} ---------------------------------- The polymorphic positions between Zicaitai and Caixin in 10 chromosomes were used as coordinates to simulate two sets of genotype data. The two genotype datasets were considered as the homologous genotypes of two homozygous parental genomes. The combination of the two sets of genotypes (two parental haplotypes) resulted in the heterozygous status of a simulated F~1~ genome. Based on this F~1~ genome, we further simulated recombination events between the two parental haplotypes. Besides the random combinations of different chromosomes between the haplotypes, we simulated \~25 recombinations with 20% bias for both the pollen and egg haplotypes. It means 2.5 recombination events for each chromosome in average. We simulated 200 such haplotypes in total for both types. One egg haplotype was then randomly combined with one pollen haplotype to generate genotypes for the 200 F~2~ individuals. Six loci (different to that of Zicaitai and Caixin population) were further set to simulate trait-related genetic factors, with one locus showing dominant suppression (epistasis) on the other five loci. While the other five loci directly contribute to the phenotype level of the simulated trait and follow the recessive genetic model, with one locus designated as the major factor contributing 60% to the phenotype, and each of the other four loci contributing 10% to the phenotype. Based on these rules and the simulated genotype data of the 200 F~2~ individuals, we obtained the phenotype levels for all F~2~ individuals. These simulated data were then submitted to both QTL-Seq and linkage analysis. The perl script for the population simulation is available through the link: <http://brassicadb.org/brad/datasets/pub/Zct_Cx/simCode/>. Determination of candidate genes {#Sec9} -------------------------------- Based on the comparative genomic information between B. rapa and A. thaliana, we identified the pairwise syntenic-gene relationships between the two species. With the functional information of syntenic orthologs in A. thaliana, as well as our previous work on the determination of anthocyanin biosynthesis genes in B. rapa, we predicted the potential functions of candidate genes located in the QTL regions detected by QTL-Seq. We selected the intersection region of five sliding windows (200 kb sliding window with a 20 kb increment) that have the highest ∆SNP-index as the candidate interval, which are the peak signal in A07. Since the size of the sliding window was 200 kb and the increment was 20 kb, the intersection of five neighbor windows is 120 kb. The start and stop positions of the interval were then refined according to the locations of SNPs around its boundaries. It was found that there was one anthocyanin-synthesis related gene, BrMYBL2.1, located at this QTL region of A07. Experimental verification of a large-sequence insertion {#Sec10} ------------------------------------------------------- DNA was extracted from the fresh leaves of both Zicaitai and Caixin, and a pair of primers was designed to amplify the sequence that contains the insertion variant from the two DNA samples. The PCR system constituted 20 μL of PCR mixture with 2 μL of DNA (120 ng/μL), 10 μL of 2 × Rapid Taq Master Mix from the Vazyme Biotech Co., Ltd. (Nanjing, China), 0.8 μL (10 μM) of both forward and reverse primers, and 6.4 μL of ddH~2~O. The reaction mixture was incubated in a thermal cycler (9700, ABI, USA) at 95 °C for 3 min, followed by 35 cycles of 95 °C denaturation with 15 s for each cycle, following by primer annealing at 60 °C for 15 s, and a final extension at 72 °C for 5 min. The PCR products were further separated by electrophoresis on 1% agarose gel, running at 150V for 10 min, and then submitted to silver staining for band analysis. The PCR products were then sent to Majorbio Pharmaceutical Technology Co., Ltd. (Shanghai, China) for Sanger sequencing. The sequencing results were analyzed in BioEdit (<http://www.mbio.ncsu.edu/bioedit/bioedit.html>) and MUSCLE^[@CR31]^. Results {#Sec11} ======= BSA-Seq determines different major loci to those of linkage mapping {#Sec12} ------------------------------------------------------------------- An accession of Zicaitai was crossed with an accession of Caixin to create an F~2~ population of 200 individual plants showing segregation of the trait of anthocyanin enrichment (i.e., phenotypic variation in the purple color observed on the stem and leaf). We further selected two groups of F~2~ plants exhibiting an intense purple color, with one group containing 30 individuals showing intense purple and the other containing 30 samples lacking any purple. The DNA of the two groups of samples was extracted and combined into two respective pools. The two DNA pools together with two DNA samples of the two parents Zicaitai and Caixin were sequenced on the Illumina Solexa sequencing platform. The genomes of the two parents were sequenced to \~10× coverage for each (\~5-Gb Illumina re-sequencing data), while the two pools were sequenced to \~20× coverage for each (\~10-Gb data; Table [1](#Tab1){ref-type="table"}). Using the genome sequences of "Chiifu" as the reference, the reads were aligned and SNP and insertion/deletion (InDel) variants from the genomes of the two parents were called. Polymorphic loci of the SNPs and InDels between the two parents were selected for further analysis. The resequencing data of the two mixed pools were then aligned to the reference genome, and the genotypes of these polymorphic loci between the parents were called out from each of the two pools.Table 1The statistics on sequencing data generated for QTL-Seq and draft-genome assembly.StudiesSampleLibrary size (bp)Clean data (Gb)QTL-SeqZicaitai3505.31Caixin3505.42Pool 135010.99Pool 235011.01AssembleZicaitai18021.26Zicaitai5005.46Caixin18021.40Caixin5005.62 A significant signal was detected in chromosome A07, which differs from our previous linkage mapping results^[@CR3]^. A total of 1.19 million SNPs were identified between the two parental lines. In the F~2~ pool, approximately 1.02 million SNPs were detected at the polymorphic positions between the parental lines. Using the genotype datasets of the two parents and two pools, the SNP-index and Δ(SNP-index) were calculated to locate the QTL locus (see Methods). Figure [1](#Fig1){ref-type="fig"} shows a major locus that was detected in close proximity to the end of chromosome A07, which harbors the genetic factor that regulates the formation of extreme levels of purple color, i.e., contributes the major effect of anthocyanin enrichment. We previously performed linkage mapping with the same F~2~ population on the trait of anthocyanin variation and determined a major QTL on chromosome A09^[@CR3]^, which differs from the current major location identified by QTL-Seq. Additionally, a weak QTL signal was also detected in A07, suggests that these is a genetic factor located at A07 which also contributes to the anthocyanin enrichment in Zicaitai. Together with the information of physical positions of these genetic markers used in linkage mapping, we confirmed that the two methods located distinct major loci for anthocyanin enrichment in one F~2~ population---one locates at A07, while the other one locates at A09. The results suggest the existence of different genetic factors that may adopte different models of inheritance in the regulation of anthocyanin formation in this Zicaitai accession.Figure 1Major QTL loci of anthocyanin enrichment determined from an F~2~ population of Zicantai/Caixin. (a) QTL-Seq located at the major QTL locus at chromosome A07; the grey points are the Δ(SNP-index) of each SNP, the purple points are the average values of the Δ(SNP-index) in a 200-kb sliding window with a 20-kb increment across all chromosomes, and the red lines show the confidence intervals of the Δ(SNP-index). (b) The LOD scores generated by linkage analysis with the MEM algorithm using software MapQTL6.0. The major QTL locus was located at chromosome A09. Simulation experiments for assessing the inconsistency between the QTL-Seq and linkage analysis {#Sec13} ----------------------------------------------------------------------------------------------- We generated simulation data to explore the potential possibility of the inconsistent chromosomal locations detected by QTL-Seq and linkage analysis. Firstly, two sets of genotype data were simulated on the same positions of the polymorphic SNPs between Zicaitai and Caixin and were considered as genotypes of two in silico homozygous parents. Secondly, we simulated recombination events (cross-overs) between the two parental haplotypes (i.e., half of the homozygous genotypes), as occurs in the sex cells in F~1~ plants, with one paternal haplotype (pollen) or one maternal haplotype (egg) having \~25 cross-overs (\<0.2 bias). With this rule, we generated 200 such paternal and maternal haplotypes. Thirdly, we randomly selected one paternal haplotype and one maternal haplotype, and combined them into the genotype data of one simulated F~2~ plant---the process of selfing of F~1~ plants. We repeated this process and simulated 200 such F~2~ plants. Furthermore, a trait and its causal loci were simulated in one of the parents, and the information of their inheritance in the F~2~ population was surveyed and recorded for further mapping analysis. To design multiple factors regulated trait, we simulated six loci (Supplementary Table [S1](#MOESM1){ref-type="media"}) that were associated to the final trait, with five loci contributing directly to the level of the trait. Among the five loci, one contributed 60% to the final phenotype of the trait and followed the recessive genetic model, while each of the other four loci contributed 10% to the phenotype of the trait and also followed the recessive model. The last locus does not contribute directly to the trait, but shows epistatic effects on all the other five loci and follows the dominant genetic model. Under these rules, one of the parents shows 100% of the trait, while the other one shows 0% F~1~ also shows 0%, which is similar to the purple phenotype in the Zicaitai-Caixin population. Combining these rules with the simulated genotypes of the F~2~ population, we further calculated the trait level in the 200 simulated F~2~ individuals. We performed linkage mapping and QTL-Seq analysis with these simulated genotype datasets and the trait information, following the same analysis pipeline as that used in studies on the real F~2~ population of Zicaitai-Caixin. Figure [2a](#Fig2){ref-type="fig"} shows the inheritance patterns of the parental genomic fragments in each of the 200 F~2~ individuals. Linkage analysis identified a major locus on chromosome A02 (Fig. [2c](#Fig2){ref-type="fig"}), which corresponds to the simulated epistatic locus. Another smaller signal was detected on A04, which corresponds to the simulated major locus (60% contribution) that directly regulates the simulated trait. Additionally, we selected two groups from the F~2~ population, with one group of samples showing \>70% of the trait and the other showing 0% of the trait. Each group contained 25 samples and were recorded as a pool. We further simulated \~30 reads to cover the simulated SNP loci of the 25 samples in each pool and performed QTL-Seq analysis using the same method as that used in the real Zicaitai-Caixin population. The result showed that the QTL-Seq located a major locus at chromosome A04 (Fig. [2b](#Fig2){ref-type="fig"}), corresponding to the simulated major locus (60% contribution)---which is different to the major locus (chromosome A02) mapped by linkage analysis. This simulation analysis generated similar inconsistence between QTL-seq and linakge mapping analysis as that observed in the mapping of anthocyanin enrichment trait in the Zicaitai-Caixin population, supporting that there are two different loci following different genetic models relating to purple color regulation in Zicaitai.Figure 2QTL mappings in a simulated population with 200 F~2~ individuals. (a) The inheritance of parental genomic fragments in each of the F~2~ individuals; green fragments are from parent one, orange fragments are from parent two, while grey fragments are heterozygous genotype from both parents. (b) QTL--Seq analysis located the major QTL locus at chromosome A04. The colors denote similar objects as that in Fig. [1](#Fig1){ref-type="fig"}. The six black triangles denote the locations of the simulated loci. The bigger triangle at chromosome A04 is epistatic to the other loci. (c) Linkage analysis mapped the major QTL locus at chromosome A02. Draft-sequence assembly and fine-mapping of the causal gene {#Sec14} ----------------------------------------------------------- We screened candidate genes in the mapped regions with the annotation information and functional mutations of both genes. We previously predicted the anthocyanin biosynthesis genes in the genome of B. rapa by comparison with genes in A. thaliana^[@CR32]^. As reported in the linkage analysis, there are two such genes, BrEGL3.1 (BraA09g015240.3C) and BrEGL3.2 (BraA09g013410.3C), in the QTL region of A09^[@CR3]^ (Table [2](#Tab2){ref-type="table"}). They are two homeologs of the EGL gene, which is a positive regulator of anthocyanin biosynthesis in A. thaliana. Furthermore, we found one gene, BrMYBL2.1 (BraA07g036130.3C), located almost at the midpoint of the major locus in A07 (Table [2](#Tab2){ref-type="table"}), which was detected by QTL-Seq. BrMYBL2.1 is a homeolog of the MYBL2 gene, which, in contrast to EGL3, is a negative regulator of anthocyanin biosynthesis. We further assessed the sequencing depth of the BrMYBL2.1 gene to determine if there was any structural variant that could not be determined by variant calling based on the re-sequencing reads mapping. By comparing the differences in sequencing depth between the two pools of purple and non-purple plants, we detected that in the third exon of BrMYBL2.1, a position in the purple pool exists that shows much lower sequencing depth than the non-purple pool (Fig. [3a](#Fig3){ref-type="fig"}).Table 2Candidate anthocyanin-biosynthesis genes in major QTL loci.QTL lociCandidate genesChromosomeStartStopIDStartStopA0725,112,90325,230,669BrMYBL2.125,180,91825,181,774A097,230,3969,242,789BrEGL3.18,904,7648,908,450BrEGL3.27,821,1827,824,280Figure 3A large sequence insertion was found in gene BrMYBL2 in Zicaitai. (a) The depth the reads covered on each nucleotide around the genomic region of gene BrMYBL2; purple points denote the reads depth from the sequencing data of purple pool in the QTL-Seq analysis, while green points denote reads depth from that of the green pool. (b) The gene model of BrMYBL2 and the positions of a sequence insertion and four non-synonymous mutations in the third coding exon of this gene in Zicaitai. We assembled draft genome sequences of Zicaitai and Caixin to investigate the structural mutations in the candidate genes of anthocyanin enrichment between the parents. For both Zicaitai and Caixin, we generated \~80× coverage sequencing data in total (Table [1](#Tab1){ref-type="table"}). These data was used to assemble the two genomes. Finally, we obtained draft genome sequences for both Zicaitai and Caixin with assembled sizes of \~403-Mb and \~350-Mb, and scaffold N50 values of 4-kb and 6-kb, respectively (Table [3](#Tab3){ref-type="table"}). Although the two assembled genomes do not possess sufficient sequence contiguity, they are useful for the detection of structural variations in the coding sequence regions of the genes. Assembled sequences were then compared with candidate genes to assess for structural variations. We BLASTed the scaffolds of both Zicaitai and Caixin against the coding sequences of the candidate genes. The structural variations between Zicaitai and Caixin on each of these candidate genes were then observed on these aligned sequences. By doing this, we determined a large sequence insertion (with "N" as a gap in it) in the third exon of BrMYBL2.1 in Zicaitai compared to that of Caixin, which overlapped with the region that shows lower re-sequencing depth in the purple pool than the non-purple pool (Fig. [3b](#Fig3){ref-type="fig"}). No such structural variations were found in the other candidate genes.Table 3The draft-genome-assembly information of the two parents.SampleAssembled size (Mb)Max length (bp)N50 (bp)Zicaitai403.9598,1824,576Caixin350.8195,9176,543 Mutation verification through Sanger sequencing {#Sec15} ----------------------------------------------- To verify the insertion variation in BrMYBL2.1 of Zicaitai, we performed polymerase chain reaction (PCR) amplification of the sequence containing the insertion. The electrophoresis products indicated that the size of the amplified sequences differed between the two parents. The sequence amplified from Zicaitai was \~100-bp longer (estimated based on the size of the DNA ladder) than that of Caixin (Fig. [4a](#Fig4){ref-type="fig"}), which is consistent with previous results of the comparisons between the assembled sequences. The PCR products were further submitted to Sanger sequencing, and the results showed that there was a sequence insertion containing "GGGAATCGATCCAACTTTGTTTC" linked with poly-"A"s to the gene in Zicaitai (Fig. [4b](#Fig4){ref-type="fig"}) in comparison with Caixin. This confirmed the sequence insertion detected by the sequence assembly. Sanger sequencing peaks denoting different types of nucleotides were observed to be overlapping with each other following the poly-A and could not be easily recognized (high level of heterozygous peaks) for determining the accurate size and sequences of the insertion variation. However, the \~100-bp sequence insertion in the third coding exon of BrMYBL2.1 changed the coding protein in Zicaitai. The gene BrMYBL2.1 in chromosome A07 is a negative regulator of anthocyanin biosynthesis, while BrEGL3.1, located at the QTL region of A09, is a paralog of EGL3, which is a positive regulator of anthocyanin biosynthesis. The functional mutations of the two genes both contributed to the variation in anthocyanin enrichment of Zicaitai. The different roles of the two genes in the biosynthesis of anthocyanins explains the different results of the QTL-Seq and linkage mapping, even when using the same population.Figure 4Experimental verification of a sequence insertion in gene BrMYBL2. (a) The electrophoretogram of the PCR products of the target sequence that contains the insertion variant. Three replicates for both Caixin and Zicaitai were labeled as one to three. (b) The alignment of the target sequences from the Chiifu reference genome (Chiifu_V3.0), draft sequence assembly of Caixin (Caixin-ass) and Zicaitai (Zicaitai-ass), and Sanger sequencing results (Caixin-F and Caixin-R, Zicaitai-F and Zicaitai-R, F denotes forward sequence, R denotes reverse sequence). The red color denotes the inserted sequence. Discussion {#Sec16} ========== QTL-Seq and linkage analysis using the same segregating population can identify different major loci. In this work, we performed QTL-Seq analysis on the purple phenotype (i.e., the trait of anthocyanin enrichment) in an F~2~ population of Zicaitai-Caixin and compared the result with our previous linkage-analysis study using the same population. The two methods located different major anthocyanin enrichment loci in Zicaitai, which indicated that two major genetic factors contribute to the trait and have different roles and/or follow different genetic models in the regulation of anthocyanin enrichment. In order to test this, we generated a simulated F~2~ population with two major loci regulating a trait under different genetic models. The first directly contributes to the phenotype under a recessive model, while the second is epistatic to the first locus and follows a dominant model. QTL-Seq and linkage mapping were then conducted on the simulated population using the same pipeline as that applied to that of the Zicaitai-Caixin population. The results showed that QTL-Seq and linkage mapping located different major loci from that observed in the Zicaitai/Caixin population. QTL-Seq located the major locus that regulates the trait directly, while the linkage mapping located the locus that is epistatic to the previous locus. It is reasonable that the sampling of two pools with opposite and extreme traits will cause the QTL-Seq to detect a locus associated with a high level of the trait, while linkage analysis will locate the gene that shows a clearer inheritance pattern. Inspired by the simulation analysis, we speculated that these two major factors identified through QTL-Seq and linkage mapping in Zicaitai-Caixin F~2~ populations should have different regulation pathways and/or follow different genetic models relating to anthocyanin biosynthesis. Actually, the BrMYBL2 gene detected by QTL-Seq function as a negative regulator and may contribute directly to the purple trait in Zicaitai. While in the linkage analysis, the positive regulator gene EGL3 with potentially epistatic function was localized to mediate the anthocyanin biosynthesis. The complexity of the regulation mechanisms on anthocyanin biosynthesis in Brassicas were highlighted by recent studies. Many efforts have been devoted to dissect the genes that contribute to the purple traits in different Brassica crops. In Brassica napus, BnAPR2, encoding an adenosine 5'-phosphosulfate reductase, at the end of A03 chromosome was identified as an incomplete dominant regulatory gene through map-based cloning^[@CR33]^. Another gene BrMYB73 was also mapped to the end of A03 in B. rapa^[@CR34]^. BrMYB73 was predicted to encode a R2R3-MYB transcription factor, and one deletion and one SNP were found in this gene in purple-leaf parent. Recently, Li et al. identified a bHLH49 transcription factor (BrbHLH49) on A07 that might positively regulate the anthocyanin accumulation in Zicaitai based on a specific-locus amplified fragment sequencing method^[@CR19]^. These studies revealed the genetic diversity of anthocyanin biosynthesis related genes in Brassica crops. QTL mapping together with draft genome assembly is an efficient means of fine-mapping causal genes and mutations from segregating populations. In QTL analysis, considering the limited recombination events in a given population, the mapping resolution is not always sufficient to locate the causal genes, and dozens or hundreds of candidate genes can make the fine-mapping of the gene difficult. Furthermore, with variants called from the re-sequencing data, only small-scale variants, such as SNPs and short InDels, can be analyzed in these candidate genes, which results in the major loss of large-scale functional mutations during the screening of candidate genes. However, the draft genome assembly of parental genomes and pairwise comparisons of coding sequences of candidate genes between parents will capture these large-scale mutations, thus resulting in the rapid fine-mapping of the causal genes and mutations of the traits. In this work, we performed draft genome assembly of Zicaitai and Caixin. Together with the mapping results from the QTL-Seq, we located a large sequence insertion in the coding exon of an anthocyanin biosynthesis-related gene. The insertion changed the sequence of the translated protein of the genes in Zicaitai in comparison to Caixin. Considering that high-throughput sequencing is becoming increasingly affordable, QTL mapping accompanied by the draft sequence assembly of parental genomes could constitute an efficient means of fine-mapping the genes in segregation populations. Low-depth re-sequencing of individual samples from a segregating population is a superior option for analyzing a trait that is regulated by more than one major locus. In this study, we located two genomic regions that contribute to the enrichment of anthocyanins in Zicaitai; one being a positive regulator of anthocyanin biosynthesis and the other being a negative regulator of anthocyanin biosynthesis. The two genes have different roles in anthocyanin enrichment in Zicaitai. In similar cases, QTL-Seq as well as linkage mapping may not be able to locate all of these major loci in one experiment. Therefore, we propose a strategy of low-depth re-sequencing of all F~2~ individuals. With the re-sequencing data, a genetic map with a high density of bin markers (combined markers in a local region) can be constructed, allowing for efficient and simple linkage mapping. More importantly, different samples can be repeatedly combined into two extreme pools based on different rules, and QTL-Seq can be performed as many times as needed. This strategy should help to locate multiple major loci that contribute to one or more traits in a segregating population. Supplementary information ========================= {#Sec17} Supplementary Table S1. Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Xin Zhang and Kang Zhang. Supplementary information ========================= is available for this paper at 10.1038/s41598-020-58916-5. The work was supported by the National Program on Key Research Project (2016YFD0100307), the National Natural Science Foundation of China (NSFC grants 31301771 and 31722048), the Science and Technology Innovation Program of the Chinese Academy of Agricultural Sciences, and the Key Laboratory of Biology and Genetic Improvement of Horticultural Crops, Ministry of Agriculture, P.R. China. F.C. and X.W. designed the experiments. F.C. and X.Z. analyzed the data. X.Z., J.W., N.G. and J.L. carried out the experiment. K.Z. participated in data analysis and interpretation. F.C., X.Z. and K.Z. wrote this manuscript. The authors declare no competing interests.	{ "pile_set_name": "PubMed Central" }
Introduction ============ Severe Acute Respiratory Syndrome (SARS)-CoV and Middle East Respiratory Syndrome (MERS)-CoV are two closely related zoonotic coronaviruses. Both have successfully crossed the species barrier to allow animal-to-human transmission, and further to allow human-to-human transmission ([@evw246-B42]; [@evw246-B37]). The SARS outbreak in 2003 had a mortality rate of 10% ([@evw246-B1]), and SARS-CoV was considered the most aggressive coronavirus compared to other human coronaviruses that commonly cause mild to moderate infection in their hosts ([@evw246-B45]). MERS-CoV is the cause of an ongoing outbreak of the respiratory illness MERS ([@evw246-B10]). At the time of writing, 1791 MERS cases have been confirmed with a mortality rate of approximately 35% ([@evw246-B48]). Both MERS and SARS have higher mortality rates in elderly and immunosuppressed populations ([@evw246-B19]). The host changes by MERS-CoV and SARS-CoV suggest that other coronaviruses can potentially cross the species barrier, become zoonotic, and enable human-to-human transmission, ultimately causing high morbidity and mortality. SARS-CoV and MERS-CoV exploited mechanistically different approaches to overcome the human species barrier, but these two viruses have a lot in common ([@evw246-B29]). Here, we aim to identify the vulnerable regions in the proteomes of coronaviruses that neither SARS-CoV nor MERS-CoV nor their contemporary and forthcoming relatives can proliferate without, and address how to mobilize a defense against the present and future coronaviruses by targeting these regions. SARS-CoV and MERS-CoV are positive (+)-strand RNA viruses encoding approximately 25 protein products. The MERS-CoV proteome is primarily composed of two polyproteins, ORF1a and ORF1ab; the latter is generated by a -1 ribosomal slippage frameshift. These proteins are cleaved into 16 nonstructural proteins (NSPs). NSPs 1--10 are products of both polyproteins, whereas NSPs 12--16 are only yielded by ORF1ab. NSP11 is unique to ORF1a ([@evw246-B44]). Structural proteins envelope (E), spike (S), membrane (M), and nucleocapsid (N) are elements of the physical structure that encloses the viral genome and come from distinct reading frames, unlike ORF1a and ORF1ab, which come from overlapping reading frames. Additionally, the structural proteins are the product of subgenomic mRNAs that are joined during discontinuous negative RNA strand synthesis ([@evw246-B44]). Finally, NS3 protein (NS3), NS4A protein (NS4A), NS4B protein (NS4B), NS5 protein (NS5), and Orf8b protein encompass the remainder of the proteome and also arise from distinct reading frames ([@evw246-B44]). Our approach utilizes genomic sequence data, which is readily available for viruses known to cause disease. However, because most viruses pose no major threat to their host, they pass by unnoticed leaving the majority of virus genome space uncharted. With the availability of cost-efficient genome sequencing technology, and recent developments in the field of viral metagenomics, large-scale identification of viral genome space is on the rise ([@evw246-B39]; [@evw246-B34]). By exploring viral diversity, critical components constituting a viral genus' fitness can be evaluated. Examples such as the common influenza virus illustrate the rapidity of viral gene mutation and in order to maintain immune protection, an annual flu vaccination is recommended. Underway efforts aim to generate broadly neutralizing vaccines whose design accounts for the genomic sequences of multiple types of influenza virus to eliminate frequent re-vaccination against the flu ([@evw246-B16], [@evw246-B17]). Development of broadly neutralizing vaccines often relies on the consensus or ancestral sequences of extant viral sequences in order to provide greater coverage for related viruses ([@evw246-B26]). Unfortunately, consensus sequences can be misleading, and ancestral sequence reconstruction is error-prone for quickly diverging sequences ([@evw246-B32]). In addition, viruses with compact genomes often express proteins with structural disorder that may undergo structural transformations. Although these transformer proteins, like VP40 in Ebola, are masters at changing their structure, and thus expanding their functional repertoire as needed for the life cycle of the virus ([@evw246-B4]), flexible regions are potentially important in rewiring protein--protein interactions between the virus and its host ([@evw246-B5]; [@evw246-B35]; [@evw246-B18]). The flexibility trait of many viral proteins is a complicating factor in vaccine development. For instance, Dengue virus exhibits serotype-specific antibody affinity that causes antibody-dependent enhancement, an obstacle in the development of Dengue vaccines that protects against all four serotypes ([@evw246-B15]). To overcome the hurdle posed by structural flexibility, we propose an additional screening step in identifying potential vaccine or antiviral targets that considers the structural flexibility of the viral proteins. The Structural Genomics Initiatives increased their success rate by excluding proteins predicted to be structurally disordered ([@evw246-B41]). A similar approach can perhaps benefit vaccine development. Furthermore, to make this approach robust to potential mutations, minimizing loss in efficacy or resistance, the evolutionary context of sequence and structure must be considered. Thus, we suggest expanding the concept of broadly neutralizing vaccines/antivirals by increasing the diversity of viruses considered if possible. Sites conserved for sequence, structure, and with low disorder propensity among diverse virus protein homologs are very likely to be constrained from 1) changing sequence on evolutionary time scales and 2) undergoing real-time structural transitions. These sites have potential as targets for broad-specificity antivirals or vaccines because conservation makes them broad-specificity and low dynamics avoids targeting a conformational ensemble, which is not only difficult ([@evw246-B51]), but that may change as the sequence diverges ([@evw246-B40]). A recent large-scale study of structural disorder in \>2,000 viral genomes in 41 viral families found the amount of disorder in different virus families varying from 2.9% to 23.1% ([@evw246-B36]). It was reported that Coronaviridae has very low disorder content (mean disorder 3.68%) ([@evw246-B36]). Coronaviridae contains two subfamilies: Coronavirinae and Torovirinae. SARS-CoV and MERS-CoV are part the Coronavirinae subfamily, from here on referred to as coronavirus (CoV). The lack of disorder is intriguing because it may be important for rewiring interactions between viral proteins and host proteins ([@evw246-B35]) and providing opportunities to acquire novel functional sequence motifs ([@evw246-B18]). Structural disorder has also been proposed to be important for viral viability, enabling multifunctionality and vigor in response to changes in the environment ([@evw246-B49]). Given the low fraction of structural disorder reported across Coronaviridae, we set out to investigate the conservation of structural disorder and secondary structure across CoV. Sites identified as conserved for structure and lacking disorder can be considered to be vulnerable and druggable in the proteomes of coronaviruses. The structural divergence capacity of these regions is limited, leaving a wider range of the present and emergent coronaviruses susceptible to the effects of potential broadly neutralizing anti-CoV therapies targeting these sites. We will refer to these sites as target sites. Materials and Methods ===================== Protein Family Reconstruction ----------------------------- Protein sequences were identified by individual BLAST searches with MERS-CoV (Taxonomy ID: 1335626) proteins ORF1ab (YP_009047202.1; polyprotein), S protein (YP_009047204.1), M protein (YP_009047210.1), E protein (YP_009047209.1), and N protein (YP_009047211.1) against coronaviruses. BLAST searches of the ORF1ab protein were performed, using start and end positions as detailed in the ORF1ab NCBI Reference Sequence file, against the refseq_protein database. The sequences retrieved from the BLAST output maintained the following cutoff: \>30% sequence identity and \>50% coverage relative to MERS-CoV sequence query. The 30% sequence identity and 50% query coverage cutoff strikes a balance between alignment quality and at least 10 sequences for most protein families. NSP1 (YP_009047202.1; 1-193), NSP2 (YP_009047202.1; 194-853), NS3 (YP_009047205.1), NS4A (YP_009047206.1), NS4B (YP_009047207.1), NS5 (YP_009047208.1), ORF8b protein (YP_009047212.1), and NSP11 (YP_009047203.1; 4378-4391) are not included in this study due to \<10 BLAST hits. Multiple sequence alignments were constructed for the selected BLAST hits using MAFFT ([@evw246-B25]). Phylogenetic trees were constructed using MrBayes 3.2.2 with a four category gamma distribution and the mixed model for amino acid substitution ([@evw246-B20]; [@evw246-B38]). Each tree ran for five million generations, with a sample frequency of 100. The final tree was constructed from the last 75% of samples, discarding the first 25% of samples as the default burnin, and using the half-compatible parameter, to avoid weakly supported nodes (i.e., with a posterior probability \<0.5). All trees were midpoint rooted. For every protein family, the amino acid substitution rate per site in its multiple sequence alignment was calculated using empirical Bayesian estimation as implemented in Rate4Site ([@evw246-B31]). Substitution rates were calculated using 16 gamma categories, the JTT substitution matrix ([@evw246-B24]), and the reconstructed phylogenies. The rates were normalized per protein family with an average across all sites equal to zero and SD equal to 1. This means that sites with a rate \<0 are evolving slower than average, whereas sites with a rate \>0 are evolving faster than average. Prediction of Intrinsic Disorder Propensity and Secondary Structure ------------------------------------------------------------------- Intrinsic disorder propensity was inferred using two different predictors: IUPred (default settings; "long" option) ([@evw246-B11], [@evw246-B12]) and DISOPRED2 ([@evw246-B46]) for all proteins. For IUPred, the site-specific continuous disorder propensities for each protein were mapped onto their corresponding position in the multiple sequence alignment as raw disorder propensities and as binary states, order or disorder, using two cutoffs of 0.4 and 0.5. Disorder propensities below the cutoff were assigned order and disorder propensities at the cutoff or above were assigned disorder. For the DISOPRED2 predictions that were inferred using the nr database, the continuous disorder propensities for every site in a protein were mapped onto their corresponding position in the multiple sequence alignment as raw disorder propensities and as binary states, order or disorder, using a cutoff of 5. Consequently, for every protein family (a multiple sequence alignment and its corresponding phylogenetic tree), two continuous matrices and three binary matrices resulted: IUPred 0.4, IUPred 0.5, and DISOPRED2. An additional matrix was generated to indicate sites where the binary order and disorder assignments differ between IUPred 0.4 and DISOPRED2. A similar methodology was employed to analyze secondary structure predicted by PSIPRED ([@evw246-B33]) and JPred ([@evw246-B13]). For both predictors, the uniref90 database was used and sites were classified as loops, alpha helices, or beta strands and mapped back onto their corresponding sites in the multiple sequence alignment. This resulted in two three-state matrices for each protein family alignment, one for each predictor, and two binary matrices displaying secondary structure elements (alpha helix and beta strand) or loops. An additional matrix was generated to indicate sites where the secondary structure assignments differ between PSIPRED and JPred. For every protein family, the binary matrices resulting from the different disorder predictions and from the different secondary structure predictions were analyzed in the corresponding evolutionary context using GLOOME. GLOOME (Gain-Loss Mapping Engine) analyzes binary presence and absence patterns in a phylogenetic context ([@evw246-B8]). In this study, the Rate4Site option in GLOOME was used to analyze the binary matrices (IUPred 0.4, IUPred 0.5, DISOPRED2, PSIPRED, and JPred) with the corresponding phylogenetic trees to map change of state across sites in each individual protein phylogeny ([@evw246-B9]; [@evw246-B8]). GLOOME was run with 16 gamma categories and a substitution matrix set to equal rates within each state and transitions between states treated equally. From the binary disorder and order matrices, transition rates between disorder and order or vice versa (DOT) were estimated. From the binary structure and loop matrices, transition rates between structure and loop or vice versa (SLT) were estimated. Similar to Rate4Site, the rates were normalized per protein family with an average across all sites equal to zero and SD equal to 1. This means that sites with a rate \<0 are evolving slower than average, while sites with a rate \>0 are evolving faster than average. Protein Family Visualization ---------------------------- Protein families were visualized in an integrative manner with a phylogenetic tree, any matrix (multiple sequence alignment or predictor based) displayed as a heatmap, and site-specific sequence transition rates using Python packages ETE3 ([@evw246-B21]) and Matplotlib ([@evw246-B22]). Statistical Analysis of Amino Acid Evolutionary Rate Distributions ------------------------------------------------------------------ Amino acid evolutionary rates (SEQ) for all sites across all alignments were aggregated and binned into four possible categories characterized by the distribution of PSIPRED predicted secondary structure at each site. Sites predicted to have a loop across all sequences are "conserved loops; C(L)" and sites predicted to have a helix across all sequences or a strand across all sequences are "conserved helix-strand; C(HS)" ([table 3](#evw246-T3){ref-type="table"}). Sites predicted to have all three states (helix, strand, and loop) or any combination of loop and one other state are "non-conserved helix, loop, strand; NC(HLS)" and sites predicted to have a mixture of helix and strand are "non-conserved helix-strand; NC(HS)" ([table 3](#evw246-T3){ref-type="table"}). In all cases, gaps were ignored when classifying combinations of secondary structure at a site or if secondary structure conservation exists at a particular site. Results ======= Phylogenies ----------- Phylogenies were built for all protein products encoded in the MERS-CoV single-stranded RNA genome, except for NSP1, NSP2, NS3, NS4A, NS4B, ORF8b protein, and NSP11, all of which had insufficient sequence data (\<10 sequence hits with BLAST). NSP12 is often used as a measure for newly identified coronaviruses. According to the International Committee of Taxonomy of Viruses, a major criterion in determining if a coronavirus is considered novel is pairwise sequence identity below 90% for NSP12 in all comparisons to previously known coronaviruses ([@evw246-B3]). Four main clades, alphacoronavirus, betacoronavirus, gammacoronavirus, and deltacoronavirus ([fig. 1](#evw246-F1){ref-type="fig"}), are identified in agreement with the taxonomic classifications described by the ICTV ([@evw246-B23]). Coronaviruses not listed by the ICTV are assumed to be a part of the clade in which representatives with known classifications are situated in our NSP12 phylogeny. ![CoV representative phylogeny. (A) NSP12 is a representative for the CoV protein phylogenies, colored by clade (alphacoronavirus or FIPV, green; betacoronavirus has four different subclades: SARS, blue; MERS, gray; HKU1, pink; EQU, purple; gammacoronavirus or SW1, yellow; deltacoronavirus or HKU19, white.) Posterior probability indicating node support is shown in red. (B) Protein family distribution across coronavirus based on the given cutoff (\>30% sequence identity and \>50% coverage relative to MERS-CoV sequence query). Clade color applied throughout the remaining figures. Areas shaded in gray with an arrow indicate that the protein family is not identified for that clade with the given cutoffs, but is found from the arrow tip. See [Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online, for the remaining phylogenetic trees.](evw246f1p){#evw246-F1} The MERS clade and SARS clade are sister clades in the NSP12 phylogeny. The HKU1 clade and EQU clade are also sister clades. Together these four clades form the Betacoronavirus clade, in accordance with the ICTV classification ([@evw246-B23]). Betacoronavirus is represented in all phylogenies although the order of the individual subclades varies. Alphacoronavirus is often found as the sister clade or outgroup to betacoronavirus. Deltacoronavirus or gammacoronavirus are the most distantly related to the betacoronavirus. In the nucleocapsid phylogeny, gammacoronavirus is the first outgroup clade to betacoronavirus, and alphacoronavirus is the most distant outgroup. Most NSP trees exhibit some unresolved nodes at junctures immediately preceding terminal nodes. As an effect of the 50% majority rule, most of the 546 resolved nodes are well supported with posterior probability \>0.9 for 82% and \>0.99 for 68% ([Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online). Most trees follow the NSP12 topology for the main clades, with minor clade rearrangements. It should be noted that for NSP5, the entire alphacoronavirus clade is placed within the betacoronavirus clade, as a sister clade to the MERS clade ([Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online). This may be due to increased sequence divergence rates or due to recombination. Recombination events are rather frequent in coronaviruses ([@evw246-B43]), and the MERS clade potentially underwent multiple recombination events as part of the host change ([@evw246-B52]). The phylogenies for membrane protein, spike protein, NSP5, and NSP8--NSP16 demonstrate (with the given BLAST cutoffs) recoverable protein homologs such that all coronaviruses are represented (i.e., all coronaviruses represented in the NSP12 phylogeny). Nucleocapsid, NSP4, and NSP7 have recoverable homologs in all clades except deltacoronavirus. NSP3 and NSP6 homologs are too divergent in deltacoronavirus and/or gammacoronavirus relative to MERS-CoV. Envelope appears specific to betacoronavirus ([fig. 1](#evw246-F1){ref-type="fig"}), but it is a short protein that has been found to diverge rapidly and is likely present outside betacoronavirus ([@evw246-B14]). Because different protein families yield slightly different phylogenies, for the remaining evolutionary analyses, every protein family was analyzed in the context of its own phylogeny. Intrinsic Disorder Is Rarely Conserved -------------------------------------- For all protein families, structural disorder propensities were predicted using IUPred ([@evw246-B11], [@evw246-B12]) and DISOPRED2 ([@evw246-B46]). To verify the robustness of the binary IUPred and DISOPRED2 predictions, the binary assignments were compared on a site-by-site basis ([table 1](#evw246-T1){ref-type="table"}). When converted to binary (i.e., two states per site disordered or ordered) IUPred 0.4 and IUPred 0.5 are in good agreement with the larger differences seen for NSP8, NSP9, and nucleocapsid (7.5%, 6.5%, and 19.0%, respectively) ([table 1](#evw246-T1){ref-type="table"}). Comparing IUPred 0.4 or IUPred 0.5 to DISOPRED2, large differences are in particular seen for nucleocapsid (38.7% and 29.7% respectively) and NSP8 (23.5% and 25.9%, respectively) ([table 1](#evw246-T1){ref-type="table"}). For nucleocapsid, regions that are found to be disordered by IUPred 0.4 are found to be ordered by IUPred 0.5 and DISOPRED2 ([fig. 2](#evw246-F2){ref-type="fig"} and [Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online). For NSP8, regions that are only slightly disordered in a few sequences according to IUPred 0.4 and IUPred 0.5, DISOPRED2 predicts disorder to be conserved for all sequences ([fig. 3](#evw246-F3){ref-type="fig"}). ###### Protein Family Wide Disagreement of Disorder and Secondary Structure Predictions --------------------- ![](evw246ie1p.jpg) --------------------- Tukey boxplot constructed using the IUPred 0.4 predicated disorder fraction (number of disordered sites/total sites) per sequence per protein. Green dots represent outliers; red diamond are the mean and red lines are the median values. ![The evolutionary context of intrinsic disorder in nucleocapsid. The phylogenetic tree was built using the multiple sequence alignments for nucleocapsid. Here, the multiple sequence alignment is colored by disorder propensity (with gaps in gray): (A) IUPred 0.4, blue-to-white-to-red shows disorder propensity according to the scale for IUPred 0.4. (B) IUPred 0.5, blue-to-white-to-red shows disorder propensity according to the scale for IUPred 0.5. (C) DISOPRED2, blue-to-white-to-red shows disorder propensity according to the scale for DISOPRED2. Above the heat maps, the normalized evolutionary rates per site for amino acid substitution (SEQ) and the DOT for the binary transformations of A--C are shown. Heat maps visualized with the Python packages ETE3 ([@evw246-B21]) and Matplotlib ([@evw246-B22]).](evw246f2p){#evw246-F2} ![The evolutionary context of intrinsic disorder in NSP8. The phylogenetic tree was built using the multiple sequence alignments for NSP8. (A) The multiple sequence alignment is colored by amino acid according to scale, arranged based on TOP-IDP disorder promoting propensity of the amino acids ([@evw246-B7]), and gray denotes gaps. (B) IUPred disorder propensity per site in the multiple sequence alignment. Blue-to-white-to-red shows disorder propensity according to the scale for IUPred 0.4. (C) IUPred disorder propensity per site in the multiple sequence alignment. Blue-to-white-to-red shows disorder propensity according to the scale for IUPred 0.5. (D) DISOPRED2 disorder propensity per site in the multiple sequence alignment. Blue-to-white-to-red shows disorder propensity according to the scale. Above the multiple sequence alignment, the normalized evolutionary rates per site for amino acid substitution (SEQ) and the DOT for the binary transformations of B--D are shown. Heat maps visualized with the Python packages ETE3 ([@evw246-B21]) and Matplotlib ([@evw246-B22]). See [Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online for additional graphics for every protein family.](evw246f3p){#evw246-F3} To quantify the fraction of disordered sites per protein family, we report the IUPred 0.4 results only for simplicity ([table 1](#evw246-T1){ref-type="table"}). In general, IUPred 0.4 predicts more disorder than DISOPRED2, but several protein families have almost no disordered sites. NSP3 and NSP8-10 have some variation in disorder content for different viruses. Based on the fraction of disorder, nucleocapsid is the only highly disordered protein among the CoVs in this study, even if NSPs 8--10 have outliers that are \>20% disordered. To compare the disorder-to-order transition rates (DOT) for all protein families where the binary matrices of disorder and order include both states, the quadrant count ratio (QCR) was estimated as a measure of association in assigning slower than average vs. faster than average transition rates. For IUPred 0.4 vs. IUPred 0.5, for IUPred 0.5 vs DISOPRED2, and for IUPred 0.4 vs. DISOPRED2, the QCRs were 0.76, 0.69, and 0.63, respectively. This shows a strong positive association for site-specific DOT for all methods and cutoffs, with IUPred 0.4 vs. IUPred 0.5 being the strongest ([table 2](#evw246-T2){ref-type="table"}). For nucleocapsid and NSP8, the positive associations are weaker, suggesting that many sites have IUPred disorder propensity in the 0.4 to 0.5 range and large differences between IUPred and DISOPRED2, in accordance with the large disagreement between the binary assignment of these predictors ([tables 1](#evw246-T1){ref-type="table"} and [2](#evw246-T2){ref-type="table"}). ###### QCR[^a^](#evw246-TF2){ref-type="table-fn"}for DOT and SLT Rate -------------- -------------------------------------------- ------ ------ ------ NSP3 0.75 0.68 0.61 0.51 NSP4 N/A[^b^](#evw246-TF3){ref-type="table-fn"} N/A 0.58 0.61 NSP5 0.72 0.84 0.73 0.65 NSP6 N/A N/A N/A 0.70 NSP7 N/A N/A 0.9 0.66 NSP8 0.86 0.43 0.38 0.67 NSP9 0.76 0.55 0.62 0.67 NSP10 0.93 0.67 0.68 0.42 NSP12 0.96 0.89 0.86 0.57 NSP13 0.76 0.70 0.59 0.36 NSP14 0.87 0.85 0.8 0.58 NSP15 0.82 0.85 0.67 0.53 NSP16 0.93 0.86 0.85 0.59 Envelope N/A N/A 0.58 0.53 Membrane 0.54 0.57 0.67 0.55 Nucleocapsid 0.33 0.43 0.31 0.61 Spike 0.81 0.62 0.49 0.55 All 0.76 0.69 0.63 0.55 QCR: Quadrant Count Ratio measures the association for the same site-specific rate with different predictors or cutoffs. N/A: at least one of the rates in the comparison could not be estimated due to the lack of any disordered state in the binary state matrix ([Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online). Secondary Structure Prediction and Structure-to-Loop Transitions ---------------------------------------------------------------- For all protein families, secondary structure elements were predicted using PSIPRED ([@evw246-B33]) and JPred ([@evw246-B13]). For most protein families, the disagreement between secondary structure predictors is greater than for the disorder predictors ([table 1](#evw246-T1){ref-type="table"}). In fact, 15 of the 17 protein families compared disagree at more than 10% of alignment sites, and two of these disagree at more than 20% of sites. To compare the binary structure-to-loop transitions (SLT), QCR was estimated as a measure of association for SLT based on the different predictors. In general, there is a moderate positive association between SLT for PSIPRED vs. SLT for JPred that is weaker than for the different DOT comparisons ([table 2](#evw246-T2){ref-type="table"}). It should be noted that SLT does not differentiate between alpha helix and beta strand, but considers both as "structure." This is a correct assumption if protein structure is conserved and consistently predicted, but for some protein families that is not the case. Four protein families (NSP3, NSP12, NSP13, and SPIKE) have more than 40% of their sites found within the NC(HLS) category with non-conserved helix, strand, and loop (two or three states present at the same site) ([table 3](#evw246-T3){ref-type="table"}). For NSP13, JPred predicts 72% of all sites to be a mixture of helix, strand, and loop, or any combination of loop and one other structural element ([fig. 4](#evw246-F4){ref-type="fig"}). Envelope and NSP6 have 13% and 12% of their respective sites in the NC(HS) category. Considering only the PSIPRED predictions, the NC(HS) category has 245 sites across all 17 protein families. That is one-tenth the size of the next smallest set which is C(HS) with 2275 sites. Next, C(L) has 3344 sites, and the largest category is NC(HLS) with 4257 sites. Comparing the evolutionary sequence rates for the sites in the different categories, based on PSIPRED predictions only, reveals that sites in the C(HS) category are evolving at a slower rate than all other categories. NC(HS) is only just significantly different (P = 4.62E−03) from C(HS), and is not significantly different from NC(HLS) and C(L) (P = 1.85E−02 and P = 8.33E−01, respectively). However, NC(HLS) and C(L) are significantly different from each other, and both are significantly different from C(HS) (P = 1.82E−46 and P = 2.33E−21, respectively) ([fig. 5](#evw246-F5){ref-type="fig"}). ![The evolutionary context of secondary structure in NSP13. The phylogenetic trees were built using the multiple sequence alignments for NSP13. (A) The multiple sequence alignment is colored as in [fig. 3](#evw246-F3){ref-type="fig"}. (B) PSIPRED secondary structure prediction per site in the multiple sequence alignment, color coded according to the scale. (C) JPred secondary structure prediction per site in the multiple sequence alignment, color coded according to the scale. Above the multiple sequence alignment, the normalized evolutionary rates per site for sequence substitution (SEQ) and SLT based on the binary transformations of B-C are shown. Heat maps visualized with the Python packages ETE2 ([@evw246-B21]) and Matplotlib ([@evw246-B22]). See [Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online for a complete set of graphics for every protein family.](evw246f4p){#evw246-F4} ![Comparison of SEQ at sites characterized by secondary structure. All pairwise rate distributions, except NC(HS) vs. NC(HLS) and NC(HS) vs. C(L), are significantly different (P \< 0.05, after Bonferroni correction: P \< 0.008). For a summary of the U statistic and two-tailed P values for each pairwise comparison see [Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online.](evw246f5p){#evw246-F5} ###### Structural Conservation of Sites Per Protein Family --------------------- ![](evw246ie2p.jpg) --------------------- Identifying Target Sites ------------------------ For regions with five or more consecutive sites that were 100% conserved in sequence across 1) all CoV or 2) across the MERS and SARS clades, the information of structural disorder prediction from IUPred and DISOPRED2 was used to identify all ungapped sites that were consistently predicted to have 100% conserved order. Next, the information of secondary structure prediction from PSIPRED and JPred was used to narrow down this list further by only including sites that are not changing their predicted secondary structure state for both predictors. Applying the aforementioned filters to the initial 10,000 sites resulted in one (1) region of five residues or more conserved across all CoV within the N-terminal domain of NSP12: DNQDL ([table 4](#evw246-T4){ref-type="table"}). Interestingly, this region is in the vicinity of sites found important for nucleotidylating activity across the order Nidovirales ([@evw246-B27]). ###### Sites Conserved in Sequence and Structural Property Protein family PfamA domain Conserved sites in the MSA[^a^](#evw246-TF4){ref-type="table-fn"} ------------------------------------------------------- ----------------- ------------------------------------------------------------------- NSP5 Peptidase_C30 149-GSCGS-153[^b^](#evw246-TF5){ref-type="table-fn"} 213-AWLYAA-218[^b^](#evw246-TF5){ref-type="table-fn"} NSP7 Replicase 7-KCTSVVLL-14[^b^](#evw246-TF5){ref-type="table-fn"} 16-VLQQL-20[^b^](#evw246-TF5){ref-type="table-fn"} NSP12 RPol N-term 228-L[DNQDL]{.ul}NG-235 239-DFGDF-243 RdRP_1 521-DKSAG-525 588-MTNRQ-592 677-LANECAQVL-685 800-[GGT]{.ul}SSGD-706 C-term 853-YPDPSR-858 871-KTDGT-875 889-[YPL]{.ul}TK-893 NSP13 N-term 10-SQTSLR-15 AAA_30 362-NALPE-366 402-DPAQLP-407 AAA_12 539-SSQGS-543 NSP14 NSP11 281-AHVAS-285[^b^](#evw246-TF5){ref-type="table-fn"} 290-MTRCLA-295[^b^](#evw246-TF5){ref-type="table-fn"} 438-HAFHT-442[^b^](#evw246-TF5){ref-type="table-fn"} 494-CNLGG-499[^b^](#evw246-TF5){ref-type="table-fn"} Sites conserved across all clades in the protein family are underlined and in BOLD font. All other sites are conserved across the SARS and MERS clades. Experimentally determined structures are available in Protein Data Bank ([@evw246-B2]).. Considering only the sequences in the SARS and MERS clades, 21 sequence regions of five residues or more were found in seven protein families ([table 4](#evw246-T4){ref-type="table"}). For NSP5, NSP7, and NSP14, experimentally determined structures show that most regions are surface accessible ([fig. 6](#evw246-F6){ref-type="fig"}). Some of the identified target sites are known for their functional importance. For instance, C145 in the middle of GSCGS in NSP5 is part of the catalytic dyad in the NSP5 protease ([@evw246-B50]). For NSP12 and NSP13, which have the majority of all sites, no structures are available. The sites adjacent to DNQDL are also conserved in the SARS and MERS clades, and five additional target sites, conserved for the SARS and MERS clades, are found in the C-terminal direction relative to the DNQDL motif ([table 4](#evw246-T4){ref-type="table"}). Continuing into the RNA-dependent RNA polymerase domain (RdRP) in NSP12, four additional regions of target sites are found, and the last three regions are found in the C-terminal part. Importantly, in RdRP and in the C-terminal part are sites that are also conserved across all CoVs in this study. NSP13 has four regions of target sites distributed across the protein. ![Target sites shown in 3D context. (A) NSP5 dimer, based on PDB id 1UK4 ([@evw246-B50]). (B) NSP7, based on PDB id 5F22 (unpublished). (C) NSP14, based on PDB id 5C8T ([@evw246-B30]). Protein structure visualized with Bioviva Discovery Studio .](evw246f6p){#evw246-F6} Discussion ========== We have analyzed the protein evolution of the genetic components that make up the MERS-CoV proteome. As previously established, MERS-CoV has the same genomic makeup as HKU4-CoV and HKU5-CoV in the MERS clade ([@evw246-B47]). Some protein products are only found in the MERS clade, and these were excluded from this study due to insufficient data. Furthermore, for other protein products, some clades may not be represented in our protein families if their proteins were too divergent. This was an important factor in determining the applied BLAST hit cutoffs, as relaxing cutoffs produced alignments with more gaps and increasing stringency reduced the representative pool. Because alignment quality is important due to the sensitivity of both Rate4Site and for phylogenetic reconstruction, the chosen cutoffs are suitable. We note some clade-specific differences in recoverable homologs between different CoV, but many components are shared among them ([fig. 1](#evw246-F1){ref-type="fig"}). Viral proteins often possess multifunctionality, mediated by a conformational change in response to environment-specific factors ([@evw246-B49]). Although conformational flexibility is important for function, it also offers flexibility in what sequence motifs are on display. If these sequences are rapidly diverging, different sequence motifs will be displayed, reinforcing the notion that flexible regions are potentially important in rewiring protein--protein interactions between virus and host ([@evw246-B18]). Although most CoV proteins have almost no intrinsic disorder, several CoV protein families have homologous sites that display loop in some sequences, helix in others and strands in some ([table 3](#evw246-T3){ref-type="table"}, [Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online). These sites are not necessarily disordered but they may be conformationally flexible in real-time (with secondary structure transitions in the same sequence, making them difficult to predict) or on evolutionary time-scales (so that different secondary structure elements actually are present in different sequences). The C(HS) and C(L) sites make up approximately 50--80% of most multiple sequence alignments. With the common expectation that protein structure is more conserved than sequence these numbers are surprisingly low. Neither PSIPRED nor JPred consistently predicts the same state for 20--50% of all sites in these multiple sequence alignments. The accuracy of PSIPRED and JPred's secondary structure predictions are about 80% ([@evw246-B6]; [@evw246-B13]). PSIPRED has been found to rarely predict an alpha helix instead of a beta strand and vice versa, and most of the PSIPRED errors are due to secondary structure not being predicted ([@evw246-B28]). When secondary structure is not conserved for the same site in a multiple sequence alignment, it suggests that the secondary structure prediction may be 1) inaccurate, 2) not predicted with high confidence, or 3) the regions are indeed metamorphic; they can transition from one element to another. Although (1) is difficult to address without experimentally determined structures for all sequences, (2) and (3) are not necessarily incompatible interpretations because low confidence secondary structure prediction could indicate metamorphic secondary structure regions. Metamorphic secondary structure regions have interesting consequences for conformational and functional flexibility. It should be noted that, despite the low amount of disordered sites in most CoV proteins, several regions are not conserved in disorder propensity across all sequences, but sometimes the different predictors disagree as in the case of NSP8. Clade-specific disordered regions resulting from indel events suggest that they are not essential to the critical functions of the protein, but could cause gain-and-loss of interactions with its hosts. However, when disorder propensity is only mildly fading for a region that is present across the protein family, it may be important for the fundamental function of the protein. The virus structural proteins that interact to form the virion commonly include an envelope protein, a membrane protein, and a capsid protein that together form the machinery that encases, transports, and releases the virus. The interactions between the structural proteins are often regulated by conformational changes like VP40 in Ebola ([@evw246-B4]) and Envelope protein from Dengue virus ([@evw246-B53]). Conformational changes in these proteins are needed for the virus life cycle. For CoVs, nucleocapsid is the only structural protein that is highly disordered. Yet, rapid evolutionary dynamics of disorder is present in nucleocapsid using two different IUPred cutoffs (0.4 and 0.5) and with DISOPRED2. Even if the different predictors and cutoffs disagree somewhat where regions with rapid evolutionary dynamics are present, these patterns suggest that nucleocapsid may be rapidly changing from one virus to another. It should also be noted that two MERS clade specific inserts around position 241 and toward the C-terminal are consistently predicted to be highly disordered. With inserts and changing structural dynamics between clades or viruses, the questions become 1) which sequence motif are displayed and 2) to what extent are these sequence motifs displayed? Furthermore, based on the inconsistent prediction of secondary structure elements, the possibility that CoVs are more conformationally flexible than their intrinsic disorder content implies is noteworthy. Altogether, this suggests that various mechanisms for rewiring conformational and functional space are operating in the coronaviruses studied here. If regions symptomatic of conformational and functional flexibility can be avoided in order to identify broad-specificity antiviral targets with potential to be effective against coronaviruses of today and in the future, coronaviruses as a group may become more attractive drug targets for the pharmaceutical industry in the event an additional coronavirus changes host to include humans or increase its virulence. Supplementary Material ====================== ###### Click here for additional data file. We thank Joseph Ahrens, Janelle Nunez-Castilla, and Helena Gomes Dos Santos for assistance in the lab and for helpful discussions. The authors would also like to acknowledge the Instructional & Research Computing Center (IRCC) at Florida International University for providing HPC computing resources that have contributed to the research results reported within this article, web: <http://ircc.fiu.edu>. Supplementary Material ====================== [Supplementary Data](#sup1){ref-type="supplementary-material"} and [Supplementary Data](#sup1){ref-type="supplementary-material"} are available at Genome Biology and Evolution online ([http://www.gbe. oxfordjournals.org/](http://www.gbe.oxfordjournals.org/)). [^1]: Associate editor: Dr. Chantal Abergel	{ "pile_set_name": "PubMed Central" }
Introduction ============ The latest Global Burden of Disease Study showed that 1.25 billion people had migraine in 2017.[@ref1] Migraine was responsible for 45.1 million years of life lived with disability.[@ref2] In the 15-49 year age group, migraine was the top cause of life lived with disability.[@ref3] Although some authors have suggested that prophylactic drugs should be recommended for migraineurs who have at least four headache days per month, only 13% of patients reported current use of preventive drugs.[@ref4] Ineffectiveness of and/or contraindications to migraine medication affect 10-15% of people with migraine.[@ref5] Hence, a need exists to investigate non-drug interventions. Acupuncture has emerged as a prophylactic treatment for migraine because of its potential efficacy.[@ref6] [@ref7] [@ref8] Clinically, two acupuncture interventions are widely used: manual acupuncture and electrical acupuncture. Manual acupuncture is commonly used by traditional acupuncturists and is highly effective for pain relief.[@ref9] Besides insertion of penetrating needles into specific acupuncture points, manual manipulation (that is, intermittent rotation as well as lift and thrust) of the needle may enhance its clinical benefit.[@ref10] [@ref11] However, clinical evidence for the benefit of manual acupuncture for migraine prophylaxis remains scarce. Several randomised clinical trials found no differences between manual and sham acupuncture.[@ref12] [@ref13] [@ref14] [@ref15] [@ref16] These negative results might be ascribed to inappropriate placebo control settings. Blinding is challenging in trials of non-drug treatments. An appropriate placebo acupuncture design should be both physiologically inert and indistinguishable from true acupuncture. So far, only one randomised clinical trial of acupuncture in migraine has reported successful blinding.[@ref15] Furthermore, some authors have suggested that sham acupuncture interventions were not totally inert, often being associated with moderately large non-specific effects.[@ref17] [@ref18] [@ref19] This could make detection of additional specific effects of manual acupuncture difficult. In particular, sham acupuncture involving penetrating needles should be avoided in clinical trials.[@ref20] However, penetrating sham acupuncture has been commonly used in previous randomised clinical trials in migraine.[@ref12] [@ref13] [@ref14] [@ref15] [@ref16] Therefore, we designed a randomised clinical trial among patients with migraine who were naive to acupuncture, using non-penetrating sham control and assessment of blinding to determine the efficacy of manual acupuncture and quantify the true placebo response in the prophylaxis of episodic migraine without aura. Methods ======= Study design ------------ This randomised, single blind, three arm clinical trial was conducted in seven centres in China between 5 June 2016 and 15 November 2018. The protocol and the statistical analysis plan are available in appendix 1. The total trial period was 24 weeks, including four weeks of baseline assessment, eight weeks of treatment after randomisation, and 12 weeks of follow-up. We obtained written informed consent from each patient before randomisation. Participants ------------ A neurologist made the diagnosis of episodic migraine without aura on the basis of the International Classification of Headache Disorders, 3rd edition β version (ICHD-3β).[@ref21] Other inclusion criteria were age between 15 and 65 years, history of migraine without aura for more than 12 months, initial onset of migraine before the age of 50 years, between two and eight migraine attacks during the baseline phase, naivety to acupuncture, and ability to give informed consent. Exclusion criteria included all other types of primary and secondary headaches, history of a clinically significant disorder (for example, severe mental illness), pregnancy or breast feeding, and non-adherence to the baseline diary. In addition, all patients were instructed not to take any other analgesics and to avoid starting other interventions. Randomisation and masking ------------------------- After the baseline assessment, an independent investigator randomised eligible patients in a 2:2:1 ratio to receive manual acupuncture, sham acupuncture, or usual care. We used the 2:2:1 ratio to facilitate recruitment and increase patients' adherence. Treatments were allocated in a centralised manner via an interactive web response system with stratification by centre and a block size of five. To maintain blinding in patients, outcome assessors, and statisticians, the allocation sequence was concealed until the end of the study. Study interventions ------------------- Fourteen licensed acupuncturists delivered treatments. They each had more than five years of clinical experience and attended centralised training before recruitment. All patients received 20 sessions of 30 minute acupuncture treatments or usual care over eight weeks. They were treated every other day to fulfil a 10 session treatment course and received another 10 session course after a nine day break. In both the manual and sham acupuncture group, acupuncture treatment began after randomisation. In the usual care group, patients received acupuncture for free after waiting 24 weeks. Details of acupuncture (location of acupuncture points/non-acupuncture points, depth of insertion) are shown in tables 1 and 2 of appendix 1. Acupuncturists applied manual acupuncture at 10 obligatory acupoints, included bilateral L14, LR3, EX-HN5, GB20, and GB8. Additional acupoints were selected on the basis of meridian diagnosis and the patient's symptoms: bilateral ST8 for Yangming meridian headache, BL10 for Taiyang meridian headache, and DU20 for Jueying meridian headache. Streitberger acupuncture needles (0.30 mm in diameter and 30 mm in length) were used. After sterilisation, the sharp needles were inserted into the deep tissue layers of acupoints. Acupuncturists then tried to elicit acupuncture de-qi sensation by the manual manipulation of needles. Over a 30 minute period, manual manipulation for each acupoint lasted 10 seconds and was repeated four times with intervals of 10 minutes. Non-penetrating sham acupuncture was performed at four bilateral non-acupuncture points, which were all located on the back and in different segments from the headache area. After sterilisation, Streitberger placebo needles with blunt tip were used.[@ref22] When they were fixed on the skin through plastic rings, patients felt a pricking sensation, simulating a puncture of the skin. However, instead of penetrating the skin, the needles retracted up into the shaft when they were pressed against the skin. We formulated and followed standardised step-by-step instructions and operations to use the same rituals in the manual acupuncture and sham acupuncture groups as far as possible (appendix 2). In accordance with the migraine guidelines developed by the Canadian Headache Society Prophylactic Guidelines Development Group in 2012,[@ref23] we provided usual care to all three groups in terms of lifestyle changes and migraine self-management. We taught patients about lifestyle factors that may contribute to an increased frequency of migraine, identifying migraine triggers or precipitants through the keeping of a detailed headache diary, pacing, pain communication, sleep hygiene, and regular exercise. In case of severe pain (visual analogue score \>8), diclofenac sodium enteric coated tablets (25 mg/tablet; maximal tolerated dose 200 mg/day) were allowed as a rescue medication. Outcome measures ---------------- Independent research assistants supervised patients to complete their headache diaries in paper and pencil format from baseline to week 20 and evaluated their outcomes. The primary outcomes were change in the mean number of migraine days and migraine attacks per four week cycle during weeks one to 20 after randomisation compared with baseline (the four weeks before randomisation). Secondary outcomes included the proportion of patients achieving at least a 50% reduction in the mean number of migraine days or migraine attacks during weeks 17 to 20 and changes in the severity of migraine as measured by a visual analogue scale, the Migraine-Specific Quality-of-Life Questionnaire (MSQ),[@ref24] the Pittsburgh Sleep Quality Index (PSQI),[@ref25] the Migraine Disability Assessment Score (MIDAS),[@ref26] the Beck Anxiety Inventory (BAI) scale,[@ref27] the Beck Depression Inventory II (BDI-II) scale,[@ref28] and the mean dose of used rescue medication from baseline to week 20. We also measured patients' expectations of of positive outcomes from acupuncture by using the Acupuncture Expectancy Scale.[@ref29] At the end of the study, we determined the maintenance of blinding of patients by asking them whether they thought the needles had penetrated the skin. Patients recorded adverse events after each treatment, including bleeding, subcutaneous haemorrhage, serious pain, palpitation, fainting, and local infection. Statistical analyses -------------------- On the basis of a pilot study, we anticipated a mean difference of 1.8 (SD 2.8) days and 1.2 (1.8) attacks for manual versus sham acupuncture and a difference of 2.2 (2.8) days and 1.5 (1.8) attacks for manual acupuncture versus usual care. To drive all the primary hypothesis tests, we used PROC POWER in SAS to calculate that a sample of 135 evaluable patients (54 in manual acupuncture, 54 in sham acupuncture, and 27 in usual care) would be needed to provide 90% power at a two sided significance level of 5%. Therefore, we planned to recruit 150 patients into this study, with an anticipated 10% dropout rate. We did efficacy analyses in the full analysis set, which included all randomised patients who received at least one session of treatment and had at least one efficacy measure. We imputed missing data by using the last observation carried forward method.[@ref30] We did sensitivity analyses in the per protocol set, which included all randomised patients who had no major protocol deviation. We did safety analyses in the safety set, which included all randomised patients who received at least one session of acupuncture. We assessed the poolability of treatment effects across centres by including an interaction term of treatment group by centre in the model for the primary analysis. If P values associated with the interaction term of both primary endpoints were greater than 0.15,[@ref31] the treatment effect was homogenous across centres. We presented continuous variables as the mean with standard deviation or median and interquartile range. For tests across groups, we used analysis of variance or the Kruskal-Wallis test when relevant. We described categorical variables as numbers and percentages and used the χ^2^ test or Fisher's exact test. For comparisons of the primary endpoints across the groups, we used analysis of covariance with the baseline migraine days or migraine attacks adjusted. Superiority of manual acupuncture over sham acupuncture or usual care could be claimed only when both co-primary endpoints were statistically significant. We used a fixed sequence procedure for multiple comparisons among groups, which would not inflate the type I error, and comparisons were made with an α value of 0.05. We also evaluated multivariate adjusted treatment effects across the groups on the basis of a linear mixed effects model. For the secondary endpoints, we made pairwise multiple comparisons on the basis of Bonferroni adjusted P values and confidence intervals, if a significant difference was noted across the groups. An independent statistician used SAS statistical software for all statistical analyses. For both continuous and categorical variables, 95% confidence intervals were calculated as appropriate. All the statistical comparisons were two sided with P\<0.05 considered as significant. Patient and public involvement ------------------------------ No patients were involved in the design or conduct of the study, in setting the research question, or in determining the outcome measures of this study, nor did they have any input on data analysis, interpretation, or writing up of results. Results of the trial will be shared with all participants by web or email. Results ======= Study participants ------------------ Between 5 June 2016 and 15 November 2018, we screened 210 patients, of whom 150 were randomised ([fig 1](#f1){ref-type="fig"}). The main reason for exclusion was not meeting the inclusion criterion of between two and eight migraine attacks during the baseline phase (table A in appendix 3). The baseline characteristics of all randomised patients were well balanced among the three groups ([table 1](#tbl1){ref-type="table"}). Similar baseline characteristics were observed in the full analysis set population (table B in appendix 3). ![Flowchart showing trial group assignments, loss to follow-up, treatment completion, and protocol deviations](xus053060.f1){#f1} ###### Baseline characteristics of randomised population. Values are mean (SD) unless stated otherwise Characteristics Manual acupuncture (n=60) Sham acupuncture (n=60) Usual care (n=30) --------------------------------------------------- --------------------------- ------------------------- ------------------- Age, years 36.6 (12.0) 36.0 (10.9) 37.3 (11.7) No (%) women 47 (78) 50 (83) 26 (87) Median (IQR) body mass index 20.4 (19.5-23.1) 22.0 (20.3-23.4) 20.8 (19.5-22.4) Median (IQR) disease duration, years, 10.0 (5.0-19.5) 10.0 (6.0-14.0) 16.5 (7.8-23.0) Accompanying symptoms: No (%) nausea or vomiting 53 (88) 51 (85) 25 (83) No (%) photophobia or phonophobia 52 (87) 52 (87) 23 (77) Patient personality: Neuroticism 31.8 (7.1) 31.9 (6.6) 31.3 (7.1) Extraversion 37.4 (7.0) 38.0 (6.6) 37.0 (6.4) Openness 37.6 (4.9) 37.4 (4.8) 35.9 (2.8) Agreeableness 43.2 (4.6) 43.1 (4.9) 43.6 (5.2) Conscientiousness 41.2 (5.6) 41.8 (6.2) 41.5 (5.5) No (%) chronic migraine 1 (2) 2 (3) 0 (0) Acupuncture Expectancy Scale score 10.9 (3.7) 11.1 (3.5) NA Days with migraine 5.8 (2.6) 6.3 (3.8) 5.8 (3.0) No of migraine attacks 3.8 (1.4) 4.1 (2.6) 3.9 (1.7) Mean VAS score 5.1 (1.3) 5.3 (1.3) 5.2 (1.8) Migraine-Specific Quality-of-Life Questionnaire: Role restrictive subscale 55.2 (17.6) 54.4 (15.5) 56.0 (15.0) Role preventive subscale 60.6 (23.3) 59.9 (19.5) 64.0 (19.0) Emotional subscale 61.6 (23.3) 65.3 (20.3) 62.9 (21.6) Pittsburgh Sleep Quality Index 5.7 (2.7) 5.6 (2.7) 5.3 (3.5) Median (IQR) Migraine Disability Assessment Score 27.5 (12.0-49.8) 33.5 (15.3-57.3) 27.5 (13.8-44.8) Median (IQR) doses of rescue medication 0.0 (0.0-1.8) 0.0 (0.0-3.5) 0.0 (0.0-2.0) Median (IQR) Beck Depression Inventory II 6.5 (3.0-13.0) 7.0 (2.3-14.0) 7.0 (1.0-13.0) Median (IQR) Beck Anxiety Inventory 10.0 (4.0-15.0) 7.5 (4.0-14.0) 7.0 (3.8-9.5) IQR=interquartile range; NA=not applicable; VAS=visual analogue scale. After randomisation, six (4%) patients were lost to follow-up (three in the manual acupuncture group, one in the sham acupuncture group, and two in the usual care group; P=0.3025, χ^2^ test). Of these patients, three were lost without one primary outcome measure and another three did not complete all the visits. Therefore, 147 patients were included in the full analysis set and 144 in the per protocol set. Primary outcomes ---------------- [Table 2](#tbl2){ref-type="table"} and [figure 2](#f2){ref-type="fig"} show the primary analyses. The reduction of migraine days and migraine attacks per four weeks from baseline was significantly greater in the manual acupuncture group than in the usual care group during weeks 1 to 20. Compared with sham acupuncture, manual acupuncture resulted in a significantly greater reduction in migraine days during weeks 13 to 20, with a group difference of −1.4 (95% confidence interval −2.4 to −0.3; P=0.005) days at weeks 13 to 16 and −2.1 (−2.9 to −1.2; P\<0.001) days at weeks 17 to 20, and a significantly greater reduction in migraine attacks at weeks 17 to 20, with a group difference of −1.0 (−1.5 to −0.5; P\<0.001) attacks. Sham acupuncture also resulted in a significantly greater reduction in migraine attacks during weeks 5 to 20 compared with usual care. We observed similar results in the per protocol analysis population (table C in appendix 3). In addition, we developed mixed effects models including participants, centres, acupuncturists, baseline, study visits, treatment groups, and the interaction term of treatment groups by study visits, the results of which were consistent with those of the primary analyses (table D in appendix 3). The poolability of both primary efficacy endpoints across centres was verified (P values for the interaction effects were \>0.15). ###### Primary outcome measurements of full analysis set population. Values are mean (SD) unless stated otherwise Time point Manual acupuncture (n=58) Sham acupuncture (n=60) Usual care (n=29) P value\* Pairwise comparisons---mean (95% CI) --------------------------------------------------------------- --------------------------- ------------------------- ------------------- ----------- -------------------------------------- --------- --------------------- --------- --------------------- ------- Change from baseline in mean number of days with migraine Weeks 1-4 −2.2 (2.0) −1.6 (3.0) −0.7 (2.9) 0.003 −0.9 (−1.8 to 0.0) 0.052 −1.5 (−2.5 to −0.4) 0.004 −0.6 (−1.7 to 0.5) 0.589 Weeks 5-8 −3.1 (2.6) −2.7 (3.6) −1.3 (2.5) 0.002 −0.7 (−1.6 to 0.2) 0.198 −1.7 (−2.8 to −0.5) 0.001 −1.0 (−2.1 to 0.2) 0.119 Weeks 9-12 −3.5 (2.6) −2.9 (3.8) −1.3 (2.7) \<0.001 −1.0 (−2.0 to 0.1) 0.077 −2.1 (−3.4 to −0.8) \<0.001 −1.1 (−2.4 to 0.1) 0.100 Weeks 13-16 −3.5 (2.5) −2.4 (3.4) −1.0 (2.4) \<0.001 −1.4 (−2.4 to −0.3) 0.005 −2.4 (−3.7 to −1.2) \<0.001 −1.1 (−2.3 to 0.2) 0.136 Weeks 17-20 −3.9 (3.0) −2.2 (3.2) −1.4 (2.7) \<0.001 −2.1 (−2.9 to −1.2) \<0.001 −2.4 (−3.5 to −1.4) \<0.001 −0.4 (−1.4 to 0.7) 0.495 Change from baseline in mean number of migraine attacks Weeks 1-4 −1.1 (1.3) −1.0 (2.5) −0.2 (1.5) 0.032 −0.3 (−0.9 to 0.3) 0.643 −0.8 (−1.6 to −0.1) 0.026 −0.5 (−1.3 to 0.2) 0.305 Weeks 5-8 −1.8 (1.4) −1.7 (2.5) −0.7 (1.3) 0.002 −0.3 (−0.9 to 0.2) 0.526 −1.0 (−1.7 to −0.4) \<0.001 −0.7 (−1.4 to 0.0) 0.032 Weeks 9-12 −1.9 (1.3) −1.8 (2.7) −0.4 (1.3) \<0.001 −0.4 (−1.0 to 0.2) 0.410 −1.5 (−2.3 to −0.8) \<0.001 −1.1 (−1.9 to −0.4) 0.001 Weeks 13-16 −2.1 (1.5) −1.8 (2.7) −0.7 (1.3) \<0.001 −0.6 (−1.2 to 0.1) 0.092 −1.3 (−2.1 to −0.6) \<0.001 −0.8 (−1.5 to 0.0) 0.043 Weeks 17-20 −2.3 (1.7) −1.6 (2.5) −0.4 (1.3) \<0.001 −1.0 (−1.5 to −0.5) \<0.001 −1.8 (−2.4 to −1.2) \<0.001 −0.8 (−1.4 to −0.2) 0.008 P values, differences, and confidence intervals were based on analysis of covariance adjusted for days with migraine (or number of migraine attacks) at baseline. According to fixed sequence procedure planned for primary analyses, pairwise comparisons at weeks 17-20 were made at nominal α level without further adjustment. For visits other than weeks 17-20, multiple pairwise comparisons were adjusted by Bonferroni approach. ![Primary outcomes throughout trial. Top: number of migraine days per 4 weeks through study period. Bottom: frequency of migraine attacks per 4 weeks through study period](xus053060.f2){#f2} Secondary outcomes ------------------ [Table 3](#tbl3){ref-type="table"} summarises the secondary analyses. Overall, the responder rates in the manual acupuncture group were significantly higher than in the two control groups during weeks 17 to 20. All subscales of the Migraine-Specific Quality-of-Life Questionnaire were improved significantly more in the manual acupuncture group than in the two control groups at weeks 20. The visual analogue scale showed a greater reduction in score for the manual acupuncture group than for the two control groups at weeks 20. ###### Secondary outcome measurements of full analysis set population Time point Manual acupuncture (n=58) Sham acupuncture (n=60) Usual care (n=29) P value\* Pairwise comparison -------------------------------------------------------------------------------------- --------------------------- ------------------------- ------------------- ----------- --------------------- --------- ----------------------- --------- ---------------------- --------- Proportion of participants with ≥50% reduction in number of migraine days (%) Weeks 17-20 82.5 45.8 17.9 \<0.001 36.7 (17.0 to 56.4) \<0.001 64.6 (43.5 to 85.7) \<0.001 27.9 (4.6 to 51.2) 0.012 Proportion of participants with ≥50% reduction in number of migraine attacks (%) Weeks 17-20 78.9 44.1 14.3 \<0.001 34.9 (14.7 to 55.0) \<0.001 64.7 (44.2 to 85.1) \<0.001 29.8 (7.6 to 51.9) 0.004 Mean (SD) change from baseline in visual analogue scale Weeks 20 −2.2 (2.5) −0.9 (1.9) −0.0 (1.4) \<0.001 −1.3 (−2.2 to −0.3) 0.005 −2.1 (−3.3 to −1.0) \<0.001 −0.9 (−2.0 to 0.3) 0.224 Mean (SD) change from baseline in MSQ, role restrictive subscale Weeks 20 24.4 (21.4) 12.8 (16.0) 5.8 (14.2) \<0.001 11.6 (3.5 to 19.8) 0.002 18.6 (8.5 to 28.7) \<0.001 7.0 (−3.1 to 17.0) 0.285 Mean (SD) change from baseline in MSQ, role preventive subscale Weeks 20 22.5 (25.1) 10.1 (14.1) 4.5 (15.5) \<0.001 12.5 (3.7 to 21.2) 0.002 18.1 (7.2 to 29.0) \<0.001 5.6 (−5.2 to 16.4) 0.630 Mean (SD) change from baseline in MSQ, emotional subscale Weeks 20 19.5 (26.8) 8.4 (19.0) 8.8 (20.9) 0.020 11.2 (0.9 to 21.4) 0.027 10.7 (−2.0 to 23.4) 0.129 −0.4 (−13.1 to 12.2) \>0.999 Mean (SD) change from baseline in global score of PSQI Weeks 20 −1.8 (2.3) −0.7 (2.5) −0.1 (2.6) 0.007 −1.1 (−2.2 to 0.0) 0.058 −1.7 (−3.0 to −0.3) 0.011 −0.6 (−2.0 to 0.8) 0.894 Median (IQR) change from baseline in score of MIDAS Weeks 20 −15.0 (−36.0-−5.0) −6.0 (−30.0-2.0) −0.5 (−16.0-6.5) 0.007 −8.0 (−17.0 to 2.0) 0.139 −14.0 (−28.0 to −3.0) 0.006 −6.0 (−22.0 to 4.0) 0.523 Median (IQR) change from baseline in doses of rescue medication Weeks 17-20 0.0 (−1.0-0.0) 0.0 (−1.0-0.0) 0.0 (−1.5-0.0) 0.745 NA NA NA NA NA NA Median (IQR) change from baseline in BDI- II Weeks 20 −2.0 (−6.0-0.0) −1.0 (−5.0-1.0) 0.0 (−2.5-1.0) 0.081 NA NA NA NA NA NA Median (IQR) change from baseline in BAI Weeks 20 −4.0 (−8.0-0.0) −1.0 (−5.0-1.0) 0.0 (−3.5-1.5) 0.005 −2.0 (−5.0 to 0.0) 0.051 −4.0 (−7.0 to −1.0) 0.007 −1.0 (−4.0 to 1.0) 0.980 BAI=Beck Anxiety Inventory; BDI-II=Beck Depression Inventory II; IQR=interquartile range; MIDAS=Migraine Disability Assessment Scores; MSQ=Migraine-Specific Quality-of-Life Questionnaire; NA=not applicable; PSQI=Pittsburgh Sleep Quality Index. Pairwise comparison include differences in means, medians, and proportions and corresponding confidence intervals. For continuous variables, analysis of variance or Kruskal-Wallis test was used for comparisons among treatment groups. If Kruskal-Wallis test was significant, pairwise comparisons were made using Wilcoxon rank sum test. Difference in median is estimated on basis of Hodges-Lehmann estimator of location shift. Asymptotic (Moses) confidence intervals are provided. For categorical variables, linear probability model was used for comparisons among treatment groups. For all pairwise comparisons, P values and confidence intervals were adjusted by Bonferroni approach. The total scores on the Pittsburgh Sleep Quality Index and Migraine Disability Assessment Score were significantly lower in the manual acupuncture group than in the usual care group at week 20. However, we found no significant difference in these scores between manual acupuncture and sham acupuncture. We also found no significant difference in the mean dose of rescue medication or in Beck Anxiety Inventory and Beck Depression Inventory II scores among the three groups at week 20. At the end of the study, we assessed the credibility of blinding of treatment. We found no significant difference between the manual acupuncture and sham acupuncture groups for patients' ability to correctly guess their allocation status (P=0.891) (table E in appendix 3). Adverse events -------------- In the manual acupuncture group, five (8%) patients reported at least one acupuncture related adverse event compared with no patients in the sham acupuncture group. One patient withdrew from the study because of an adverse event. No patients in either the manual or sham acupuncture group had severe adverse events (table F in appendix 3). Discussion ========== In this study in acupuncture naive patients with episodic migraine without aura, 20 sessions of manual acupuncture produced a relatively long lasting reduction in migraine days and migraine attacks compared with sham acupuncture and usual care. We found that, compared with the usual care group, both the reductions in migraine days and migraine attacks per four week period from baseline in the manual acupuncture group were significantly larger from the first four week period after the start of treatment, and the significant reductions lasted until the last four week period of follow-up (weeks 17-20), with an apparent increasing trend. Compared with the usual care group, only in migraine attacks per four week period from baseline did the sham acupuncture group show significant reductions from the second four week period after treatment to the last four week period of follow-up. In the last four week period, that reduction fell a little. In comparisons between the manual acupuncture group and the sham acupuncture group, reductions in migraine days were significantly larger in the manual group at weeks 13-16 and 17-20, and reductions in migraine attacks were significantly larger in the manual group at weeks 17-20. Overall, the therapeutic effects in the manual acupuncture group occurred earlier, were larger, and might last longer. Comparison with other studies ----------------------------- The latest Cochrane meta-analysis,[@ref32] published in 2016, found that up to both eight weeks/two months and three to four months after randomisation, acupuncture was associated with a small but statistically significant reduction in migraine attacks compared with sham. The effect size was −0.35 (95% confidence interval −0.57 to −0.13) at up to eight weeks/two months after randomisation and −0.32 (−0.53 to −0.10) at up to three to four months after randomisation. Since then, results have been published for only one randomised clinical trial.[@ref30] It found that electrical acupuncture resulted in a long term reduction in migraine attacks. At week 16 after randomisation, a greater reduction was observed in the true electrical acupuncture group than in the sham acupuncture group (difference of 1.1 (0.4 to 1.9) attacks; P=0.002) and in the true electrical acupuncture group than the no acupuncture group (difference of 1.8 (1.1 to 2.5) attacks; P\<0.001). The effect size for manual acupuncture over sham acupuncture observed in our study was similar to the effect size seen with electrical acupuncture,[@ref30] but larger than the effect size seen in the latest Cochrane meta-analysis.[@ref32] We also updated the meta-analysis to include our results (appendix 4). Additionally, the effect size for manual acupuncture was larger than the effect size seen with prophylactic drugs, such as monthly subcutaneous injections of fremanezumab (225 mg or 675 mg) or galcanezumab (120 mg or 300 mg).[@ref33] [@ref34] However, our findings are not completely consistent with those of several randomised clinical trials,[@ref12] [@ref13] [@ref14] [@ref15] [@ref16] which found no significant differences between true acupuncture and sham acupuncture in reducing migraine headaches. The inconsistency might be driven predominantly by differences in treatments received by the control group. Comparison between manual and electrical acupuncture ---------------------------------------------------- In acupuncture practice, manual acupuncture and electrical acupuncture are the two most common interventions for the treatment of pain conditions. Manual acupuncture emphasises the occurrence of "de-qi" feelings, which can be induced by proper and effective manual manipulation of the needle. The intensity of de-qi feelings plays a key role in the analgesic effects of manual acupuncture.[@ref35] The importance of manipulation of the needle is no less than that of the selection of the appropriate acupoint.[@ref36] The effects of electrical acupuncture depend on current parameters (frequency, intensity, and duration of pulses). Different manual manipulations or different electrical parameters may produce different therapeutic effects. The peripheral and central mechanisms for pain relief produced by manual acupuncture and electrical acupuncture are homologous, but some differences exist. The effects of manual acupuncture may be mediated by all types of afferents, whereas the effects of electrical acupuncture might be preferentially mediated by Aβ and parts of Aδ type afferents.[@ref9] In addition, subtle differences exist between the areas of the brain activated by manual and electrical acupuncture.[@ref37] So far, only a few small trials for several conditions have directly compared the effects of manual and electrical acupuncture.[@ref38] [@ref39] [@ref40] [@ref41] [@ref42] The findings of these studies were inconsistent and inconclusive. The choice of different manual manipulations or different electrical parameters may need to be based on a specific disease or condition and should be guided by more high quality randomised controlled trials. Strengths and limitations of study ---------------------------------- An advantage of this trial is the design of a reasonable sham comparator. Previous acupuncture research has often used penetrating sham acupuncture, involving needling non-acupuncture points, needling irrelevant acupuncture points, or superficial needling.[@ref43] [@ref44] However, whether the needle is inserted into an acupuncture point or a non-acupuncture point, it could produce a physiological effect.[@ref44] This is in part due to the activation of the pain inhibiting system in the spinal cord and diffuse noxious inhibitory controls.[@ref45] [@ref46] [@ref47] In addition, sham acupuncture applied at points beside the real acupoints might be an active control, because the area of the acupoint could be enlarged by increased expression of nociceptive substances during various pain conditions.[@ref48] In this study, we applied non-penetrating sham acupuncture at heterosegmental non-acupuncture points, thereby avoiding segmental analgesia and minimising any physiological effect in the sham acupuncture group. Another strength of this trial is the successful blinding. Previous trials lacked assessment of blinding or had ineffective blinding,[@ref12] [@ref13] [@ref14] [@ref16] so comparisons between true acupuncture and sham acupuncture might be biased by unsuccessful blinding. Ideal blinding should be psychologically credible yet physiologically inert. To ensure successful blinding, we recruited acupuncture naive patients, used non-penetrating needles as the control, and designed the same procedures to perform the same rituals as far as possible in the manual and sham acupuncture groups. These factors made the sham acupuncture as close to a real placebo as possible. At the end of the study, we asked patients to guess whether they thought the needles penetrated the skin. Satisfyingly, we found no significant difference between the two treatment groups in patients' ability to correctly guess their allocation status. Successful blinding showed the reliability of the difference between manual acupuncture and sham acupuncture, as well as between sham acupuncture and usual care. Furthermore, successful blinding increased the adherence of patients, especially those allocated to the sham acupuncture group. Only one patient in the sham acupuncture group was lost to follow-up. The study does, however, have limitations. Firstly, the lack of baseline prophylaxis is not typical, but the latest meta-analysis indicated that acupuncture reduced the frequency of migraine significantly more than did drug prophylaxis and was less likely to lead to dropout due to adverse effects and to reporting of adverse effects.[@ref32] In addition, we should respect the opinions of patients who are reluctant to use drug treatment; the decision to treat with drugs and the choice of a prophylactic drug should be made together with the patient.[@ref49] Secondly, the timeframe of this study may be not have been long enough to see a long lasting effect, although we observed that the treatment effects of manual acupuncture could persist for 12 weeks after treatment. In the future, trials with longer follow-up periods are needed to assess how long the effects of manual and sham acupuncture last. Clinical relevance ------------------ A considerable number of patients do not respond well to drug treatment, cannot tolerate the adverse effects of drugs, or have contraindications, which can lead to low adherence, chronification of headache, and acute drug overuse.[@ref50] Our results show that acupuncture can be recommended as an effective alternative treatment. Although sham acupuncture produced a minor benefit in reducing migraine attacks, the differences between manual acupuncture and sham acupuncture support the specific effects of manual acupuncture. On the basis of our results, clinical acupuncturists should pay careful attention to the specific components of acupuncture, including acupuncture points, needle manipulation, and acupuncture sensations, and the non-specific effects of acupuncture (acupuncture ritual or expectation for efficacy[@ref51] [@ref52]) should also not be neglected in clinical practice. In randomised controlled trials of acupuncture, the sham acupuncture group should not use penetrating needles. Conclusion and policy implications ---------------------------------- Among acupuncture naive patients with episodic migraine without aura, treatment with manual acupuncture, compared with sham acupuncture or usual care, resulted in a significantly greater reduction in the frequency of migraine days and migraine attacks. Acupuncture can be recommended as a prophylactic treatment when updating the guidelines and insurance coverage for migraine and its related substance misuse. When discussing prophylactic treatment strategies with patients, clinicians should provide them with information about acupuncture as an option. What is already known on this topic =================================== 1. Clinical evidence for the benefit of manual acupuncture in migraine prophylaxis remains scarce 2. Several recent clinical trials found that manual acupuncture was superior to no acupuncture 3. Comparisons between manual acupuncture and sham acupuncture found only minor or even no differences What this study adds ==================== 1. Manual acupuncture was more effective than sham acupuncture and usual care in reducing migraine headaches We thank Fengcai Zhu and Jingxin Li of the Jiangsu Provincial Center for Disease Control and Prevention, for guidance of the statistical analyses. We also thank the research assistants, acupuncturists, supporters of this study, and all of the patients who participated in this study. Extra material supplied by authors ###### Appendix 1: protocol ###### Appendix 2: acupuncture ritual ###### Appendix 3: supplementary tables ###### Appendix 4: meta-analysis Contributors: SX, LY, and XL contributed equally to the study. SX, LY, XL, HW, and WW conceived and designed the study. SX, LY, XL, HW, and WW had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. SX, LY, and XL drafted the manuscript. SX, LY, XL, MW, GC, QZ, ZZ, JS, HJ, HW, and WW contributed to the acquisition, analysis, and interpretation of data. All authors contributed to the critical revision of the manuscript for important intellectual content and gave final approval of the version to be published. WL did the statistical analysis. HW and WW obtained funding. GC, ZZ, JS, HJ, GH, FL, HW, and WW provided administrative, technical, or material support. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. HW and WW supervised the study and are the guarantors. Funding: This study was supported by a grant from the National Natural Science Foundation of China (61327902), and a project grant from Hubei University of Chinese Medicine/Hubei Provincial Collaborative Innovation Center of Preventive Treatment by Acupuncture and Moxibustion. Study sponsors were not involved in study design, data interpretation, writing the article, or the decision to submit the article for publication. Competing interests: All authors have completed the ICMJE uniform disclosure form at [www.icmje.org/coi_disclosure.pdf](http://www.icmje.org/coi_disclosure.pdf) and declare: funding for the submitted work as described above; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work. Ethical approval: This study was approved by the Clinical Trial Ethics Committee of Tongji Medical College, Huazhong University of Science and Technology (2016S009), the Medical Ethics Committee of Wuhan No.1 Hospital (WuWeiYiYuanLunShen\[2016\]14), the Ethics Committee of Hubei Province Hospital of Traditional Chinese Medicine (HBZY2016-C25-01), the Ethics Committee of Xiangyang No.1 Hospital (YN20171220), the Ethics Committee of The Second Hospital of Huangshi (HEYLWJY1701), and the Ethics Committee of Hubei University of Chinese Medicine (\[2018\] IEC No. 2). The trial was overseen by an independent trial steering committee and data monitoring and ethics committee. All patients gave written informed consent. Data sharing: The raw trial data after de-identification can be shared on individual request to the corresponding author at wwang\@vip.126.com. Transparency: The lead author affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained. Dissemination to participants and related patient and public communities: Participants will be informed of the results of the study by telephone. Dissemination to the public will be achieved through media outreach.	{ "pile_set_name": "PubMed Central" }
Introduction {#Sec1} ============ Percutaneous sacroiliac (SI) screw fixation is a minimally invasive and effective method for the stabilization of the posterior pelvic ring \[[@CR1]--[@CR6]\]. Screw placement under two-dimensional (2D) fluoroscopic control is the gold standard \[[@CR4], [@CR6]--[@CR8]\]. CT-controlled and navigated SI screw fixation techniques are demanding procedures and, therefore, not widely available beyond trauma centers \[[@CR6], [@CR9]--[@CR15]\]. Major complications of fluoroscopically controlled percutaneous SI screw fixation include nerve and vessel lesions following screw misplacement. The relative risk of lesions of the fifth lumbar nerve (L5), for example, is up to 8 % \[[@CR6], [@CR16]--[@CR20]\]. Variations in the morphology of the upper sacrum, widely known as ''sacral dysmorphism'', \[[@CR21], [@CR22]\] primarily account for this relatively high complication rate. Three-dimensional (3D) models of the upper sacrum have been described in the literature to assess safe corridors for SI screw placement even in the presence of sacral dysmorphism \[[@CR23]--[@CR25]\]. These techniques, however, require particular planning software and are typically not applicable to the surgeon during preoperative assessment. The aim of this study, therefore, was to present a simple method for the preoperative assessment of the sacral morphology using CT scans with widely available tools. In addition, it was our goal to classify the morphology of the upper sacrum based on the risk of iatrogenic nerve lesions during percutaneous SI screw fixation. Materials and methods {#Sec2} ===================== CT scans of a consecutive series of 1000 patients were analyzed in this study. The patients were included irrespective of gender, health conditions, or premorbidities. Exclusion criteria were:Age \<18 years.Recent or consolidated pelvic fractures.Osteolytic pelvic lesions. All patients underwent abdominal CT scans, which include the osseous pelvis, in a supine position with hip and knee joints extended. CT scans were performed during clinical routine. Indications for the CT scans were different from the aim of this study. Accordingly, the patients were not exposed to additional radiation for the purpose of this study. A 64-section CT scanner (LightSpeed VCT, GE Healthcare, Chalfont St. Giles, UK) was used to perform the CT scans. The CT images were taken with 0.6 mm slice depth and saved in slices of 1.25 mm in thickness. 2D reconstructions and measurements were performed with both a standard picture archiving and communication system (PACS) and software (Impax EE R20 XIV SU2, AGFA diagnostic software, Ridgefield Park, NJ). True inlet, outlet, and lateral views of the sacrum were obtained by using a two-dimensional reconstruction tool in a standardized and reproducible manner. In a first step, the midsagittal plane was selected in the lateral view (Fig. [1](#Fig1){ref-type="fig"}a). The axes were adjusted parallel to the lower endplate of S1 (red lines) and along the posterior boarder of S1 (blue line). In a next step, both axes were adjusted in the outlet view (Fig. [1](#Fig1){ref-type="fig"}b) to the vertical axis of the sacrum (green line) and parallel to a line connecting the cranial borders of the SI joint (red line). In a last step, the adjustment in the inlet view (Fig. [1](#Fig1){ref-type="fig"}c) included alignment in a sagittal axis from the middle of the promontorium to the spinous process (green line) and in the coronal plane parallel to posterior wall of the body of S1 (blue line). The same procedure was performed in S2 as well.Fig. 1a Alignment in the lateral view in a midsagittal plane: parallel to the lower endplate of S1 in the upper, middle, and lower quartiles of the body S1 (red lines) and along the posterior boarder of S1 (blue line). b Alignment in the outlet view: along the vertical axis of the sacrum (green line) and parallel to a line connecting the upper borders of the SI joint (red line). c Alignment in the inlet view: from the middle of the promontorium to the spinous process (green line) and parallel to the posterior wall of the body of S1 (blue line). d Assessment of the width of the corridor in the inlet view at the SI joint (diameter 1), in the middle of the distance between SI joint and midsagittal plane (diameter 2) and in the midsagittal plane (diameter 3) The assessment of the width of the corridor in the inlet view (inlet corridor) is shown in Fig. [1](#Fig1){ref-type="fig"}d. A line parallel the posterior wall of S1 (blue line) served as a reference for the measurement. The width of the corridor was defined as the perpendicular distance between the anterior cortex of the sacrum and this line. The corridor was assessed on both sides at the SI joint (diameter 1), in the middle of the distance between SI joint and midsagittal plane (diameter 2) and in the midsagittal plane (diameter 3) on the right side and left side. In addition, the corridor was assessed in the upper quartile, in the middle and in the lower quartile of the craniocaudal extension of the body S1 by parallel translation of the red axis in the sagittal view (Fig. [1](#Fig1){ref-type="fig"}a). The same procedure was performed in S2 as well. Subsequently, the diameters were used to calculate different types of inlet corridors. In a first step, the corridors were either defined as ascending, horizontal, or descending corridors. In an ascending corridor, the diameter 1 (SI joint) is smaller than the diameter 3 (midsagittal, Fig. [2](#Fig2){ref-type="fig"}a). In a descending corridor, the diameter 1 is larger than the diameter 3 (Fig. [2](#Fig2){ref-type="fig"}b). In a horizontal corridor, the diameters 1 and 3 are equal (Fig. [2](#Fig2){ref-type="fig"}c). A threshold of 3 mm was applied to eliminate marginal differences between the diameters, which are not clinically relevant. Accordingly, the diameter 3 in an ascending corridor should be minimum more than 3 mm larger than the diameter 1 (and vice versa in a descending corridor).Fig. 2a Ascending type. b Descending type. c Horizontal type In a second step, the presence of a ''notch'' was assessed. A notch is an indentation of the anterior cortex of the sacrum between the SI joint and the midsagittal plane with proximity to the nerve root L5, which, therefore, may be harmed during screw misplacement. The arithmetic mean between diameter 1 and 3 was calculated, which is equivalent to diameter 2 in the presence of a straight-lined anterior cortex between SI joint and the midsagittal plane. If the measured diameter 2 is more than 3 mm smaller than the calculated diameter 2, the presence of a notch was defined (Fig. [3](#Fig3){ref-type="fig"}a). Figure [3](#Fig3){ref-type="fig"}b shows an example of a horizontal corridor without notch.Fig. 3a Horizontal type with notch. b Horizontal type without notch SPSS Statistics 21.0 (SPSS, Chicago, IL, USA) was used for the statistical analysis. Metric scaled data are reported as arithmetic mean ± standard deviation and categorical data as absolute frequency and percentage distribution. Correlations for metric scaled data were quantified using the Pearson coefficient and for non-parametric samples using the Spearman coefficient. The Kolmogorov--Smirnov test was used for the determination of the distribution form. The probability level was set at p \< 0.05. Results {#Sec3} ======= CT scans of 1000 patients were analyzed in this study. Measurements were performed both on the right side and left side resulting in 2000 measurements. The mean age of the patients was 68.8 ± 21.5 years. There were 623 female (62.2 %) and 387 male patients (37.8 %) included. The results of the assessment of the different corridor types are shown in Table [1](#Tab1){ref-type="table"}. While the ascending type was the most frequent corridor type in the upper quartile of S1 (71 %), its relative frequency declined to 5 % in the lower quartile. There were no ascending types in S2. Descending types were found in 15 % in the upper quartile of S1 and in 86 % in the lower quartile. In S2, descending types were, by far, the most frequent corridor type. Horizontal types were less frequent with a relative incidence between 2 and 14 % (Table [1](#Tab1){ref-type="table"}). The results of the assessment of the presence of a notch are shown in Table [2](#Tab2){ref-type="table"}. We found a notch in the majority of cases with relative frequencies ranging from 69 % (upper quartile of S1) to 95 % (upper quartile of S2) (Table [2](#Tab2){ref-type="table"}). The mean depths of the notch are shown in Table [3](#Tab3){ref-type="table"}. There were no significant differences between the different groups.Table 1Distribution of the relative frequency of the different corridor types in the upper, middle, and lower quartiles of S1 and S2![](402_2016_2528_Figa_HTML.gif){#d29e530}S1 upper quartile (%)S1 middle quartile (%)S1 lower quartile (%)S2 upper quartile (%)S2 middle quartile (%)S2 lower quartile (%)Ascending type71215000Horizontal type14129225Descending type156686989895 Table 2Distribution of the relative frequency of the presence of a notch in the upper, middle, and lower quartiles of S1 and S2S1 upper quartileS1 middle quartileS1 lower quartileS2 upper quartileS2 middle quartileS2 lower quartileNumber (n = 2000)138014951878189716041711Percent69 %75 %94 %95 %80 %86 % Table 3Depth of the notch in the upper, middle, and lower quartiles of S1 and S2 (in mm)Mean (±SD)S1 upper quartile9.9 (±7.0)S1 middle quartile8.5 (±5.2)S1 lower quartile9.4 (±4.3)S2 upper quartile11.8 (±7.6)S2 middle quartile6.1 (±2.3)S2 lower quartile7.5 (±5.1) Discussion {#Sec4} ========== Percutaneous SI screw fixation under fluoroscopic control is an effective and widely used method for the stabilization of the posterior pelvic ring. Screw misplacement with consecutive nerve lesions, mainly of the nerve root L5, is a major complication of this technique \[[@CR6], [@CR16]--[@CR20]\]. Interindividual variations in the morphology of the upper sacrum, the so-called sacral dysmorphism, particularly account for these findings. Sacral dysmorphism has been described by several authors. Most of these studies, however, are anatomical studies describing the morphological variance of the upper sacrum without assessing their relevance for percutaneous SI screw fixation \[[@CR26]--[@CR28]\]. Other authors assess the sacral morphology considering its clinical relevance for SI screw fixation. Carlson, for example, described the three-dimensional shape of the corridor for SI screw fixation (vestibule concept) in a CT-based study, \[[@CR23]\], while Mendel showed a CT-based 3D model of secure bone corridors (safe zones) and optimal trajectories for sacroiliac screws \[[@CR24]\]. The authors, additionally, provided recommendations for the angulation of SI screws in different planes. These recommendations, however, differ significantly between the authors, which, obviously, is not surprising given the variations in sacral morphology. The angle in a caudocranial direction (outlet view), for example, ranged from 2° to 18°, while the angle in a posteroanterior direction (inlet view) ranged from 2° to 16° \[[@CR23], [@CR24]\]. In conclusion, a ''one-size-fits-all'' approach to percutaneous SI screw fixation is not feasible given the diversity of the sacral morphology. To avoid screw misplacement in percutaneous SI screw fixation, CT guided and navigated procedures have been developed \[[@CR9]--[@CR12], [@CR14], [@CR15], [@CR29]\]. These procedures allow for secure screw placement even in the presence of narrow osseous corridors and sacral dysmorphia, which was demonstrated in the above-mentioned studies. Besides a higher radiation exposure \[[@CR14]\], the major drawback of these techniques, however, is their limited availability beyond trauma centers. Accordingly, SI screw fixation under fluoroscopic control is and will continue to be an important technique for minimally invasive stabilization of the posterior pelvic ring. SI screw fixation under fluoroscopic control not only relies on sound intra-operative imaging, but also on the surgeon´s ability to interpret the two-dimensional images in a 3D context. Several authors have provided different approaches for the interpretation of these images in an inlet, outlet, and lateral views \[[@CR4], [@CR30]--[@CR32]\]. Routt, for example, reported that abnormal morphological patterns of the upper sacrum could be easily identified using mainly the outlet view and the lateral view \[[@CR30]\]. We agree that the outlet view allows for secure screw placement above the first sacral foramina in S1 even in the presence of dysmorphic foramina. The lateral view allows for the assessment of the upper, lower, and posterior borders of the osseous corridor. However, we disagree that the lateral view is valuable for the assessment of the anterior border of the corridor. First, the anterior border is not strict parallel to the projection in the lateral view. Our data show that the anterior border may be divergent (ascending type), horizontal, or convergent (descending type) relative to the posterior border, which highly influences its radiological appearance in the lateral view. Second, indentations of the anterior cortical border of the corridor (notches) may not be assessable in a view axial to this border. The techniques of corridor assessment described by Mendel \[[@CR31]\] and Noojin \[[@CR32]\] neither assess the ''notch'' problem as well. Miller \[[@CR33]\] summarized the three key points of fluoroscopically controlled SI screw fixation in the presence of sacral dysmorphism in a recent review article as follows: First, dysmorphic sacral neuroforamina are assessed in the outlet view. Second, the posterior wall and variances in the sacral alar slope are assessed in the lateral view. Third, notches of the anterior cortex are assessed in the inlet view. According to our own experience, however, the inlet view may not be appropriate to reliably detect notches of the anterior cortex in all cases. The inlet view is a two-dimensional projection with an overlap of the anterior cortex of the vertebral bodies S1 and S2. Our data, however, show that the anterior cortex of the upper sacrum is not a flat surface in a three-dimensional space, which would result in a single line in a two-dimensional projection. The corridor type, for example, changes very frequently within the vertebral body of S1 (Table [1](#Tab1){ref-type="table"}). In addition, the presence or absence of a notch is not detectable in an inlet view in the majority of cases and further limits the assessment of the anterior cortex of the upper sacrum in this view. The fluoroscopic control during SI screw fixation is adequate when drilling in a strict transverse direction. Transverse corridors of adequate width, however, are present in a minority of patients only. Ascending corridors with a notch have a very narrow intraosseous corridor in a transverse direction. These types require a more posterior entry point (even posterior to the posterior wall of S1 in a lateral view) and a more ascending drilling direction, which increases the risk of nerve lesions particularly in the additional presence of a notch. The aim of our study, therefore, was to develop a simple method for the preoperative assessment of the corridor type and the presence of a notch with widely available tools as shown above. We consider horizontal and descending corridor types without a notch to be feasible for fluoroscopically controlled SI screw fixation, while ascending types with notch bear a higher risk of iatrogenic nerve lesions. In these cases, we recommend navigated SI screw fixation. One interesting finding of our study was the high frequency of ascending corridors in the upper quartile of S1 with a much lower frequency in the middle quartile and lower quartile of S1. An ascending corridor in the upper quartile is mainly determined by the inclusion of the sacral promontory in the corridor analysis. In the upper quartile, there was an additional notch in 69 % of the cases as well. Accordingly, we consider a screw trajectory to the promontory to be associated with a higher risk of L5 nerve root lesions than more flat screw trajectories. One limitation of our study has to be noted. It was our goal to classify the morphology of the upper sacrum based on the risk of iatrogenic nerve lesions during percutaneous SI screw fixation. Accordingly, we focused on the corridor morphology in the inlet view and defined different inlet corridor types as well as the presence of a notch. For a safe overall screw placement additional parameters, such as the absolute diameter of the corridor in mm and corridors in the outlet view are relevant as well. The assessment of these parameters, however, was not the goal of this study. In conclusion, our concept obviously is not a "one-size-fits-all'' concept neither. It aims at three goals: (a) to increase the awareness of the surgeon for the problem of sacral dysmorphism, (b) to emphasize the importance of a preoperative assessment of the osseous corridor, and (c) to provide a simple method for the preoperative assessment with widely available tools. A correction to this article is available online at <https://doi.org/10.1007/s00402-018-2928-7>. Open access funding provided by Medical University of Innsbruck. Conflict of interest {#FPar1} ==================== The authors declare that they have no conflict of interest.	{ "pile_set_name": "PubMed Central" }
All relevant data are within the paper. Introduction {#sec005} ============ Epilepsy and intellectual/developmental disabilities (ID/DD) are both common pediatric neurological disorders. ID/DD is one of the main comorbidities of epilepsy \[[@pone.0141782.ref001]\] with about a quarter of epileptic children having ID/DD. The prevalence is much higher in children presenting with early-onset (before three months after birth) epilepsy \[[@pone.0141782.ref002]--[@pone.0141782.ref004]\]. Although frequent and refractory seizures may cause cognitive and motor regression, common pathophysiological mechanisms may be responsible for the high rate of co-occurrence of epilepsy and ID/DD \[[@pone.0141782.ref005]\]. The etiologies of these two disorders are complex and diverse with the majority being unknown. Genetic factors play a major role in the etiologies of epilepsy and ID/DD, especially in pediatric patients, who are highly heterogeneous. Defects in many genes have been reported as shared underlying mechanisms of epilepsy and ID/DD \[[@pone.0141782.ref006]--[@pone.0141782.ref009]\]. These genes are involved in different pathways. Ion channel genes, which are particularly relevant to epilepsy, account for a significant proportion \[[@pone.0141782.ref010]\]. However, the phenotypes related to these genes are difficult to differentiate clinically, and the detection of mutations in suspected genes is always a challenge. Seizures were always intractable and indicated a poor prognosis when co-occurring with ID/DD. Therefore, knowing the genetic background and pathogenesis of epilepsy and ID/DD is valuable not only for diagnosis and prognosis, but also for genetic counseling and treatment. Many genes related to epilepsy and ID/DD have been reported. However, for individual patients with non-syndromic epilepsy, it has been difficult to isolate the causative gene mutations from a large number of possible candidate genes using conventional Sanger sequencing. The rapidly developing targeted next-generation sequencing (NGS) has been proved to be a fast, economic and accurate approach for screening gene mutations in disorders with both genetic and phenotypic heterogeneity, including epilepsy and ID/DD \[[@pone.0141782.ref011]\]. In this study, we used targeted NGS to investigate 300 candidate genes related to epilepsy and ID/DD in 253 Chinese children with unexplained epilepsy and ID/DD. We aim to make genetic diagnosis for these patients and find clues to help us explain the common genetic background of epilepsy and ID/DD. Methods {#sec006} ======= Ethics statement {#sec007} ---------------- Written informed consent was obtained from the parents of all the patients. This study was approved by the Institutional Review Boards of Peking University First Hospital. All data of this study were analyzed anonymously. Patients {#sec008} -------- A total of 253 Chinese children with unexplained epilepsy and ID/DD were recruited from the Department of Pediatrics, Peking University First Hospital from 2006 to 2014. All patients were clinically diagnosed as having epilepsy and ID/DD of unknown origin, including 65 patients diagnosed as early-onset (before three months after birth) epilepsy. Nevertheless, it was strongly suspected that the etiology of these patients' diseases was genetic, owing to the following evidences: (1) no definite perinatal brain injury; (2) no hypoxia, ischemia, infection of the central nervous system or cranial trauma; (3) no evidence of typical inherited metabolic disorders or specific neurodegenerative disorders based on clinical features, neuroimaging or blood/urinary metabolic diseases screening; (4) normal routine karyotyping and (5) the detection of chromosome sub-telomeric rearrangements with multiplex ligation-dependent probe amplification (MLPA) showing no abnormalities \[[@pone.0141782.ref012]\]. 253 cases of our cohort included cases from 246 trios, of which father or mother of four trios (1649, 5165, 5168, and 5237) had epilepsy history, and parents of all the other 242 trios had no epilepsy and any related history. Other seven cases were probands of seven families with more than one patient, of which six families (3604, 5750, 6047, 6364, 6526, and 6636) consisted of two affected children and their unaffected parents, and one family (5240) with five affected patients (proband, father, uncle, aunt and grandfather) accorded with autosomal dominant inheritance. All genomic DNA used in experiments were extracted from peripheral leukocytes. Targeted next-generation sequencing {#sec009} ----------------------------------- In accordance with the literatures searched within online databases, a total of 300 candidate genes associated with epilepsy and ID/DD were selected as the genes of interest. We used a custom-designed gene panel, synthesized using the Agilent Sure-Select Target Enrichment technique (Zhongguancun Huakang Gene Institute, China), to capture the coding regions from the 300 genes, including their exons and exon-intron boundaries (1.285M bp in total). The following NGS was performed on an Illumina GAIIx platform (Illumina, San Diego, CA, U.S.A.) using paired-end sequencing of 110 bp. Bioinformatics analysis of the raw data included the following steps: (1) image analysis using RTA software (real-time analysis, Illumina); (2) base calling using CASAVA software v1.8.2 (Illumina); (3) filtered out duplicate and low base quality score reads using the Genome Analysis Tool kit (GATK); (4) aligned clean paired-end reads to the human reference genome build hg19 using BWA software (Pittsburgh Supercomputing Center, Pittsburgh, PA, U.S.A.); (5) identified insertion-deletions (indels) and single-nucleotide polymorphisms (SNPs) using the GATK and annotated using ANNOVAR; (6) performed in silico pathogenicity prediction of novel missense variations using Polyphen2 \[[@pone.0141782.ref012], [@pone.0141782.ref013]\]. The sequencing depth was more than 5X (range of 5X-185X; average of 136X), and the mean coverage was 98.56%. On average, 423 variations within the 300 genes were found in each patient. We then formulated the following filtering criteria to determine every possible pathogenic variation from the large amount of initial variations: (1) insertion/deletion variations; (2) premature/delayed termination codon variations; (3) splice site variations including substitution at nucleotide +1/-1 of intron; (4) missense variations predicted by polyphen2 as probably/possibly damaged or benign. The variation meeting any one of the above criteria was considered to be a candidate for pathogenic variations and was selected for further analysis \[[@pone.0141782.ref012]\]. According to the HGMD Professional database, the 1000 Genomes Browser, PubMed and the UCSC database, we marked the reported pathogenic mutations and excluded known polymorphisms. Finally, on average 17 possible pathogenic variations were identified in each patient (range 5--27). We chose variations which to validate according to the known inheritance pattern of the involved genes. Heterozygous variations of genes with autosomal or X-linked dominant heredity, homozygous or compound heterozygous variations of genes with autosomal recessive heredity, and hemizygous variations of genes with X-linked recessive heredity were regarded as likely causative variations. We performed validation and parental origin analyses for these variations by conventional Sanger sequencing, and confirmed causative mutations according to parental origin of the variations and clinical features of the patients \[[@pone.0141782.ref012]\]. Protein structure modelling {#sec010} --------------------------- The homology modelling server SWISS-MODEL \[[@pone.0141782.ref014]--[@pone.0141782.ref016]\] was used to predict the tertiary structure of KCNAB1 protein. We chose the protein crystal structure 3EAU \[[@pone.0141782.ref017]\] of KCNAB2 from Protein Data Bank (PDB) \[[@pone.0141782.ref018]\] as the template. For the homology of KCNAB1 and KCNAB2, the structure from 89 to 413 amino acids of KCNAB1 protein (UniProtKB ID: Q14722) were predicted. Results {#sec011} ======= We identified causative mutations within 24 genes in 46 patients of our cohort, including two likely pathogenic mutated genes in two patients. The total detection rate of our study was 18% (46/253) in the whole group and 26% (17/65) in the early-onset epilepsy group. The detected mutations included 32 novel and 16 reported mutations. Nineteen of the mutations were severe, including eight premature/delayed termination codon mutations, ten insertion/deletion mutations and one splicing site mutation. The remaining 29 mutations were non-synonymous, including 27 mutations predicted to be "probably damaging" and two mutations (AFF2 p.Gly547Asp and RELN p.Val3426Ile) that were predicted to be "possibly damaging" by Polyphen2. In two patients (5240 and 6189) with distinct epilepsy and ID/DD phenotypes, two mutated genes were regarded as likely pathogenic due to the unmatched phenotype or the unavailability of segregation analysis in the family. An overview of the clinical features of patients and their mutations is described in [Table 1](#pone.0141782.t001){ref-type="table"}. 10.1371/journal.pone.0141782.t001 ###### Overview of clinical features of the patients and the mutations. ![](pone.0141782.t001){#pone.0141782.t001g} Gene Case Sex Study age (y) Seizures (onset age) ID/DD Nucleotide substitution Amino acid substitution Parental origin Novel/ reported Final diagnosis ----------- ------------------------------------------- ----- --------------- ------------------------------------------------------- -------------------- ------------------------- -------------------------- ----------------- ----------------- ----------------- SCN1A 152 M 8 GTCS, FS (6 m) Severe c.4547C\>A p.Ser1516Ter De novo Reported DS 2038 M 10 GTCS, PS (4 m) Severe c.2134C\>T p.Arg712Ter De novo Reported DS 5791 M 3 PS, GTCS (3 m) Severe c.4942C\>T p.Arg1648Cys De novo Reported DS 6047 M 3 PS, FS, Myoclonus, Absences (8 m) Severe c.2589+3A\>T \- De novo Reported DS 6207 F 15 PS. GTCS, FS (3 m) Severe c.3733C\>T p.Arg1245Ter De novo Reported DS 6222 F 0.5 PS, SE (16 d) Severe c.659T\>A p.Val220Asp De novo Novel MMPSI 6300 F 3 PS, Myoclonus, GTCS, SE, FS (6 m) Severe c.3372delT p.Phe1124LeufsTer4 De novo Novel DS 6492 M 4 PS, FS (7 m) Severe c.2488G\>T p.Glu830Ter De novo Novel DS SCN8A 3129[\](#t001fn002){ref-type="table-fn"} M 9.5 GTCS, FS (11 m) Severe c.2668G\>A p.Ala890Thr De novo Novel 5487[\](#t001fn002){ref-type="table-fn"} M 3.5 GS (6 m) Severe c.4850G\>A p.Arg1617Gln De novo Reported EE 6219[\](#t001fn002){ref-type="table-fn"} F 1.5 PS, GTCS, Myoclonus, IS (3 d) Severe c.1221G\>C p.Leu407Phe De novo Novel EIEE 6325[\](#t001fn002){ref-type="table-fn"} F 1 PS, GTCS (2.5 m) Moderate to severe c.2549G\>A p.Arg850Gln De novo Novel EIEE 6908[\](#t001fn002){ref-type="table-fn"} M 1.5 GTCS (4 m) Moderate to severe c.4787C\>G p.Ser1596Cys De novo Novel EE YL F 3 PS (6 m) Severe c.4935G\>A p.Met1645Ile De novo Novel KCNQ2* 597 M 3 GS (3 d) Severe c.365C\>T p.Ser122Leu De novo Reported EIEE 2202 M 2 PS, GTCS, FS (3 d) Severe c.956A\>C p.Lys319Thr De novo Novel EIEE 2312 M 1.5 PS, FS (4 h) Severe c.830C\>T p.Thr277Ile De novo Novel EIEE 5630 M 1 PS, IS (3 d) Severe c.1655A\>C p.Lys552Thr De novo Reported EIEE IQSEC2 3481 M 9 PS (3 y 6 m) Severe c.88_90delATC p.Ile30del De novo Novel 3604 M 7 PS, FS (6 m) Severe c.1049C\>A p.Ala350Asp Mother Novel 5292 M 2 PS, IS, Myoclonus (1 y 4 m) Severe c.2846_2852delCCCAGGT p.Ser949CysfsTer7 De novo Novel CDKL5 1613 F 1 PS (40 d) Severe c.2314delA p.Lys772ArgfsTer12 De novo Novel EIEE 5057 F 1.5 PS (1 m) Severe c.528G\>A p.Trp176Ter De novo Novel EIEE DYRK1A 2091 M 7 GTCS, FS, SE (1 y 6 m) Severe c.859G\>T p.Asp287Tyr De novo Novel 2959 F 3 GS (2 y) Severe c.946C\>T p.Gln316Ter De novo Novel GABRB3 1538 M 1 PS, GTCS (6 m) Severe c.914C\>T p.Ala305Val De novo Novel SYH M 1.5 PS (9 m) Severe c.509T\>G p.Leu170Arg De novo Novel PCDH19 6526 F 9 PS (5 m) Severe c.1091delC p.Pro364ArgfsTer4 Father Reported LXX F 3 PS, FS (9 m) Moderate c.370G\>A p.Asp124Asn Father Novel STXBP1 527 M 2 IS (7 d) Severe c.568C\>T p.Arg190Trp De novo Reported EIEE MYS M 2 IS (3 d) Severe c.568C\>T p.Arg190Trp De novo Reported EIEE AFF2 6636 M 11 PS (6 y) Moderate c.1640G\>A p.Gly547Asp Mother Novel ALDH7A1 5921 M 4 PS (3 m) Moderate to severe c.1553G\>C c.1061A\>G p.Arg518Thr p.Tyr354Cys FatherMother Novel Reported PDE ATP1A2 5871 M 1.5 PS, FS (1 m) Mild c.2563G\>A p.Gly855Arg De novo Reported CASK 2584 F 6 IS (1 y) Severe c.2141delC p.Ala714GlufsTer13 De novo Novel MICPCH FOXG1 2539 M 2 GTCS, PS, Laugh attack (6 m) Severe c.738C\>A p.Tyr246Ter De novo Novel GRIN2A 6245 F 6 PS, ESES (5 y) Severe c.2191G\>A p.Asp731Asn De novo Reported GRIN2B 1503 M 2 PS, IS, Myoclonus, Tonic, Startle (6 m) Severe c.1985A\>C p.Gln662Pro De novo Novel KCNAB1 HY F 4 PS, GTCS (10 d) Severe c.1062dupCA p.Leu355HisfsTer5 De novo Novel EIEE PRRT2 5240 F 1 GS, IS (3 m) Severe c.649dupC p.Arg217ProfsTer8 Father Reported RELN 6235 F 7 Myoclonus (3 y) Severe c.10276G\>A c.2252A\>C p.Val3426Ile p.Lys751Thr Father Mother Novel Novel SHANK3 ZXT M 3 IS (1 y 10 m) Severe c.3598G\>C p.Ala1200Pro De novo Novel SLC2A1 1649 M 5.5 GTCS (4 y 8 m) Mild c.1477T\>C p.Ter493ArgextTer56 Mother Novel SYNGAP1 5828 F 5 Atypical absences, Myoclonus, Atonic, GTCS, FS (10 m) Severe c.829dupC p.Pro277ProfsTer7 De novo Novel UPF3B 6189 M 2 IS (8 m) Severe c.883T\>A p.Leu295Met Mother Novel ZEB2 4620 M 3 PS (?) Severe c.1426_1427insA p.Met476AsnfsTer6 De novo Reported M, male; F, female; y, years; m, months; d, days; h, hours; PS, partial seizures; FS, febrile seizures; SE, status epilepticus; DS, Dravet syndrome; EIEE, early infantile epileptic encephalopathy; EE, epileptic encephalopathy; IS, infantile spasms; GS, generalized seizures; PDE, pyridoxine-dependent epilepsy; MMPSI, malignant migrating partial seizures in infancy; ESES, electrical status epilepticus in sleep; MICPCH, mental retardation and microcephaly with pontine and cerebellar hypoplasia \have been reported in \[[@pone.0141782.ref012]\]. Among the 46 cases with a causative mutation, patients with an SCN1A* mutation accounted for the largest proportion of 17% (8/46), followed by SCN8A, KCNQ2 and IQSEC2 of 13% (6/46), 9% (4/46) and 7% (3/46) respectively. A total of 38% (9/24) of the mutated genes (SCN1A, SCN8A, KCNQ2, IQSEC2, CDKL5, DYRK1A, GABRB3, PCDH19, and STXBP1) reoccurred at least two times in this study, and 63% (15/24) of the mutated genes (ALDH7A1, AFF2, ATP1A2, CASK, FOXG1, GRIN2A, GRIN2B, KCNAB1, PRRT2, RELN, SHANK3, SLC2A1, SYNGAP1, UPF3B and ZEB2) occurred only one time. All mutated genes detected in this study along with the number of patients in whom each gene was detected were shown in [Fig 1](#pone.0141782.g001){ref-type="fig"}. ![Pathogenic (red) and likely pathogenic (green) mutated genes identified in 253 patients with unexplained epilepsy and intellectual/developmental disabilities.](pone.0141782.g001){#pone.0141782.g001} Notably, we detected a novel de novo heterozygous mutation (c.1062dupCA p.Leu355HisfsTer5) within KCNAB1 in one patient with EIEE. This mutation led to a premature termination codon. The patient with the mutation was a four-year-old girl with early-onset seizures (onset at ten days after birth). The frequent partial seizures were followed by intractable generalized tonic-clonic seizures (GTCS). Until now, she has used seven antiepileptic drugs (AEDs). Sodium Valproate, Levetiracetam, Topiramate and Clonazepam had no obvious effect, while Oxcarbazepine, Lamotrigine and Zonisamide reduced the seizures; among them Zonisamide add-on had the best effect with a 75% reduction of the seizures. Her parents are planning to consent for her to receive vagal nerve stimulation (VNS) to treat her intractable epilepsy. She also has severe ID/DD now (non-verbal, limited interaction with her parents, but can walk independent from 2 years 3 months), with occasional panic attacks. Her prenatal history was normal and neurological examination was unremarkable. One of her electroencephalograph (EEG) recordings showed a slow spike-wave in the left frontotemporal region, while another EEG recorded frequent sharp waves in the left temporooccipital region. Her cranial magnetic resonance imaging (MRI) was normal. KCNAB1 encodes the beta-1 member of the shaker-related family of voltage-gated potassium channels. This member includes three isoforms (Kvβ1.1- Kvβ1.3) of the KCNAB1 gene \[[@pone.0141782.ref019]--[@pone.0141782.ref023]\]. Kvβ1.1, the longest isoform, is restricted expressed in brain \[[@pone.0141782.ref024]\]. The shaker-related voltage-gated K+ (Kv) channels consist of alpha and beta subunits \[[@pone.0141782.ref025]\]. The beta subunits modulate the gating properties of the alpha-subunit potassium channels. Voltage-dependent potassium channel proteins are responsible for the electrical properties of excitable cells and play physiological roles in non-excitable cells \[[@pone.0141782.ref026]\]. To further study the pathogenicity of p.Leu355HisfsTer5 mutation, protein tertiary structures of wild type and p.Leu355HisfsTer5 mutation of KCNAB1 were predicted using the SWISS-MODEL. Because the template 3EAU was a homotetramers crystal structure, we predicted both monomer and tetramer of KCNAB1 protein. According to the monomer prediction ([Fig 2A](#pone.0141782.g002){ref-type="fig"}), the mutated KCNAB1 protein lost C-terminal helices when compared to wild type. According to the tetramer prediction ([Fig 2B](#pone.0141782.g002){ref-type="fig"}), the wild type was able to bind to NADP^+^ by sharing the same binding domains of KCNAB2 protein (data from: <http://www.uniprot.org/uniprot/Q14722>). But the mutated KCNAB1 protein was not able to bind to NADP^+^ for losing a NADP^+^ binding domain (375--381 amino acids), though might still be able to form a tetramer. Previous study of experimental point mutation within NADP^+^ binding domain of KCNAB1 protein showed significant effects on Kv1 channel trafficking and axonal targeting \[[@pone.0141782.ref027]\]. Therefore, the p.Leu355HisfsTer5 mutation that we identified is probably pathogenic. ![Structure modeling of wild type and p.Leu355HisfsTer5 mutation of KCNAB1 with SWISS-MODEL.\ A: Monomer view: comparing with wild type, the mutant lost C-terminal structure of the protein; B: Tetramer view: comparing with wild type, the mutant might lose the ability of binding NADP^+^.](pone.0141782.g002){#pone.0141782.g002} To understand a possible common genetic mechanism of epilepsy and ID/DD, we classified 24 mutated genes identified in our study into groups according to gene function. We found that ion channel genes had the largest percentage of occurrence, 33% (8/24) with genes related to synapse coming second at 21% (5/24). Other genes also identified were classified as having functions in transcriptional regulation, protein kinase modulation, cell metabolism and cell-cell interaction. The classification and the number of genes of each group were shown in [Table 2](#pone.0141782.t002){ref-type="table"}. 10.1371/journal.pone.0141782.t002 ###### The classification of mutated genes. ![](pone.0141782.t002){#pone.0141782.t002g} Gene Function Gene (n) Mutated Genes -------------------------- ---------- ----------------------------------------------------------------------------- Ion channel 8 SCN1A, SCN8A, KCNQ2, GABRB3, ATP1A2, GRIN2A, GRIN2B, KCNAB1 Synapse 5 IQSEC2, STXBP1, PRRT2, SHANK3, SYNGAP1 Transcription regulation 4 AFF2, FOXG1, UPF3B, ZEB2 Protein kinase 3 CDKL5, DYRK1A, CASK Cell metabolism 2 ALDH7A1, SLC2A1 Cell-cell interaction 2 PCDH19, RELN Discussion {#sec012} ========== In this study, we discovered 32 novel and 16 reported mutations within 24 genes in 46 patients of our cohort, including two likely pathogenic mutated genes in two patients. The total detection rate of our study was 18% (46/253) in the whole group and 26% (17/65) in the early-onset epilepsy group. Early-onset epilepsy had the relatively higher detection rate. We made genetic diagnosis for these 46 patients. This was critical for them to improve further management and genetic counseling for their epilepsy. We expanded the phenotype and mutation spectrum of the 24 genes identified in our study. This provided more information for further understanding of these disease-causing genes. Patients with an SCN1A mutation accounted for the largest proportion, 17% (8/46), of which seven patients were diagnosed as Dravet syndrome and one patient was diagnosed as MMPSI. MMPSI cases with a SCN1A mutation have been reported and MMPSI is regarded as the most severe phenotype of SCN1A to date \[[@pone.0141782.ref028], [@pone.0141782.ref029]\]. Our MMPSI patient with a novel SCN1A mutation provided further evidence that SCN1A defects play an important role in MMPSI. A total of 63% (15/24) of the mutated genes occurred only one time in our study; therefore, it seems that epilepsy and ID/DD are phenotypes that occur as a consequence of brain dysfunction caused by highly diverse mutated genes, most of which are isolated and fit the rule of common disease rare variations. In addition, the spectrum of mutated genes in our study is rather different from those reported in other similar studies \[[@pone.0141782.ref011], [@pone.0141782.ref030]\]. Population diversity and different inclusion criteria (both for patients and candidate genes) may account for this inconsistency. In this study, we identified a novel de novo heterozygous mutation (c.1062dupCA p.Leu355HisfsTer5) within KCNAB1 in one patient with EIEE. KCNAB1 has been reported as a susceptibility gene for epilepsy, particularly temporal lobe epilepsy (TLE), but no pathogenic mutation has been reported. An association study of 2717 epileptic patients reported that numerous SNPs located within KCNAB1 contributed to the susceptibility to epilepsy. These patients manifested various forms of epilepsy \[[@pone.0141782.ref031]\]. Furthermore, KCNAB1 was regarded as a candidate gene for lateral temporal epilepsy (LTE) because of its functional interaction with LGI1 \[[@pone.0141782.ref032]\], the disease-causing gene of autosomal dominant LTE (ADLTE). However, sequencing of KCNAB1 in ADLTE families without LGI1 mutations failed to identify any mutations. This suggested that KCNAB1 does not act as a major disease-causing gene in ADLTE \[[@pone.0141782.ref033]\]. Nevertheless, another association study of 142 LTE patients suggested that KCNAB1 may be a susceptibility gene of LTE \[[@pone.0141782.ref034]\]. In addition; a genome-wide scan study was conducted on a TLE family. Linkage analysis identified a locus on chromosome 3q25-q26. KCNAB1 was one of the highest priority genes in this region, but sequencing of KCNAB1 was unable to identify any mutations \[[@pone.0141782.ref035]\]. In summary, although several previous studies have supported the association of KCNAB1 with epilepsy, no KCNAB1 mutations have been reported in patients with this disease previously. However, a mouse model of KCNAB1 disruption showed significant alterations in hippocampal learning and memory functions \[[@pone.0141782.ref036]\], supporting a possible relationship between KCNAB1 defects and brain dysfunction. Here we reported an epileptic patient with a KCNAB1 mutation, which supports the relationship between KCNAB1 dysfunction and epilepsy, and interestingly, the epileptic discharges of this patient located mostly at temporal region. We also found four patients with a KCNQ2 mutation. Because KCNQ2 \[[@pone.0141782.ref037]\] and KCNT1 \[[@pone.0141782.ref038]\] have already been reported to be common and important genes for epileptic encephalopathy, adding the recently reported KCNA2 \[[@pone.0141782.ref039]\], KCNH1 \[[@pone.0141782.ref040]\], KCNC1 \[[@pone.0141782.ref041]\] genes and now the KCNAB1 gene, this led us to pay more attention to the potassium channel genes as a group in epilepsy, especially epileptic encephalopathy. Two mutated genes were regarded as likely having pathogenic mutation in our study. First, we detected a heterozygous mutation (p.Arg217ProfsTer8) within PRRT2 in one family (5240). The proband had severe ID/DD and infantile spasms, but other individuals with the same mutation in this family had benign epilepsy during infancy and normal intelligence during adulthood. We hypothesized that the PRRT2 mutation in the proband may only increase the risk of epilepsy, while another undiscovered mutated gene may instead contribute to the severe phenotype. In addition, we discovered a hemizygous mutation (p.Leu295Met) in UPF3B which have been reported as a causative gene of X-linked recessive mental retardation. The previously reported cases with mutation in UPF3B had no seizures. The mutation we found is novel, but probably damaging predicted by Polyphen2. Unfortunately, DNA samples of other male maternal family members were unavailable to make sure the pathogenicity of this mutation. To understand a possible common genetic mechanism of epilepsy and ID/DD, we classified the mutated genes identified in our cohort according to gene function. We have found that ion channel genes had the largest percentage of occurrence. This suggests that ion channels play a vital role in the pathogenesis of epilepsy and ID/DD. Activation of neurotransmitter receptor ion channels at synapses promotes synaptic plasticity during brain development. Therefore, abnormal ion transport may affect neural excitability and brain development, resulting in epilepsy and ID/DD \[[@pone.0141782.ref042]\]. Further, synapse formation and normal function are essential in the signaling and the formation of neural networks. Genes related to synapse formation and function were also closely related to epilepsy and ID/DD. In addition, some factors in transcriptional regulation, protein kinase modulation, cell metabolism and cell-cell interaction may also participate in the common pathogenesis of epilepsy and ID/DD. However, relevant details remain unclear. We believe further study of the common pathogenesis of epilepsy and ID/DD are urgently needed. In our study, with the detection rate of 18%, the targeted NGS is certainly supposed to be an efficient and precise approach to screen monogenic mutations in patients with highly heterogeneous disorders such as epilepsy and ID/DD. However, according to our experience, some limitations of this approach and tips for best performance should be discussed here. First, owing to false positive results, conventional Sanger sequencing is definitely required for the validation of the variations supposed to be significant, especially when the targeted regions have insufficient coverage. Second, on the other hand, false negative results may also occur and may lead to the loss of crucial data. This might be one of the reasons that our study failed to detect gene mutations in the other 207 patients. Third, DNA samples of the parents and other affected or even unaffected members of families are essential to analyze the pathogenicity of the variations. Availability of almost all parental DNA samples in our study played a significant role in data analysis. However, the unavailability of other members in a few families hampered further confirmation of their etiology. Finally, precise clinical data is a prerequisite, without which the genetic diagnosis cannot be made. For example, patient 5871 who carried a de novo mutation in ATP1A2 also had an inherited homozygous mutation (p.Ile105Val) in CLN3. Although the CLN3 nonsynonymous mutation was predicted to be "probably damaging", we still excluded its pathogenicity according to his phenotype, not like neuronal ceroid lipofuscinoses clinically. In summary, we used targeted NGS to investigate causative gene mutations in Chinese children with unexplained epilepsy and ID/DD. We established genetic diagnosis for 46 patients of our cohort and expanded the phenotype and mutation spectrum of 24 genes associated with epilepsy and ID/DD. This study is the first to identify a KCNAB1 mutation in a patient with EIEE. More cases with mutations in this gene are needed to confirm and clarify its role in epilepsy. We thank our patients for participating in this study. We thank Dr. Macdonald (Department of Neurology, Vanderbilt University Medical Center) for his kindly critical reading and valuable suggestion to this manuscript. [^1]: Competing Interests:The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: YJ. Performed the experiments: YZ WK. Analyzed the data: YZ YJ WK YG KG HX. Contributed reagents/materials/analysis tools: YJ XL YW YZ JW FG XW. Wrote the paper: YJ YZ.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1-animals-08-00027} =============== Food guarding (FG) is a natural behavior for Canids, yet can result in adverse consequences when domestic dogs exhibit this behavior in animal shelters. FG can occur when a dog has a consumable item that it wants to retain. Guarding behaviors, such as growling and snapping towards a perceived competitor, can be seen in puppies as young as two to three weeks of age \[[@B1-animals-08-00027]\]. Most, if not all, standardized behavior assessments currently used in shelters \[[@B2-animals-08-00027],[@B3-animals-08-00027],[@B4-animals-08-00027]\] include an evaluation of FG behavior. These assessments typically evaluate the extent of FG by using a fake hand, made of rubber or plastic on a dowel, to pull the food bowl away from the dog and touch their cheek while eating. Different assessments vary the time and degree of contact between the dog and the fake hand. In 2005, The American Society for the Prevention of Cruelty to Animals^^®^^ (ASPCA^^®^^) \[[@B5-animals-08-00027]\] conducted a survey of US animal shelters. Eighty-nine percent of responding organizations reported they conducted a behavior assessment in their shelter and almost all assessed for guarding around a food bowl. For the sake of concision in this manuscript, the specific FG component of the behavior assessment is referred to as the FG item. Two recent studies indicated that assessing for FG in the shelter was a poor predictor of FG behavior in the home. Marder et al. \[[@B6-animals-08-00027]\] looked at 97 dogs adopted from an animal shelter and compared their FG behavior during assessment and in-home. All dogs were assessed using Match-Up II Shelter Dog Rehoming program \[[@B4-animals-08-00027]\]. Twenty (21%) dogs exhibited FG in the assessment and of those, 11 (55%) showed FG in-home. Interestingly, of the 77 dogs that did not display FG during the assessment, 17 (22%) did display FG in-home. Whether or not dogs guarded food during the shelter assessment had little bearing on the FG behavior they displayed in-home. The second study \[[@B5-animals-08-00027]\] followed 96 dogs that exhibited FG during the Safety Assessment for Evaluating Rehoming (SAFER™) assessment \[[@B2-animals-08-00027]\]. Dogs that exhibited FG were allowed free-access to dry dog food in their kennel, and then subsequently adopted into homes. Three weeks after adoption, only one of the 96 dogs was reported to exhibit FG behavior in the home. By three months, no dogs were reported showing FG behavior in-home. The discrepancy between these two studies in the rate of FG in-home could be a result of the free feeding regimen the dogs experienced while in the shelter, substantially different durations of follow-up, support given to adopters, fundamental differences in the shelter dog populations, or other unidentified variables. However, the conclusions regarding the inability of the FG item to reliably predict FG in-home is consistent across the two studies. Both of these studies call into question the value of assessing FG while the dog is in the shelter. It appears that dogs showing FG guarding during the assessment are no more or less likely to exhibit FG in a home. Equally as important, the Marder study \[[@B6-animals-08-00027]\] found that adopters did not perceive FG behavior as a problem in keeping their dog, which suggests that standardized FG assessments may be unnecessary and irrelevant. In 2016, Patronek and Bradley \[[@B7-animals-08-00027]\] considered the usefulness of standardized behavior assessments in shelters and noted that "reliably predicting problematic behaviors in future adoptive homes is vanishingly unlikely." They argued that if a dog bit or threatened to bite during a standardized assessment that whether or not the dog would do so in an adoptive home "would be at best, not much better than flipping a coin". Although the authors were referring to aggression assessments in general, and not just the FG portion of the assessment, their primary argument is that standardized assessments result in an unacceptably high incidence of false positives (which are those dogs that exhibit aggression during the assessment and not in-home). A high rate of false positives is particularly alarming when it comes to FG, because FG was one of the top reasons shelter staff reported a dog would be deemed unadoptable \[[@B5-animals-08-00027]\] and euthanized \[[@B6-animals-08-00027]\]. Even when a shelter chose to not euthanize dogs with FG behavior, those dogs were more likely to have adoption restrictions (such as experienced owners only or families without children) thereby reducing the pool of potential adopters for the dogs \[[@B6-animals-08-00027]\]. Since the publication of the Marder et al. \[[@B6-animals-08-00027]\] and the Mohan-Gibbons et al. \[[@B5-animals-08-00027]\] studies, members of the animal welfare field have shared their interest and concerns with the ASPCA^^®^^ about omitting the FG item from their assessment process. Their primary concerns were that injuries to staff or adopters would increase and that dogs would be returned more often. Thus, the primary goals of this study were to compare the outcomes, injuries, and returns after the FG item was removed from the assessment. 2. Materials and Methods {#sec2-animals-08-00027} ======================== 2.1. Inclusion Criteria {#sec2dot1-animals-08-00027} ----------------------- In order to be included in this study, shelters had to already be using a standardized behavior assessment that included a FG item and be willing to discontinue the FG item for the investigation phase. The FG item involved presenting the dog with a bowl of food and then using a fake hand on a dowel to touch the dog and the bowl while the dog was eating the food. Only FG toward people was addressed; FG directed towards other dogs was not a part of any standardized assessment in this study. Any dog exhibiting FG was included, regardless of breed or age (as long as they were at least 6 months old, per assessment guidelines). Shelters refrained from making any other program or adoption changes during the study that would have affected the research. 2.2. Recruitment of Shelters {#sec2dot2-animals-08-00027} ---------------------------- Animal shelters were recruited for this study through a convenience sample. The authors had prior relationships with some of the shelters and spoke directly with the staff to ask for their participation. Other organizations volunteered themselves for the study following a presentation on guarding behaviors in the shelter given by Drs. Emily Weiss and Amy Marder at the Society of Animal Welfare Associations conference in 2014. The authors also sought recommendations from other leaders in the field to identify organizations that might be interested in participating. Any organization that expressed interest in participating was sent an email providing a summary of the project. A total of 11 shelters enrolled in the study. In the first month, 2 shelters were excluded: 1 shelter pre-screened dogs for FG as part of their relocation program, rendering them inadmissible and the other shelter withdrew because they felt they could not meet the data collection requirements. The shelters were all non-profit humane societies except for one large municipal animal control organization. These shelters were located in seven states: Colorado, Florida, Mississippi, New York, North Carolina, and South Carolina. There was one shelter in each of these states except for Florida and New York, which each had two. Seven shelters used the SAFER™ assessment and two shelters used their own standardized assessment. 2.3. Protocol {#sec2dot3-animals-08-00027} ------------- Nine shelters completed the 5-month study from 1 September 2015 to 31 January 2016. In the first 2 months (baseline period), shelters were instructed to continue with their regular process for assessing behavior, then they were to omit the FG item for the next 2 months (investigation phase). In the investigation phase, shelters were advised to conduct all other components of their assessment (minus the FG item) and to proceed with adoptions as per normal. A data collection form was issued to each shelter to collect the following information for each month of the baseline phase and the investigation phase for the entire dog population: (1) total intake, (2) total returns of dogs adopted within the past 30 days, (3) average length of stay, (4) total number of dogs adopted, transferred, and euthanized; and the total number of bites and injuries (5) observed while in-shelter and (6) reported in-home post-adoption. Those data were also collected for the dogs that were classified as food guarders in both the baseline and investigation phases. During the last month of the study, in January, the only data collected were the number of returns and reports of bites or injuries that occurred in-home (\#2 and \#6 above). Bites and injuries were collected as a single category, defined as any type of injury to a human, in any context, while in the shelter or after adoption. This information was gleaned from reports in the shelter and from conversations when owners contacted the shelter to discuss their dog post-adoption. Shelters were asked to report all incidents for any reason and, as a result, an incident could occur during any interaction with the dog and were not restricted to interactions around food. For example, a dog in the FG group may have caused harm by pulling someone off their feet while on a walk was still counted as an injury, even though no food was involved. This ensured a conservative approach and that all incidents were captured for this research. On a monthly basis, completed forms were collected from each shelter by email. At the end of the study, the authors conducted an informal conference call with all of the shelters to learn more about their experiences during the study. 2.4. Identification of FG {#sec2dot4-animals-08-00027} ------------------------- Shelters were instructed to identify FG in one of three ways: (1) during the FG item of the assessment in the baseline phase, (2) from interviews conducted with the person who brought the animal to shelter, or (3) through observations made by staff, volunteers, and potential adopters while the dog was housed at the shelter (separate from the assessment). As noted in [Section 2.3](#sec2dot3-animals-08-00027){ref-type="sec"}, the FG item was performed only during the baseline phase but the other ways of identifying FG were utilized over the entire study period. Once staff saw a dog exhibiting FG behavior, they recorded on the standardized data collection form how that behavior was identified. The researchers provided shelters with descriptions of FG to enable them to categorize the level of severity of the guarding behavior. These descriptions were labeled as mild, moderate or severe ([Appendix A](#app1-animals-08-00027){ref-type="app"}). The behaviors could be seen in various contexts including during the assessment, in their kennel, or reported from the owner. Staff documented the level of severity for each dog regardless of how the behavior was identified. 2.5. Grants {#sec2dot5-animals-08-00027} ----------- Once enrolled in the study, each shelter was issued a \$2500 grant to support the work needed to complete the study. Grant funds were used to train staff, create literature on FG for adopters and staff, start a behavior hotline, or recruit shelter volunteers. 2.6. Statistical Analysis {#sec2dot6-animals-08-00027} ------------------------- Characteristics of the groups were described using frequencies and percentages calculated with Microsoft Excel. Chi-square tests were performed to compare the baseline and investigation phase rates of bites/injuries and returns for the entire general dog population, as well as specifically for the group of FG dogs. Chi-square was also used to compare the percentage of food guarders and non-food-guarders euthanized during the four-month study period. StataSE 13.1 (StataCorp LP, College Station, TX, USA) was used for these analyses, and p \< 0.05 was considered to be statistically significant. 3. Results {#sec3-animals-08-00027} ========== 3.1. Population {#sec3dot1-animals-08-00027} --------------- A total of 14,180 dogs entered the nine shelters across the four-month study period. The highest intake at a participating shelter was 3247 dogs and the lowest intake was 596 dogs. During the baseline phase, 7112 dogs entered the shelters and 7068 dogs entered during the investigation phase. During the baseline phase, only 49% of the dogs were assessed upon intake. The other dogs were returned to their owners before an assessment could be completed or were considered adoptable without being assessed, due to their desirable appearance, size, or breed. Over the whole study period, 5% (778/14,180) of the dogs either exhibited FG or had a history of FG. Further details of which dogs were identified with FG during each phase can be found in [Figure 1](#animals-08-00027-f001){ref-type="fig"}. 3.2. Risk of Injury {#sec3dot2-animals-08-00027} ------------------- [Figure 1](#animals-08-00027-f001){ref-type="fig"} shows total intake of dogs between the two study phases and color groupings for those who displayed FG (green) vs non-FG (orange). The number of bites/injuries was low in both phases. FG dogs, either in the shelter or in the home, were no more likely to inflict injuries once the assessment item was stopped. A within-phase study analysis was also conducted. In the home, adopters were significantly more likely to report bites/injuries from dogs in the FG group compared to the non-FG group, in both baseline (p = 0.006) and investigate stages (p = 0.01). In the home, there were only 8 reported bites/injuries for the FG group and 35 for the non-FG group. In the shelter, there was no association of bites/injuries related to the FG or non-FG groups, in baseline (p = 0.5) or investigative phases (p = 0.6). 3.3. Outcomes for FG Dogs {#sec3dot3-animals-08-00027} ------------------------- Out of the 14,180 dogs entering the shelter, 14% (2051/14,180) were still in the shelter at the termination of the study and could not be assigned an outcome. Over the course of the study, 56% (7961/14,180) of all dogs were adopted and 18% (2576/14,180) were transferred to other organizations for adoption. Fifty-four percent of dogs (3867/7112) were adopted during the baseline phase and 58% (4094/7068) during the investigation phase. Adoption of FG dogs was 39% in the baseline (223/571) and 45% the investigation phase (93/207) (p = 0.14, X^2^ = 2.17, 1 df). Across both study phases, dogs that showed FG behavior stayed in the shelter longer, a mean of 13.6 days (range = 4.7--44 days), than the general dog population, a mean of 9.9 days (range = 3.6--20 days). A total of 11.2% of dogs (1592/14,180) were euthanized during the study. Significantly more FG dogs (15.4%, 120/778) were euthanized than non-FG dogs (10.9%, 1472/13,402), p \< 0.001, X^2^ = 0.09, 1 df. Of the 120 dogs euthanized in the FG group, 74% (n = 89) had behavior listed as a reason for euthanasia and 45% of those had FG listed as a reason for euthanasia. The other 55% were euthanized for other behavioral issues such as aggression toward other animals or people, high arousal, or problematic behavior in the kennel. ### Degree of FG and Impact If dogs exhibited FG in-shelter, the majority of the time it was mild in severity ([Table 1](#animals-08-00027-t001){ref-type="table"}). Another key finding is the percentage of dogs identified with severe FG was unchanged, even once the FG item was omitted. [Figure 2](#animals-08-00027-f002){ref-type="fig"} shows outcomes of adoption, euthanasia, and transfers, categorized by the severity of FG. During both baseline and investigation phases, dogs with severe FG behavior were less likely to be adopted and more likely to be euthanized than moderate or mild food guarders. 3.4. Returns {#sec3dot4-animals-08-00027} ------------ Dogs in the FG group were no more likely to be returned (p = 0.7, X^2^ = 0.11, 1 df) in the investigation phase (11%, 10/93) than they were during the baseline phase (9%, 21/223). There were significantly more returns of all dogs (p \< 0.001, X^2^ = 18.33, 1 df) during the investigation phase (13%, 526/4094) than during the baseline phase (10%, 379/3867). However, even though statistically significant, a difference of 3% in returns is too small to be meaningful to the field. 4. Discussion {#sec4-animals-08-00027} ============= This study documented the impact of omitting the FG item from standardized shelter behavior assessments. During the baseline phase, 8% of dogs were classified as food guarders. During the investigation phase, when dogs were no longer assessed for FG, only 3% were classified as food guarders. Other studies \[[@B5-animals-08-00027],[@B6-animals-08-00027]\] reported higher percentages of FG (14--21%), however they looked at the number of FG dogs based on the number assessed, not on overall intake numbers. If this study were to calculate FG dogs based on those assessed (which could only be done during baseline phase), the percentage aligns with the other studies (16.5%). However, since many shelters assess less than half their dog population, these reported percentages likely over-represent the prevalence of FG behavior in the shelter dog population. When designing this study, it was not anticipated that such a large number of dogs would not be assessed during baseline. While it is not known what each individual staff person did within each shelter (relative to the dogs not assessed), shelters did not always assess dogs who were easily adoptable (i.e., healthy, small, obviously social, and friendly dogs). It is not clear how much bias this brings and in what direction the bias would be, due to the lack of assessments among a subset of the population of dogs. One key factor used to improve validity was each shelter was its own baseline. This study was designed to be applied research in functioning shelters, constrained by shelter realities. Since this was a convenience sample, there could be a bias in those shelters that chose to participate. The shelters varied in size (number of dogs and staff), one was a municipal shelter while most were non-profit organizations, and they varied in their location around the country. We deliberately aimed for a large sample size from a variety of shelters to minimize biases in the results and improve external validity. The safety of staff, volunteers, and adopters was identified as a critical concern when shelters considered removing the FG item from their assessments. However, few bites or injuries were reported during this study, whether or not dogs were assessed for FG. It is interesting that the rate of bites and injuries was so low, even in the baseline phase, when a significant number of dogs were not assessed. This further confirms the conclusion that testing for FG does not improve safety in a shelter setting. Reported injuries in the home did not increase once the FG item was stopped, however there was a higher percentage of injuries by FG dogs compared to the non-FG dog group. However, there was a low prevalence of bites/injuries in the home for all dogs ([Figure 1](#animals-08-00027-f001){ref-type="fig"}) and the reporting included all bites and injuries, even when not involving food. Since many shelters advised adopters that these dogs might FG in the home, it is possible that new owners were more likely to report bites or injuries thus biasing the numbers. It is plausible that dogs who FG in the shelter may also have other behavior problems that make them more likely to cause injuries post-adoption, but this study did not yield that data. Another limitation of this study, which was also beyond the scope, was a lack of systematic follow-up in the home, like the protocol used in the Mohan-Gibbons study \[[@B5-animals-08-00027]\]. As a result, we could not ask for more information on the situations where the bite or injury occurred in the home. We did not ask shelters to separate actual bites from other types of injuries, nor did we ask for detail about the incidences. Although collecting that information would have provided more behavioral context, the concern was it would introduce unwanted bias because staff and adopters would need to correctly interpret every situation if FG was present or not. By tracking all incidences of bites and other injuries, our conclusions are more conservative. In this study, dogs with food guarding behavior stayed in the shelter longer than the general population and were more likely to be euthanized. An earlier survey \[[@B5-animals-08-00027]\] found that FG behavior was one of the top reasons why shelters did not place dogs into adoption. In this study, while only 15.4% of FG dogs were euthanized, over half of these were euthanized for behavior concerns unrelated to FG (i.e., medical problems and other severe behavior challenges). Most of the dogs that showed FG behavior in both the baseline and investigation phases were classified as mild FG. In the investigation phase, even though fewer dogs were identified with food guarding behavior, the percentage of dogs classified as severe remained unchanged (17% in both phases). Thus, it appears that dogs exhibiting more serious FG behavior are likely to be identified without using a standardized assessment. Relinquishing owners may report that their dog displays FG when it is more serious and staff may be more likely to observe severe guarding behavior during daily interactions in the shelter. This confirms the assertion made by Patronek & Bradley \[[@B7-animals-08-00027]\] that more serious forms of aggression are highly likely to be identified without putting the dog through a behavior assessment. The animal welfare community had expressed concerns that returns to the shelter would increase if the FG item was eliminated. Our study showed that return rates for FG dogs were not significantly different from the baseline phase, however the return rate for the general population was. Even though those returns presumably included dogs that guarded food in the home, a difference of 3% is not large enough to be meaningful in the context of shelter returns. The statistical significance is driven by the large number of dogs adopted from the nine shelters. A dog returned from the FG group may not have been returned for food guarding. The reasons given for returns were insufficient to provide any level of detail. It is challenging to gather quality data when those data are based on owners relaying details to shelter personnel who then capture that information into restrictive shelter software. Future work could consider learning more about the reasons for dogs being returned to the shelter. If shelters wish to further reduce returns, we recommend that they have conversations with all adopters about normal canine behavior, such as FG, and that any dog might exhibit this behavior in the home. Staff should encourage adopters to call the shelter if they encounter problems and need help. Other research has shown that when pet owners are given realistic expectations, they are more likely to keep their pets \[[@B8-animals-08-00027]\]. Perhaps providing adopters with basic information about FG and offering post-adoption support will prove to be a key to successful adoptions for this at-risk population. Shelters did not make large organizational changes during the course of the study. Shelter staff was aware when the FG was stopped. We encouraged shelters to talk with owners relinquishing dogs to obtain a history of FG, and some shelters changed their intake forms to reflect this. We encouraged adoption counselors to tell potential adopters that any dog might guard their food and some added this information to their adoption paperwork. Many shelters reported that staff became more diligent about recognizing and reporting FG behavior in the investigative phase because the dogs were not going to be assessed for FG. One shelter even uncovered that staff would not report observations of a dog guarding food in the kennel (during the baseline phase) because they believed "the behavior staff would catch it in the FG item during the assessment". It is crucial for shelters to establish a clear communication system for staff and volunteers to share pertinent behavior information about the dogs in their care, especially when considering omitting the FG item from their assessment. In previous research, a significant number of dogs that were observed guarding food during the assessment did not exhibit this behavior in the home \[[@B5-animals-08-00027],[@B6-animals-08-00027]\]. These "false positive" dogs are more likely to stay longer in the shelter and to be euthanized. In addition, some dogs do not guard food while in the shelter but will at home. These "false negative" dogs are often incorrectly deemed as "safe" by shelter staff and sent home with adopters who are not adequately prepared to encounter this behavior. Any dog can exhibit food guarding behavior. Our research, paired with previous research referenced in this manuscript, substantiate that the existence of food guarding behavior in the shelter does not always predict the same behavior in the home. The incidence of injury to shelter staff and adopters is low and assessing dogs for FG does not further reduce the already low injury rate. 5. Conclusions and Recommendations {#sec5-animals-08-00027} ================================== Previous research suggested that many dogs exhibiting FG in the shelter can be safely placed into homes because FG is often not exhibited in the home nor was it a problem for adopters when it was seen. In light of the current results, the authors recommend that shelters need not conduct the FG item of a standardized assessment because it results in a longer stay in the shelter, increased likelihood of euthanasia, and the incidence of false positives is likely unacceptably high. FG was not prevalent in the participating nine shelters and, of the dogs that did exhibit the behavior, most showed only mild FG behavior. Once the FG item was omitted, more of those dogs were adopted and fewer were euthanized. Severe FG behavior was found regardless of the assessment. Discontinuing the FG assessment item did not result in an increase in injuries to shelter staff or adopters, and there was no meaningful change in returns post-adoption. Our recommendation is for shelters to collect as much information about the dog from all sources (such as prior owner and staff observations) to best support each dog in finding a home. For example, if a dog shows guarding behavior over his food while in the kennel, the shelter should have procedures to ensure staff are trained so that they can stay safe while interacting with the dog. However, they should not assume the dog will show this behavior in his new home. Shelters should not ignore FG behavior. Rather, staff should be transparent with adopters about what they know from prior history and behavior in-shelter. Staff should set adopters up for success by explaining that FG is a normal behavior for dogs, it has a low occurrence, and if seen, they should contact the shelter promptly for assistance. There are techniques for managing and modifying FG behavior in the home when the owners are both concerned as well as able to implement these techniques. The authors deeply appreciate the following shelters for their participation in the study and for their time in the data collection process; Humane Society of Boulder Valley (CO), SPCA Tampa Bay (FL), Pet Alliance of Greater Orlando (FL), Southern Pines Animal Shelter (MS), SPCA of Wake County (NC), Nebraska Humane Society (NE), Erie County SPCA (NY), Animal Care Centers of New York (NY), Charleston Animal Society (SC). Processing costs for this paper were covered by the Special Issue sponsors, Maddie's Fund®, Found Animals, and The Humane Society of the United States (HSUS). Emily Weiss, Margaret R. Slater, Pamela Reid, Heather Mohan-Gibbons, and Hugh Mulligan conceived and designed the project; Margaret R. Slater, Hugh Mulligan, and Heather Mohan-Gibbons analyzed the data; Heather Mohan-Gibbons, Emily D. Dolan, Margaret Slater, Emily Weiss, and Pamela Reid wrote the manuscript. The authors declare no conflict of interest. Below are three categories with descriptions of guarding behaviors that may be observed during the food portion of an assessment. Please choose the broad category that is the best fit for each dog. Vocal behaviors, such as growling or barking, are not listed, as they can be observed in any category. ![Behavioral Categories for Food Guarding.](animals-08-00027-g0A1){#animals-08-00027-f0A1} Appendix A.1. Mild ================== Dog's head, neck and/or body become stiff/rigid but the dog does not escalate to Moderate or Severe categories. May show one or more of these behaviors: eats more quickly as food assessment progressesholds a crouched position over bowl when person is nearmoves body and/or head to block the bowl from the assessor's approach or the fake handmomentarily freezes or stops eating (without lifting head from bowl) when assessor approaches or reaches with fake hand Appendix A.2. Moderate ====================== The guarding behavior remains in or near the food bowl. Dogs may show any of the behaviors in the Mild category plus one or more of these behaviors: may lift lips/bare teethbites in the direction of the food or at the fake hand while head is in the bowl and/or while touching the dog's cheek with the fake hand or person is nearalthough the dog may bite the fake hand, the dog does not lift his head from the bowl to bite toward the assess-a-hand or pursue the hand as it is retracted Appendix A.3. Severe ==================== The guarding behavior is directed towards the assessor or handler. Dogs may show any of the Mild or Moderate behaviors, plus at least one of the following: multiple bites to the fake hand and/or bites that move up the fake handleaves the bowl to bite the fake hand (either while the hand is advancing towards the bowl or as it is withdrawn)any attempts to lunge, snap, or bite towards the person's arm or body or the handler during the food bowl itemdogs that exhibit guarding behavior either before or after the food guarding assessment (for example, the bowl is placed down and the dog immediately shows any guarding behavior before assessor has attempted their initial approach to the bowl or dogs that continue to guard the area after the food bowl is removed). ![Total intake of dogs between the two study phases: percentage of dogs who were assessed, those dogs identified with food guarding behavior, and incidence of injury/bites in each phase. Non-FG groups are in orange and food guarding (FG) groups in green, with corresponding p-values displayed between the baseline and investigative phases. In baseline phase, there are blue boxes and dotted lines that show the percentage of dogs that had the standardized shelter assessment and of those assessed, how many were identified with FG. This was added so the reader can visualize the large percentage of dogs that did not have a standardize assessment performed and to clarify the two ways a shelter can arrive at incidence of FG (17% is only of those assessed while 8% is all intake).](animals-08-00027-g001){#animals-08-00027-f001} ![Outcomes of the FG dogs, by severity of FG behavior, are shown for the (a) Baseline phase (b) Investigation phase. Some dogs in each severity category did not have an outcome by the end of the study period (no outcome). In the baseline, 20% of FG was identified by history and 80% by the assessment. During the investigation phase, 71% of FG was identified by history and 29% from shelter observations.](animals-08-00027-g002){#animals-08-00027-f002} animals-08-00027-t001_Table 1 ###### Severity of Food guarding (FG) behavior identified in all FG dogs comparing the baseline (8%, 571/7112) and investigation (3%, 207/7068) phases. There was an overall significant difference in the categories of severity between the baseline and investigative phase (p = 0.001, X^2^ = 14.86, 2 df). In the baseline, 20% of FG was identified by history and 80% by the assessment. During the investigation phase, 71% of FG was identified by history and 29% from shelter observations. Severity of FG Behavior Baseline Investigation --------------------------------- ---------- --------------- --------- -------- Total Dogs with FG Behavior 571 8% 207 3% Mild 391 68% 117 57% Moderate 85 15% 55 27% Severe 95 17% 35 17% Total 571 100% 207 100%	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1-toxins-09-00255} =============== The genus Fusarium is implicated in devastating plant diseases. Of primary importance is the sambucinum species complex, which contains species causing Fusarium Head Blight (FHB) on a number of small-grain cereals such as wheat \[[@B1-toxins-09-00255]\]. The economic impact of this disease is exacerbated by the deposition in the cereal matrix of mycotoxins, which ultimately end up in the food and feed chain with significant health risks for humans and animals \[[@B2-toxins-09-00255]\]. Many FHB-associated species may co-occur on the host and the composition of this population is variable throughout and between growing seasons \[[@B3-toxins-09-00255]\]. F. graminearum is the predominant species within the FHB complex \[[@B1-toxins-09-00255]\]. In addition, F. poae, has been described as one of the most prominent FHB species in several studies \[[@B4-toxins-09-00255],[@B5-toxins-09-00255],[@B6-toxins-09-00255],[@B7-toxins-09-00255]\]. Its occurrence is often correlated with the incidence of other Fusarium species and its preponderance increased with the application of azole fungicides \[[@B4-toxins-09-00255],[@B8-toxins-09-00255],[@B9-toxins-09-00255]\]. The most important class of mycotoxins associated with FHB is the trichothecenes. They are produced by a large number of Fusarium species and can be divided mainly into type A (diacetoxyscirpenol (DAS), neosolaniol (NEO), and T-2 toxin, HT-2 toxin) and type B trichothecenes (deoxynivalenol (DON), nivalenol (NIV) and fusarenon-x (FUS-X), differing at the functional group occupying the carbon 8 position \[[@B10-toxins-09-00255]\]. Genes for trichothecene biosynthesis are organized in different loci and clusters. Homologs with different functionality in different Fusarium species determine which end products will be produced \[[@B11-toxins-09-00255]\]. The main determinant for type A vs. type B production resides in different alleles of the Tri1 gene \[[@B12-toxins-09-00255]\]. Surprisingly however, several authors have simultaneously detected both type A and type B trichothecenes from isolates of F. poae \[[@B13-toxins-09-00255],[@B14-toxins-09-00255],[@B15-toxins-09-00255]\]. The specific lifestyle of fungal pathogens greatly influences their epidemiology, and their potential for adaptation and evolution. Species that combine sexual and asexual reproduction, and that harbor active transposable elements (TEs) have been theorized to possess the greatest potential for genetic variation and hence pose the highest threat to durable disease management \[[@B16-toxins-09-00255]\]. The occurrence of "cryptic" stages may complicate the identification of the precise lifestyle of fungi. This is the case when a sexual cycle is presumed to occur, but has never been witnessed, such as observed for several Fusarium species \[[@B17-toxins-09-00255]\]. Recombination is a clear indication that a sexual cycle is ongoing in fungi \[[@B18-toxins-09-00255]\], and genome-driven discovery of meiosis-specific markers may uncover the footprints of a sexual cycle \[[@B19-toxins-09-00255]\]. The occurrence of repeat-induced point mutation (RIP), a defense mechanism against TEs, is another line of evidence for meiosis, as it only occurs during the pre-meiotic stage \[[@B20-toxins-09-00255]\]. Subsequently, degenerated remnants of TE copies remain in the genome after RIP passage \[[@B20-toxins-09-00255]\]. The activity of RIP has recently been extensively proven for F. poae \[[@B19-toxins-09-00255]\]. The mating locus (MAT1) is another important genetic fingerprint for a sexual cycle. This locus can be occupied by two idiomorphs MAT1-1 and MAT1-2 and the equal distribution of both idiomorphs in fungal populations has been considered as an indication of an active sexual cycle. The occurrence of both MAT1-1 and MAT1-2 has been confirmed for F. poae isolates \[[@B17-toxins-09-00255],[@B21-toxins-09-00255]\]. One of the benefits of sexual reproduction is meiosis, which introduces diversity via recombination. Many studies have been carried out which investigate the population diversity of F. poae. Kerenyi et al. \[[@B22-toxins-09-00255]\] detected a highly complicated composition of vegetative compatibility groups (VCGs) in 50 geographically diverse isolates. Somma et al. \[[@B23-toxins-09-00255]\] found a high level of intra-species variability in a geographically narrow population using amplified fragment length polymorphism (AFLP) \[[@B24-toxins-09-00255]\], which was also found for a worldwide F. poae collection \[[@B21-toxins-09-00255],[@B25-toxins-09-00255]\] with AFLP and inter simple sequence repeats (ISSR). Consistently, geographic origin could only partially explain this large diversity \[[@B22-toxins-09-00255]\]. Contrary to the results in genome-wide approaches, multi-locus barcoding consistently leads to the detection of rather low intra-species variability in F. poae \[[@B26-toxins-09-00255],[@B27-toxins-09-00255]\]. This discrepancy may be in part explained by the specific genome composition of F. poae. Isolates of this species have four core chromosomes (the "core genome") and a set of supernumerary chromosomes, highly variable in size, composition, and number between individuals \[[@B19-toxins-09-00255],[@B28-toxins-09-00255]\]. It has been shown that their presence leads to significant rearrangements within the core chromosomes \[[@B19-toxins-09-00255]\]. Two major translocations from the supernumerary genome into core chromosome 3 of F. poae isolate 2516 were found, each exceeding several 100 kb. Finally, the supernumerary genome harbors numerous active TEs (where they are protected from RIP), which can migrate from the supernumerary to the core genome. Species that are capable of combining both sexual and asexual reproduction, and species that possess an active arsenal of TEs, are considered to be the biggest challenge for durable disease management \[[@B16-toxins-09-00255]\] and a key trait of a successful pathogens \[[@B29-toxins-09-00255]\]. We set out to investigate both features in this study. We started by assessing the genetic diversity using a genome wide approach in a representative F. poae collection from different geographic origins. We searched for hallmarks of sexual and asexual reproduction, and indications for how genomes evolve in a population. In addition, we investigated the genetic variability in the trichothecene gene clusters and tried to link this genetic variability with the highly variable mycotoxin profiles of F. poae comprising both type A and type B trichothecenes. 2. Results {#sec2-toxins-09-00255} ========== 2.1. Genetic Diversity of Fusarium poae {#sec2dot1-toxins-09-00255} --------------------------------------- For this study, we used 69 available F. poae isolates ([Table 1](#toxins-09-00255-t001){ref-type="table"}) and AFLP analysis was performed to assess their genetic diversity. A total of 247 markers were scored of which 201 were polymorphic. As a control F. poae isolate 9125 was scored independently three times in the analysis. The clustering of these three fingerprints shows the reproducibility of the AFLP method but also illustrates the variability obtained for a perfectly clonal isolate. The genotyping error rate calculated from these three repetitions was 3.3--3.8% similar to levels described in literature (2--5%) \[[@B30-toxins-09-00255],[@B31-toxins-09-00255]\]. Based on the AFLP analysis, the F. poae isolates are divided into sub-clusters with no clear correspondence to either host or geographic origin. This is exemplified by isolates from maize (L24, K46, S46, Q57, F49), isolates from barley (6114, 175, 182, 185, bfb0176) and isolates from Italy (designation 9\#\#\#) which are scattered throughout the tree. However, the only Chinese isolate in the study (bfb0173) is clearly distinct from all other F. poae isolates analyzed ([Figure 1](#toxins-09-00255-f001){ref-type="fig"}). Isolates which share \>90% genetic similarity as determined by the Dice coefficient are grouped by the same color in [Figure 1](#toxins-09-00255-f001){ref-type="fig"}. Several other genetic markers were included in [Figure 1](#toxins-09-00255-f001){ref-type="fig"}, including the translocations of supernumerary sequences to chromosome 3 \[[@B19-toxins-09-00255]\]. In isolate 2516, two sequence blocks, each several 100 kb in size, translocated from the supernumerary chromosomes to core chromosome 3. Seven isolates in the collection are found to have the first insertion, and three of these also have the second insertion. This likely occurred in a sequential fashion: starting with a \>204 kb translocation closest to the telomere of chromosome 3, followed by a 464 kb event further away from the telomere (883,738 bp). [Figure 1](#toxins-09-00255-f001){ref-type="fig"} shows that an entire sub-cluster of the AFLP tree is made up of isolates containing the first insertion of supernumerary sequence. 2.2. Fusarium poae Likely Combines Sexual and Asexual Reproduction {#sec2dot2-toxins-09-00255} ------------------------------------------------------------------ It has been shown that RIP is associated with sexual reproduction \[[@B20-toxins-09-00255]\]. Since RIP functions on the core genome of F. poae \[[@B19-toxins-09-00255]\], it stands to reason that meiosis should be ongoing in this species. Moreover, the presence and conservation of the "meiotic toolbox" was shown \[[@B19-toxins-09-00255]\]. We expanded these authors' investigation into the distribution of MAT1-1 and MAT1-2 in F. poae. These results are visualized in [Figure 1](#toxins-09-00255-f001){ref-type="fig"} and [Table 1](#toxins-09-00255-t001){ref-type="table"}. The MAT1-1 idiomorph is predominantly present (81% MAT1-1, 19% MAT1-2; one undetermined, but both idiomorphs do co-occur in the same location ([Table 1](#toxins-09-00255-t001){ref-type="table"})). The presence of detectable recombination events dictates that F. poae should be a sexual species, even if the sexual stage has never been observed. All F. poae isolates were screened for the presence of two major translocations from the supernumerary genome to chromosome 3 (see also \[[@B19-toxins-09-00255]\]). Eight isolates containing the first translocation of supernumerary sequence (closest to the telomere of chromosome 3) were collected from different locations across Flanders, Belgium ([Table 1](#toxins-09-00255-t001){ref-type="table"}). It can be assumed that this genotype has spread clonally from one location to the other two. 2.3. Transposable Element Proliferation between Near-Clonal Isolates {#sec2dot3-toxins-09-00255} -------------------------------------------------------------------- To uncover if TE insertions are shared between isolates, the location of TE insertions in isolates 2516 and 2531 were compared. According to [Figure 1](#toxins-09-00255-f001){ref-type="fig"}, these isolates are near-clonal (comparable to the AFLP pattern of the three repetitions of isolate 9125). Reads from an Illumina HiSeq run of DNAs from five different isolates, including isolate 2531 were mapped onto the reference assembly of isolate 2516 and 73 of the 135 TE copies on the core genome of 2516 did not have read support from the DNA read mix, including absence of coverage by reads originating from isolate 2531 ([Table 2](#toxins-09-00255-t002){ref-type="table"}). These TEs represent copies uniquely present in isolate 2516, and they are absent in the same location in isolate 2531. 2.4. Assessing the Genetic Chemotype of F. poae Isolates {#sec2dot4-toxins-09-00255} -------------------------------------------------------- All three thrichothecene loci were positioned on the assembly of F. poae isolate 2516 within their exact genomic coordinates. The Tri1 gene is located on chromosome 1 (55,207--56,959 bp). The organization of the Tri1 region in isolate 2516 is identical to what was described previously by Proctor et al. \[[@B32-toxins-09-00255]\]. The main trichothecene biosynthesis cluster (Tri5) is located on chromosome 2 (5,545,692--5,575,616 bp) and the Tri101 gene can be found on chromosome 4 at (4,906,101--4,907,486 bp). These positions are similar to those for the three loci in F. graminearum. To study the organization of the main trichothecene cluster, we applied the diagnostic PCR developed by Dinolfo et al. \[[@B33-toxins-09-00255]\] to predict the genotype of all F. poae isolates towards NIV production. This assay is based on the Tri7 gene. Using this approach, all isolates were genetically chemotyped to be of the NIV genotype. In a next step, the main trichothecene gene cluster was extracted from chromosome 2 of isolate 2516 (NCBI accession LYXU01000002.1). [Figure 2](#toxins-09-00255-f002){ref-type="fig"} depicts the organization of the main cluster and the comparison with the gene clusters in F. sporotrichioides (isolate NRRL 3299, T-2 toxin chemotype, NCBI accession AF359360.3) and F. graminearum (isolate 88-1, NIV chemotype, NCBI accession AF336365.2). The overall organization of the main trichothecene biosynthesis cluster is highly similar to that previously described in F. sporotrichioides and F. graminearum. The largest difference is the intergenic distance between Tri5 and Tri6 that is expanded in isolate 2516. A low complexity (low GC %) region with similarity to a pogo TE accounts for an additional 2.1 kb of sequence in this region. The best Blastx hit for this element is "pogo transposable element with KRAB domain from F. oxysporum f. sp. cubense race 1" (NCBI accession ENH68388.1). The genome sequences of isolates 2548, 7555 and bfb0173 were consulted to find differences in the organization of the main trichothecene biosynthesis cluster with isolate 2516 and these isolates. The main trichothecene biosynthesis cluster of isolates 2548, 7555 and bfb0173 can be found on contigs 77, 419 and 361 respectively. The RIPped pogo TE between Tri5 and Tri6 is also present in isolate bfb0173 and is highly conserved (three SNPs (single-nucleotide polymorphisms) between 2516 and bfb0173). It is not present in isolates 2548 and 7555. The F. poae collection was screened for the presence of this element ([Figure S1](#app1-toxins-09-00255){ref-type="app"}). Thirty-eight isolates contain the pogo element at the same location as isolates 2516 and bfb0173. [Figure 1](#toxins-09-00255-f001){ref-type="fig"} visualizes the distribution of these isolates across the AFLP tree. Downstream of Tri14 a RIPped retrotransposon occurs in isolates 2548 and bfb0173. In isolate bfb0173, it consists of 3.7 kb of low complexity sequence, while in isolate 2548 it is only partially assembled, with 1.2 kb of sequence that is almost identical to the sequence of isolate bfb0173 (10 SNPs). The best Blastx hit for this retrotransposon is a protein from Claviceps purpurea (NCBI accession CCE29311.1). In isolates 2516 and bfb0173, there is a 400 bp repetitive element without Blastx hits between Tri7 and Tri3 that is not present in isolates 2548 and 7555. The three insertions in and around the main trichothecene biosynthesis cluster (between Tri7 and Tri3, between Tri6 and Tri5 and bordering Tri14) show a specific pattern, as different combinations of two insertions are not shared between the same isolates. Finally, variation at the Tri1 locus, on chromosome 1, is known to be responsible for chemotype diversity in Fusarium species \[[@B12-toxins-09-00255],[@B34-toxins-09-00255]\]. The Tri1 genes in isolates 2516 and bfb0173 are identical to one another and to the published sequence for FpTri1 (NCBI accession GQ915520), and 97% sequence identity to the Tri1 gene of isolates 2548 and 7555 (respectively 51 and 52 (=51 + 1) SNPs across the 1753 bp gene). Respectively 24 and 25 of these SNPs occur in the exons of the gene, while 87% of the gene is exonic (1527 bp). Remarkably, only two (out of 24) and three (out of 25) of these SNPs are non-synonymous mutations. Two of these amino acid substitutions occur at the N-terminal end of the protein within the first 14 amino acids. [Figure S2](#app1-toxins-09-00255){ref-type="app"} shows the read mapping of HiSeq reads from isolate bfb0173 and isolate 2548 on the published FpTri1 type, and shows that a disproportionate amount of SNPs occurs in the introns of the genes. A subset of 34 isolates of the F. poae collection was selected for sequencing of a 1100 bp fragment of Tri1. The existence of two major Tri1 types within the collection was confirmed, with additional variation at the level of individual isolates ([Figure 3](#toxins-09-00255-f003){ref-type="fig"}). 2.5. Assessing the Phenotypic Chemotype of F. poae Isolates {#sec2dot5-toxins-09-00255} ------------------------------------------------------------- [Figure 4](#toxins-09-00255-f004){ref-type="fig"} shows for 61 F. poae isolates the trichothecenes that were detected after growth in the trichothecene biosynthesis inducing medium. The values were log transformed as the amounts of DAS were higher than those of the other toxins. DAS is also the most common mycotoxin, with 59 of 61 isolates being positive for DAS, ranging from the LOD to 22804 ng/mL. Of these 59 isolates, 55 produced NEO in the range of the LOD to 375 ng/mL. Of this set, 24 isolates produced FUS-X in the range of the LOD to 172 ng/mL and, of these FUS-X producers, 14 isolates also produced NIV in the range of 88 ng/mL to 122 ng/mL. The hierarchical nature of these chemotypes is visualized in [Figure 4](#toxins-09-00255-f004){ref-type="fig"}A. Finally, there is no significant effect of the presence/absence of the pogo element on the chemotype, for any of the four detected trichothecenes (Mann--Whitney U test, p values \> 0.05) ([Figure 1](#toxins-09-00255-f001){ref-type="fig"} and [Table 1](#toxins-09-00255-t001){ref-type="table"}). [Figure 4](#toxins-09-00255-f004){ref-type="fig"} summarizes the trichothecene production in vitro for all isolates. DAS production is about 50-fold higher compared to the production of NEO, FUS-X or NIV. Note that two isolates show outlier values for DAS and NEO production (isolate 30702 and isolate 9194). Concordantly, [Figure 4](#toxins-09-00255-f004){ref-type="fig"}B shows that production of DAS and NEO is correlated. [Table S1](#app1-toxins-09-00255){ref-type="app"} shows the result of two biological repetitions of the chemotyping experiment for 28 isolates. Compounds with a plus sign were detected consistently, while compounds indicated with "X" were only detected in one of the two repetitions. For 20 of the isolates, the chemotype previously shown in [Table 1](#toxins-09-00255-t001){ref-type="table"}, moved "further downstream", with FUS-X or even NIV production where there was none in another iteration. This hierarchical nature of production was never violated in any of the strains in either of the repetitions. The chemotype of seven isolates was identical in the different repetitions: for example isolate 7555 only produced DAS, and isolate 2569 produced all four compounds. 3. Discussion {#sec3-toxins-09-00255} ============= In this paper, we assessed genetic variation, sexual and asexual reproduction, the trichothecene clusters and the chemotype of a set of F. poae isolates. AFLP analysis on the F. poae isolates to assess genetic variation among isolates from geographic different areas was performed. The isolates show \>75% similarity, suggesting that they are part of a single monophyletic lineage, which is similar to what was found before for F. poae \[[@B35-toxins-09-00255],[@B36-toxins-09-00255]\]. No obvious connection was found between geography and the AFLP sub-clustering, which is in accordance with previous studies \[[@B21-toxins-09-00255],[@B25-toxins-09-00255]\]. Still, isolate bfb0173 from China which is the most distant, clustered separately from all other isolates and the dominant genotype among the Italian isolates did not occur in Belgium. In accordance, Kerenyi, Taborhegyi, Pomazi and Hornok \[[@B22-toxins-09-00255]\] reported a link between geographic origin and VCG/RAPD profile (Random amplification of polymorphic DNA). We subsequently researched whether there was a link between the position of isolates in the AFLP tree and the host plant. Five isolates from maize were included in the study, aside from isolates from wheat and barley. The maize isolates are scattered throughout the tree, suggesting that no significant host specialization has occurred within F. poae. This is in agreement with literature reports that suggest that most isolates have a broad host range \[[@B22-toxins-09-00255]\]. However, a host preference cannot be excluded. As species capable of combining both sexual and asexual reproduction, and species with an active arsenal of TEs, are considered to be the biggest challenges for durable disease management \[[@B16-toxins-09-00255]\] and key trait of successful pathogens \[[@B29-toxins-09-00255]\], we investigated both features in the F. poae isolates. A sexual cycle has never been shown for F. poae, but is highly likely to occur. However, the skewed mating type distribution encountered among the Belgian F. poae isolates indicates that meiosis does not occur at high frequency \[[@B19-toxins-09-00255]\]. Although in oomycetes, such as Phytophtora infestans, an unbalanced distribution of the two mating types has been related to fitness costs, in F. poae there is no evidence for a similar link. Opposed to what was described recently by Dinolfo, Castanares and Stenglein \[[@B21-toxins-09-00255]\], we did not find any evidence for homothallism in F. poae. To confirm the presence of a sexual cycle, RIP was studied in isolate 2516. This analysis showed that meiosis and RIP likely did not happen very recently, allowing the accumulation of intact TE on the core genome during asexual propagation \[[@B19-toxins-09-00255]\]. The tight correlation between RIP and meiosis should offer opportunities for understanding the balance between clonal and sexual reproduction, by using TE insertions as markers for "track and trace" studies and monitoring when/if they become RIPped. We showed that near-clonal isolates can differ heavily in TE proliferation, which may occur over very short evolutionary periods. It seems likely that such TE mobilization is obscuring the nearly clonal nature of some isolates. The occurrence of TE integrations during asexual reproduction, such as in in vitro cultures, may also account for the instability of single spore cultures reported previously for F. poae \[[@B22-toxins-09-00255]\]. Moreover, novel genotypes that are generated through the presence of the supernumerary genome show large genetic divergence from other F. poae isolates when analyzed by a genome-wide method such as AFLP. These findings may explain the discrepancy in intra-species variability between genome-wide and multi-locus barcoding approaches that has been witnessed for F. poae. Genome-wide studies such as those employing RAPD \[[@B22-toxins-09-00255]\], AFLP \[[@B21-toxins-09-00255],[@B23-toxins-09-00255]\] and ISSR \[[@B21-toxins-09-00255],[@B25-toxins-09-00255]\] consistently detect medium to high genetic diversity, which is in sharp contrast to multi-locus barcoding analyses \[[@B26-toxins-09-00255],[@B27-toxins-09-00255]\]. The genome dynamics and the active TEs arsenal that are present in F. poae, may accelerate its evolution and lead to elevated levels of genetic diversity, without interfering with barcode diversity. In Fusarium species within the F. sambucinum clade, trichothecene production is orchestrated by three loci comprising the Tri 5 gene cluster, the Tri1-Tri16 cluster and the Tri 101 gene. Within the trichothecene type B producing FGSC (Fusarium graminearum species complex), it is common to define the chemotype of an isolate by its genotype, determined by a diagnostic PCR of a differential trichothecene biosynthesis gene \[[@B37-toxins-09-00255]\]. A similar method has been developed for the NIV chemotype in F. poae using the Tri7 gene \[[@B33-toxins-09-00255]\]. All isolates in the study of Dinolfo et al. \[[@B21-toxins-09-00255]\], all isolates in the study by Covarelli et al. \[[@B7-toxins-09-00255]\], and all isolates in our study were genetically determined to have the NIV chemotype. Another important locus determining trichothecene production is Tri1. Variations in Tri1 sequence are responsible for recent shifts in trichothecene production within F. graminearum this species \[[@B34-toxins-09-00255]\]. Remarkably in F. poae, different haplotypes of the FpTri1 gene can also be found although most SNPs are synonymous mutations. The existence of two major Tri1 types within the collection was confirmed, with additional variation at the level of individual isolates. As both types code for highly similar enzymes, this variability cannot explain the variable phenotypic chemotype. Finally, we studied the Tri5 gene cluster in F. poae. In F. poae isolate 2516, this main trichothecene cluster and neighboring genes are similar in organization to the trichothecene clusters in F. graminearum and F. sporotrichioides \[[@B38-toxins-09-00255]\]. However, uniquely for a Fusarium species, but in line with the unique genome dynamics of F. poae, the main trichothecene biosynthesis cluster and its environment are invaded by TEs. The insertion of repetitive sequences into the genome is a useful tool for track and trace studies in fungal populations, as every such insertion can be considered a unique event. In fact, all sequenced genome contain a unique pattern of TE insertions in the trichothecene biosynthetic gene cluster. Using the three insertions across the trichothecene biosynthesis cluster, a recombination event was uncovered illustrating the presence of a sexual cycle in F. poae. In F. graminearum, it was shown that the main trichothecene cluster, with similar genomic coordinates as in F. poae, is situated in a region of the genome with an elevated recombination rate \[[@B39-toxins-09-00255]\]. Ward et al. \[[@B40-toxins-09-00255]\] have shown that the composition of the main trichothecene cluster in type B producing species is the result of multiple recombination events. O'Donnell et al. \[[@B41-toxins-09-00255]\] have shown that the main trichothecene biosynthesis cluster in F. aethiopicum (a member of the FGSC) is the result of inter-chemotype recombination. That cluster is the result of recombination between different haplotypes in these species independently, but it remains to be seen what the functional consequences for trichothecene production have been. The localization of a RIPped pogo element between Tri6 and Tri5 in isolates 2516 and bfb0173 is remarkable. No intact copies of this element could be found in the core genome, and it was likely RIPped to extinction in isolate 2516 \[[@B19-toxins-09-00255]\]. In the F. equiseti complex, this space is occupied by an additional Zn2Cys6 transcription factor \[[@B32-toxins-09-00255]\]. There is a strong link between transposable elements and transcription factors (TF). Many TF-associated DNA binding sites are derived from ancient transposable elements and TEs have played a defining role in the formation of complex regulatory networks \[[@B42-toxins-09-00255]\]. Moreover, some TF families are derived from ancient TE domestication \[[@B43-toxins-09-00255]\]. There is no similarity between the pogo element in F. poae and the Zn2Cys6 TF in F. equiseti (data not shown), so an evolutionary link between the two seems unlikely. Nevertheless, insertions into the Tri6-Tri5 intergenic region have been shown to influence trichothecene production \[[@B44-toxins-09-00255]\]. To investigate whether the variability in trichothecene genes influenced the production of trichothecenes (i.e., the phenotypic chemotype), we monitored the production of both type A (DAS and NEO) and type B (FUS-X and NIV) trichothecenes. Remarkably, there was no correlation between any of the variations at the level of the Tri genes (Tri1, pogo element between Tri5 and Tri6) and the production of type A and Type B trichothecenes. Other researchers addressed this phenotypic chemotype of F. poae. Pasquali et al. \[[@B45-toxins-09-00255]\] used agmatine rather than L-arginine as a nitrogen source in their trichothecene biosynthesis inducing medium and detected FUS-X and NIV production from four F. poae isolates, with large inter-isolate differences. Previous studies have utilized autoclaved cereal substrates \[[@B15-toxins-09-00255],[@B23-toxins-09-00255]\] and in planta inoculations \[[@B46-toxins-09-00255]\]. The discrepancies in chemotype depending on the matrix or medium (and nitrogen source) used, may lead to an underestimation of certain toxicological aspects of F. poae. The Italian isolates in our study were previously chemotyped by growth on autoclaved wheat. Three that were at that time found to produce no type A trichothecenes, produced both DAS and NEO in the mineral medium. This is an important point to make, as, at least in certain assays, type A trichothecenes are more toxic than type B trichothecenes \[[@B47-toxins-09-00255],[@B48-toxins-09-00255]\]. Despite all the variability in trichothecene production between isolates, the chemotype was shown to be strictly hierarchical in all isolates, implying that when NIV is detected in F. poae, Type A trichothecenes are also synthesized. Inoculations in cereal matrices seemingly shift the balance more towards type B production (NIV), while the in vitro mineral medium with [l]{.smallcaps}-arginine mainly stimulated DAS production \[[@B46-toxins-09-00255]\]. However, trichothecenes of the DAS-related group such as mono-acetyl scirpenol and scirpenol are often overlooked in mycotoxin surveys, while they may be important compounds in nature, particularly associated with F. poae \[[@B49-toxins-09-00255],[@B50-toxins-09-00255]\]. Finally, although F. poae is often reported as T-2 toxin and HT-2 toxin producing species, our data on the TRI genes detected in F. poae confirmed previous reports on the lack of the key gene related to the production of both mycotoxins in the trichothecene cluster \[[@B32-toxins-09-00255]\]. This achievement is extremely important for a correct risk assessment of the F. poae occurrence on agro-food important crops, since T-2 toxin and HT-2 toxin are the most acute poisoning mycotoxins produced by Fusarium species and both are under evaluation at European level for their possible limits in food products. 4. Conclusions {#sec4-toxins-09-00255} ============== Using a combined approach of AFLP and MAT locus analysis, we identified hallmarks of both sexual recombination and clonal spread of isolates in the population. Based on Tri7 sequence analysis, all F. poae isolates had the NIV chemotype, although not all isolates produced NIV in vivo. Tri gene clusters showed two layers of genetic variability. Firstly, the Tri1 locus was highly variable with mostly synonymous mutations and mutations in introns. Secondly, in a subset of isolates, the main trichothecene gene cluster was invaded by a transposable element between Tri5 and Tri6. Nevertheless, these sources of variability could not explain variations in the phenotypic chemotype. 5. Materials and Methods {#sec5-toxins-09-00255} ======================== 5.1. Fusarium Collection {#sec5dot1-toxins-09-00255} -------------------------- For the purpose of this study, a broad collection of Fusarium isolates was gathered ([Table 1](#toxins-09-00255-t001){ref-type="table"}). Of the 69 F. poae isolates, forty-one isolates were collected from fields in Flanders, Belgium. Ten Italian isolates were previously described \[[@B23-toxins-09-00255]\]. Four isolates (Norwegian isolates) were donated by Dr. Anne van Diepeningen. One isolate (PD93/1780) originated from the Fusarium collection at Wageningen university (WUR). Twelve isolates came from the MUCL culture collection (MUCL, Louvain-La-Neuve, Belgium) and originated from maize and wheat. One Chinese isolate was included in the study (bfb0173). Eight F. graminearum isolates and five F. culmorum isolates were obtained from Belgian fields as described above and the F. graminearum 8/1 isolate was kindly provided by Dr. Karl-Heinz Kogel. The F. sporotrichioides isolate (MUCL6133) and F. langsethiae isolate (MUCL34988) were purchased from the MUCL collection. One F. langsethiae isolate (CBS11324) was purchased from the CBS collection (Utrecht, The Netherlands). The other F. langsethiae isolates were kindly donated by Dr. Simon Edwards (Harper Adams College, UK) and Dr. Ingerd Hofgaard (Bioforsk, Norway) respectively. For a detailed overview of these isolates, we refer to [Table S3](#app1-toxins-09-00255){ref-type="app"}. Field samples, presumed to consist of a mix of species and genotypes, were obtained and confirmed to contain F. poae as described in Audenaert, Van Broeck, Bekaert, De Witte, Heremans, Messens, Hofte and Haesaert \[[@B5-toxins-09-00255]\]. These were purified to a single spore level according to a method by Dr. Susanne Vogelgsang (Agroscope, Zürich, Switzerland, personal communication). Mycelium plugs were placed on a PDA (potato dextrose agar) plate for 7 days under a light regime of UV/darkness for sporulation (12 h (365 nm 10 W)/12 h). Sporulated mycelium was resuspended in 9 mL of distilled water. After vortexing and diluting, the resulting conidial dilutions were poured on plates containing water agar. Conidia were allowed to settle briefly after which the suspension was poured off. Plates were incubated for 16 h in the dark in a slanted position. Finally, typical F. poae spores were isolated using a Pasteur pipette. DNA extraction of isolates was performed as described by Audenaert et al. \[[@B5-toxins-09-00255]\]. Isolates confirmed to be F. poae with primers Fp82F/R ([Table S2](#app1-toxins-09-00255){ref-type="app"}) \[[@B51-toxins-09-00255]\] were preserved at −80 °C as spores in a 20% glycerol solution. 5.2. Phylogenetic Analyses {#sec5dot2-toxins-09-00255} -------------------------- DNA for AFLP analyses was extracted with the Invisorb Spin Plant MiniKit (Invitek, Berlin, Germany) according to the manufacturer's instructions. DNAs were quantified with the Nanodrop 1000 system (ThermoFisher Scientific, Merelbeke, Belgium) and AFLP analysis was performed as described by Vos et al. \[[@B24-toxins-09-00255]\]. For selective amplification, four primer pairs, EcoRI-AC/MseI-CC, EcoRI-AC/MseI-CA, EcoRI-AC/MseI-CG and EcoRI-GG/MseI-CA were used. Fragments were analyzed on a CEQ 2000 Genetic Analysis System (Beckman Coulter, Fullerton, California) and visualized with the Genographer 1.6.0 software (Benham, Montana State University, Bozeman, Montana). Bands were scored visually and data was processed as described in Scauflaire et al. \[[@B52-toxins-09-00255]\]. From the visual scoring, a binary matrix was constructed and UPGMA cluster analysis was performed based on the Dice similarity coefficients between the isolates. To simulate the AFLP pattern of a clonal isolate, and to determine the genotyping error rate, three DNA subsamples from one isolate were included in the experiment. 5.3. Fusarium poae Genome Data {#sec5dot3-toxins-09-00255} -------------------------------- The sequence of the trichothecene biosynthesis loci in isolates bfb0173, 2548 and 7555 was extracted from their assemblies and compared with the genome sequence of isolate 2516 (NCBI accession LYXU00000000.1 \[[@B19-toxins-09-00255]\]). The annotation of isolate 2516 was checked to ascertain that genes were correctly annotated, in accordance with literature on related Fusarium species. Similarities between amino acid sequence of the predicted proteins, and between nucleotide sequence of the intergenic regions, were determined with ClustalW2 \[[@B53-toxins-09-00255]\]. To visualize differences between the different Tri1 types, the HiSeq reads of isolates bfb0173 and 2548 were mapped to the only Tri1 type already published (NCBI accession GQ915520) with CLC Genomics Workbench 7.5. Reads were mapped at 0.5 length fraction and 0.8 similarity fraction. DNA from five F. poae isolates (6127, 6114, 42824, 30702 and 2531) was pooled in a 1:1:1:1:1 ratio. This DNA mix was sequenced with Illumina technology as outlined by Vanheule et al. (2016) \[[@B19-toxins-09-00255]\]. The resulting HiSeq reads cannot be separated and are therefore a hybrid pool of five isolates. Investigation into identical integration of TEs was performed as outlined by Vanheule et al. \[[@B19-toxins-09-00255]\]. 5.4. Trichothecene Production Analyses {#sec5dot4-toxins-09-00255} -------------------------------------- A trichothecene biosynthesis inducing medium described by Gardiner et al. \[[@B54-toxins-09-00255]\] was used. The general composition of the medium was identical as described in their study, but Phytagel was excluded and NaNO~3~ was exchanged with [l]{.smallcaps}-arginine (Duchefa Biochemie, Haarlem, The Netherlands) at 5 mM as Gardiner et al. \[[@B54-toxins-09-00255]\] showed that this nitrogen source induces trichothecene biosynthesis in F. graminearum. Medium was prepared in double concentration. Conidia were harvested by adding distilled water amended with 0.01% Tween80 (Merck, Darmstadt, Germany) to the fully grown PDA plates and by rubbing the mycelium with a spatula. Conidia were counted with a Bürker counting chamber and diluted to a final concentration of 2 × 10^6^ conidia/mL. Finally, 0.5 mL of double concentrated medium and 0.5 mL of double concentrated conidial suspension were added together in 24 well plates (leading to 1× concentration of medium and 1× concentration of conidia). The 24 well plates were closed with a lid and the fungus grew stationary under 16 h light/8 h dark regime at 22 °C. The medium and fungal mass was removed from the wells. After seven days, supernatant was extracted from the cultures by centrifugation at 2750 g for 10 min and immediately processed for LC-MS/MS analysis. Sample preparation and LC-MS/MS followed an "evap and shoot" principle, without extensive sample cleanup. Individual mycotoxin solid standards (1 mg) of DON, NIV, NEO, FUS-X, T-2 toxin, HT-2 toxin, DAS and deepoxydeoxynivalenol (DOM) were supplied by Coring System Diagnostics (Gernsheim, Germany) as certified solutions. All mycotoxin solid standards were dissolved in acetonitrile (1 mg/mL), and were storable for a minimum of 1 year at −18 °C \[[@B55-toxins-09-00255]\]. Working solutions of 10 ng/μL for DON, NIV, NEO, FUS-X, T-2 toxin, HT-2 toxin, DAS and DOM were prepared in methanol and stored at −18 °C. From the individual working solutions, a mixture was prepared in methanol, stored at −18 °C and renewed monthly with the following concentrations: the mycotoxin mix (mycotoxins, 10 ng/μL) and the internal standard (DOM, 10 ng/μL). An aliquot of the trichothecene biosynthesis inducing medium was transferred in a 10 mL glass tube and the internal standard was added (DOM, 10 μL, final concentration 100 ng/mL). To construct a calibration curve, five blank aliquots were spiked with the mycotoxin mix in an increasing concentration range (2 μL (20 ng/mL), 5 μL (50 ng/mL), 10 μL (100 ng/mL), 15 μL (150 ng/mL) and 20 μL (200 ng/mL), respectively). Samples were vortexed for 2 min (Labinco, Breda, The Netherlands) and spiked medium was evaporated to dryness under a gentle N~2~-stream at 60 °C using the Turbovap^®^ LV Evaporator (Biotage, Uppsala, Sweden). The dried residue was redissolved in 100 μL injection solvent, which consisted of 70% mobile phase A (water/methanol (95/5, v/v) + 5 mM ammonium acetate, 0.1% glacial acetic acid) and 30% mobile phase B (water/methanol (95/5, v/v) + 5 mM ammonium acetate, 0.1% glacial acetic acid). Prior to injection, the sample was vigorously vortexed for 2 min, collected in a 0.22 μm Ultrafree-MC centrifugal device (Millipore, Bedford, MA, USA) and centrifuged for 10 min at 10,000 g. LC-MS/MS methodology was as detailed in Vanheule, Audenaert, De Boevre, Landschoot, Bekaert, Munaut, Eeckhout, Hofte, De Saeger and Haesaert \[[@B8-toxins-09-00255]\]. LODs and LOQs for the trichothecenes produced by F. poae were, respectively, 30 and 61 ng/mL (NIV), 31 and 63 ng/mL (FUS-X), 39 and 78 ng/mL (NEO) and 41 and 82 ng/mL (DAS). Certain samples showed peaks which according to the calibration curve corresponded to concentrations between the LOD and the LOQ for DAS, NEO and FUS-X. These samples were assigned the LOD concentration as they could not be reliable quantified. Samples with concentrations out of the range of the calibration curve were diluted and reanalyzed. To circumvent the possible matrix effect issue, a new calibration curve was constructed with a blank sample. This blank sample was also diluted, using the same protocol as the unknown diluted samples. Internal standards and spiked concentrations were then added anew. Data acquisition and processing was performed with MassLynx 4.1 and QuanLynx 4.1 software respectively (Micromass, Manchester, UK). Pearson correlations between the levels of trichothecenes were determined with Microsoft Excel 2013. The concentration range of the trichothecenes was visualized in Box-Whiskers plots, constructed with SPSS22 (IBM, Armonk, NY, USA). The effect of the pogo insertion on trichothecene production was analyzed with the Mann--Whitney U test in SPSS. 5.5. Diagnostic PCRs and Amplicon Sequencing {#sec5dot5-toxins-09-00255} -------------------------------------------- All primers are listed in [Table S2](#app1-toxins-09-00255){ref-type="app"}. Mating type was determined as described by Kerenyi et al. (2004). For the two insertions of supernumerary sequence into the core chromosomes (INS1 and INS2), primers were used as described in Vanheule et al. (2016). A 1100 bp fragment of the Tri1 gene was amplified as described in Proctor et al. (2009) for a set of 34 F. poae isolates selected from different backgrounds and chemotypes. The presence of a pogo TE in the Tri cluster was investigated with three primers as detailed in [Table S2](#app1-toxins-09-00255){ref-type="app"}. PCR reactions were done with Promega GoTaq G2 Polymerase according to the manufacturer's instructions (Promega, Leiden, The Netherlands). Annealing temperatures were chosen, based on Promega's Biomath Calculator. PCRs were performed in an Applied Biosystems GeneAmp PCR System 9700, amplicons were separated on 1.5% (w/v) agarose gels stained with 0.1 μg/mL ethidium bromide, and visualized with a Biorad Gel Doc XR+. PCR products were purified with the EZNA Cycle-Pure kit (VWR Chemicals, Haasrode, Belgium) and sequenced in both directions by Macrogen Inc. (Amsterdam, The Netherlands). Forward and reverse sequences were curated with CLC Main Workbench 7. Maximum parsimony analysis was performed with PAUP version 4.0b10 \[[@B56-toxins-09-00255]\]. The Tri1 fragment from Fusarium sp. NRRL 36351 (NCBI accession GQ915523.1) was used as the outgroup, however for construction of the phylogenetic tree mid-point rooting was used. Bootstrapping was performed with PHYLIP based on 1000 replicates \[[@B57-toxins-09-00255]\]. We would like to acknowledge the financial support received for this study from MYCOKEY, a project funded within the Horizon2020 Research and Innovation program of the European Commission (project number 678781). The following are available online at [www.mdpi.com/2072-6651/9/9/255/s1](www.mdpi.com/2072-6651/9/9/255/s1), Figure S1: Diagnostic PCR of the pogo insertion between Tri5 and Tri6, Figure S2: Read mapping of reads from isolate bfb0173 (A and C) and isolate 2548 (B and D) on NCBI accession GQ915520, which is the only sequence of FpTri1 available \[[@B32-toxins-09-00255]\], Table S1: Chemotypes of 28 F. poae isolates in two biological repeats, Table S2: List of all primers used in this study, Table S3: Isolates from additional Fusarium species collected for this study. ###### Click here for additional data file. K.A., C.W. and T.v.d.L. conceived and designed the experiments; A.V., M.D.B., A.M., J.S., F.M., B.B. performed the experiments; A.V., T.v.d.L. and M.D.B. analyzed the data; J.S. and F.M. contributed reagents/materials/analysis tools; A.V., S.D.S., G.H., C.W., T.v.d.L. and K.A. wrote the paper. The authors declare no conflict of interest. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results. ![AFLP tree of F. poae isolates, and superposition of genetic markers. UPGMA (Unweighted Pair Group Method with Arithmetic Mean) cluster analysis was performed based on the Dice similarity coefficients between the isolates. In total, 247 markers were scored of which 201 were polymorphic. Three technical replicates of isolate 9125 are also included in the graph (blue bar). Clusters of isolates that share more than 90% genetic similarity according to the Dice similarity coefficient are grouped with the same colors. Bootstrap values exceeding 70% are given at the nodes, based on 1000 replications. Four genetic markers that were determined throughout this study are listed (MAT1, pogo, INS1, and INS2). White entries are markers for which no data could be collected. MAT1: yellow is MAT1-1, green is MAT1-2. Pogo: green indicates presence of the pogo element between Tri5 and Tri6 in the major gene cluster, red indicates absence. INS1 and INS2 are the major insertions of supernumerary sequence into chromosome 3 of a subset of isolates. Yellow indicates the absence of the insertion, green indicates the presence of the insertion. The isolate for which reference genomes is available is shown with asterisks.](toxins-09-00255-g001){#toxins-09-00255-f001} ![Organization of the main trichothecene biosynthesis cluster in F. poae (isolate 2516; NCBI accession LYXU01000002.1), which is identical to the organization in F. graminearum (isolate 88-1, NIV chemotype, NCBI accession AF336365) and F. sporotrichioides (isolate NRRL 3299, NCBI accession AF359360). The difference in Tri6-Tri5 intergenic region between F. poae and F. graminearum/F. sporotrichioides is explained by the presence of a RIPped pogo transposable element (yellow). This element is also present at the same location in isolate bfb0173. In isolates 2516 and bfb0173, there is a 400 bp repetitive element between Tri7 and Tri3 (blue). Finally, in isolates 2548 and bfb0173, there is a RIPped retrotransposon downstream of Tri14 (shaded box).](toxins-09-00255-g002){#toxins-09-00255-f002} ![Maximum parsimony phylogenetic tree that was built from the 1100 bp fragments of the Tri1 gene with PAUP \* 4.0b10. The tree is based on 74 parsimony informative characters. Bootstrap values were calculated with the PHYLIP package and those exceeding 70% are shown on the tree. Fusarium sp. NRRL 36351 (NCBI accession GQ915523.1) was used as outgroup, however, for construction of the phylogenetic tree, mid-point rooting was used.](toxins-09-00255-g003){#toxins-09-00255-f003} ![Summary of trichothecene biosynthesis in F. poae isolates. (A), simplified representation of the chemotype of DAS producing F. poae isolates. The chemotypes detected were strictly hierarchical, i.e., when compounds lower on the scheme are detected, the compounds higher on the scheme have been produced as well. (B), Pearson correlations between the four trichothecenes that are produced by F. poae isolates. (C), overview as Box-Whisker plots of the concentration ranges of the four trichothecenes. Note that the range of DAS is two orders of magnitude higher than the range for NEO, FUS-X and NIV concentrations. The boxes for every mycotoxin show the lower and upper quartile (delineating the boxes) of the measured concentrations (ng/mL) and the median (thick line within the boxes). The whiskers represent the minimum and maximum values. Circles and asterisks are outliers and extreme values which fall respectively outside of one-and-a-half additional box lengths and three additional box lengths counted from the upper quartile limit.](toxins-09-00255-g004){#toxins-09-00255-f004} toxins-09-00255-t001_Table 1 ###### List of F. poae isolates (n = 69) used in this study. Location, host, year of isolation and mating type are shown. The ID is the identification number in our local collection. Concentrations of mycotoxins are expressed as ng/mL. ND: not detectable (\<LOD). \>LOD: amount above the LOD but below the LOQ. NCA: no chemotype available (not determined). DAS: diacetoxyscirpenol; NEO: Neosolaniol; FUS-X: Fusarenon-X; NIV: Nivalenol. Absence (-) or presence (+) of two supernumerary sequence insertions in chromosome 3 (INS1: \>204 kb and INS2: 464 kb) are indicated in the table. Availability of an AFLP profile is indicated by "+" or "-". -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ID Location Host Year Reference Mating Type INS1 INS2 AFLP DAS NEO FUS-X NIV ----------- ----------------------------- ---------------- ------ ------------------------------ ------------------- ---------------------------- ---------------------------- ---------------------- ----------------------------------------------------------------- ------ ------- ----- 175 Aas, Norway barley 1996 Sundheim L. unpubl. MAT1-2 \- \- \- 355 55.1 ND ND 177 Norway wheat 1996 Sundheim L. unpubl. MAT1-1 \- \- \+ NCA 182 Norway barley 1996 Sundheim L. unpubl. MAT1-1 \- \- \+ NCA 185 Norway barley 1996 Sundheim L. unpubl. MAT1-1 \- \- \+ ND ND ND ND 1879 Bottelare, Belgium wheat 2010 this study MAT1-1 \- \- \+ 2160 66 25 88 2004 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 89 ND ND ND 2019 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 231 54 ND ND 2022 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 1588 62 18 ND 2023 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 44 12 ND ND 2028 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 3650 79 77 93 2031 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 109 ND ND ND 2033 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 1306 59 ND ND 2041 Zwevegem, Belgium wheat 2010 this study MAT1-2 \- \- \+ 60 12 \>LOD 108 2043 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 3971 72 ND ND 2044 Zwevegem, Belgium wheat 2010 this study MAT1-2 \- \- \+ 144 9 ND ND 2056 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ NCA 2371 Bottelare, Belgium wheat 2011 this study MAT1-2 \- \- \+ 6923 101 ND ND 2375 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 15 10 ND ND 2377 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 31 11 \>LOD 103 2380 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 1743 64 ND ND 2381 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 222 54 ND ND 2390 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 3027 67 ND ND 2392 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 196 14 ND ND 2395 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 193 55 ND ND 2410 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 9562 84 ND ND 2411 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 238 54 ND ND 2424 Koksijde, Belgium wheat 2011 this study MAT1-1 \- \- \+ 2098 59 ND ND 2476 Poperinge, Belgium wheat 2011 this study MAT1-1 \- \- \+ 488 75 172 103 2491 Poperinge, Belgium wheat 2011 this study MAT1-2 \- \- \+ 12,336 99 ND ND 2514 Zwevegem, Belgium wheat 2011 this study MAT1-1 \- \- \+ 2800 73 83 116 2516 Zwevegem, Belgium wheat 2011 this study MAT1-1 \+ \+ \+ 1126 57 ND ND 2517 Zwevegem, Belgium wheat 2011 this study MAT1-1 \- \- \+ 641 58 ND ND 2519 Zwevegem, Belgium wheat 2011 this study MAT1-2 \- \- \+ 9766 139 ND ND 2521 Zwevegem, Belgium wheat 2011 this study MAT1-1 \+ \- \- ND ND ND ND 2524 Zwevegem, Belgium wheat 2011 this study MAT1-1 \- \- \+ 380 17 ND ND 2525 Zwevegem, Belgium wheat 2011 this study MAT1-1 \- \- \+ 2195 61 ND ND 2531 Zwevegem, Belgium wheat 2011 this study MAT1-1 \+ \+ \+ 803 59 22 92 2532 Zwevegem, Belgium wheat 2011 this study MAT1-1 \- \- \+ 1940 66 25 ND 2547 Zwevegem, Belgium wheat 2011 this study no amplicon \- \- \+ NCA 2548 Zwevegem, Belgium wheat 2011 this study MAT1-1 \- \- \+ 67 ND ND ND 2565 Zuienkerke, Belgium wheat 2011 this study MAT1-1 \+ \- \+ 69 18 \>LOD 115 2569 Zuienkerke, Belgium wheat 2011 this study MAT1-1 \+ \- \+ 4910 44 \>LOD 122 2570 Zuienkerke, Belgium wheat 2011 this study MAT1-1 \+ \- \+ 6630 66 25 ND 2571 Zuienkerke, Belgium wheat 2011 this study MAT1-1 \+ \- \+ NCA 2671 Linter, Belgium wheat 2011 this study MAT1-1 \+ \+ \+ 2457.8 75 37 100 6127 Wageningen, The Netherlands wheat 1964 MUCL MAT1-2 \- \- \- 10,864 303 75 ND 6114 Denmark barley 1964 MUCL, Hennebert G.L. unpubl. MAT1-2 \- \- \+ 135 55 ND ND 7555 Heverlee, Belgium wheat 1965 MUCL MAT1-1 \- \- \- 1067.8 ND ND ND 9125 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-1 \- \- \+ 5096 73 ND ND 9139 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-2 \- \- \+ 8 ND ND ND 9181 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-2 \- \- \+ NCA 9186 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-1 \- \- \+ 8024 77 24 ND 9189 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-1 \- \- \+ 6342 98 ND ND 9192 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-1 \- \- \+ 10,415 132 46 91 9194 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-1 \- \- \+ 13,404 270 81 ND 9196 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-2 \- \- \+ 3610 115 34 ND 9203 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-1 \- \- \+ 115 55 ND ND 9209 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-1 \- \- \+ 5502 115 28 ND 11456 Heverlee, Belgium barley 1968 MUCL, Meyer J.A. unpubl. MAT1-2 \- \- \+ 264 20 \>LOD 103 30702 unknown in vitro plant 1990 MUCL, Marchand D. unpubl. MAT1-2 \- \- \+ 22,804 375 57 ND 15926 Quebec, Canada wheat 1970 MUCL, Hennebert G.L. unpubl. MAT1-1 \- \- \+ NCA 42824 Belgium wheat 2000 MUCL MAT1-1 \- \- \- 2120 73 ND ND bfb0173 China barley 2005 Yang et al. (2008) MAT1-1 \- \- \+ NCA F49 Ath, Belgium maize 2007 MUCL (Scaufflaire J.) MAT1-1 \- \- \+ 5720 76 ND ND K46 Ath, Belgium maize 2007 MUCL (Scaufflaire J.) MAT1-1 \- \- \+ 5339 100 ND ND L24 Buissenal, Belgium maize 2007 MUCL (Scaufflaire J.) MAT1-1 \- \- \+ 162 55 ND ND Q57 Buissenal, Belgium maize 2007 MUCL (Scaufflaire J.) MAT1-1 \- \- \+ 916 56 ND ND S46 Villeroux, Belgium maize 2007 MUCL (Scaufflaire J.) MAT1-1 \- \- \+ 1208 59 ND ND PD93/1780 The Netherlands carnation 2003 WUR, Waalwijk et al. (2003) MAT1-1 \- \- \+ 4008 66 23 ND n = 69 1 not determined\ 8 isolates containing INS1 3 isolates containing INS2 AFLP for 64 isolates chemotype for 61 isolates; no chemotype analysed for 8 isolates 13 MAT1-2\ 55 MAT1-1 -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- toxins-09-00255-t002_Table 2 ###### Transposable elements at unique locations in isolate 2516. List of TEs found on the core genome of isolate 2516, as well as those that do not have any read support from reads of a DNA mix consisting of five isolates, including isolate 2531. These are TE copies that are not present in the same location in isolate 2531, and this is a lower bound as there is no way to discern in which isolate(s) the remaining TEs are present (reads cannot be traced back to isolates). \#, indicates the number of copies. \# On the Core Genome of 2516 \# Present in 2516 but Not in 2531 (No Read Support) ------------------ ----------------- ---------------- ------------------------------- ------------------------------------------------------ Retrotransposons RLG\_Maggy Gypsy/Ty3 like 27 1 RLG\_Skippy Gypsy/Ty3 like 5 2 DNA transposons DTF\_Fot4 Pogo 1 0 DTF\_Fot5-A Pogo 40 26 DTF\_ESP4-B Pogo 12 0 DTF\_Drogon Pogo 33 33 DTF\_Viserion Pogo 8 3 DTM\_Hop7 Mutator 8 8 DTM\_Hop4 Mutator 1 0 Sum 135 73 [^1]: These authors have equal contributions.	{ "pile_set_name": "PubMed Central" }
1.. INTRODUCTION ================ Vector-borne diseases like Chagas disease and leishmaniasis are caused by parasitic human pathogens and are a major health burden in many developing countries. Current therapies are not very effective and suffer from toxic side effects \[[@R1]-[@R2]\]. In addition, the emergence of drug-resistant strains has been reported \[[@R3]-[@R6]\]. These vector-borne diseases have been recognized by the World Health Organization (WHO) as Neglected Tropical Diseases (NTD) -- chronic infectious diseases endemic mainly in underdeveloped countries, and even though millions of people are affected and thousands die every year, there are no effective cures \[[@R7]\]. In recent years, significant research efforts have been focused on NTD due to policies and research programs implemented by the WHO and other governmental and private organizations \[[@R8]\]. Fungi from Aspergillusspecies cause a series of broncho-respiratory infections collectively known as aspergillosis \[[@R9]-[@R10]\]. Infections by Aspergillus fumigatus are the most common in immuno-compromised individuals. Once infection has been established, the mortality rate can be close to 50% \[[@R11]\]. Therefore, new effective anti-fungal drugs are urgently needed. A possible mode of intervention against these parasitic and fungal pathogens is to inhibit the activity of enzymes that aid in cell wall biosynthesis and/or host-pathogen interactions \[[@R12]-[@R13]\]. It has recently been shown that galactofuranose (Galf), a sugar not found in humans, plays an important role in cell wall biosynthesis in A. fumigatus and many bacteria and is a major component of the cell surface matrix of Trypanosoma cruzi and Leishmania major, the causative agents of Chagas disease and leishmaniasis, respectively. In these parasites, Galf plays a major role in virulence \[[@R14]\]. The flavoenzyme UDP-galactopyranose mutase (UGM) is a unique enzyme not present in humans and is essential in the biosynthesis of Galf. Here, we provide an overview of the biosynthesis of Galf and its role in pathogenesis in eukaryotic pathogens with a focus on recent studies on eukaryotic UGMs from T. cruzi, L. major, and A. fumigatus. 1.1. Neglected Tropical Diseases Caused byT. cruziandL. major ----------------------------------------------------------------- Chagas disease (or American trypanosomiasis) is endemic throughout Central and South America. It is caused by the protozoan parasite T. cruzi and is usually transmitted through a sylvatic cycle from an infected triatomine ("kissing bug") vector that lays parasite-laden feces on wounds and mucous membranes, as well as in conjunctivas \[[@R15]\]. In addition, there have been reports of infection via blood transfusion and orally through ingesting infected mother's milk, raw and undercooked meat, or other food infected by triatomines and/or their feces \[[@R16]\]. Very often in its early stages, Chagas disease is asymptomatic. If left untreated, parasite invasion becomes a serious health risk; symptoms can develop 10-20 years later when the disease becomes chronic and has high mortality rates, typically due to the parasitosis of the heart, causing myocarditis \[[@R17]\]. It is estimated that approximately 16-18 million people have Chagas disease and approximately 50,000 of them die annually; however, these numbers could be higher, since infections are often misdiagnosed due to the very limited, or sometimes complete lack of symptoms \[[@R15],[@R18]\]. Leishmaniasis is a vector-borne systemic disease caused by a trypanosomatid protozoa from the Leishmania spp., which invade human macrophages and replicate intracellularly after being transmitted to humans by infected sandflies (genera Phlebotomus and Lutzomyia) \[[@R19]\]. Depending on the particular parasitic species, leishmaniasis can develop into three forms: cutaneous, mucocutaneous, or visceral leishmaniasis \[[@R20]\]. L. major is the causative agent of cutaneous leishmaniasis, which manifests as a severe skin infection that often causes disfigurement and is endemic in developing countries in the tropics, subtropics, and the Mediterranean basin, with thousands reported new cases annually \[[@R7]\]. 1.2. Infections Caused by A. fumigatus ---------------------------------------- Fungi of the genus Aspergillus are responsible for several human diseases ranging from allergic reactions and lung infections to sepsis and death \[[@R9]\]. There are hundreds of members of the Aspergillusgenus,and some are pathogenic to humans, with A. fumigatusandA. nigerbeing the most common \[[@R21]-[@R23]\]. Among the diseases related to Aspergillus infection, allergic bronchopulmonary aspergillosis (ABPA) and invasive pulmonary aspergillosis (IPA) represent a significant health threat to both immuno-competent and immuno-compromised persons \[[@R9],[@R24]\]. IPA infections are commonly observed in patients receiving chemotherapy, organ transplants, and in late-stage AIDS \[[@R25]-[@R26]\]. An increase from 0.3% to 5.8% in IPA infections in patients admitted to intensive care units (ICUs) has been reported in recent years, and has been accompanied by a high mortality rate (50-70%) \[[@R27]-[@R28]\]. This demonstrates the need for new anti-fungal drugs to combat Aspergillus infections. 2.. ROLE OF GALACTOFURANOSE IN VIRULENCE ======================================== Galactose is a hexose and a C-4 epimer of glucose (Fig. [1](#F1){ref-type="fig"}). In mammals, galactose exists only in the pyranoside form (Galp) linked to other carbohydrates as an essential component of glycolipids and glycoproteins \[[@R29]\]. The main source of galactose in humans comes from consumption of dairy products and its metabolism occurs through the Leloir or Isselbacher pathways \[[@R30]-[@R31]\]. Galactose in the furanoside (Galf) form is not found in mammals; however, Galf is an important building block of glycans of the cell wall and cell surface in several pathogenic organisms and, therefore, its biosynthesis is a strategic target in the discovery of anti-microbial treatments \[[@R14]\]. The specific role of Galf-containing molecules in Leishmania spp., T. cruzi, and A. fumigatus is described in this section. 2.1. Galactofuranose in T. cruzi ---------------------------------- In T. cruzi, β-Galf is found in glycoinositolphospholipids (GIPLs) and glycosylphosphatidylinositol (GPI) anchor proteins \[[@R32]-[@R33]\]. These glycoconjugates are highly expressed throughout the life cycle of T. cruzi and are the main component of the parasite dense surface coat, which has a protective function in parasite survival in the hydrolytic and digestive environment inside their hosts and are important for proliferation \[[@R34]-[@R36]\]. For instance, a 45 kDa GPI-mucin is expressed only in invasive trypomastigotes and not in non-invasive amastigotes \[[@R37]\]. Using specific monoclonal antibodies against this protein prevented adhesion of T. cruzi to heart myoblasts \[[@R37]\]. These results suggest that Galf-containing glycoconjugates are involved in the mechanism of myocardial invasion by T. cruzi. 2.2. Galactofuranose in Leishmania spp ---------------------------------------- In L. major, Galf is found in the oligosaccharide core of lipophosphoglycans (LPG) and glycoinositolphospholipids (GIPL) that are essential for parasite survival in the midgut of the vector insect and for parasite transmission to the mammalian host \[[@R38]-[@R40]\]. GIPL-1 from L. major has been shown to contribute to the infection process \[[@R41]-[@R42]\]. LPG deletion mutants in L. major showed LPG involvement in resistance to oxidative stress and evasion of the human immune system \[[@R39]-[@R40]\]. 2.3. Galactofuranose in A. fumigatus -------------------------------------- Of the vast Aspergillusgenus that includes over 185 species,A. fumigatus and A. niger are among the \~20 reported human fungal pathogens that cause a variety of opportunistic diseases facilitated by the suppression of the immune system \[[@R43]\]. Galf has been identified in both organisms and is an important component in the fungal cell wall assembly, where it was found in galactomannan, glycoproteins, sphingolipids, and lipid-linked glycans \[[@R44]-[@R48]\]. In A. fumigatus, Galf accounts for up to 5% of the dry weight, and is important for fungal growth and cell wall biosynthesis, cell morphogenesis and wall architecture, hyphal adhesion, spore development, and pathogenesis \[[@R22]-[@R23],[@R49]\]. 3.. UDP-GALACTOPYRANOSE MUTASE: AN ATTRACTIVE DRUG TARGET AGAINST EUKARYOTIC HUMAN PATHOGENS ============================================================================================ UDP-Galactopyranose mutase (UGM) is a flavin-dependent enzyme that catalyses the isomerization of UDP-Galp to UDP-Galf through a unique type of flavin-dependent catalysis (Fig. [2](#F2){ref-type="fig"}) \[[@R13],[@R50]-[@R52]\]. The gene encoding for UGM (glf) was first identified in prokaryotes in 1996 while studying theEscherichia coli K12 O antigen \[[@R53]\]. In the following years, it was identified in other pathogens including the eukaryotes L. major, T. cruzi, and A. fumigatus \[[@R54]\]. Deletion of the UGM gene leads to attenuated virulence in L. major \[[@R55]\]. In T. cruzi, the role of Galf in binding to mammalian cells has been shown; however, deletion of the UGM gene in this parasite have not been performed. Deletion of the UGM gene in A. fumigatus, in addition to causing attenuated virulence, leads to cell-wall morphology defects, increased sensitivity to anti-fungal drugs, and growth reduction \[[@R21]-[@R22]\]. These results validate UGM as a potential target for the development of drugs against these eukaryotic pathogens. 3.1. Primary Structure of UGMs ------------------------------ The polypeptide chain lengths of eukaryotic UGMs are generally about 100 amino acid residues longer than those of the prokaryotic enzymes (Fig. [3](#F3){ref-type="fig"}). Sequence alignment reveals a moderate identity (47-60%) among the eukaryotic UGMs from A. fumigatus (AfUGM), L. major (LmUGM), and T. cruzi (TcUGM), and a slightly lower (37-44%) sequence identity among the prokaryotic homologs from Escherichia coli (EcUGM), Mycobacterium tuberculosis (MtUGM), Klebsiella pneumoniae (KpUGM), and Deinococcus radiodurans (DrUGM). However, the sequence identity between eukaryotic and prokaryotic UGM groups is surprisingly low (14-18%) (Tab. [1](#T1){ref-type="table"}). Conserved among all UGMs is the GxGxxG motif that is necessary for FAD binding. Only partial conservation of active site residues is observed (Fig. [3](#F3){ref-type="fig"}). The low amino acid conservation and the extra amino acid sequence in eukaryotic UGMs endows these enzymes with unique structural features that are important for enzyme function; these are discussed in the next section. Interestingly, an obvious NAD(P)H binding domain or motif is not found in this family of enzymes. This is intriguing since this class of enzymes has been shown to function only in the reduced state. 3.2. 3-Dimensional Structure of UGMs ------------------------------------ Whereas several crystal structures of bacterial UGMs have been determined \[[@R56]-[@R58]\], among the eukaryotic enzymes, only the structure of AfUGM is known at this time (Tab. [2](#T2){ref-type="table"}) \[[@R59]-[@R60]\]. AfUGM is a mixed α/β fold protein containing three structural domains (Fig. [4](#F4){ref-type="fig"}). Domain 1 includes a Rossmann fold core and participates in FAD binding. Domains 2 and 3 function in substrate binding \[[@R59]\]. This general 3-domain architecture is also found in the bacterial enzymes; however, the eukaryotic enzymes have extra structural elements that are important in oligomerization and substrate recognition, as summarized below. The conformations of the flavin and flavin-protein interactions are highly conserved between prokaryotic and eukaryotic UGMs (Tab. [3](#T3){ref-type="table"}). The isoalloxazine ring of the oxidized enzyme is planar, which is typical for flavoenzymes. Characterization of the reduced FAD conformation is important for understanding the chemical mechanism because the reduced FAD functions as a nucleophile in the UGM reaction. This function places certain structural restrictions on the flavin isoalloxazine. In particular, steric considerations suggest that the reduced isoalloxazine should be nonplanar with the wings of the isoalloxazine bending away from the substrate. Indeed such a conformation is observed in reduced AfUGM and DrUGM. In both cases, the isoalloxazine exhibits a butterfly-like deviation from planarity in which the pyrimidine ring bends \~7º away from the substrate site such that the si face is concave \[[@R56],[@R59]\]. Curiously, bending of the isoalloxazine by \~13º in the opposite direction is observed in reduced KpUGM; the relevance of this conformation is uncertain since it appears to be inconsistent with nucleophilic attack \[[@R61]-[@R62]\]. Comparison of the structures of AfUGM and bacterial UGMs complexed with UDP-Galp reveals conserved motifs and important differences. In all the complex structures (Tab. [2](#T2){ref-type="table"} and Tab. [3](#T3){ref-type="table"}), the OH-4 of the Galp moiety interacts with the flavin O-4 through hydrogen bonding, and the anomeric carbon of the sugar is placed within a short distance from the flavin N-5 (Fig. [5](#F5){ref-type="fig"}). Also, several Arg and Tyr residues are conserved and participate in electrostatic interactions with the pyrophosphate portion of UDP-Galp (Tab. [3](#T3){ref-type="table"}). In contrast to bacterial UGMs, in AfUGM the OH-6 of Galp is rotated by 110º. There is also a variation in the conformation of bound UDP. In AfUGM, UDP is displaced by \~4 Å and rotated by about 90º with respect to bacterial KpUGM and DrUGM. This allows for the hydrogen bonding of uracil with Gln107, a residue that is not present in bacterial UGMs. These structural differences in substrate recognition between bacterial and eukaryotic UGMs could have implications for inhibitor discovery. In particular, it seems unlikely that compounds that target the uridine site of bacterial UGMs will be effective against eukaryotic UGMs. Large protein conformational changes (\>10 Å movements) accompany substrate binding in UGMs. AfUGM, and presumably other eukaryotic UGMs, exhibit larger conformational changes. Comparison of the structures of the substrate-free and substrate-bound forms revealed two flaps (residues 179--187 and 203--209) that close down over the substrate like the flaps of a box top (Fig. [5](#F5){ref-type="fig"}). Whereas the 180s flap is analogous to the mobile loop of bacterial UGMs, the 200s flap is unique to eukaryotic UGMs. The dramatic closing of the active site in bacterial and eukaryotic UGMs is an important aspect of the catalytic mechanism. These movements result in the assembly of the constellation of residues that position the substrate for nucleophilic attack by the FAD. Furthermore, the closing of the active site prevents diffusion of intermediates, such as UDP, out of the active site during the catalytic cycle. Various oligomeric states have been observed for UGMs in solution. The oligomeric states of several UGMs have been determined from size exclusion chromatography, small-angle X-ray scattering (SAXS), and analysis of protein-protein interfaces in crystal lattices (Tab. [4](#T4){ref-type="table"}). Bacterial UGMs tend to form dimers in solution. EcUGM, KpUGM, and MtUGM form a semicircular dimer \[[@R58],[@R63]\]. The fact that this structure is formed by multiple UGMs in different crystal lattices attests to its veracity. The oligomeric state of DrUGM is less certain. The classic UGM semicircular dimer is not found in the DrUGM lattice, and solution studies of the oligomeric state have not been performed on the enzyme. The DrUGM crystal lattice implies decameric and dimeric assemblies, but clearly additional work is needed to determine the oligomeric state and quaternary structure of DrUGM. In contrast, the oligomeric state and quaternary structure of AfUGM have been unequivocally determined using a combination of SAXS and X-ray crystallography \[[@R59]\]. These studies have shown that AfUGM is unique among UGMs in that it forms a tetramer in solution \[[@R59],[@R64]\]. The AfUGM tetramer is a dimer-of-dimers assembly (Fig. [6](#F6){ref-type="fig"}). Unique structural features of AfUGM that are absent in the bacterial enzymes enable tetramerization. These extra elements include a longer C-terminus, an extra helix in domain 2, and extension of another helix of domain 2 (Fig. [4](#F4){ref-type="fig"}). Because the reduced FAD is essential for catalysis, the mechanism by which the enzyme is activated by flavin reduction is an important aspect of UGM biochemistry. Insight into the structural underpinnings of this mechanism has been obtained by comparing crystal structures of oxidized and reduced UGMs. Inspection of the bacterial enzyme structures reveals little difference between the oxidized and reduced conformations, aside from the bending of the isoalloxazine described above. Initial results for AfUGM potentially reveal a much more complex mechanism for activating eukaryotic UGMs (Fig. [7](#F7){ref-type="fig"}). Two crystal forms of AfUGM have been described, a P6~5~22 form reported by us \[[@R59]\], and a P1 form reported by Sanders' group \[[@R60]\]. Although the interpretation of these structures is complicated by the binding of sulfate ion in theP6~5~22 form, and weak electron density in the P1 form, the structures tantalizingly imply large conformational changes involving the conserved histidine loop (G61-G62-H63). The structures show that, in the oxidized enzyme, conserved His63 is near the pyrimidine ring of the isoalloxazine and the carbonyl of Gly62 points away from the isoalloxazine (Fig. [7A](#F7){ref-type="fig"} and [7B](#F7){ref-type="fig"}), which is unprecedented for UGMs. The structures further indicate that flavin reduction induces a crankshaft rotation of the loop backbone, which reverses the orientation of the Gly62 carbonyl bond vector and moves the imidazole of His63 by over 5 Å. These changes bring the carbonyl of Gly62 within hydrogen bonding distance of the N5 of the reduced flavin and move His63 to the si face of the isoalloxazine where it stacks in parallel with the isoalloxazine and forms a hydrogen bond with the OH-2' of the ribityl (Fig. [7C](#F7){ref-type="fig"}). These interactions between the histidine loop and the flavin help stabilize the reduced state of the enzyme and are found in all other UGM structures. The presence of two Gly residues in the loop is unique to eukaryotic UGMs and probably accounts for the large conformational changes seen in AfUGM compared to the bacterial enzymes. More research is needed to validate these conformational changes for AfUGM and determine whether other eukaryotic UGMs exhibit analogous movements. 3.3. Chemical Mechanism of Eukaryotic UGMs ------------------------------------------ Despite structural differences, the unique chemical mechanism utilized by UGMs is conserved among the members of this enzyme family. For all UGMs, only the reduced form of the enzyme is active \[[@R63]\], although the reaction does not involve a net gain or loss of electrons, which is common among many other classes of FAD-dependent enzymes \[[@R65]-[@R67]\]. The reported steady-state values with UDP-Galf as substrate and dithionite as the reductant show only minor differences in k~cat~and k~cat/~K~M~ among members of the UGM family (Table [5](#T5){ref-type="table"}). The enzymatic reaction has been shown to involve cleavage of an anomeric bond and the formation of a Galp-FAD adduct (Fig. [8](#F8){ref-type="fig"}) \[[@R68]\]. This process was initially thought to involve one of three mechanisms: a single-electron transfer from the reduced flavin to a postulated oxocarbenium intermediate of Galp \[[@R69]-[@R70]\] or a nucleophilic substitution via an S~N~1 or S~N~2 mechanism, both leading to the formation of a Galp-FAD adduct \[[@R71]-[@R72]\]. While the Galp is bound to the FAD it undergoes ring opening and closing rearrangement and, after nucleophilic attack by UDP^-^, the UDP-Galf is produced. Formation of the FAD-sugar adduct has been demonstrated by chemical quenching, trapping, and characterization by mass spectrometry in both eukaryotic and prokaryotic UGMs \[[@R71],[@R73]\]. Rapid reaction kinetic analysis with reduced TcUGM and UDP-Galf did not show the presence of a transient flavin semiquinone, inconsistent with a single electron transfer step. Instead, absorbance changes consistent with the formation of a flavin iminium ion, were observed and occur very fast \[[@R73]\]. The structures of AfUGM and prokaryotic UGMs in complex with UDP-Galp clearly show that the substrate binds in a conformation optimal for direct attack by the flavin N5. Furthermore, linear free energy relationship (LFER) studies with prokaryotic UGM, reconstituted with various FAD analogs, show changes in k~cat~ values that correlate linearly with changes in the nucleophilicity of the flavin N5 (slope of ρ = --2.4 ± 0.4), which is consistent with an S~N~2 mechanism \[[@R72]\]. Viscosity effect studies showed that product release was not rate limiting in the case of TcUGM \[[@R73]\]. In vivo, all UGMs function in an aerobic environment. Therefore, a system for the generation and maintenance of the reduced flavin must exist in the cell. Despite not having found a canonical NAD(P)H binding motif in the primary sequence of eukaryotic UGMs, NAD(P)H was identified as an effective electron donor for the reduction of the flavin cofactor in TcUGM \[[@R73]\]. Kinetic analyses show that there is preference for NADPH, as it reduces the flavin 7 times faster and binds 5 times tighter than NADH (Tab. [6](#T6){ref-type="table"}). In contrast, MtUGM is unable to effectively react with reduced coenzymes \[[@R73]\]. It has been previously reported that the activity of KpUGM was enhanced by the addition of NADH or NADPH \[[@R74]-[@R75]\]. The binding affinities or rates of reduction were not reported, however, the rate enhancement was observed at concentrations greater than 20 mM NADH and at incubation times longer than 10 minutes \[[@R74]\]. Taking into account that NADPH is capable of reducing TcUGM with a rate constant in the second time scale and it binds with micromolar affinity, it is clear that relative to eukaryotic UGM, the bacterial enzymes are not effective NAD(P)H oxidases. The mechanism shown in (Fig. [8](#F8){ref-type="fig"}) was recently proposed for TcUGM \[[@R73]\]. Although, as mentioned above, the initial steps in the catalysis, NAD(P)H binding and subsequent FAD reduction, occur much less effective in prokaryotic UGMs, the steps leading to the conversion of Galf are conserved in these enzymes \[[@R71],[@R73]\]. 4.. METHODS FOR HIGH THROUGHPUT SCREENING FOR UGM INHIBITORS ============================================================ With the exponential advance of robotics, data collection, and analysis methods, high throughput screening (HTS) provides an effective and relatively fast preliminary analysis of chemical libraries composed of thousands of chemical compounds for the search of potential chemotherapeutics \[[@R76]-[@R80]\]. Whether the goal is to find an effective inhibitor for a well-explored enzyme or to match existing drugs to new macromolecular targets, HTS provides the rational starting point in the drug discovery process. Elimination of ineffective drug candidates early on using HTS is essential and saves time and resources during later stages of drug development, since libraries can contain thousands of compounds with a 0.1-0.2% probability of identifying positive hits \[[@R81]-[@R82]\]. Thus, the development of a successful assay for HTS is extremely important. Standard methods used to assay UGMs include: HPLC analysis, UV/Vis and stopped-flow spectroscopy, redox potentiometry, fluorescence polarization, and radiochemical detection \[[@R64],[@R69],[@R71],[@R83]-[@R84]\]. The HPLC method has been adopted by many groups, as it easily allows one to measure the activity of UGMs both qualitatively and quantitatively. In general, the assay monitors the reverse reaction, UDP-Galf to UDP-Galp conversion. Both the substrate and the product are easily detected at 262 nm, which corresponds to the absorbance maxima of UDP. Despite the broad utilization of the HPLC method by many research groups, this assay is not suitable for screening large chemical libraries because of the lengthy HPLC run times and because it is not suitable for running multiple measurements at once. A radioactive assay based on the generation and monitoring of tritiated formaldehyde, from the radioactive UDP-Galf degradation product, was used in the screening of 1,300 potential inhibitors against prokaryotic MtUGM. However, the poor sensitivity of the assay due to the equilibrium of the reaction not favoring the formation of UDP-Galf isomer was an essential drawback of this approach for high throughput screening applications \[[@R84]\]. Other reported assays used in HTS against UGMs are based on fluorescence polarization (FP) \[[@R83],[@R85]-[@R86]\]. FP relies on changes in the tumbling of a chromophore as it transitions from the enzyme-bound to the free-state due to competition by an inhibitor. This assay is simple, fast, and can be done on a small scale. Various fluorescent probes based on UDP were synthesized to develop a FP assay for both prokaryotic and eukaryotic UGMs (Fig. [9](#F9){ref-type="fig"}). Chromophore 1 was shown to effectively bind prokaryotic UGMs from M. tuberculosis and K. pneumoniae and used in HTS with the library strategically derivatized from a thiazolidinone core \[[@R85]\]. Contrary to prokaryotic UGMs, the fluorescein fluorophore was not as effective with AfUGM, and TAMRA analog 2 was developed instead. Chromophore 2 was shown to bind to AfUGM with relative high affinity, thus, yielding a potential tool for HTS in search of UGM inhibitors in eukaryotes \[[@R83]\]. The binding of UDP-chromophore to other eukaryotic UGMs is much less effective. For instance, the K~d~ value of chromophore 2 for TcUGM is \>20 μM, and similar low affinity is observed for LmUGM (Qi and Sobrado, unpublished results). However, since the active sites of eukaryotic UGMs are highly conserved, inhibitors of AfUGM might also be effective against the other eukaryotic UGM homologs. 5.. IN SILICO DRUG DESIGN AND ITS APPLICATIONS TO UGM ======================================================= Protein crystal structures serve as the blueprints for computer-guided molecular recognition, design, and virtual-screening of drug-like molecules and diagnostic probes \[[@R87]\]. There are three extensively used strategies for in silico drug design: de novo design, fragment-based drug discovery \[[@R88]-[@R89]\], and virtual screening \[[@R90]\]. The first two are very similar in their algorithms and concepts: both are based on design "from scratch", involving the screening of small pharmacophoric chemical blocks (or fragments) within the three-dimensional active site of the target enzyme. At later stages of the experiment, these fragments are further expanded ("grown") upon other moieties or directly joined together through a chemical bond or a linker. Virtual screening deals with the vast libraries of small chemical compounds utilizing high-throughput docking and pharmacophore-based searching algorithms and can be classified into two broad categories: ligand-based or structure-based docking and scoring \[[@R90]-[@R91]\]. Recently, a virtual screening using various computational tools toward the identification of inhibitors against EcUGM (also KpUGM and MtUGM) was applied to a small-molecule library comprised of 84,000 compounds (LeadQuest,Tripos, Inc.) \[[@R57]\]. A total of 13 compounds (0.015% of the library) were identified as positive hits and tested for inhibitory activity toward KpUGM and MtUGM. Only three compounds were shown to be effective inhibitors and had comparable affinity with the best previously published prokaryotic UGM inhibitors (IC~50~ 7.2 - 62 µM, \[[@R84]-[@R86],[@R92]-[@R93]\]). The effective application of in silico screening to bacterial UGMs suggests that a similar approach can be applied to eukaryotic enzymes. Furthermore, it is expected that an in silico screening approach with the structure of AfUGM will also identify potential inhibitors for TcUGM and LmUGM. 6.. CONCLUDING REMARKS ====================== Galf is important for cell wall biosynthesis and cell surface glycan structures in bacteria, fungi, and parasites. Galf is either essential for growth or important for pathogenesis, making enzymes in its biosynthetic pathway potential drug targets. In this pathway, UGM is an ideal target for drug discovery because this enzyme is absent in humans, and its structure and chemical mechanism are unique. During the past decade, the catalytic mechanism was fully elucidated and the structural differences between prokaryotic and eukaryotic UGM characterized. Recent discoveries in the field of eukaryotic UGMs set the stage for the identification of inhibitors that might lead to drugs for the treatment of neglected diseases like Chagas disease, leishmaniasis, and fungal infections caused by Aspergillus spp. In principle, every conformation along the catalytic cycle is a potential design target, including both the active, reduced enzyme and the inactive, oxidized one. Strategies toward developing effective drugs can include the design of a small-molecule competitive inhibitor with much higher binding affinity to UGM with respect to UDP-Galp/f, or even molecules that do not bind to the active site but interact with the mobile loops to prevent proper binding of the substrate. This work was supported by NIH grant R01 GM094469. CONFLICT OF INTEREST ==================== The authors confirm that this article content has no conflicts of interest. UDP : = Uridine diphosphate UGM : = UDP-galactopyranose mutase Galp : = Galactopyranose Galf : = Galactofuranose WHO : = World Health Organization NTD : = Neglected Tropical Diseases A. fumigatus : = Aspergillus fumigatus A. niger : = Aspergillus niger T. cruzi : = Trypanosoma cruzi L. major : = Leishmania major Leishmania spp. : = Leishmania species ABPA : = Allergic bronchopulmonary aspergillosis IPA : = Invasive pulmonary aspergillosis AIDS : = Acquired Immune Deficiency Syndrome ICU : = Intensive care unit GIPLs : = Glycoinositolphospholipids GPI : = Glycosylphosphatidylinositol FAD : = Flavin adenine dinucleotide NAD(P)H : = Nicotinamide adenine dinucleotide (phosphate) SAXS : = Small-angle X-ray scattering LFER : = Linear free energy relationship HTS : = High throughput screening HPLC : = High performance liquid chromatography UV/Vis : = Ultraviolet/visible FP : = Fluorescence polarization TAMRA : = Tetramethylrhodamine ![Structures of β-D-galactopyranose (A) and β-D-galactofuranose (B).](CPD-19-2561_F1){#F1} ![Reaction catalyzed by UDP-galactopyranose mutase.](CPD-19-2561_F2){#F2} ![Sequence alignment of eukaryotic (group 1) and prokaryotic (group 2) UGMs. The residues conserved in all of the sequences are shown in red shaded boxes. Those conserved only in one group are shown in red color. Active site residues are marked with a star, and those interacting with flavin are marked with triangles. The α-helix (spiral) and β-sheets (arrows; TT -- strict β-turns) of AfUGM are depicted on top. ClustalW was used to generate the alignment and ESPript 2.2 to create the figure. (The color version of the figure is available in the electronic copy of the article).](CPD-19-2561_F3){#F3} ![Protomer structure of reduced AfUGM complexed with UDP-Galp. FADH^-^ and UDP-Galp are colored gray and green, respectively. The flexible active site flaps are colored red. (The color version of the figure is available in the electronic copy of the article).](CPD-19-2561_F4){#F4} ![Close-up view of the active site of reduced AfUGM complexed with UDP-Galp highlighting flap closure. The flaps of the ligand-free reduced enzyme are colored magenta, while those of the UDP-Galp complex are colored cyan. The black arrows denote the direction of flap closure induced upon substrate binding. The red arrow denotes the direction of nucleophilic attack by the flavin N5 on the anomeric C atom of the substrate. (The color version of the figure is available in the electronic copy of the article).](CPD-19-2561_F5){#F5} ![Structure of the AfUGM tetramer. Each protomer has a different color. The yellow protomer has the same orientation as the protomer in Fig. [4](#F4){ref-type="fig"}. (The color version of the figure is available in the electronic copy of the article).](CPD-19-2561_F6){#F6} ![Conformations of the histidine loop in AfUGM structures. A, oxidized P6~5~22 form (PDB code 3UTE); B, oxidized P1 form (PDB code 3UKH); C, reduced enzyme (PDB code 3UTF).](CPD-19-2561_F7){#F7} ![Proposed mechanism for TcUGM. The oxidized enzyme binds and reacts with NADPH. UDP-Galp binds to the reduced enzyme and a flavin-sugar adduct is formed rapidly by the direct attack of the flavin. Formation of the flavin iminium ion leads to opening of the sugar ring. Attack of the UDP to form the UDP-Galf and its release occur rapidly. The reaction can occur for several more cycles (\~1000) before the enzyme is oxidized by molecular oxygen. The rate limiting step is proposed to be the closing of the sugar ring \[[@R73]\].](CPD-19-2561_F8){#F8} ![Fluorescent probes used in HTS assays for the identification of UGM inhibitors.](CPD-19-2561_F9){#F9} ###### Primary Structure Comparison of UGMs from Different Organisms. TcUGM LmUGM EcUGM KpUGM MtUGM DrUGM --------- ---------- ---------- --------- ---------- ---------- ---------- AfUGM 47.0 49.4 14.0 15.3 15.2 17.8 TcUGM 60.1 16.8 18.3 14.5 15.2 LmUGM 15.8 16.8 14.2 15.4 EcUGM 38.7 44.4 37.4 KpUGM 42.1 39.0 MtUGM 37.6 Percentage Identity Shown in Bold Corresponds to the UGMs from the Same Class (Prokaryotic or Eukaryotic). ClustalW Program was Used to Calculate Percentage Identity ###### Available UGM Crystal Structures in the Protein Data Bank (PDB) UGM Active site ligand PDB code \[[@R94]\] Ref. ----------------- ----------------------- --------------------- -------------------------------------- AfUGM~red~ 3UTF \[[@R59]\] AfUGM~red~ UDP 3UTG \[[@R59]\] AfUGM~red~ UDP-Galp 3UTH \[[@R59]\] AfUGMox sulfate 3UTE \[[@R59]\] AfUGM~red~ UDP-Galp 3UKF \[[@R60]\] AfUGM~red/ox~ UDP-Galp 3UKH \[[@R60]\] AfUGM~ox~ UDP 3UKL \[[@R60]\] EcUGM~ox~ 1I8T \[[@R63]\] KpUGM~ox~ FMN 3KYB n.a.[\](#T2F1){ref-type="table-fn"} KpUGM~ox~ UMP, UDP-Glcp* 3GF4 \[[@R61]\] KpUGM~ox~ UDP-Galp 3INR \[[@R62]\] KpUGM~red~ UDP, UDP-Galp 3INT \[[@R62]\] KpUGM~red~ 1WAM \[[@R58]\] KpUGM~ox~ 2BI7 \[[@R58]\] KpUGM~red~ 2BI8 \[[@R58]\] MtUGM~ox~ 1V0J \[[@R58]\] DrUGM~ox~ UDP 3HE3 \[[@R56]\] DrUGM~red/ox~ UDP-Galp 3HDY \[[@R56]\] DrUGM~ox~ UDP-Galp 3HDQ \[[@R56]\] DrUGM~ox~ UDP, UDP-CH~2~-Galp 3MJ4 \[[@R95]\] Gruber TD, Dimond MC, Kiessling LL, Forest KT, Structure of UDP-galactopyranose mutase bound to flavin mononucleotide. Unpublished results. ###### Ligand Interactions with UGMs ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Type of interaction AfUGM~red~ EcUGM~ox~ KpUGM~red~ DrUGM~red~ --------------------------------------------------- -------------------------- ---------------------------- -------------------------------------- ---------------------------- ---------------------------- UDP-Galp contacts π-π stacking with uracil Y104, F158 n.a.[\](#T3F1){ref-type="table-fn"} F152, Y155 F176, Y179 H-bonding with uracil F106, Q107 N270 F175, N296 Interactions with diphosphate R182, Y317\ R174, Y185, R280, Y314 R198, Y209\ R327, Y419\ R305, Y335, Y370 Y453 H-bonding with Galp* R182, N207\ N84, Y349 H109, R305 N457 Other important amino acids for substrate binding N163, W167 W160 T180, W184 FAD contacts H-bonding with ribose H63, G456, S461 N39, Y347 H60, L350, T355 H85\ Y371 Interaction with pyrophosphate T18, L46\ F12, N39\ F13, S14, N41, R343 F39, A40\ R447 R340 N67, R364 π-π stacking with isoalloxazine H63 H56 H60 H85 Interaction with adenine D38, S39, V242 E31, K32, D212,F213 F219 R60, D242\ Y243 H-bonding with isoalloxazine V64, Q458 I57, M349, Y346 I61, M352 I86, M373 H-bonding with N5 G62 A55 P59 P84 H-bonding with ribose D38 E31 D59 FAD-substrate contacts OH-4(Galp) and CO-4(FAD) n.a.[\](#T3F1){ref-type="table-fn"} OH-4(Galp) and CO-4(FAD) OH-4(Galp) and CO-4(FAD) ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Not available. ###### Amino Acid Composition and Molecular Weight of UGMs from Different Organisms Organism Oligomeric state in solution Number of amino acids MW of monomer, Da Ref. ------------------- ------------------------------------------------------------------------- ----------------------- ------------------- --------------------------- A. fumigatus* Tetramer[a](#T4F1){ref-type="table-fn"},[b](#T4F2){ref-type="table-fn"} 510 56,820 \[[@R59], [@R64]\] T. cruzi Monomer[a](#T4F1){ref-type="table-fn"} 480 54,690 \[[@R73]\] L. major Monomer[a](#T4F1){ref-type="table-fn"} 491 54,970 \[[@R96]\] E. coli Dimer[c](#T4F3){ref-type="table-fn"},[d](#T4F4){ref-type="table-fn"} 367 42,970 \[[@R63], [@R97]\] K. pneumoniae Dimer[d](#T4F4){ref-type="table-fn"} 384 44,460 \[[@R58], [@R61]-[@R62]\] M. tuberculosis Dimer[d](#T4F4){ref-type="table-fn"} 399 45,820 \[[@R58]\] D. radiodurans Not determined 397 45,700 \[[@R56], [@R98]\] Determined by size exclusion chromatography. Determined by SAXS. Determined by light scattering. Inferred from analysis of protein-protein interfaces in the crystal lattice. ###### Steady State Kinetic Parameters of UGMs from Different Organisms ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Organism k~cat~, s^-1^ K~M~, µM k~cat~/K~M~, µM^-1^ s^-1^ Ref. ------------------- ----------------------------------------------- --------------------------------------------- -------------------------------------------------- -------------------- A. fumigatus 72 ± 4[a](#T5F1){ref-type="table-fn"} 110 ± 15[a](#T5F1){ref-type="table-fn"} 0.65 ± 0.09[a](#T5F1){ref-type="table-fn"} \[[@R59], [@R64]\] T. cruzi 13.4 ± 0.3[a](#T5F1){ref-type="table-fn"};\ 140 ± 10[a](#T5F1){ref-type="table-fn"};\ 0.093 ± 0.006[a](#T5F1){ref-type="table-fn"};\ \[[@R73]\] 11.5 ± 0.4[b](#T5F2){ref-type="table-fn"};\ 200 ± 20[b](#T5F2){ref-type="table-fn"};\ 0.056 ± 0.005[b](#T5F2){ref-type="table-fn"};\ 8.4 ± 0.9[c](#T5F3){ref-type="table-fn"} 690 ± 150[c](#T5F3){ref-type="table-fn"} 0.012 ± 0.001[c](#T5F3){ref-type="table-fn"} L. major 5 ± 0.2[a](#T5F1){ref-type="table-fn"} 87 ± 11[a](#T5F1){ref-type="table-fn"} 0.057 ± 0.006[a](#T5F1){ref-type="table-fn"} \[[@R96]\] E. coli 27[a](#T5F1){ref-type="table-fn"} 22[a](#T5F1){ref-type="table-fn"} 1.22[a](#T5F1){ref-type="table-fn"} \[[@R99]\] K. pneumoniae 5.5 ± 0.7[a](#T5F1){ref-type="table-fn"} 43 ± 6[a](#T5F1){ref-type="table-fn"} 0.12 ± 0.02[a](#T5F1){ref-type="table-fn"} \[[@R100]\] M. tuberculosis 8 13 0.62 \[[@R101]\] D. radiodurans 66 ± 2.4[a](#T5F1){ref-type="table-fn"} 55 ± 7[a](#T5F1){ref-type="table-fn"} 1.18[a](#T5F1){ref-type="table-fn"} \[[@R56]\] ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Reduced with 5-20 mM dithionite. Reduced with 0.5 mM NADPH. Reduced with 2.5 mM NADH. ###### Kinetic Parameters of TcUGM Reduction with NAD(P)H \[[@R73]\]. Substrate k ~red~, s^-1^ K ~d~, µM k ~red~/K~d~, µM^-1^ s^-1^ ----------- ------------------ ------------- -------------------------------- NADH 0.085 ± 0.0006 550 ± 10 0.00015 ± 0.000002 NADPH 0.600 ± 0.006 98 ± 3 0.0061 ± 0.0001	{ "pile_set_name": "PubMed Central" }
I. INTRODUCTION {#acm20151-sec-0001} =============== Multi‐channel CT has been widely accepted for diagnostic applications as well as for radiotherapy virtual simulation purposes. The effect of CT acquisition parameters on image quality is important to all clinical applications, though the goals of the applications are quite different. A few studies^(^ [^1^](#acm20151-bib-0001){ref-type="ref"} ^--^ [^6^](#acm20151-bib-0006){ref-type="ref"} ^)^ regarding optimizing acquisition parameters for certain applications in diagnostic imaging (DI) using multi‐channel CT scanners partially addressed the effects of machine settings on image quality. However, a more thorough study is needed regarding image quality for newer scanners for diagnostic and therapeutic applications. CT technology continues to develop and there is a vast difference in acquisition parameter options between manufacturers. Though there is some information available regarding technical details, such as image thickness for CT radiotherapy simulation,[^(7)^](#acm20151-bib-0007){ref-type="ref"} the optimal CT image acquisition parameter settings for radiotherapy simulation have not yet been thoroughly investigated. The purpose of this study was to address this gap, and determine the optimal image acquisition parameter settings for conventional CT simulation protocols, including pitch (defined as couch movement per gantry rotation divided by beam width), detector configuration, and rotation time using several different CT scanners in our radiation oncology (RO) department. Routine radiation dose values were also collected in order to help understand image quality characteristics between scanners. II. MATERIALS AND METHODS {#acm20151-sec-0002} ========================= A. Rationale for CT Simulation in Radiation Oncology {#acm20151-sec-0003} ---------------------------------------------------- The implementation of CT simulators in radiation oncology (RO) is somewhat different from CT scanners in DI, though image quality is a main concern for both. Unlike the regular practice in DI, breath holding, contrast injection, variable image thickness, and gantry tilt are generally discouraged in RO. Because most patients undergoing CT simulation in RO have diagnosed cancers, and generally only one scan is used for CT simulation, radiation dose is of less concern than it is in DI. Instead, image quality has singular importance, as accuracy in RO treatment planning is particularly vulnerable to artifacts. In order to simulate radiation treatment conditions, free or gated breathing is required during the CT scan. Fast scanning is generally not desirable because a snapshot of one breathing phase does not represent the average anatomy position during treatment delivery. A scan time of 1 to 2 minutes is considered acceptable. For CT simulation, patient skin surface must be included in the image for treatment planning. Image truncation is intolerable; therefore a larger display field of view (DFOV) is used, usually 50 cm or 65 cm. For a head scan, a body size DFOV is generally used, because part of the shoulders must be scanned as part of patient positioning for most procedures. For CT simulation, the choice of image thickness involves balancing the conflicting considerations of the increased resolution of digitally reconstructed radiographs (DRR) obtained from thin images and increased physician effort required to contour multiple images. The standard choice of the physicians and physicists in our institution is 2.5 mm or 3 mm image thickness and 2.5 mm interval, which agrees with general recommendations.[^(7)^](#acm20151-bib-0007){ref-type="ref"} B. Phantoms and Scanners {#acm20151-sec-0004} ------------------------ A Catphan phantom[^(8)^](#acm20151-bib-0008){ref-type="ref"} (The Phantom Laboratory, Salem, New York) was primarily used to evaluate image quality. Three modules were used in this study: 1) CTP528, high‐contrast resolution bar patterns; 2) CTP515, 1.0% low‐contrast objects in the Supra Slice portion; 3) CTP401, artifact evaluation. The high‐contrast resolution section of the GE CT daily QA phantom[^(9)^](#acm20151-bib-0009){ref-type="ref"} (GE Medical Systems, Milwaukee, Wisconsin) was used for additional artifact evaluation. The high contrast region was chosen to evaluate artifacts because objects with high‐edge gradients are more likely to produce helical artifacts. Both phantoms were positioned by suspending each from an appropriate bracket, following the manufacturers\' standard setup recommendations. Five different CT scanners were evaluated: General Electric (GE) LightSpeed (LS) RT (4‐channel, 80 cm bore)LS RT16 (16‐channel, 80 cm bore)LS16 (16‐channel, 70 cm bore) Philips Medical Systems AcQsim CT (1‐channel, 80 cm bore)MX8000 (16‐channel, 70 cm bore). C. Evaluation Methods {#acm20151-sec-0005} --------------------- Various pitch, detector configuration, and gantry rotation time (1 sec and less) settings were used to acquire images, while maintaining the following constant settings: 120 kVp, 240 effective mAs (defined as tube current $\left( \text{mA} \right) \times \text{rotation}$ time (sec) / pitch; where $\text{pitch} = \text{table}$ movement per gantry rotation / total beam collimation), 1s rotation time unless specified otherwise, 25 cm display field of view (DFOV), standard reconstruction filter, 2.5 mm image interval, and 2.5 mm image thickness for GE or 3 mm image thickness for Philips. On multi‐channel CT scanners, there are multiple data channels that route the signals collected on the detector surface to the reconstruction computer. These data channels can be utilized in a flexible manner to produce more than one thin image during one rotation of the gantry, or fewer thicker images during one rotation of the gantry. The combination of the number of active channels and the detector width (z‐direction) assigned to each channel is called the detector configuration. The product of these two values represents the nominal x‐ray beam width (z‐direction). In general, a combination of fine detector spacing (also called small channel width) and a wide beam width are the most desirable in CT imaging. Fine detector spacing produces a denser helical dataset from which to interpolate, and a wide beam allows for higher table speeds and faster exams. The phantom images were evaluated from three aspects: 1) high‐contrast resolution; 2) low‐contrast resolution; and 3) the presence of helical artifacts. A single observer primarily evaluated all the images. Zoom, display window and level were adjusted to best display the image region of interest for high‐contrast and low‐contrast resolution evaluation. For low‐contrast resolution, since CTP 515 module spans 70 mm, the images with best resolution were selected for evaluation. Three levels were used to score the presence of artifact: "Yes", "Slight", and "No". "Yes" meant the artifact was clearly visible, "No" meant no artifact was present, and "Slight" meant the artifact was present in the phantom images, but it was very subtle and most likely invisible in clinical images. To examine the effect of gantry rotation time, 1 sec and sub‐second rotations were evaluated for scanners that had these options. The fastest gantry rotation for GE RT and Philips AcQsim was 1 sec. Therefore they were not included in the evaluation of gantry rotation time effect. The single‐channel Philips AcQsim CT is limited by prolonged scan time and the effects of excessive tube heating, so only pitches greater than 1 were evaluated for this single‐channel CT. In order to interpret results such as low‐contrast resolution, the standard annual radiation dose assessment made for each scanner was also reported on a normalized (per 100 mAs) basis. These dose evaluations were obtained by using the standard CT Dose Index (CTDI) approach.[^(10)^](#acm20151-bib-0010){ref-type="ref"} III. RESULTS {#acm20151-sec-0006} ============ A. Artifact evaluation {#acm20151-sec-0007} ---------------------- For all CT scanners, helical artifacts were more severe when the channel width was equal to the image thickness and a relatively high pitch was used. For example, for the GE LS16 (Fig.[1](#acm20151-fig-0001){ref-type="fig"}), a detector configuration of $8 \times 2.5\,\text{mm}$ and a pitch of 1.675 resulted in 2.5 mm thick images with severe artifacts Fig. [1(a)](#acm20151-fig-0001){ref-type="fig"}, while a $16 \times 1.25\,\text{mm}$, 1.75 pitch setting showed slight artifacts Fig. [1(b)](#acm20151-fig-0001){ref-type="fig"}, and a $16 \times 1.25\,\text{mm}$, 0.562 pitch had no artifacts in the images Fig. [1(c)](#acm20151-fig-0001){ref-type="fig"}. ![Artifacts in 2.5 mm thick images of the Daily QA phantom for GE LS16: (a) $8 \times 2.5\,\text{mm}$, pitch 1.675; (b) $16 \times 1.25\,\text{mm}$, pitch 1.75; (c) $16 \times 1.25\,\text{mm}$, pitch 0.562.](ACM2-9-151-g001){#acm20151-fig-0001} For GE scanners, a channel width setting equal to half of the image thickness was sufficient to produce artifact‐free images for pitch values close to 1. For GE LS RT, $4 \times 1.25\,\text{mm}$ and a pitch of 0.75 Fig. [2(a)](#acm20151-fig-0002){ref-type="fig"} offered the best image quality for 2.5 mm thick images, while $4 \times 2.5\,\text{mm}$ and 1.5 pitch Fig. [2(d)](#acm20151-fig-0002){ref-type="fig"} settings caused severe helical artifacts. There was no substantial difference in the severity of artifacts for the settings $4 \times 1.25\,\text{mm}$ and pitch 1.5 Fig. [2(b)](#acm20151-fig-0002){ref-type="fig"}, and $4 \times 2.5\,\text{mm}$ and pitch 0.75 Fig. [2(c)](#acm20151-fig-0002){ref-type="fig"}, though the former produced better low‐contrast resolution. ![Artifacts in 2.5 mm thick images of the Catphan phantom for GE LS RT: (a) $4 \times 1.25\,\text{mm}$, pitch 0.75; (b) $4 \times 1.25\,\text{mm}$, pitch 1.5; (c) $4 \times 2.5\,\text{mm}$, pitch 0.75; and (d) $4 \times 2.5\,\text{mm}$, pitch 1.5.](ACM2-9-151-g002){#acm20151-fig-0002} The Philips CT scanners behaved quite differently from the GE CT scanners (Fig. [3](#acm20151-fig-0003){ref-type="fig"}). For the Philips MX8000, $16 \times 1.5\,\text{mm}$ detector configuration for 3 mm image thickness setting (i.e., a channel width equal to half image thickness) caused severe artifacts for all available pitches Fig. [3(D)](#acm20151-fig-0003){ref-type="fig"}. A channel width of 1/4 of the image thickness with pitch less than 1 was needed to eliminate helical artifacts Fig. [3(a)](#acm20151-fig-0003){ref-type="fig"} on this scanner. The other two settings, $16 \times 0.75\,\text{mm}$, pitch 1.438 Fig. [3(b)](#acm20151-fig-0003){ref-type="fig"} and $16 \times 1.5\,\text{mm}$, pitch 0.688 Fig. [3(c)](#acm20151-fig-0003){ref-type="fig"}, resulted in less severe and different artifacts; the setting for Fig. [3c](#acm20151-fig-0003){ref-type="fig"} were better than those for Fig. [3b](#acm20151-fig-0003){ref-type="fig"}. For Philips AcQsim CT, helical artifacts increased with pitch (Fig. [4](#acm20151-fig-0004){ref-type="fig"}). ![Artifact evaluation of 3 mm thick images of the Catphan phantom for Philips MX8000: (a) $16 \times 0.75\,\text{mm}$, pitch 0.688; (b) $16 \times 0.75\,\text{mm}$, pitch 1.438; (c) $16 \times 1.5\,\text{mm}$, pitch 0.688; (d) $16 \times 1.5\,\text{mm}$, pitch 1.313. Note: Windows and levels were adjusted separately for optimal display.](ACM2-9-151-g003){#acm20151-fig-0003} ![Artifact evaluation of 3 mm thick images of the Catphan phantom for Philips AcQsim single‐channel CT. (a) pitch 1.0; (b) pitch 1.5.](ACM2-9-151-g004){#acm20151-fig-0004} Helical artifacts appeared to be scanner/model specific, not only different across manufacturers, but also different across models made by a single manufacturer. In Fig.[5](#acm20151-fig-0005){ref-type="fig"}, we compared the images from GE LS RT16 Fig. [5(a)](#acm20151-fig-0005){ref-type="fig"}, [5(b)](#acm20151-fig-0005){ref-type="fig"} with those from GE LS16 Fig. [5(c)](#acm20151-fig-0005){ref-type="fig"}. For the same pitch (1.375) and detector configuration ($16 \times 1.25\,\text{mm}$), the GE LS RT16 produced more severe artifacts than the GE LS16 Fig. [5(b)](#acm20151-fig-0005){ref-type="fig"}, [5(c)](#acm20151-fig-0005){ref-type="fig"}. The combinations of the largest channel width and largest pitch that produced artifact‐free images for all scanners are listed in Table [1](#acm20151-tbl-0001){ref-type="table-wrap"}. ![GE 16‐channel CT comparison, 2.5 mm thick images, for detector configuration: $16 \times 1.25\,\text{mm}$. GE LS RT16, pitch 0.562 (a), pitch 1.375 (b), and GE LS16, pitch 1.375 (c).](ACM2-9-151-g005){#acm20151-fig-0005} ###### Combination of the largest channel width and largest pitch that produced artifact‐free images. Model Image thickness (mm) Pitch Detector configuration (mm) Head $\text{CTDI}_{W}/100\text{mAs}$ (mGy) Body $\text{CTDI}_{W}/100\text{mAs}$ (mGy) ---------------- ------------------------ --------- -------------------------------------------------------------------- ------------------------------------------------ ------------------------------------------------ Philips MX8000 3 0.688 $16 \times 0.75 = 12$ 13.2 4.2 Philips AcQsim 3 1 $1 \times 3.00 = 3$ 13.0 3.5 GE LS16 2.5 1.375 $16 \times 1.25 = 20$ 14.4 7.3 GE LS RT16 2.5 0.938 $16 \times 1.25 = 20$ 12.4 6.8 GE LS RT 2.5 0.75 $4 \times 1.25 = 5$ 23.8 10.4 2.5 0.75 $4 \times 2.5 = 10$ [^\^](#acm20151-tbl-note-0001){ref-type="fn"} 17.7* 6.4 Not artifact‐free for helical scan, but for dose comparison. The helical streak artifact appears to depend on the combination of detector configuration and pitch, and that the relationships between these characteristics are not consistent between or even within vendors. The severity of artifacts increased with increasing pitch. The artifact evaluation in Tables [2](#acm20151-tbl-0002){ref-type="table-wrap"}, [3](#acm20151-tbl-0003){ref-type="table-wrap"} and [4](#acm20151-tbl-0004){ref-type="table-wrap"} based on the high‐contrast section of the Catphan phantom gives additional information for specific scanner settings. ###### Image quality results for GE LS RT16. Rotation Time (s) Mode / Pitch High Contrast Resolution (lp/cm) 1% Low Contrast Diameter (mm) Artifact Present? --------------------- ---------------- ------------------------------------ --------------------------------- --------------------- 0.5 Axial 7 4 No 1.0 Axial 7 4.5 No 0.5 0.562 7 4 No 0.5 0.938 7 4 No 0.8 0.938 8 4 No 0.5 1.375 7 4.5 Yes 0.8 1.375 7 4 Yes Technique: effective $\text{mAs} = 240$, image thickness 2.5 mm, detector configuration $16 \times 1.25\,\text{mm}$. Low contrast was defined as the diameter of the smallest circular target detectable. ###### Image quality results for GE LS16. Rotation Time (s) Pitch High Contrast Resolution (lp/cm) 1% Low Contrast Diameter (mm) Artifact Present? --------------------- --------- ------------------------------------ --------------------------------- --------------------- 0.5 0.562 8 3.5 No 0.8 0.562 8 4 No 0.8 0.938 7 4 No 0.8 1.375 7.5 3.5 Slight 1 1.75 7 3 Yes Technique: effective $\text{mAs} = 240$, image thickness 2.5 mm, detector configuration $16 \times 1.25\,\text{mm}$. Low contrast was defined as the diameter of the smallest circular target detectable. ###### Image quality results for Philips MX8000. Rotation Time (s) Pitch High Contrast 1% Resolution (lp/cm) Low Contrast Diameter (mm) Artifact Present? --------------------- --------- --------------------------------------- ------------------------------ --------------------- 0.4 0.564 8 4 No 0.5 0.563 8 3.5 No 0.75 0.563 8 3 No 1.5 0.563 8 3 No Technique: effective $\text{mAs} = 240$, image thickness 3 mm, detector configuration $16 \times 0.75\,\text{mm}$. Low contrast was defined as the diameter of the smallest circular target detectable. B. Resolution evaluation {#acm20151-sec-0008} ------------------------ Image resolution (low and high contrast) was evaluated using the phantom sections shown in Fig. [6](#acm20151-fig-0006){ref-type="fig"}. For these GE scanners (Tables [2](#acm20151-tbl-0002){ref-type="table-wrap"} and [3](#acm20151-tbl-0003){ref-type="table-wrap"}), the image resolution did not appear to be affected by pitch and rotation time, provided that mA was adjusted appropriately. For high‐contrast resolution, the variation was within 1 line pair (lp) /cm around the average value of $7.3\text{lp}/\text{cm}$. For low contrast resolution, the variation was within 0.6 mm over the average diameter of 3.9 mm for 1% detectable low‐contrast objects. These variations may be due to quantum and statistical fluctuations. However, GE LS16 had systematically better high‐ and low‐contrast resolution than GE LS RT16. GE LS16 has a maximum tube current limit of 440 mA, thus a pitch of less than 1 had to be used to achieve 240 effective mAs for 0.5s gantry rotation. Of these two GE models, the larger bore CT scanner had lower image quality and was more prone to helical artifacts. ![Sample image sections used for resolution evaluation: (a) Catphan CTP515 mode, the marked area is 1% contrast which was used for low‐contrast resolution evaluation; (b) Catphan CTP528 mode, use for high‐contrast resolution evaluation. Techniques: GE LightSpeed RT16, pitch 0.938, rotation time 0.8 sec, $16 \times 1.25\,\text{mm}$ detector configuration, 2.5 mm image thickness (display not optimized).](ACM2-9-151-g006){#acm20151-fig-0006} A similar approach was used for images obtained using Philips MX8000 (Table [4](#acm20151-tbl-0004){ref-type="table-wrap"}). A clear trend was observed that short rotation time (0.5 sec or less) caused slight low‐contrast resolution degradation, though high‐contrast resolution remained unaffected. The image resolution was best for Philips MX8000, followed by GE LS16, and GE RT16. C. Radiation Dosimetry {#acm20151-sec-0009} ---------------------- The CTDIW was normalized for 100 mAs using the standard head and body phantoms [^(11)^](#acm20151-bib-0011){ref-type="ref"} and represents the most recent annual dosimetry evaluations. CTDIW / 100 mAs varied by nearly a factor of 2 from 12.4 mGy to 23.8 mGy among the five scanners for the head phantom. For the head CTDI phantom, normalized dose (CTDIW / 100 mAs) was highest for GE RT, followed by GE LS16, Philips MX8000, and Philips AcQsim, and was lowest for GE LS RT16. For the body CTDI phantom, CTDIW / 100 mAs ranged from 3.5$- 10.4$mGy; this value was highest for GE LS RT, followed by GE LS16, GE LS RT16, and Philips MX8000, and was lowest for Philips AcQsim. For this study, the head CTDI dose results would be more relevant than the body CTDI values. The normalized body CTDI results were included for completeness. IV. DISCUSSION {#acm20151-sec-0010} ============== We found that for GE scanners, a channel width setting equal to half of the image thickness was sufficient to produce artifact‐free and good quality images for pitch values close to 1. Our result with GE RT 4‐channel CT agrees with Takahashi, et al.[^(3)^](#acm20151-bib-0003){ref-type="ref"} who determined that for this scanner model, the settings that achieved the best IQ were found to be $4 \times 1.25\,\text{mm}$ detector configuration and a pitch of 0.75 (though that study was only focused on depicting arterial stenosis, and imaging artifact was not specifically evaluated). For GE RT, the detector setting $4 \times 1.25\,\text{mm}$ resulted in a high CTDIW of 2.38 cGy (Table [1](#acm20151-tbl-0001){ref-type="table-wrap"}), while $4 \times 2.5\,\text{mm}$ setting lowered CTDIW to 1.77cGy due to improved dose efficiency. If a faster scan speed is needed, we would suggest using $4 \times y2.5\,\text{mm}$ and pitch 0.75 rather than $4 \times 1.25\,\text{mm}$ and pitch 1.5, though they have the same table speed and similar image quality. For Philips MX8000, a channel width of 1/4 of the image thickness with a pitch of less than 1 was needed to eliminate helical artifacts. However, the MX8000 had overall somewhat better high‐ and low‐contrast resolution than GE scanners (Tables [2](#acm20151-tbl-0002){ref-type="table-wrap"}, [3](#acm20151-tbl-0003){ref-type="table-wrap"}, [4](#acm20151-tbl-0004){ref-type="table-wrap"}). Differences between these two manufacturers may be due to the different reconstruction algorithms or different dose values (matching technique parameters does not necessarily result in equivalent radiation output or quality between scanners). We observed a larger than expected difference in normalized dose (CTDIW / 100 mAs) among these scanners; when using the head CTDI phantom, there was a range of a factor of 2 in CTDIW / 100 mAs. When using the body CTDI phantom, we observed a difference between vendors by a factor of 3. In general, when spatial resolution was enhanced, artifacts present in the image were also enhanced. In helical CT, data re‐binning and interpolation are used to generate planar transverse images from helical attenuation data. This process yields better quality images if helical data is more finely spaced. The choice of detector configuration is generally a much more powerful method of reducing helical artifact than decreasing scan pitch. Sometimes, the selection of the maximum number of active channels and the finest channel width must be balanced against the need for a faster couch speed. The small differences in spatial resolution shown in Tables [2](#acm20151-tbl-0002){ref-type="table-wrap"} to [4](#acm20151-tbl-0004){ref-type="table-wrap"} are not likely to be detectable in clinical images, because the DFOV of 50 cm or greater over the $512 \times 512$ matrix used in CT simulation limits the achievable spatial resolution. Some difference in contrast resolution is expected due to the different bore sizes, detector designs and reconstruction algorithms. Increasing exposure, or effective mAs, will increase low‐contrast resolution across all scanners. Though phantom studies are appropriate for comparative evaluations, they are not accurate simulations of patient scans. A major difference between a phantom and a patient is that a patient often moves (voluntarily or involuntarily) during a CT acquisition, especially for those of the chest and abdomen. The resulting blur is not present on the stationary phantom images. Faster rotation time reduces motion‐blurring artifact[^(12)^](#acm20151-bib-0012){ref-type="ref"} and improves patient image quality, though it should not affect the phantom image quality. Because fast scan time (breath‐hold) is not desired in CT simulation, a reasonably low pitch, fast rotation and fine detector configuration should be used to improve image quality and also to acquire patient images in a more averaged position rather than a snapshot of one moment. In general, we recommend using fast gantry rotation for chest and abdomen scans, 0.5$- 0.8$ sec for example, if the options are available. Image resolution evaluation can be used as a reference when setting up scan protocols for radiotherapy simulation. Tube current modulation should not affect the appearance of helical artifacts. However, tube current modulation may affect low‐contrast resolution, because effective mAs will vary in the x/y plane, along the axis, or both. V. CONCLUSIONS {#acm20151-sec-0011} ============== In conclusion, we have quantified the effects of changes in various parameters used for RO CT simulation on specific image metrics. Given the differences we found between CT machines in our own study, choosing optimal image acquisition parameter settings for CT simulation purposes may require a phantom evaluation for each scanner model to be used.	{ "pile_set_name": "PubMed Central" }
In the past 5 years, as methicillin-resistant Staphylococcus aureus (MRSA) has emerged as a community pathogen, awareness of the role of animal exposure from pets or farming as sources of MRSA has increased ([@R1]--[@R3]). We identified a clone of S. aureus previously associated with outbreaks of infections in animals and in humans who work with animals in 2 unique collections of S. aureus isolates. The first was from a population-based study of S. aureus colonization among residents of northern Manhattan in New York, NY, USA; the second was from isolates obtained from the Dominican Republic. This clone does not digest with the restriction enzyme SmaI, which is generally used for pulsed-field gel electrophoresis (PFGE). Consequently, the clone is identified by multilocus sequence typing as sequence type 398 (ST398). Both methicillin-resistant and methicillin-susceptible isolates of S. aureus have been reported ([@R4]). ST398 has been found primarily in Europe, where it has been isolated from pigs and pig farmers in the Netherlands and France and from dogs, pigs, horses, and humans in Germany and Austria ([@R5]--[@R8]). Colonization with MRSA ST398 has recently been reported in pigs and pig farmers in Canada ([@R9]). The Study ========= The community-based study was conducted from 2004 through 2007 in the northern section of Manhattan, a borough of New York City. Northern Manhattan contains a large, medically underserved population that has close ties to the Dominican Republic. Participants were recruited by using random-digit dialing. Consenting persons and household members were subsequently interviewed and screened for S. aureus colonization. A total of 321 eligible households containing 914 household members participated. In 9 households, 13 participants were found to be colonized with S. aureus isolates that were SmaI resistant. Digestion with the Cfr9I, an isoschizomer of SmaI, yielded identical PFGE profiles ([Figure](#F1){ref-type="fig"}). Subsequent multilocus sequence typing confirmed the ST398 identification (allelic profile 3--35--19--2-20--26--39). All strains were methicillin susceptible. A representative strain was spa-typed as type t571 (allelic profile 8--16--2-25--2-25--34--25, eGenomics type 109); it was Panton-Valentine leukocidin negative. ![Pulsed-field gel electrophoresis profiles of sequence type 398 isolates from the Dominican Republic (D); northern Manhattan (M), New York, NY, USA; and Canada (C) (provided by Scott Weese). Strains within households in which \>1 person was colonized are identified numerically. The dendrogram shows the percent similarity of the isolates. A similarity [\>]{.ul}70% indicates closely related or identical strains.](08-0609-F){#F1} Characteristics of persons colonized with ST398 were similar to those of persons in the community-based study and with northern Manhattan census characteristics ([Table](#T1){ref-type="table"}). The 13 isolates were from 9 different families; 1 family had 4 members colonized with ST398 at either nasal or axillary sites. The mean age of those colonized was 33.4 years; only 1 child (7 years of age) was colonized. Two persons from different families were colonized with ST398 at multiple sites, none of which were confirmed as infections. ###### Characteristics of persons colonized with Staphylococcus aureus ST398, northern Manhattan, New York City, NY, USA, 2004--2007, compared with study population and 2000 census population for area\* Characteristics ST398 subset, no. (%) Overall study population, no. (%) 2000 census population, no. (%) ----------------------- ----------------------- ----------------------------------- --------------------------------- Race/ethnicity Latino 11 (84.6) 813 (89) 173,755 (68) Non-Hispanic white 2 (15.4) 90 (9.8) 65,449 (25.6) African American 0 11 (1.2) 53,514 (20.9) Asian 0 0 5,370 (2.1) Sex Male 4 (30.8) 362 (39.6) 120,866 (47.3) Female 9 (69.2) 552 (60.4) 134,723 (52.7) Age group, y \<5 0 84 (9.2) 17,878 (7.0) 5--17 1 (7.6) 238 (26.2) 49,196 (19.3) 18--44 10 (76.9) 297 (32.5) 112,195 (44.0) [\>]{.ul}45 2 (15.4%) 290 (31.9) 76,320 (29.9) Occupational exposure 5 (38.5) 58 (6.4) NA Travel outside USA 7 (53.8) 171 (18.7) NA Daycare exposure 2 (15.4) 87 (9.5) NA Total population 13 (100) 914 (100) 255,589 (100) \ST, sequence type; NA, not available. No household reported owning pets, although 2 reported animal contact. Of the 12 adults, 5 (41.7%) reported possible job exposure to S. aureus, including 1 who worked in a healthcare-associated field. No household reported patronizing viveros, or live poultry markets, which are common in the Latino communities of northern Manhattan and the Bronx. Two households reported having children who attended day care, although none of these children were colonized with S. aureus. Although 15% of the Dominican population in the study reported travel to the Dominican Republic within 6 months of their interview, none of the colonized participants reported recent travel to the Dominican Republic. No contact among the different households was reported. A second collection of S. aureus* isolates was gathered during 2007 and 2008 from a convenience sample of 89 anonymous infection and colonization isolates received from the Dominican Republic. Six isolates were identified as methicillin-susceptible S. aureus clone ST398. Strains were provided by 1 hospital in Santo Domingo, Dominican Republic (n = 53), and 1 microbiology laboratory, the Laboratorio Referencia (n = 28), which serves as a reference laboratory for the country. Four isolates from the hospital and 2 from Laboratoria Referencia were identified as methicillin-susceptible ST398. Sociodemographic data on these persons were limited. Two of the ST398 isolates were from women with infections living in Santo Domingo, and the remaining 4 were colonization samples. Of the 6 isolates, 5 were found to be spa-type t571 (eGenomics type 109), and 1 was found to be type t3625 (eGenomics untyped). Pairwise similarity scores for the isolates were calculated by the Dice coefficient, and an overall similarity score was calculated by using the unweighted pair group method with arithmetic mean. Comparing the isolates by using a dendrogram-based similarity score [\>]{.ul}70%, we found that the strains from northern Manhattan and from the Dominican Republic were closely related, although they contrasted with ST398 isolates from Canada (provided by Scott Weese, Ontario Veterinary College, University of Guelph) ([Figure](#F1){ref-type="fig"}) ([@R10]). Conclusions =========== Identification of the S. aureus clone ST398 in northern Manhattan and in the Dominican Republic suggests its introduction into the United States by travelers between the 2 countries. The largely Latino population of northern Manhattan is composed mainly of immigrants or first-generation families from the Dominican Republic; travel between the 2 regions is common. Alternatively, northern Manhattan may contain reservoirs, such as live poultry markets, which may serve as a means of strain transmission. Colonization or infection with the S. aureus clone ST398 has been associated with exposure to pigs, pets, and other animals ([@R5],[@R7]--[@R9]), and the S. aureus clone ST398 has been isolated from meat products ([@R2]). However, transmission is not limited to animal exposures. Person-to-person spread has occurred among household members and in the hospital setting ([@R6],[@R8],[@R11]). For example, a dramatic increase in persons colonized as well as infected with MRSA clone ST398 was recently reported in a Dutch hospital ([@R12]). The presence of this strain among several household members in our study reinforces earlier observations of the potential for horizontal transmission of this clone after it is introduced into an appropriate setting. Although information about the persons from the Dominican Republic was limited, the 2 groups provided identical strain profiles, suggesting a possible link between the 2 countries. Given ST398's history of rapid dissemination in the Netherlands, its potential for the acquisition of methicillin resistance, and its ability to cause infections in both community and hospital settings, monitoring the prevalence of this strain in northern Manhattan and the Dominican Republic will be important to understand more about its virulence and its ability to spread in these communities. Suggested citation for this article: Bhat M, Dumortier C, Taylor B, Miller M, Vasquez G, Yunen J, et al. Staphylococcus aureus ST398, New York City and Dominican Republic. Emerg Infect Dis \[serial on the Internet\]. 2009 Feb \[date cited\]. Available from <http://www.cdc.gov/EID/content/15/2/285.htm> We gratefully acknowledge the advice and suggestions of Scott Weese. This study was supported by a grant from the US Centers for Disease Control and Prevention. C.D. was supported by a Fulbright Scholar grant made possible by the US Department of State, by the Franco-American Commission for Educational Exchange, and by a grant from the Bourse Collery de l\'Académie Nationale de Médecine. Ms Bhat is a researcher in the Infectious Diseases Division, Department of Medicine, Columbia University. She is currently investigating the molecular epidemiology of S. aureus transmission in communities and hospitals.	{ "pile_set_name": "PubMed Central" }
Introduction ============ The father of modern orthodontics, Edward Hartley Angle, in 1899 classified malocclusions in Class I, Class II, and Class III based on permanent first maxillary and mandibular molars relationship and alignment (or lack of it) of teeth with reference to the line of occlusion.[@b1-ccide-10-099] Gradually, Angle's classification was modified and additional information such as jaw relationship and the pattern of growth were also included. Thus, a Class III jaw relationship suggests that the mandible has acquired a more mesial position in relation to the maxilla and/or cranial base.[@b2-ccide-10-099],[@b3-ccide-10-099] Occasionally, due to dental compensation, sometimes there is Class I dental relationship on the Class III skeletal base. Further, Charles Henry Tweed classified Class III malocclusions as a pseudo Class III malocclusion with normal mandible and underdevelopment of maxilla (category A) and skeletal Class III malocclusion with prognathic mandible or an underdevelopment of maxilla (category B).[@b4-ccide-10-099] Class III problems may arise due to deficient growth of maxilla in the downward and forward direction and more forward growth or reduced downward growth of mandible. Hence, a hypodivergent growth pattern accentuates the Class III problem due to more growth rotation of the mandible in the upward and forward direction, while a vertical growth pattern alleviates it due to downward and backward rotation, provided that excessive facial height does not become the problem instead.[@b5-ccide-10-099] Nonsurgical treatment of Class III problems remains a challenge in our profession. However, prompt diagnosis and early intervention of Class III malocclusion may be helpful to reduce the extent of burden for severe Class III malocclusion in late adolescence.[@b6-ccide-10-099] This paper presents an overview of developing Class III malocclusion, with the emphasis on challenges and their solutions based on the best current available evidence. Prevalence ========== Existing literature regarding the global prevalence of Class III malocclusions has shown that its prevalence varies greatly among and within different races, ethnic groups, and geographic studied regions ([Tables 1](#t1-ccide-10-099){ref-type="table"} and [2](#t2-ccide-10-099){ref-type="table"}). There is a wide range of reported prevalence, even with conflicting results, and the discrepancies in the prevalence rate might be attributed to the variation among samples, the timing of investigation, and type of analysis performed.[@b7-ccide-10-099] A recent systematic review[@b7-ccide-10-099] reported a global prevalence of Angle Class III malocclusion within the interval of 0%--26.7% for different populations. Prevalence rates of 15.80%, 15.69%, and 16.59% were revealed for Southeast Asian countries, Chinese, and Malaysian groups, respectively. Among Japanese it was around 14%, for Koreans 9%--19%, and about 1.65% for Taiwanese. For Indian children aged from 5 to 15 years, the prevalence varied within 0%--4.76%.[@b7-ccide-10-099] Further, from a global viewpoint, Indians had the lowest prevalence of 1.19% among all other racial groups. A prevalence of 10.18% was reported for Middle Eastern populations, and among them, for Israeli Arabs it was 1.3%, Iranians about 15.2%, Turkish about 10.30%--11.5%, and Egyptians showed a rate from 4% to 11.38%. Regarding African countries,[@b7-ccide-10-099] the prevalence rate was found to be 4.59% and varying for Kenya, Tanzania, and Nigeria (between 1% and 16.8%). Class III malocclusions have been found to be more prevalent in Hispanic than in African or Caucasian groups. Prevalence of about 9.1% and 8.3% were reported for Americans and Mexican Americans, respectively.[@b8-ccide-10-099] Factors such as the method of malocclusion study and the age group studied may influence the varying prevalence in Caucasians between 3% and 5%.[@b9-ccide-10-099]--[@b15-ccide-10-099] Prevalences of \~5% and from 2% to 6% have been found in Latin and European populations, respectively.[@b8-ccide-10-099] Furthermore, the White population in United Kingdom and Scandinavia had a Class III incidence of about 3%--5%,[@b16-ccide-10-099] and about 6% for Sweden.[@b17-ccide-10-099] For Americans, the prevalence was found to be about 5%.[@b15-ccide-10-099],[@b18-ccide-10-099] Studies on US African-American population groups found the prevalence in the range of 3%--6%.[@b19-ccide-10-099]--[@b21-ccide-10-099] Similar studies conducted on other nationalities revealed a Class III malocclusion prevalence of about 3% for Brazilian,[@b22-ccide-10-099] 14% for Syrian,[@b23-ccide-10-099] and 9.4% for Saudi Arabian individuals.[@b24-ccide-10-099] Etiology of Class III malocclusion ================================== Similar to most of the malocclusions and dentofacial deformities, the etiology of Class III malocclusion is multifactorial. It results from a distortion of normal development, rather than from any pathological process. Expressions of Class III malocclusion are results of interaction between innate factors or genetic hereditary with environmental factors.[@b25-ccide-10-099]--[@b27-ccide-10-099] Studies of human inheritance have provided sufficient evidence to establish the fact that mandibular growth is mainly affected by heredity.[@b27-ccide-10-099]--[@b31-ccide-10-099] Familiar genetic inheritance has a strong influence on skeletal craniofacial dimensions contributing to Class III malocclusion and a significantly higher incidence of this malocclusion has been found to have a familial occurrence between members of many generations.[@b32-ccide-10-099],[@b33-ccide-10-099] The best known example of familial inheritance is Habsburg Jaw, in which mandibular prognathism recurred over multiple generations in the European royalty.[@b34-ccide-10-099],[@b35-ccide-10-099] The pattern of transmission of Class III malocclusion still remains an issue of controversy. According to some authors, the transmission is autosomal recessive, and according to others, it is autosomal dominant with complete or incomplete penetrance; yet, some others support the polygenic transmission mode.[@b36-ccide-10-099],[@b37-ccide-10-099] Environmental factors known to contribute and influence this malocclusion include wrong postural habits of the mandible which pathologically alter the mandibular condyle positioning within the fossa and as a result the final mandibular spatial position expressed with a forward slide of the mandible. Various factors such as growth stimulus, history of prolonged sucking or resting tongue habits, atypical swallowing, nasal airway obstruction, mouth breathing, functional mandibular shifts because of respiratory needs, tongue size and pharyngeal airway shape and size altered (enlarged tonsils, large tongue, adenoids), hormonal imbalances and disturbances such as gigantism or pituitary adenomas, trauma, premature loss of primary teeth, congenital anatomic defects (ie, cleft lip, cleft palate), and muscle dysfunction alone or in combination with other environmental factors play a definitive etiological role.[@b38-ccide-10-099]--[@b43-ccide-10-099] Component of Class III malocclusion =================================== Class III malocclusion represents a complex three-dimensional facial skeletal imbalance between maxillary and mandibular growth along with varying degrees of dentoalveolar and soft tissue compensations which can be expressed in many morphological ways.[@b44-ccide-10-099] Class III malocclusion may be associated with maxillary growth deficiency (and/or maxillary retrognathia), mandibular growth excess (and/or mandibular prognathism), or a combination of both along with vertical and transverse malformations.[@b45-ccide-10-099]--[@b49-ccide-10-099] Based on the position of the maxilla relative to the craniofacial skeleton, Park and Baik classified Class III malocclusions into three basic types: true mandibular prognathism type A -- individual with normal maxilla and prognathic mandible; type B -- individual with excessive growth of maxilla and mandible, but with relatively more growth of mandible; type C -- individual with maxillary hypoplasia, obtuse nasolabial angle, and concave facial profile. Type C individuals can easily be camouflaged orthodontically by dentoalveolar compensation.[@b50-ccide-10-099] Common skeletal features such as shortened anterior (N-S) and posterior cranial base (S-Ar/Ba), reduced saddle angle (N-S-Ar), and an increased gonial angle (Ar-Go-Gn) were identified to lead to a more forward positioning of the glenoid fossa resulting in Class III malocclusion.[@b51-ccide-10-099]--[@b53-ccide-10-099] Studies about the skeletal and dental components of Class III malocclusions have revealed the establishment of a facial pattern at early childhood which has a tendency to worsen with growth.[@b54-ccide-10-099]--[@b58-ccide-10-099] Skeletal Class III malocclusions can be a result of various factors: prognathic and/or macrognathic mandible with a normal maxilla both in position and in size;retrognathic and/or micrognathic maxilla with a normal mandible both in position and in size;combination of retrognathic and/or micrognathic maxilla with prognathic and/or macrognathic mandible;normal skeletal jaw relationship with reverse overjet in the presence of centric relation (CR)--centric occlusion (CO) discrepancy, also known as a "pseudo" Class III relationship. Dental features of Class III individuals include Class III molar and canine relationship, maxillary incisors protrusion and mandibular incisors retrusion with edge-to-edge bite or anterior crossbite. Based on various combinations of skeletal components, patients with Class III malocclusion exhibit a wide range of underlying skeletal and craniofacial features similar to the prevalence of Class III malocclusion, which can vary among different racial and ethnic groups as shown by comparative studies. For example, Mongoloid populations (Japanese, Koreans, and Chinese) with Class III phenotypes present with characteristic features such as acute anterior cranial base angle and a prominent and elongated mandible with a short and hypoplastic maxilla, while normal maxillary size and position were observed for Caucasians.[@b51-ccide-10-099],[@b59-ccide-10-099] Differential diagnosis of Class III malocclusion ================================================ The specter of problems of Class III malocclusion ranges in gravity from dentoalveolar problems with functional anterior shift of the mandible to true skeletal problems with serious maxillomandibular discrepancies, which leads to its highly challenging diagnosis.[@b60-ccide-10-099] Differential diagnosis of Class III malocclusions and anterior crossbite in a step by step approach has been presented here based on the work of authors Ngan et al,[@b6-ccide-10-099] Battagel,[@b61-ccide-10-099] and Turley.[@b62-ccide-10-099] History of the mandibular prognathism or anterior crossbite ----------------------------------------------------------- The first question that should be asked to the patient or parents (caregiver) should concern mandibular prognathism or anterior crossbite in the family and close relatives. A history of it indicates a genetic cause of Class III malocclusion. Examination of any functional shift ----------------------------------- The second most important step is to access the relationship of maxilla and mandible in CR and CO to determine any functional shift. Molar and incisor relationship in this group of malocclusion is one of the most important diagnostic criteria. Class III malocclusion with positive overjet or edge-to-edge incisors relation accompanied by lingually inclined mandibular incisors represents a compensated form of Class III malocclusion. Anterior repositioning of the mandible may be due to abnormal tooth contact in CR that forces the mandible forward in CO.[@b63-ccide-10-099] These pseudo Class III malocclusion individuals have a Class I skeletal pattern, orthognathic profile, and Class I molar relation in CR, but a Class III skeletal and dental pattern in CO.[@b64-ccide-10-099] Elimination of the CO or CR discrepancy should show whether it is a simple Class I malocclusion or a compensated Class III malocclusion.[@b65-ccide-10-099] In these individuals early correction proves to be a favorable environment for future growth. Clinical assessment ------------------- This should be carried out while the patient is sitting upright in natural head position (NHP) to evaluate the sagittal and vertical facial proportions. The profile of the patient should be evaluated in NHP using "a line down from the bridge of the nose to the base of the upper lip and a second one extending from that point downward to the chin. A straight or concave profile in young patients indicates a skeletal Class III jaw relationship."[@b66-ccide-10-099] Similarly, the transverse dimension should be assessed to evaluate any facial or dental asymmetries. Examination of the temporomandibular joint, oral musculature, and intraoral soft and hard tissue should also be performed. Evaluation of the anterior crossbite aimed to differentiate a true Class III malocclusion from a pseudo Class III malocclusion has been described by Ngan et al,[@b60-ccide-10-099],[@b65-ccide-10-099] and the given diagnostic scheme can be adapted ([Figure 1](#f1-ccide-10-099){ref-type="fig"}). Lateral cephalogram analysis ---------------------------- Best analyses for cephalometric assessment of Class III malocclusion are those that correlate the maxilla to the mandible and each of them to the anterior cranial base. These are ANB (2°), Wits (0 mm), maxillomandibular differential (linear measurement from condylion to point A and condylion to gnathion: 23 mm for 12 years old), nasion perpendicular to point A (+2.3 mm), and nasion perpendicular pogonion (0 mm). Values of each of the cephalometric measurements presented here are for normal individuals. It has been found in discriminant analysis that "Wits" appraisal is the most important factor in the decision making from orthodontic camouflage treatment to surgical options.[@b66-ccide-10-099] A "Wits" appraisal from 0 to −5 mm may be suggestive of a Class III problem being resolved by means of orthodontic camouflage treatment with facemask or chin cup therapy.[@b67-ccide-10-099] A "Wits" appraisal between −4 and −12 mm requires further growth treatment response vector (GTRV) analysis using serial cephalometric radiographs before a decision can be made of whether to camouflage or wait for comprehensive growth before surgical treatment.[@b68-ccide-10-099] Growth trends of Class III malocclusion ======================================= For effective treatment planning and realistic prediction of stability of treatment outcomes in Class III growing patients, knowledge of growth trends is of paramount importance. Concern regarding early treatment and the need for interceptive care in the case of Class III treatment has always been a dilemma for clinicians as well as patients, knowing that not all problems will be solved in these cases until growth is further completed, and the outcomes in long term irrespective of treatment approaches may depend on the growth tendency of an individual. Turpin[@b69-ccide-10-099] in his editorial published in the American Journal of Orthodontics and Dentofacial Orthopaedics has emphasized that a deep understanding of the individual patient's growth weighs much more than the treatment timing and treatment mechanics used on a specific Class III malocclusion growing patient whose timing of treatment is considered to be controversial. In essence, he summarized the studies of Mitani et al[@b70-ccide-10-099] on Class III growth assessment over the years, as "the basic pattern of mandibular prognathism is established before puberty and does not change fundamentally. However, their total growth increments were about the same as those with a normal mandible after the pubertal growth peak." The points raised by Turpin suggest that in spite of a controversial treatment timing the clinician should continue to have an interest in early interceptive treatment of Class III malocclusion as and when required. Assessment of growth and prognosis in Class III patients ======================================================== The prognosis of orthopedic treatment for skeletal Class III malocclusion is favorable when treatment is administered before the pubertal growth peak.[@b71-ccide-10-099]--[@b74-ccide-10-099] However, a Class III malocclusion may worsen due to growth if a patient is left untreated. Therefore, early treatment is recommended for skeletal Class III malocclusion to obtain a balanced skeletal relationship and, by doing this, there is a possibility to minimize the need of further future complicated treatment such as orthognathic surgery.[@b71-ccide-10-099]--[@b75-ccide-10-099] Studies have suggested that the posttreatment outcome of orthopedic treatment may not be stable depending on the residual growth.[@b76-ccide-10-099] When unfavorable growth is expected, treatment would not be initiated in the early phase or would be delayed until completion of growth, because a discrepancy between maxillary, mandibular, and skeletal base growth during the pubertal phase can result in relapse of the corrections which have been achieved earlier,[@b77-ccide-10-099] and as a result some patients ultimately may require orthognathic surgery at a later stage. This is a disappointing situation, not only for patients but also for clinicians, and may be prevented if accurate prediction of the eventual prognosis of early orthopedic treatment for skeletal Class III malocclusion would have been possible before commencing and executing treatment. As a result, patient selection and decision making regarding reliability of decisions and the timing of treatment would be much easier.[@b78-ccide-10-099] Thus, the million dollar question which arises here is: Is the prediction of growth of mandible in growing Class III patient possible with optimal accuracy and reliability? Several investigators have taken this challenge and attempted to predict the prognosis of Class III malocclusion based on the evaluation of the patient's single cephalogram for morphological characteristics and cephalometric analysis.[@b79-ccide-10-099]--[@b83-ccide-10-099] Björk,[@b79-ccide-10-099] based on single cephalogram analysis, defined seven structural signs of extreme growth rotation of mandible in the early developmental stage. These seven signs are "the inclination of the condylar head, curvature of the mandibular canal, shape of the lower border of the mandible, width of the symphysis, interincisal angle, intermolar angle, and anterior lower face height." Even though seven structural signs of growth rotation have been reported,[@b79-ccide-10-099] growth prediction of the mandible still remains a challenge, probably because mandibular growth shows wide variations in terms of amount, direction, and time or because the prediction procedure is based on individual reasoning. Several cephalometric variables have been identified based on mandibular morphology by various studies to predict the results of early treatments.[@b77-ccide-10-099],[@b80-ccide-10-099]--[@b83-ccide-10-099] These studies reported long-term success rates of 50.0%--71.4% for orthopedic treatment of skeletal Class III malocclusion. Because of treatment success criteria, patient characteristics, and time points of outcome evaluation differed among studies; the reported success rates ranged widely and cannot be generalized to all patients treated for skeletal Class III malocclusion. In a recent study, Choi et al[@b78-ccide-10-099] evaluated the long-term success of orthopedic treatment in growing skeletal Class III malocclusions, and verified previously reported success rates and prediction models. This research group found the same results published by previous researchers and concluded that no particular method or factor can predict the long-term success of orthopedic treatment for growing skeletal Class III malocclusions. Assessment of direction and magnitude of mandibular and maxillary growth was proposed by Musich by using a serial cephalograms (D Musich, Growth treatment response vector analysis, personal communication, November 1, 2001). He proposed the GTRV analysis to predict the possibility of excessive mandibular growth after early interceptive orthopedic treatment in Class III patients. Based on immense clinical and extensive research experience in Class III malocclusion, Ngan described the use of serial cephalometric radiographs for GTRV analysis to predict excessive mandibular growth.[@b84-ccide-10-099]--[@b86-ccide-10-099] GTRV ==== According to Ngan,[@b84-ccide-10-099]--[@b86-ccide-10-099] "the horizontal growth changes of the maxilla and mandible are determined by locating the A and B points on the posttreatment lateral cephalometric radiograph. Another cephalogram should be taken during 2--4 years of follow-up visits after the treatment of Class III malocclusion with facemask therapy." The incisal tip of maxillary incisor and mesiobuccal cusp tip of maxillary molar are used as a landmark to construct the occlusal plane (O). Point A and point B should be marked on the lateral cephalogram and connecting points A and B perpendicular to the occlusal plane would construct the lines AO and BO. Using stable landmarks on the midsagittal cranial structure, the nasion (N), and sella turcica (S), post-facemask treatment tracing (first tracing) is superimposed on the follow-up radiograph and the lines AO and BO are again constructed on the follow-up radiograph on the occlusal plane of the first tracing. The distance between the A points and point B of the two tracings along the occlusal plane represents the growth changes of the maxilla and mandible, respectively ([Figure 2](#f2-ccide-10-099){ref-type="fig"}). After the measurements are completed, the GTRV ratio for each patient can be determined by using the formula: GTRV = horizontal growth changes of the maxilla/horizontal growth changes of the mandible. Hence, GTRV is defined as "the horizontal growth changes at A point divided by the horizontal growth changes at B point on the post-facemask and follow-up lateral cephalogram." For an individual with the age range of 6--16 and with normal growth pattern, the GTRV ratio is 0.77. This suggests that the horizontal growth of mandible exceeds 23% as compared to the maxilla to maintain a normal skeletal relationship. In studies by Ngan[@b86-ccide-10-099] and Youssef et al,[@b87-ccide-10-099] in patients treated with early interceptive orthopedic treatment, the GTRV ratios were significantly different for successful cases and unsuccessful cases. For successful cases, GTRV was in the range of 0.33--0.88 (mean 0.45) and for unsuccessful cases it was in the range of 0.06--0.38 (mean 0.22). This suggests that subjects with mild to moderate Class III malocclusion can be camouflaged orthodontically after successful early interceptive treatment, if the GTRV ratio ranges between 0.33 and 0.88. Class III patients with a GTRV ratio between 0.33 and 0.38 can be considered as borderline cases which can either be treated successfully with facemask or turn out to be surgical cases at the end and would better be warned of future need for surgical intervention, if the GTRV ratio is \<0.38. Hence, the GTRV analysis as suggested by Ngan[@b84-ccide-10-099]--[@b86-ccide-10-099] is helpful to the clinicians to assess the growth of the mandible to predict the prognosis after early interceptive orthopedic treatment of Class III malocclusion. Treatment timing ================ It is an accepted fact that skeletal Class III malocclusion establishes itself early in life, is not a self-correcting disharmony,[@b70-ccide-10-099],[@b88-ccide-10-099] and is often associated with maxillary constriction. Intervention at an early stage, such as deciduous dentition, or prepubertal growth phase has been recommended.[@b73-ccide-10-099],[@b89-ccide-10-099] In particular, the prepubertal treatment of Class III malocclusion by means of rapid palatal expansion and facemask protraction yields favorable growth corrections both in maxilla and in the mandible.[@b73-ccide-10-099] In a controlled long-term study, it has been found that patients who have been treated before the pubertal growth phase showed a stable increment in the maxillary skeletal width, maxillary intermolar width, and lateronasal width, while patients treated after the pubertal growth phase showed only dentoalveolar effects after the follow-up of \~8 years.[@b90-ccide-10-099] The proper timing of interventions may therefore rely on chronological age[@b91-ccide-10-099] and phases of dentition[@b92-ccide-10-099] for very young patients, and on other radiological indicators, such as cervical vertebral maturation and/or hand and wrist maturation methods for older children.[@b93-ccide-10-099] A summary of the timing of interventions and main indicators for Class III malocclusion has been presented in [Table 3](#t3-ccide-10-099){ref-type="table"}. Treatment of developing Class III malocclusion ============================================== It is very critical to make a decision for developing Class III malocclusion on whether to treat or wait for further growth and dental development. Although a Class III malocclusion may be identified in the developing dentition, a decision needs to be made as to whether it is better to treat it at this stage or wait for further dental development and growth. The timing of early treatment is crucial for a successful outcome. Some studies have reported that treatment should be carried out in patients \<10 years of age to enhance the orthopedic effect.[@b71-ccide-10-099],[@b94-ccide-10-099]--[@b96-ccide-10-099] In contrast, some studies found that age of the patient had little influence on treatment response and outcome.[@b97-ccide-10-099],[@b98-ccide-10-099] There is no strong evidence to support the notion that early treatment would be beneficial. The main goals of early intervention are to create a more favorable environment for growth and to improve the occlusal relationship: for example, correcting the crossbite and facial esthetics.[@b71-ccide-10-099] Hence, interceptive treatment of Class III malocclusions should be undertaken if it prevents damage to the oral tissues and prevents or significantly reduces the amount, or severity, of future orthodontic and surgical intervention. Turpin[@b99-ccide-10-099] developed a list of positive and negative factors that helped decision making on developing Class III malocclusions ([Table 4](#t4-ccide-10-099){ref-type="table"}) and these guidelines were reviewed by Campbell for deciding when to intercept Class III malocclusion.[@b71-ccide-10-099] Turpin suggested that early interceptive treatment of Class III malocclusion should be considered for patients who presented with positive characteristics as mentioned. The author also recommended that individuals with negative characteristics should delay treatment until the completion of growth. He also suggested that patients should be warned that surgery may be needed in future, even after an early successful interceptive treatment. Common cephalometric predictors for successful Class III camouflage for the evaluation of the maxillary and mandibular position include: ANB (\<−2° to −3°);Wits appraisal (−2 to −6 mm for nonsurgical treatment, −6 to −9 mm for a compromised orthodontic result);maxillomandibular differential and gonial angle within the normal range. The most important factor is clinical assessment to evaluate the need to optimize the facial esthetic. The combination of clinical and cephalometric information will identify which type of Class III malocclusion can be treated in the mixed dentition and help in deciding the best interceptive approach. A Class III patient with mild to moderate Class III skeletal patterns with a GTRV ratio between 0.33 and 0.88 can be successfully camouflaged orthodontically later after initial interceptive treatment in mixed dentition, and a GTRV ratio \<0.38 should be warned. However, the GTRV ratio requires a serial lateral cephalogram to be obtained; so, this tool may not be suitable for decision making for new patients.[@b84-ccide-10-099] In mixed dentition from the treatment point of view, there are effectively three types of Class III malocclusions:[@b100-ccide-10-099],[@b101-ccide-10-099] dental: incorrect inclination or position of maxillary or mandibular incisors;pseudo: anterior positioning of the mandible as a result of premature dental contacts deflecting the mandible anteriorly to allow the patient to achieve full intercuspation;skeletal: true skeletal discrepancies in the maxilla and/or mandible. Treatment of simple dentoalveolar anterior crossbites ----------------------------------------------------- A simple anterior crossbite can be corrected with either a removable appliance or a fixed appliance. The percentage of success is increased if there is a minimal existing proclination of the upper incisors and adequate overbite to maintain correction at the end of treatment. Types and indications of both types of appliances are provided below. Removable appliances: there are three types of removable appliances used to intercept the developing crossbite: inclined plane;modified inclined plane;active Hawley appliance.Fixed appliance: 2 by 4 appliance. Inclined plane -------------- An inclined plane is a good treatment choice in deciduous dentition or early mixed dentition.[@b102-ccide-10-099] It is indicated in patients with: retroclined maxillary anterior teeth with an anterior crossbite with or without functional shift;well-aligned mandibular anterior teeth without proclination;normal to deep overbite;average to horizontal growth pattern patients. The inclined plane is fixed onto the lower anterior teeth with temporary cement. Appropriate angulation between the inclined plane and the upper anterior teeth in crossbite should be determined by considering the vertical discrepancy between the teeth in crossbite and the adjacent teeth, as well as the degree of overbite of the teeth in crossbite by adjusting the different contact angulations. Most anterior dental crossbites can be corrected within 3--4 weeks using an inclined plane. Modified inclined plane ----------------------- Modified inclined plane is indicated in patients with an anterior dental crossbite, with lingually inclined maxillary incisors and labially inclined mandibular incisors; for this reason, an inclined plane is contraindicated because it may cause mandibular incisors proclination.[@b103-ccide-10-099],[@b104-ccide-10-099] A modified inclined plane is similar to that of a Hawley appliance plus an inclined plane on the anterior part. The inclined plane portion covers the lower anterior teeth up to their incisal third. When the patient bites, the inclined plane portion raises the bite and proclines the upper anterior teeth labially. The metal wire parts consist of a labial bow and Adam's clasps on the first permanent molars for better appliance retention and stability. The labial bow is placed on the labial side of the lower incisors, near the cervical third, to ensure that the force application is closer to the center of resistance of the lower anterior teeth. The acrylic resin lingual to the lower incisors may be trimmed to allow lingual movement of the lower anterior teeth when the labial bow is activated. Anterior crossbite problems can be corrected within 3--4 weeks using a modified inclined plane. After the correction of anterior crossbite, the same appliance may be used as retainer by trimming of the anterior inclined plane portion. Active Hawley appliance ----------------------- A removable Hawley appliance is indicated in patients with anterior crossbite in mixed dentition stage.[@b100-ccide-10-099] This appliance has an active component anteriorly to procline the upper anterior tooth or teeth to correct the anterior crossbite. This active component can either be a palatal "Z" spring, which is activated by the clinician, or a screw, which the patient activates. Usage of the Jack expansion screw is recommended to procline maxillary anterior teeth at the rate of one turn at every third day till the correction of anterior crossbite. The turning of the screw by a quarter turn (90°) brings about 0.18 mm of linear movement depending upon the pitch of the screw. Correction of 2 mm of overjet with this appliance can be achieved in 5--6 weeks. The appliance also incorporates retentive components (preferably Adam's clasp) to keep the appliance in place and possibly posterior bite plate to disclude the occlusion to aid in uninterrupted proclination of anterior teeth. A removable appliance can only tip the teeth; so it should be only used if simple tipping movements of the upper anterior teeth are required. Fixed appliance --------------- This appliance is often referred to as a "2 by 4" or "2 by 6" appliance as it is only bonded on the two upper first permanent molars and the four upper incisors or six anterior teeth. Fixed appliances mostly use late mixed dentition or early permanent dentition. An open coil NiTi spring often compressed between the molars and the incisors to procline the incisors or a 0.016-inch stainless steel stoppered arch wire may be used to increase the arch length. Glass ionomer cement may be placed temporarily as a posterior fixed bite plane on the molars if disclusion is required. Fixed appliances allow tipping, bodily movement, and correction of rotations as and when required. It has been seen that both types of appliances work well and the results are equally stable. Fixed appliance treatment is quicker and cheaper and has less effect on the patient's speech than a removable appliance, but patients may complain of slightly more difficulty in chewing and biting initially with the fixed appliance.[@b101-ccide-10-099]--[@b107-ccide-10-099] Growth modification and orthopedic treatment -------------------------------------------- Growth modification in developing Class III malocclusion is indicated in patients with skeletal discrepancy. The basic aim of this interceptive treatment for developing Class III malocclusion is to improve or correct the skeletal discrepancy to allow future treatment of such patients by orthodontic camouflage only without the need of orthognathic surgery. This approach of growth modification in Class III patients can be achieved through functional appliances, chin cup therapy, protraction facemask, and bone-anchored appliances. A brief description of each modality based on current evidence has been given below. Functional appliances --------------------- Functional appliances have been used to modify the skeletal pattern by enhancing the growth of the maxilla and restricting or redirecting the growth of the mandible. Two commonly used functional appliances to intercept Class III malocclusion are Frankel functional regulator III appliance (FR III) and reverse twin-block appliance. FR III has maxillary vestibular shields in the depth of the sulcus. These shields are placed away from the maxilla to stretch the periosteum and encourage anterior development of the maxilla. The lower part of the appliance attempts to restrict mandibular growth or redirect it posteriorly. In the reverse twin-block appliance, the blocks are positioned so that there are posterior forces on the mandible and anterior forces on the maxilla. Current research suggests that functional appliances can improve occlusal relationships, but this is principally due to dentoalveolar changes, proclining upper incisors, and retroclining the lowers incisors.[@b108-ccide-10-099]--[@b110-ccide-10-099] Evidence from a recent systematic review suggests that the FR III might restrict mandibular growth but not stimulate forward movement of the maxilla.[@b111-ccide-10-099] Current evidence suggests that functional appliances can successfully correct a developing Class III malocclusion, but they have principally dentoalveolar effects, with minimal or no effects on the underlying skeletal pattern. Functional appliances especially FR III can be challenging to wear in the mouth and are subject to breakage; hence, a simpler method like orthodontic camouflage may be used in place of functional appliances. Chin cup -------- Chin cup appliance treatment is indicated in young growing patients with mandibular prognathism. It has been found that chin cup therapy does not restrain mandibular growth but redirects the mandible growth vertically, causing a backward rotation of the mandible.[@b112-ccide-10-099] These changes in the direction of mandibular growth help to improve Class III malocclusion. Recent systematic reviews showed that there is considerable agreement between studies in that chin cup therapy may be used for interceptive treatment of growing Class III malocclusion based on short-term favorable results.[@b113-ccide-10-099],[@b114-ccide-10-099] It has been seen that these changes are not maintained in the long term and the normal growth pattern of the mandible reestablishes itself, if chin cup appliance therapy is discontinued before growth completion.[@b115-ccide-10-099] Hence, it is recommended that patients with Class III malocclusion with mandibular prognathism wear the chin cup appliance until growth is completed to maintain the treatment effects of chin cup therapy. Special care should be taken while deciding chin cup therapy in patients who present in the mixed dentition with marked mandibular prognathism, particularly if associated with increased vertical proportions, as these patients are often best treated by surgical orthognathic approach, when their growth is completed. As the long-term prognosis of chin cup therapy is unpredictable, patient caregivers should always be fully informed of this before initiating chin cup therapy. Line of force of applied force for chin cup therapy should be directed along the lines from the chin point to the condyle heads, bilaterally in the range of 400--500 g, 10--14 hours per day. While trying the chin cup appliance on patients, care should be taken to ensure that the chin cup does not impinge on the lower lips as it may cause retroclination of the lower incisors and recession of labial gingiva ([Figure 3](#f3-ccide-10-099){ref-type="fig"}). Protraction facemask -------------------- Protraction facemask also referred to as reverse headgear is one of the most commonly used interceptive tools to intercept developing skeletal Class III malocclusion.[@b116-ccide-10-099] The appliance is composed of two components: an extraoral framework (facemask) that fits on the forehead and chin, and an intraoral attachment to the maxillary dentition ([Figure 4](#f4-ccide-10-099){ref-type="fig"}). The chin and forehead part of the extraoral framework are connected by a middle bar for the connection of the elastics to the intraoral attachment to the maxillary dentition. The intraoral attachment is of various designs, including removable, banded, and acrylic-bonded versions. They all incorporate hooks bilaterally positioned near the maxillary canines. To minimize unwanted rotation of the palatal plane, Class III elastics should be attached near the maxillary canines at 30° to the occlusal plane. Bonded expansion appliance is preferred as it provides a temporary bite plane effect in hyperdivergent cases and facilitates the jumping of anterior crossbite in deep bite cases. The elastic forces are typically 400--450 g per side (14 to 16 OZ) and need to be worn 12--14 hours per day. The total treatment time is usually 6--9 months. An increased release of growth hormone and other growth promoting endocrine factors has been observed during evening and night than during the day. As a result, it is recommended to wear the appliance during evening and nighttime. One controversial area is the use of rapid maxillary expansion (RME) at the same time as the protraction facemask. Often patients with a Class III skeletal pattern have a constricted maxilla in the transverse dimension as well as the anteroposterior dimension; so this expansion is a helpful component of the treatment. It has been suggested that expansion may loosen the circummaxillary sutures and increase forward movement of the maxilla. Results of a recent randomized controlled trial (RCT)[@b117-ccide-10-099] and data from meta-analysis support the notion that facemask with and without RME treatment are both equally effective clinically in early treatment methods for skeletal Class III malocclusion. Hence, in case of no transverse discrepancy facemask without RME treatment can be undertaken. Maxillary expansion and protraction treatment for early permanent dentition is undertaken in case of transverse maxillary constriction.[@b118-ccide-10-099] The expansion appliance is activated twice daily (0.25 mm per turn) for 7--10 days. In case of severely constricted maxilla, activation of the screw is carried out for ≥2 weeks. In a randomized controlled clinical trial (CCT) comparing protraction facemask with no treatment, successful correction of the reverse overjet was observed in 70% of patients, with an average increase in overjet of 4 mm, and significant skeletal changes, principally due to forward movement of the maxilla and improvement in the ANB angle of 2.6° compared to the control at the end of treatment.[@b119-ccide-10-099] These patients were followed up 6 years later to see if favorable changes were maintained toward the end of growth and in particular to assess whether the interceptive use of a facemask in the developing dentition can help reduce the need for orthognathic surgery.[@b120-ccide-10-099] Of the patients that wore a protraction facemask, 36% needed orthognathic surgery at the age of 15, whereas 66% of patients in the control group required orthognathic surgery. Encouragingly, 68% of patients who wore the protraction facemask had a positive overjet after 6 years. To conclude, it appears that the use of a protraction facemask in the developing dentition will correct Class III malocclusion and reduce the need for orthognathic surgery in the future in the following types of cases: child under the age of 10;mild--moderate Class III;retrusive maxilla;average or reduced vertical proportions. As the mandible resumes its original downward and forward growth after Phase I (facemask) therapy, patients may have reduced corrected overjet and more Class III molar relation during follow up and the second phase of orthodontic camouflage. Hence, the use of an extraoral orthopedic appliance (eg, chin cup) is recommended after the first phase of interceptive therapy with facemask till the completion of mandibular growth. It is also recommended to take the serial posttreatment cephalogram to calculate the GTRV to make a decision on whether to opt for orthodontic camouflage or wait till the completion of growth. A recent review published in 2017, in evidence-based dentistry, by Smyth and Ryan assessed early treatment of Class III malocclusion with facemask therapy and included randomized clinical trials and controlled clinical trials in children aged 7--12 years undergoing fixed or removable orthodontic treatment with the primary outcome of correction of the reverse overjet. The authors concluded that there is a moderate amount of evidence to show that early treatment with a facemask appliance has positive improvement for both skeletal and dental effects in the short term.[@b121-ccide-10-099] Bone-anchored appliances ------------------------ Interceptive treatment of Class III malocclusions with a tooth-borne protraction appliance (eg, facemask) often poses problems of unwanted dental changes such as: buccal tilting of maxillary molars and extrusion may lead to an increase in vertical dimensions and downward and backward growth of the mandible;decreased arch length due to mesial movement of maxillary molars leading to crowding in the anterior teeth. In an attempt to overcome the limitations of tooth-borne appliances in the interceptive treatment of Class III malocclusions, bone-anchored maxillary protraction (BAMP) appliances have recently been used.[@b122-ccide-10-099] BAMP appliances typically involve the use of Class III elastics attached between miniplates placed in the infrazygomatic crest to miniplates placed in the mandibular symphysis region or attached to the extraoral facemask ([Figure 5](#f5-ccide-10-099){ref-type="fig"}). The success of these miniplates is related to the surgical technique and the thickness and quality of the bone. Particularly in the maxilla, the bone quality is often not as good until the patient is at least 11 years old; so this interceptive technique tends to be used in slightly older patients than the tooth-borne appliances. The results of an initial study on the effects of BAMP compared with growth of the untreated Class III subjects showed that the BAMP protocol induced an average increment on skeletal and soft tissue advancement of maxillary structures of about 4 mm with negligible changes in the maxillary incisor inclination and vertical skeletal pattern.[@b123-ccide-10-099] Recent research also found that a Hybrid Hyrax bone-anchored rapid palatal expansion appliance minimized the side effect encounter by tooth-borne rapid palatal expansion appliances for maxillary expansion and protraction and may serve as an alternative treatment appliance for correcting Class III patients with a hyperdivergent growth pattern.[@b124-ccide-10-099] Hence, BAMP has demonstrated promising initial results in its potential to offer greater skeletal changes, with less unwanted displacement of the dentition. However, there are unpredictable variations in individual outcomes, and high-quality research is needed to further investigate this technique. Summary ======= For accurate diagnosis and successful execution of orthodontic treatment, it is very important that children with developing Class III malocclusion must be evaluated for family history, and should undergo dental examination for molar and incisor relationships, functional assessment to evaluate CO or CR shift on mandibular closure, and cephalogram analysis to determine sagittal jaw discrepancy. Evaluation of GTRV to determine the individual growth direction and rate should always be performed in borderline cases. A summary of diagnosis and treatments of developing Class III malocclusions is presented and [Figure 1](#f1-ccide-10-099){ref-type="fig"} may be used for quick reference. Interceptive treatment of Class III malocclusions may be undertaken if it prevents damage to the oral tissues and/or prevents or significantly reduces the amount, or severity, of future orthodontic treatment.Simple anterior dental crossbites can be successfully corrected with removable or fixed appliances in mixed dentition.Treatment with chin cup or functional appliances can correct a Class III incisor relationship, but any orthopedic changes are likely to be minimal with these appliances.Optimal timing for interceptive treatment with face-mask is in the deciduous or early mixed dentition stage. Early interceptive treatment with a facemask allows for favorable sutural response for maxillary expansion and protraction; correction of any CO or CR discrepancies; and improvement in the facial profile and self-esteem in patients who are under 10, with mild to moderate Class III and a retrusive maxilla and with average or reduced vertical proportions. In case of mandibular prognathism, after facemask therapy patients should be advised to the wear the chin cup appliance until growth completion. A follow-up lateral cephalogram should be taken at 2--4 years after maxillary protraction to calculate the GTRV ratio. Information obtained from the GTRV ratio and from the cephalogram for the mandibular growth rate and direction is helpful in deciding whether Class III malocclusion can be camouflaged orthodontically or surgical intervention would be needed in the future. A case of Class III malocclusion treated successfully with facemask therapy has been presented in [Figures 6](#f6-ccide-10-099){ref-type="fig"}[](#f7-ccide-10-099){ref-type="fig"}[](#f8-ccide-10-099){ref-type="fig"}[](#f9-ccide-10-099){ref-type="fig"}[](#f10-ccide-10-099){ref-type="fig"}[](#f11-ccide-10-099){ref-type="fig"}[](#f12-ccide-10-099){ref-type="fig"}--[13](#f13-ccide-10-099){ref-type="fig"}.Bone-anchored appliances may offer the potential for more skeletal changes, but before its routine use more evidence is needed. Furthermore, bone-anchored appliances require help from surgeons. Written informed consent was obtained from the parents or legal guardians of the children, to have the images published. The authors are grateful to Professor OP Kharbanda, Chief of the Centre for Dental Education and Research, All India Institute of Medical Sciences (AIIMS), New Delhi for all clinical facilities, Dr Pradip Sangroula, Ex-Orthodontic resident, AIIMS for [Figure 5](#f5-ccide-10-099){ref-type="fig"}, and Mr Netrapal, Dental Technician, AIIMS for scanning of models. Disclosure The authors report no conflicts of interest in this work. ![Summary of clinical practice guidelines for developing Class III malocclusion.\ Abbreviations: CR, centric relation; CO, centric occlusion; GTRV, growth treatment response vector.](ccide-10-099Fig1){#f1-ccide-10-099} ![Horizontal growth changes of the maxilla and the mandible between the posttreatment and the follow-up cephalograms: (A) pretreatment, (B) post-facemask, (C) follow-up cephalogram, and (D) GTRV calculation on post-facemask and follow-up cephalogram.\ Abbreviation: GTRV, growth treatment response vector.](ccide-10-099Fig2){#f2-ccide-10-099} ![Chin cup appliance with line of force passing through the condyle.](ccide-10-099Fig3){#f3-ccide-10-099} ![Facemask therapy. (A) Occlusal view showing maxillary splint; (B) frontal view showing hooks for Class III elastics; and (C) facemask attached to maxillary splint through Class III elastics for maxillary protraction.](ccide-10-099Fig4){#f4-ccide-10-099} ![Bone-anchored maxillary protraction: (A) placement of miniplate in infrazygomatic crest; (B) miniplate in situ; (C) OPG showing miniplates; and (D) facemask attached to miniplates through Class III elastics for maxillary protraction.\ Abbreviation: OPG, orthopantomogram.](ccide-10-099Fig5){#f5-ccide-10-099} ![Pretreatment photographs showing Class III malocclusion.](ccide-10-099Fig6){#f6-ccide-10-099} ![Pretreatment digital study models showing Class III malocclusion.](ccide-10-099Fig7){#f7-ccide-10-099} ![Maxillary splint with Hyrax expander cemented for facemask therapy.](ccide-10-099Fig8){#f8-ccide-10-099} ![Posttreatment photographs.](ccide-10-099Fig9){#f9-ccide-10-099} ![Posttreatment digital study models.](ccide-10-099Fig10){#f10-ccide-10-099} ![Pretreatment and posttreatment cephalometric and panoramic radiographs.](ccide-10-099Fig11){#f11-ccide-10-099} ![Pretreatment and posttreatment superimposed occlusal view of maxillary and mandibular models showing treatment changes (green color showing pretreatment and rustic brown color showing posttreatment).](ccide-10-099Fig12){#f12-ccide-10-099} ![Posttreatment chin cup appliance therapy for mandibular growth redirection.](ccide-10-099Fig13){#f13-ccide-10-099} ###### Reported prevalence of Class III malocclusion globally and in different continents Continents Prevalence (%) ------------------- ---------------- Globally 0--26.7 East Asian 4--14 Southeast Asian 15.80 African 4.59 Middle Eastern 10.18 Indian 1.19 European 4.88 (2--6) Northern European 0.8--4.2 American 5 ###### Reported prevalence of Class III malocclusion among different nationality groups Nationality Prevalence (%) -------------------- ---------------- Chinese 15.69 Malaysian 16.59 Japanese 2.3--14 Korean 9--19 Taiwanese 1.65 Indian 0--4.76 Israeli Arabs 1.30 Iranians 15.20 Turkish 10.30 Egyptians 11.38 Tanzanian 1.81--19.72 Nigerian 1.22--11.79 United Kingdom 3--5 Scandinavian 3--5 Swedish 6 Brazilian 3--5 Saudi Arabian 9.40 African Americans 3--6 European Americans 0.80 Latino Americans 9.10 Mexican Americans 8.30 Italians 5 German 2.80 Belgian 6 British 2.9 Danish 4.30 Polynesian 5.50 Syrian 14.0 Lebanese 5.10 ###### Optimum timing of interventions and indicators for Class III malocclusion --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Problem Optimum timing of intervention Main indicators Treatment modality ------------------------------------------------------------ -------------------------------- ------------------------------------------------------ ---------------------------------------------- Class III malocclusion with or without constricted maxilla Pubertal growth phase Chronological age (up to 8 Y -- girls; 9 Y -- boys)\ Facemask with or without maxillary expansion Phases of dentition (up to mixed dentition)\ CVM (CS1--CS2)\ HWM (SMI 1--SMI 2) --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Abbreviations: CS, cervical stage; SMI, skeletal maturation indicator; CVM, cervical vertebral maturation; HWM, hand and wrist maturation. ###### Turpin's positive and negative factors for decision making for interception of developing Class III malocclusion Positive factors Negative factors -------------------------------------------------------------- -------------------------------- Convergent facial type Divergent facial type Anteroposterior functional shift No anteroposterior shift Symmetrical condyle growth Asymmetrical growth of condyle Young subject with remaining growth Mild skeletal disharmony Growth completed Severe skeletal disharmony Good cooperation expected Poor cooperation expected No familial prognathism Familial pattern established Good facial esthetics Poor facial esthetics [^1]: These authors contributed equally to this work	{ "pile_set_name": "PubMed Central" }
1. Introduction =============== Pulmonary embolism (PE) is the 3rd cause of cardiovascular death worldwide. More than 600,000 patients suffering from PE are admitted to the hospital every year in the United States but \<5% of them die within 3 months.^\[[@R1]--[@R5]\]^ Based on these epidemiologic data, experts emphasize the importance of early stratification to identify patients with PE at high risk of death, who may benefit from more intensive monitoring or aggressive therapy.^\[[@R6],[@R7]\]^ Recent international guidelines defined "high risk" population in the presence of persistent arterial hypotension (\<90 mm Hg) or shock. However, hemodynamic parameters defining shock status in the context of PE remain unclear and their correlation to outcome has never been evaluated. In critically ill patients admitted in the intensive care unit (ICU) for septic shock or myocardial infarction-related cardiogenic shock, several groups have reported that alterations of microcirculation blood flow are strongly predictive of mortality independently of arterial blood pressure or cardiac output.^\[[@R8]--[@R10]\]^ Microcirculatory perfusion could be evaluated using computer-assisted devices but these tools are not available at bedside. Tissue perfusion could also be assessed with more simple tools including arterial lactate level or noninvasive clinical parameters such as skin mottling or urinary output. In our department, we are interested in clinical evaluation of organ perfusion and we have showed that skin mottling, reflecting skin hypoperfusion, correlated with organ failure and predicted mortality in sepsis^\[[@R11]\]^ and septic shock patients.^\[[@R12]\]^ In this study, we aimed at investigating the relationship between 28-day mortality and tissue perfusion parameters, that is, arterial lactate level, skin mottling and urinary output, in patients admitted to ICU for acute PE. 2. Materials and methods ======================== We conducted a retrospective observational study in a 16-bed ICU in a tertiary teaching hospital. All consecutive patients older than 18 years admitted for acute PE from January 1993 to December 2015 were included. Patients were identified by querying the electronic health records with the following keywords (in French): "pulmonary embolism," "acute respiratory failure," "anti-coagulant," "deep venous thrombosis," "cardiac arrest," "heparin," "vitamin K antagonists," "cava filter," or "thrombolysis." General characteristics of patients were recorded: demographic data, diagnoses, severity of illness evaluated by the sequential organ failure assessment (SOFA) score and simplified acute physiology score II (SAPS II). Global hemodynamic variables, right ventricle (RV) dilation and tissue perfusion parameters were collected at ICU admission. RV dilation was defined as a right to left ventricle diameter ratio superior or equal to 1 on echocardiography or computed tomography (CT) scan. We recorded 3 tissue perfusion parameters at admission: arterial lactate level, the presence/absence of skin mottling around the knee area, and mean urinary output (within the first 6 hours). An ethical approval was not necessary because it was a monocenter retrospective study using anonymous data. 2.1. Statistical analysis ------------------------- Data were summarized as median (25th--75th percentiles) for skewed distributions and percentages as appropriate. Association of tissue perfusion parameters with 28-day mortality was tested using Mann--Whitney nonparametric test for continuous variables, Fischer exact test, and the Chi-squared test for categorical data according to sample size. Finally, we used logistic regression for multivariable analysis. All statistical analyses were performed using the R software (v 2.12.0; [http://cran.r-project.org](http://cran.r-project.org/)). Significance was defined as 2-sided P-value \<.05. 3. Results ========== 3.1. Studied population ----------------------- Over a 22-year period, we identified 317 patients with PE admitted in our intensive care department by querying the electronic records. After reviewing medical charts, 122 patients were excluded because PE was not the main diagnosis for ICU admission and 87 patients were excluded because of missing data, leaving 108 patients for study (Supplemental Figure 1). PE was routinely diagnosed using nuclear planar V/Q-imaging during the 1993 to 2008 period, whereas CT pulmonary angiography became the diagnosis test of choice between 2008 and 2015. Echocardiography use in the context of acute severe PE also increased between the 2 time periods (Supplemental Figure 2). Characteristics of the studied population are summarized in Table [1](#T1){ref-type="table"}. The median age was 72 (61--81) years with no gender predominance. The median length of stay was 4 (2--7) days in the ICU and 14 (8--21) days in the hospital. At admission, median SOFA score was 2 (0--6) and median SAPS II was 29 (16--43). Thirty-nine patients (36%) required support organ therapy including vasopressor infusion (n = 30, 28%) and mechanical ventilation (n = 31, 29%). Thirty-seven patients (34%) received thrombolytic therapy. Day 28 mortality rate was 25% (n = 27). Nonsurvivor patients had higher organ failure severity and received more frequently both support therapy and thrombolytic treatment. ###### General characteristics of the studied population. ![](medi-97-e11993-g001) 3.2. Tissue perfusion parameters -------------------------------- We analyzed tissue perfusion parameters at ICU admission according to 28-day outcome. When compared to survivors, nonsurvivor patients had higher arterial lactate level (4.5 \[2.3--10.3\] mmol/L vs 1.4 \[1--2.9\] mmol/L, P \< .001), more frequent skin mottling around the knee area (56% vs 25%, P = .003) and lower mean 6-hour urinary output (0.35 \[0--1\] mL/kg/h vs 0.88 \[0.62--1.68\] mL/kg/h, P = .0002) (Table [2](#T2){ref-type="table"}). ###### Analysis of tissue perfusion parameters at admission according to 28-day outcome. ![](medi-97-e11993-g002) To assess the cumulative predictive value of tissue perfusion parameters, we analyzed the relationship between 28-day mortality and the number of tissue perfusion abnormalities at admission, that is, skin mottling, hyperlactatemia (\>2 mmol/L), and oliguria (\<0.5 mL/kg/h). Interestingly, 28-day mortality increased with the number of tissue hypoperfusion parameters at 8%, 21%, 28%, and 85% in the presence of 0, 1, 2, and 3 alterations, respectively (P \< .0001, Chi-squared for a trend test) (Fig. [1](#F1){ref-type="fig"}). ![Day 28 mortality according to the number of tissue perfusion alterations. Tissue perfusion alterations were defined as the presence of skin mottling around the knee area, arterial hyperlactatemia \>2 mmol/L and oliguria (mean 6-hour urinary output \<0.5 mL/kg/h). Chi-squared test for a trend.](medi-97-e11993-g003){#F1} In a multivariate analysis, the relationship between the number of tissue perfusion alterations and 28-day mortality was maintained after adjustment on the presence of shock and RV dilation at admission (Table [3](#T3){ref-type="table"}) ###### Multivariable logistic regression analysis of risk factors for 28-day mortality, including tissue perfusion alterations, shock, and right ventricular dilation. ![](medi-97-e11993-g004) 4. Discussion ============= In this study, we found that 3 tissue perfusion parameters measured at admission, arterial lactate level, skin mottling, and urinary output were synergistically associated with 28-day mortality in ICU patients admitted for PE, independently of global hemodynamic status and right ventricular dilation. In patients with PE, current guidelines recommend the evaluation of systolic blood pressure and RV injury to identify patients at high risk of death.^\[[@R7]\]^ However, both parameters fail to fully identify patients who would benefit from more aggressive treatment (thrombectomy or fibrinolysis) with a substantial reduction on mortality without a major incidence of hemorrhagic complications.^\[[@R13],[@R14]\]^ Recently, the pathophysiology of severe PE has been extended beyond simple RV obstruction. Several studies highlighted the role of inflammation^\[[@R15]\]^ and endothelial dysfunction^\[[@R16]\]^ in the alteration of organ perfusion at the microcirculation level in the context of acute circulatory failure related to infection^\[[@R17]\]^ or myocardial infarction.^\[[@R18]\]^ In this study, we evaluated microcirculatory hypoperfusion parameters available at bedside and found these "easy to use and easy to learn" tools are helpful to identify high-risk patients in acute PE. First, we investigated arterial lactate level the measurement of which is widely available in routine laboratories. Hyperlactatemia is a hallmark characteristic of shock states^\[[@R19]\]^ and predicts organ failure and mortality^\[[@R20]\]^ in the ICU, even more than refractory hypotension.^\[[@R21]\]^ In patients with PE, Vanni et al reported that arterial lactate level measured in the emergency department was significantly associated with in-hospital mortality independently of systolic blood pressure and right ventricular dysfunction.^\[[@R22],[@R23]\]^ Here we confirmed these results in critically ill patients admitted to the ICU. We found that nonsurvivor patients had significantly higher lactate levels when compared to survivors. Mortality rate was higher in our study when compared to Vanni\'s study probably because ICU patients had more frequent comorbidities and more severe organ failure. Next, we analyzed skin mottling, an old clinical sign, readily available at bedside. Skin mottling is characterized by locally reduced blood flow^\[[@R24]\]^ and decreased tissue oxygen content. In several prospective observational studies in sepsis patients, we have showed that mottling extension is strongly predictive of mortality.^\[[@R11],[@R12]\]^ Here, for the first time, we found that the presence of skin mottling is significantly associated with 28-day mortality in patients with PE, corroborating data from the severe infection context. Urinary output can also be routinely measured. In sepsis, urinary output monitoring is recommended to assess hemodynamic status and to guide treatment.^\[[@R25]\]^ Guidelines are based on seminal studies that reported an association between oliguria and pejorative outcome.^\[[@R26],[@R27]\]^ Here, we showed that in critically ill patients with PE, urinary output was significantly lower over the first 6 hours in nonsurvivors when compared to survivors. We have previously reported a significant relationship between these tissue perfusion parameters. Indeed, in septic shock patients, mottling extension correlates negatively with urinary output and positively with arterial lactate levels.^\[[@R12]\]^ Here, we found that these parameters could be combined as the risk of mortality increased with the number of tissue perfusion alterations. In light of all these data, we believe that combining arterial lactate level, mottling and urinary output could be helpful at the bedside to improve risk stratification of patients with PE. Our study has several limitations. First, it is a retrospective study and the predictive value of these tissue perfusion parameters has to be confirmed prospectively in a larger cohort of patients with PE. The next step would be to evaluate if clinical detection of tissue perfusion disorders could have an impact on the therapeutic management of patients with PE. Second, we did not investigate the variations of tissue perfusion parameters during resuscitation, although a reduction of lactate arterial level or a decrease in mottling skin areas over time have been both reported to be correlated with better outcome.^\[[@R28],[@R29]\]^ Third, because a large number of patients were included before 2008 international guidelines, circulating markers of RV injury, including troponin and B-type Natriuretic peptide (BNP) were not measured and not included in the risk stratification for death. 5. Conclusion ============= In this retrospective study, we found that 3 tissue perfusion parameters, arterial lactate level, mottling, and urinary output are significantly associated with 28-day mortality in ICU patients admitted for acute PE independently of shock and right ventricular dilation. These tissue perfusion parameters could be helpful to improve risk stratification of patients with PE. Acknowledgment ============== The authors thank R. Leboursicaud for her participation and help in recovering medical files. Author contributions ==================== Study concept and design, all authors. Data collection T.U., Y.N. Drafting of the manuscript, T.U., N.B., E.M., B.G. and H.A.O. Critical revision of manuscript, all authors. Statistical analysis, T.U., N.B., PYB and H.A.O. Conceptualization: Tomas Urbina, Hafid Ait-Oufella. Data curation: Pierre-Yves Boelle, Hafid Ait-Oufella. Formal analysis: Tomas Urbina, Pierre-Yves Boelle, Bertrand Guidet, Hafid Ait-Oufella. Investigation: Tomas Urbina, Naike Bige, Yann Nguyen, Vincent Dubee, Jeremie Joffre, Idriss Abdallah, Jean-Luc Baudel, Hafid Ait-Oufella. Methodology: Pierre-Yves Boelle, Vincent Dubee, Eric Maury, Bertrand Guidet, Hafid Ait-Oufella. Project administration: Hafid Ait-Oufella. Software: Yann Nguyen, Hafid Ait-Oufella. Supervision: Hafid Ait-Oufella. Validation: Tomas Urbina, Naike Bige, Yann Nguyen, Pierre-Yves Boelle, Hafid Ait-Oufella. Visualization: Jean-Luc Baudel, Hafid Ait-Oufella. Writing -- original draft: Tomas Urbina, Naike Bige, Hafid Ait-Oufella. Writing -- review & editing: Naike Bige, Vincent Dubee, Jeremie Joffre, Jean-Luc Baudel, Eric Maury, Bertrand Guidet, Hafid Ait-Oufella. Supplementary Material ====================== ###### Supplemental Digital Content Abbreviations: ICU = intensive care unit, PE = pulmonary embolism, RV = right ventricle, SAPS II = simplified acute physiologic score II, SOFA = sequential organ failure assessment, sPESI = simplified pulmonary embolism severity index. The authors have no funding and conflicts of interest to disclose. Supplemental Digital Content is available for this article.	{ "pile_set_name": "PubMed Central" }
All relevant data is in the paper and supporting information files. Introduction {#sec001} ============ Polysaccharides are an extremely diverse family of natural biopolymers, which are industrially used as thickeners, stabilizers and gelling agents in foodstuffs \[[@pone.0171469.ref001]\]. There is currently growing interest in their biological functions, such as their antioxidant and prebiotic activity. Although polysaccharides are derived from various sources such as microorganisms, algae and higher plants, the market is dominated by polysaccharides obtained from algae (such as carrageenans, alginates and agar) and higher plants (such as starch, cellulose and pectin) \[[@pone.0171469.ref002]\]. The synthesis of microbial polysaccharides has emerged as an important source of new biopolymers for industrial use \[[@pone.0171469.ref003]\]. It is an attractive alternative as microorganisms can grow under controlled conditions and produce a large variety of polysaccharides with unique properties. Production can be carried out in large quantities by biotechnological routes and from renewable and low-cost raw materials, which are easily recovered \[[@pone.0171469.ref004]\]. One of the microbial polysaccharides of industrial interest is curdlan, a neutral exopolysaccharide composed exclusively of glucose residues joined by β-(1→3) bonds, commercially produced by bacterial species of Agrobacterium \[[@pone.0171469.ref002],[@pone.0171469.ref005]--[@pone.0171469.ref007]\]. After being discovered by Harada in 1966, curdlan has received considerable attention in both the food and non-food industry due to its physicochemical properties, which are unique when compared to other polysaccharides commonly used, such as starch. When in aqueous suspension, curdlan may form a thermoreversible gel (low-set gel) when heated to temperatures close to 55°C with subsequent cooling. This gel has similar behavior to agar and gelatin gels. In addition, curdlan can form a thermoirreversible gel (high-set gel) when its aqueous suspension is heated to temperatures above 80°C, which is very stable. Furthermore, gels with different strengths can be formed by varying the temperature, the heating time and the concentration of curdlan \[[@pone.0171469.ref001],[@pone.0171469.ref006]\]. Curdlan was approved for use in foodstuffs in Korea, Taiwan and Japan in 1989 and registered in 1996 by the FDA (Food and Drug Administration) in the USA as a food additive with the following functions: a formulation and processing aid, and a stabilizer, thickener and texturizer \[[@pone.0171469.ref008]\]. It is widely used due to its ability to form the thermoirreversible gel, which exhibits great stability during the industrial processes of autoclaving, frying and freezing-thawing cycles. It has no taste, color or odor, can form gels at a wide pH range (from 2 to 10) and can mimic the palatability of foods containing fat \[[@pone.0171469.ref009]\]. It is also considered a dietary fiber, thus improving the functionality of various food products. The versatility of its applications associated with its health benefits make curdlan a valuable tool in the development of innovative food systems \[[@pone.0171469.ref001],[@pone.0171469.ref003]\]. The dispersion and gelation characteristics of curdlan in an aqueous system, as well as its mechanical properties, are associated with several factors such as concentration, temperature, heating time, dispersal method, and the presence of ions, salts and low molecular weight sugars \[[@pone.0171469.ref010]\]. Therefore, the form of recovery from the culture medium and the purification of the polysaccharide are steps which decisively influence the physical and technological properties of curdlan, which directly affect the sensory characteristics of foods in which the polysaccharide is inserted. So, it is necessary to study these properties to define the most appropriate form of the implementation of the polysaccharide in the industry \[[@pone.0171469.ref011]\]. Considering the interest in the properties of dispersion, viscosity, gelation and the health benefits of curdlan, the present study aimed to characterize the physicochemical and technological properties of curdlan produced by Agrobacterium sp. IFO 13140 and recovered using the pre-gelation and precipitation methods, as well as commercial curdlan subjected or otherwise to pre-gelation treatment. The methodologies employed for the recovery of the curdlans were studied. The structures of all curdlans were compared by FT-Raman spectroscopy. In addition, the application of curdlans in yogurts was evaluated. Materials and methods {#sec002} ===================== Materials {#sec003} --------- The bacterial strain Agrobacterium sp. IFO 13140 was purchased in lyophilized form from the Institute for Fermentation of Osaka (Japan). Commercial curdlan was acquired from Wako Pure Chemical Industries, Ltd. (Osaka, Japan). All solvents were of analytical grade. Curdlan production by Agrobacterium sp. IFO 13140 {#sec004} --------------------------------------------------- The culture medium used to reactivate the microorganism was proposed by the supplier (g L^-1^), pH 7: polypeptone (10), yeast extract (2), MgSO~4~.7H~2~O (1). 30 mg of the lyophilized bacteria were incubated in 100 mL of the medium at 30°C and 120 rpm for 48 h. The cells were recovered through centrifugation (6000 ×g, 10 min), washed with 9 g L^-1^ NaCl and transferred to the production medium. For curdlan production, the liquid medium described by Martinez et al. \[[@pone.0171469.ref012]\], pH 7, was used (g L^-1^): glucose (50), KH~2~PO~4~ (2.7), NH~4~Cl (1.6), MgSO~4~ (0.5) and trace elements (10 mL L^-1^). The composition of the trace elements (g L^-1^) in HCl 0.1 mol L^-1^ was: FeCl~3~.6H~2~O (1), MnCl~2~.4H~2~O (1), ZnCl (1), CaCl~2~ (1) and CaCO~3~ (0.03). The reactivated microorganisms were transferred to Erlenmeyer flasks containing 100 mL of the production medium and maintained at 30°C and 150 rpm for 5 days. Recovery of curdlan from production medium {#sec005} ------------------------------------------ Two methodologies were used to recover curdlan from the medium. In pre-gelation method \[[@pone.0171469.ref012],[@pone.0171469.ref013]\], NaOH 3 mol L^-1^ was added to the Erlenmeyer flasks containing the production medium at a ratio of 1.8:1 (NaOH:medium) for curdlan solubilization. This mixture was centrifuged (18000 ×g, 15 min, 4°C) to separate the cells. HCl 3 mol L^-1^ was added to the supernatant until pH 6--7 to obtain the curdlan gel, which was recovered by centrifugation (18000 ×g, 15 min, 4°C). Subsequently, it was washed three times with distilled water and lyophilized. The commercial curdlan was subjected to the same treatment and it was named pre-gelled commercial curdlan. The precipitation method is an adaptation of the industrial purification of curdlan. Industrially produced curdlan is purified by dissolution in a strong alkaline solution and dried in a spray-dryer, then washed with water until neutralization \[[@pone.0171469.ref014]\]. NaOH 3 mol L^-1^ was added to the Erlenmeyer flasks containing the production medium at a ratio of 1.8:1 (NaOH: medium) to solubilize the curdlan and the mixture was centrifuged at 18000 ×g for 15 min at 4°C to separate the cells, before being lyophilized. The dried mixture was resuspended in water, filtered to remove larger impurities and then ultra-filtered using the system described by Fenelon et al. \[[@pone.0171469.ref015]\] with a 30 kDa membrane and pressure of 1 kgf cm^-2^. During ultrafiltration, the material was washed with water until the pH was neutralized, causing the precipitation of the curdlan, that was thereafter lyophilized. Structural analysis of curdlans by FT-Raman {#sec006} ------------------------------------------- Samples of the commercial and commercial pre-gelled curdlans and those produced by Agrobacterium sp. IFO 13140 (pre-gelation and precipitation methods) were analyzed by Fourier transform Raman scattering infrared spectroscopy (FT-Raman) using a Fourier Transform infrared spectrometer (Vertex 70v model with Ram module II, Bruker, Germany) equipped with a Germanium detector cooled with liquid nitrogen. A Nd:YAG laser was used for excitation at 1064 nm. The spectra were based an average of 200 scans with a resolution of 4 cm^-1^. Morphology of curdlans {#sec007} ---------------------- The morphology of the commercial and commercial pre-gelled curdlans, and the curdlans produced by Agrobacterium sp. IFO 13140 (pre-gelation and precipitation methods) was analyzed in a scanning electron microscope (SS-550 model, Superscan, Shimadzu, Japan), at an accelerating voltage of 15 kV. Physicochemical characterization of curdlans {#sec008} -------------------------------------------- The commercial and commercial pre-gelled curdlans and those produced by Agrobacterium sp. IFO 13140 (pre-gelation and precipitation methods) were characterized for their carbohydrate, moisture and sodium content. The carbohydrate content was determined through the phenol-sulfuric method \[[@pone.0171469.ref016]\]. The moisture content was determined by the gravimetric method \[[@pone.0171469.ref017]\] and sodium was identified using atomic absorption spectrometry \[[@pone.0171469.ref018]\]. Technological properties of curdlans {#sec009} ------------------------------------ ### Water dispersion and gel formation capacities of curdlans {#sec010} To verify the dispersal capacity in water of the commercial and commercial pre-gelled curdlans and those produced by Agrobacterium sp. IFO 13140 (pre-gelation and precipitation methods), they were subjected to stirring in a mixer and/or in a magnetic stirrer until homogeneous dispersion was achieved. The dispersions were heated at 95°C/1 h to assess the gel formation capacity. ### Rheological characteristics and gel strength of curdlans {#sec011} Dispersions of the pre-gelled commercial curdlan and the curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) were prepared at three concentrations: 20 g L^-1^, 40 g L^-1^ and 80 g L^-1^ in water. The curdlans were dispersed in water using a mixer at room temperature for 5 min, sonicated for 10 min and then stirred in a magnetic stirrer for 12 h at 40°C. The samples were analyzed in a controlled stress rotational rheometer (HAAKE MARS II model, Thermo Fisher Scientific Inc., Newington, Germany), with steel cone/plate geometry (60 mm, gap 0.052 mm). The elastic (G\') and viscous (G\") modulus, and the apparent viscosity were measured depending on temperature (20--60°C) at a frequency of 10 Hz. For gel strength evaluation, the dispersions of both samples at the concentration of 20 g L^-1^ were kept or not in a water bath at 61°C/1 h to prepare the low-set gel and at 95°C/1 h for preparation of the high-set gel \[[@pone.0171469.ref019]\]. After heating, they were cooled at room temperature. The strength of the suspensions and gels was evaluated in triplicate in a Stable Micro Systems texturometer (TA-XT Plus model, Texture Technologies Corp., UK), using a 36 mm probe for compression analysis. ### Water Holding Capacity (WHC), Oil Holding Capacity (OHC) and Water Solubility Index (WSI) {#sec012} Samples of 0.25 g of the commercial and pre-gelled commercial curdlans and curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) were diluted in 10 mL of distilled water or soya oil at 30°C, homogenized in a magnetic stirrer for 30 min and centrifuged (1500 ×g, 10 min). The WHC was expressed as g of water absorbed per g of curdlan sample and the OHC was described as g of oil absorbed per g of curdlan sample. To determine the WSI, the supernatant of the WRI analysis was oven dried at 105°C and the ratio between the weight of the solid residue present in the supernatant after drying and the weight of the curdlan sample was calculated \[[@pone.0171469.ref011]\]. Application of curdlan in yogurt {#sec013} -------------------------------- ### Preparation of yogurts {#sec014} Eight yogurt samples were prepared: A1) without curdlan or heat treatment; A2) without curdlan and with heat treatment; B1) with commercial curdlan and without heat treatment; B2) with commercial curdlan and heat treatment; C1) with pre-gelled commercial curdlan and without heat treatment; C2) with pre-gelled commercial curdlan and heat treatment; D1) with curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) without heat treatment; D2) with curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) with heat treatment. The curdlan was mixed in UHT (ultra-high temperature) whole milk at a concentration of 15 g L^-1^ using a mixer. Subsequently, the milk of the samples subjected to heat treatment was heated at 90°C for 120 seconds and cooled to 42°C. For samples without heat treatment, the milk was directly heated to 42°C. Next, all samples were inoculated with 2 g L^-1^ of mixed culture of Streptococcus thermophilus and Lactobacillus delbrueckii subsp. bulgaricus, and maintained at 42°C for 6 h. The yogurts were cooled and refrigerated for 24 h for rheological, texture and syneresis analysis \[[@pone.0171469.ref003],[@pone.0171469.ref020]\]. ### Texture Profile Analysis (TPA) {#sec015} The TPA of the yogurts was carried out in a TA-XT Plus texturometer (Stable Micro Systems, Godalming, UK), equipped with Texture Expert software (Stable Micro Systems, Godalming, UK). A 10 mm cylindrical probe was used (ref. P/0.5R, Stable Micro Systems). Two cycles were applied, at a velocity of 2 mm s^-1^ and a depth of 15 mm, producing force-time curves which were used to determine the firmness, cohesiveness, adhesiveness, springiness, gumminess and chewiness of the samples. The analysis was performed in triplicate, at 8°C. ### Syneresis {#sec016} The syneresis analysis of each preparation was performed after 28 days of storage of the yogurt at 8°C in triplicate. Yogurt samples (30--40 g) were centrifuged (222 ×g, 10 min, 4°C) and the supernatant was separated and centrifuged again (222 ×g, 10 min, 4°C), weighed and the syneresis (%) was calculated from the ratio between the mass of the supernatant and the initial mass of the yogurt \[[@pone.0171469.ref020]\]. ### Rheological analysis {#sec017} The rheological properties of flow of the yogurts were analyzed in triplicate, in a controlled stress rotational rheometer (HAAKE MARS II model, Thermo Fisher Scientific Inc., Newington, Germany), with steel cone/plate geometry (60 mm, gap 0.052 mm). The measurements were performed at 8°C. The scan of the deformation rate was performed from 0 to 116 s^-1^, obtaining the outward and return data. The flow and viscosity curves were obtained by determining the stress and viscosity versus the shear rate \[[@pone.0171469.ref020]\]. The rheological parameters K (consistency index) and n (flow behavior index) were calculated using the Ostwald de Waele model, while the parameter τ~0~ (yield stress) was calculated by the Herschel-Bulkley model. The hysteresis was calculated from the area of the flow curves of the yogurts using the RheoWin 4.10.000 software program (HAAKE software, Thermo Fisher Scientific Inc., Newington, Germany). Statistical analysis {#sec018} -------------------- Data were analyzed by analysis of variance (ANOVA), and means were compared with the Tukey Test (p\<0.05) using the Statistica 8.0/2008 software package (Stat Soft, Inc., Tulsa, USA). Results and discussion {#sec019} ====================== Structural analysis of curdlans by FT-Raman {#sec020} ------------------------------------------- [Fig 1](#pone.0171469.g001){ref-type="fig"} shows FT-Raman spectra of commercial and pre-gelled commercial curdlans and those produced by Agrobacterium sp. IFO 13140 (pre-gelation and precipitation methods). From the data, it is verified a structural similarity between all samples, which means that the curdlan produced by the microorganism has a similar structure to commercial curdlan and it also means that the method employed to recover curdlan from the medium did not modify the structure of the polysaccharide. ![FT-Raman spectra of: (A) commercial curdlan, (B) pre-gelled commercial curdlan, (C) curdlan produced by Agrobacterium sp. IFO 13140 (pre-gellation method), (D) curdlan produced by Agrobacterium sp. IFO 13140 (precipitation method).\ The dotted lines show the characteristic peaks of the samples.](pone.0171469.g001){#pone.0171469.g001} Several authors described the chemical bonds present in the structure of curdlan molecule based in FT-IR analysis \[[@pone.0171469.ref003],[@pone.0171469.ref014],[@pone.0171469.ref021]\]. They stated that, from the FT-IR data for curdlan, the bands around 890, 1080 and 1160 cm^−1^ correspond to the β-(1,3)-glucan linkages. Despite the difference in band assignments for FT-IR and FT-Raman spectroscopies, the spectra of carbohydrates in both techniques show a characteristic absorption band of β-anomeric configuration at \~890 cm^−1^ (886 cm^-1^ in this work) \[[@pone.0171469.ref022]\]. The Raman bands and shoulders at 1090 and 1137 cm^−1^ are typical for β-glucans. Intense highly overlapped Raman bands between 990 and 1200 cm^−1^ are attributed to COC and CC stretching vibrations of polysaccharides. The features between 1200 and 1440 cm^-1^ are mainly assigned to in-plane ring deformation including CH and OH bending modes. Finally, the band at 1457 cm^-1^ is assigned to CH~2~ in-plane bending in CH~2~OH of the molecule \[[@pone.0171469.ref023]\]. Morphology of curdlans {#sec021} ---------------------- [Fig 2](#pone.0171469.g002){ref-type="fig"} shows scanning electron microscopy images of the commercial curdlan ([Fig 2A and 2B](#pone.0171469.g002){ref-type="fig"}), the pre-gelled commercial curdlan ([Fig 2C and 2D](#pone.0171469.g002){ref-type="fig"}) and the curdlans produced by Agrobacterium sp. IFO 13140 recovered by pre-gelation and precipitation methods ([Fig 2E--2H](#pone.0171469.g002){ref-type="fig"}) with a range of magnifications. ![Scanning electron microscopy of: commercial curdlan---(A) 40x and (B) 1000x magnification; pre-gelled commercial curdlan---(C) 40x and (D) 1000x magnification; curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method)---(E) 40x and (F) 1000x magnification; curdlan produced by Agrobacterium sp. IFO 13140 (precipitation method)---(G) 40x and (H) 1000x magnification.](pone.0171469.g002){#pone.0171469.g002} Marchessault and Deslandes \[[@pone.0171469.ref024]\] define the shape of the granules of commercial curdlan as collapsed or invaginated, which coincides with the commercial curdlan shape illustrated in [Fig 2A and 2B](#pone.0171469.g002){ref-type="fig"}. The granules have widely differing sizes, varying from around 10 to 100 μm in diameter. Industrially produced curdlan is dried by spray-drying \[[@pone.0171469.ref014]\] and the type of drying has a major influence on the particle bead structure \[[@pone.0171469.ref025]\]. The expected range of particle size of polymers dried by ordinary spray dryers is 5--150 μm. Furthermore, spherical microspheres or those with pores/concavities, as shown in [Fig 2A and 2B](#pone.0171469.g002){ref-type="fig"}, are characteristic of the drying of products by a spray-dryer. The formation of pores or concavities is associated with the rapid evaporation of the liquid particles of this process \[[@pone.0171469.ref026]\]. The pre-gelled commercial curdlan and both the curdlans produced by the microorganism, when viewed at low magnification (40x), were in flake form and displayed irregularities. These characteristics are expected for lyophilized products, since the working conditions of the lyophilizer exert great pressure on the particles to be dried, meaning that a product dried by lyophilization remains amorphous in comparison with a spray dried product \[[@pone.0171469.ref025]\]. Furthermore, at a higher magnification (1000x), it can be seen that the particles obtained for the three samples that underwent the lyophilization process ([Fig 2D, 2F and 2H](#pone.0171469.g002){ref-type="fig"}), were considerably smaller than the commercial curdlan (obtained by spray-dryer). Therefore, it can be inferred that the lyophilization process can facilitate the dispersion of curdlan molecules in water, especially the pre-gelled commercial curdlan and the one produced by Agrobacterium sp. IFO 13140 (pre-gelation method), which have smaller particle sizes ([Fig 2D and 2F](#pone.0171469.g002){ref-type="fig"}). The smaller diameter increases the accessibility of water molecules to the inside of the particles, facilitating dispersion. The structure displayed by the commercial curdlan is maintained by a large amount of hydrogen bonds. But, when added to an alkaline solution, these bonds are broken due to their ionization, and the granule loses its structure. The neutralization of the curdlan suspension causes a reshaping of the hydrogen bonds that depends on the neutralizing agent used. With the addition of HCl (pre-gelation method), the reassociation causes the formation of a firm gel, with small structures of less than 1 μm in diameter ([Fig 2D and 2F](#pone.0171469.g002){ref-type="fig"}). With water (precipitation method), reassociation causes the precipitation of the curdlan into small particles with diameters greater than those obtained by neutralization with HCl, ranging from 2--10 μm in diameter ([Fig 2H](#pone.0171469.g002){ref-type="fig"}). Therefore, the difference between the sizes of the curdlan particles produced by the microorganism using the two methods is due to the different forms of reassociation of the hydrogen bonds, which are dependent on the recovery method employed, and can also influences the characteristics of water dispersion of the polysaccharide. Physicochemical characterization of curdlans {#sec022} -------------------------------------------- As an exopolysaccharide, curdlan is secreted to the extracellular medium in the form of biofilms, and has the advantage of being easy to recover. As a result, the polysaccharide should present a low degree of impurities, and consequently the carbohydrate content represents an indirect measure of the purity of the samples. The levels of carbohydrates and moisture found in the curdlan samples are shown in [Table 1](#pone.0171469.t001){ref-type="table"}. 10.1371/journal.pone.0171469.t001 ###### Carbohydrate, moisture and sodium content (%) of different samples of curdlan. Values indicate mean ± standard-deviation. ![](pone.0171469.t001){#pone.0171469.t001g} Curdlan sample Carbohydrate Moisture Sodium ----------------------------------------------------------------------- -------------- ----------- --------------- Commercial 94.5 ± 0.2 4.1 ± 0.4 1.2 ± 0.3 Commercial pre-gelled 82 ± 1 4.0 ± 0.3 4.4 ± 0.8 Microbial[\](#t001fn001){ref-type="table-fn"} (pre-gelation method) 70.6 ± 0.6 5.4 ± 0.9 0.024 ± 0.001 Microbial[\](#t001fn001){ref-type="table-fn"} (precipitation method) 85.4 ± 0.7 4.8 ± 0.6 \< 0.001 \Produced by Agrobacterium* sp. IFO 13140. The highest carbohydrate content was found for the commercial sample. The pre-gelled commercial sample had a carbohydrate content lower than the commercial curdlan due to the salt (NaCl) incorporated in the pre-gelling process. The two samples produced by Agrobacterium sp. IFO 13140 had different carbohydrate contents, with that produced by precipitation method having greater purity due to the successive washings performed. The impurities in the last samples consist of remnants of the production medium and the salt formed in the recovery step. The sodium content of curdlans was determined because curdlan is a glucose polymer with a low amount of inorganic salts, mainly sodium chloride \[[@pone.0171469.ref027]\]. Thus, sodium content is an indirect estimate of the salt content of the material. The sodium concentrations disposed in [Table 1](#pone.0171469.t001){ref-type="table"} directly influence the dispersion and gelling characteristics of the polysaccharide, which will be further discussed in the next section. Technological properties {#sec023} ------------------------ ### Water dispersibility and gelling capacities {#sec024} Curdlan can be used as a thickener, stabilizer and texturizer in food industry \[[@pone.0171469.ref008],[@pone.0171469.ref027]\]. However, as curdlan is insoluble in water, it is necessary to use an efficient homogenizer to obtain a homogeneous dispersion and apply a further stirring of the dispersion prior to the carrying out of analyses requiring uniformity \[[@pone.0171469.ref021]\]. The pre-gelled commercial curdlan and the curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) dispersed easily in water when subjected to stirring in a mixer for less than 5 min, and also acted as thickening agents. Both dispersions remained visibly homogeneous for days and, when heated at 95°C/1 h, formed a firm and homogeneous gel. The commercial curdlan did not easily disperse in water using the mixer. After 24 hours of stirring in a magnetic stirrer, a homogeneous dispersion formed for a brief period, displaying a phase separation after 10 min of rest. After 48 h of stirring, phase separation began to be observed at 30 min; after 72 h, at 1 hour; and after 96 h, phase separation began at 1.5 h. The gelling of commercial curdlan also occurred after heating at 95°C/1 h, but there was phase separation in the gel formed, which became more pronounced as the stirring period for formation of the dispersion was reduced. Marchessault and Deslandes \[[@pone.0171469.ref024]\] have previously identified the formation of a non-homogenous gel from curdlan in its native form. Probably, the difficulty of dispersion of commercial curdlan compared with those submitted to pre-gelation process is related to the particle size of the sample (as noted in morphological analysis), which makes difficult the accessibility of water molecules to the inside of the particles. Another important factor related to the dispersion and gelling of curdlan is the presence of salts. In the gelling of curdlan, the swelling phase of the molecule is essential and may be influenced by salts, as they affect the mobility of the water molecules during hydration \[[@pone.0171469.ref010]\]. The swelling of curdlan is facilitated with the salts present because it increases intermolecular association and consequently increases the viscosity of curdlan water dispersions. The curdlan produced by Agrobacterium sp. IFO 13140 (precipitation method) displayed greater difficulty in forming a homogenous dispersion in water and, in contrast to the other curdlans, did not form a gel when heated. This can be explained by its considerably lower sodium content, due to the countless washings with water used to achieve neutralization. Samples of the other curdlans were also washed in the ultrafiltration device and, following such process, no longer had the same dispersion and gelling capacities. This proves therefore, that as with other biopolymers, for curdlan, the rheological characteristics depend excessively on the methods of recovery employed following production. Due to the characteristics displayed, the technological characteristics of the curdlan produced by Agrobacterium sp. IFO 13140 (precipitation method) were not evaluated, and this sample was not applied in yogurt. Additionally, dispersions of commercial curdlan were not evaluated by rheological and gel strength analysis because the commercial curdlan did not act as a thickener with the methodology employed for preparation of the dispersions. ### Rheological characteristics and gel strength of curdlans {#sec025} Thermal scanning rheological measurements were made to evaluate both the apparent viscosity and the gelation behavior of curdlan dispersions. The temperature dependence of apparent viscosity and of G\' and G\" modulus for aqueous dispersions of pre-gelled commercial curdlan and of the curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) at 20, 40 and 80 g L^-1^ are shown in [Fig 3A and 3B](#pone.0171469.g003){ref-type="fig"}. ![Temperature dependency of: (A) apparent viscosity and (B) G\' (continuous line) and G\" (dotted line) modulus of the aqueous dispersions of curdlan.\ Pre-gelled commercial curdlan (empty symbol) and curdlan produced by Agrobacterium sp. IFO 13140 recovered by the pre-gelation method (full symbol) at (■) 20, (●) 40 and (▲) 80 g L^-1^.](pone.0171469.g003){#pone.0171469.g003} The curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) exhibited considerably higher viscosity than pre-gelled commercial curdlan ([Fig 3A](#pone.0171469.g003){ref-type="fig"}). For example, at 20°C, the apparent viscosity of curdlan produced by the microorganism was almost 100% higher than pre-gelled commercial curdlan at 20 and 80 g L^-1^. Equally, G\' and G\" modulus ([Fig 3B](#pone.0171469.g003){ref-type="fig"}) were also higher for the curdlan produced by the microorganism, revealing the highest thickening potential of this curdlan when compared to the pre-gelled commercial one. Both the apparent viscosity and the G\' module of both curdlans exhibited a different behavior as a function of temperature for each concentration employed. At 20 g L^-1^, both parameters increased with temperature for pre-gelled comercial curdlan, while for curdlan produced by the microorganism they decreased. This behavior is related to the swelling of curdlan because of the breakage of hydrogen bonds during heating. The difference between the curdlanas is because, for pre-gelled commercial curdlan, inter-molecular entanglements between the particles that were swollen formed pseudo-links, which contributed to the little increase in G\'. This phenomenon did not occur for curdlan produced by the microorganism. However, for higher concentrations (40 and 80 g L^-1^), the decrease in the viscosity and in G\' no longer occurred with temperature for any sample, because the high concentration favors the inter-molecular entanglements between the particles, which contribute to the increase of both parameters. Additionally, the increase in G\' and G\" modulus and in apparent viscosity of curdlan dispersions at about 50--55°C, especially at 80 g L^-1^, suggests the beginning of the formation of the thermo-irreversible gels due to hydrophobic interactions. Jin et al. \[[@pone.0171469.ref021]\] stated that G\' decreased until 50°C for curdlan suspensions at 20 g L^-1^, which is related to the swelling of curdlan because of the breakage of hydrogen bonds, as observed for curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) in this study. However, Funami et al. \[[@pone.0171469.ref028]\] stated only increasing in G\' parameter for dispersions of curdlans at 20, 40 and 100 g L^-1^ with increasing temperature. The authors attributed this behavior to inter-molecular entanglements between the particles that were swollen forming pseudo-links, being this result similar to the one obtained for pre-gelled comercial curdlan in all concentrations evaluated and for curdlan produced by the microorganism for concentrations from 40 to 80 g L^-1^ in this work. The strength data of suspensions and gels of the pre-gelled commercial curdlan and the curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) at 20 g L^-1^ are presented in [Table 2](#pone.0171469.t002){ref-type="table"}. According to the strength values of the suspensions and gels, the curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) has a better thickening capacity than the pre-gelled curdlan, as the suspension of the same without heat treatment was approximately 10% stronger. This result agrees with the results of rheological analysis. However, the gelling capacity of the pre-gelled commercial curdlan is considerably higher. 10.1371/journal.pone.0171469.t002 ###### Strength (×10^−3^ N) of pre-gelled commercial curdlan and the curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) samples after undergoing different heat treatments. Values indicate mean ± standard-deviation. ![](pone.0171469.t002){#pone.0171469.t002g} Heat treatment of the dispersion in a concentration of 20 g L^-1^ Strength (× 10^−3^ N) ------------------------------------------------------------------- -------------------------------------------------- ---------------------------------------------- Without heat treatment 69 ± 2[^d^](#t002fn001){ref-type="table-fn"} 76 ± 2[^c^](#t002fn001){ref-type="table-fn"} Low-set gel (61°C/1 h) 79 ± 1[^c^](#t002fn001){ref-type="table-fn"} 77 ± 2[^c^](#t002fn001){ref-type="table-fn"} High-set gel (95°C/1 h) 96.8 ± 0.8[^a^](#t002fn001){ref-type="table-fn"} 83 ± 1[^b^](#t002fn001){ref-type="table-fn"} ^a--d^ Means with different letters are significantly different (p\< 0.05). The strength of the low-set gel of the pre-gelled commercial curdlan was around 15% greater than that of the suspension without heat treatment, and statistically equal to that of the low-set gel of the curdlan produced by the microorganism. The high-set gel of the pre-gelled commercial curdlan had strength 40% greater than the suspension without heat treatment, and 17% greater than the high-set gel produced by the microorganism. When an aqueous dispersion of curdlan is heated around 55°C, the low-set gel formed is maintained by intramolecular hydrogen bonds and the curdlan chains adopt a predominantly single helix conformation. But when heated above 80°C, the high-set gel formed is maintained by intermolecular hydrophobic interactions and the curdlan chains adopt a predominantly triple helix conformation \[[@pone.0171469.ref014],[@pone.0171469.ref029]\], forming an organized and rigid gel configuration. As such, the increase in the strength of the high-set curdlan gel (95°C/1 h) is due to the greater presence of the triple helix conformation. The fact that the curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) have formed a considerably weaker gel than the pre-gelled commercial curdlan, as well as the fact that it does not exhibit the same behavior that pre-gelled commercial sample in rheological analysis at low concentrations (increase of G\' and apparent viscosity with temperature), are probably due to a difference of molecular weight/degree of polymerization of the polymers. A variety of physical properties of β-(1→3)-glucans, including gel strength (or gel-forming ability) is related to the molecular weight/degree of polymerization of the biopolymer. The higher the degree of polymerization of the polysaccharide, the greater is its gel forming ability with heating \[[@pone.0171469.ref030],[@pone.0171469.ref031]\]. Nakanishi et al. \[[@pone.0171469.ref032]\] studied the formation of a complex of curdlan with aniline blue dye and noted a relation between the rate of interaction of the polymer with the dye and the concentration, degree of polymerization, and gel-forming ability of the polymer. The variation in absorbance at 590 nm is proportional to the concentration of curdlan and to its gel-forming ability (and consequently to its degree of polymerization). [S1 Appendix](#pone.0171469.s001){ref-type="supplementary-material"} contains the methodology used by the authors described to evaluate curdlan interaction with aniline blue dye and the results of the relationship between absorbance variation and concentration of the pre-gelled commercial curdlan and curdlan produced by the microorganism. By comparing the results of this study to those obtained by Nakanishi et al. \[[@pone.0171469.ref032]\] and using the relation obtained by these authors between absorbance variation and degree of polymerization, it is concluded that the degree of polymerization of the pre-gelled commercial curdlan is about 45% higher than of the curdlan produced by the microorganism, which explains the lowest gel strength and the decrease in apparent viscosity and G\' in low concentration dispersions of the polymer. ### Water Holding Capacity (WHC), Oil Holding Capacity (OHC) and Water Solubility Index (WSI) {#sec026} Determining the technological properties of polysaccharides is of great importance when predicting their possible industrial applications. The values for the properties of water solubility index (WSI) water holding capacity (WHC) and oil holding capacity (OHC) of the curdlan samples are described in [Table 3](#pone.0171469.t003){ref-type="table"}. 10.1371/journal.pone.0171469.t003 ###### Water Holding Capacity (WHC), Oil Holding Capacity (OHC) and Water Solubility Index (WSI) (g g^-1^) of the commercial, pre-gelled commercial curdlans and those produced by Agrobacterium sp. IFO 13140 (pre-gelation method). Values indicate mean ± standard-deviation. ![](pone.0171469.t003){#pone.0171469.t003g} Curdlan sample WHC OHC WSI --------------------------------- --------------------------------------------------- --------------------------------------------------- ------------------------------------------------------ Commercial 4.6 ± 0.4[^a^](#t003fn001){ref-type="table-fn"} 0.62 ± 0.08[^c^](#t003fn001){ref-type="table-fn"} 0.006 ± 0.002[^a^](#t003fn001){ref-type="table-fn"} Commercial pre-gelled 2.20 ± 0.08[^c^](#t003fn001){ref-type="table-fn"} 4.4 ± 0.2[^b^](#t003fn001){ref-type="table-fn"} 0.00923 ± 7E-5[^a^](#t003fn001){ref-type="table-fn"} Microbial (pre-gelation method) 3.6 ± 0.3[^b^](#t003fn001){ref-type="table-fn"} 8.7 ± 0.1[^a^](#t003fn001){ref-type="table-fn"} 0.0068 ± 0.003[^a^](#t003fn001){ref-type="table-fn"} ^a--c^ Means in the same column with different letters are significantly different (p\< 0.05). The three curdlans had very low and statistically equal WSIs, which was expected as curdlan is insoluble in water. Despite this, however, the curdlans presented some water absorption, with the commercial curdlan achieving the highest value, followed by the curdlan produced by the microorganism and the pre-gelled commercial curdlan. Seguchi and Kusunose \[[@pone.0171469.ref033]\] found water absorption rates for curdlan between 5.244 g g^-1^ and 7.724 g g^-1^. Compared to other polysaccharides employed in food industry as gums, curdlan has low water holding capacity; the guar and xanthan gums have a WHC of 25.77 g g^-1^ and 27.33 g g^-1^, respectively \[[@pone.0171469.ref034]\]. However, both are soluble in water. The low solubility and water holding capacity of curdlans is explained by the existence of a large amount of intra/intermolecular hydrogen bonds within the polymer. This also explains the fact that commercial curdlan has a higher water absorption index than the other types as the recovery by the pre-gelation process employed in the latter two types favors the formation of large quantities of hydrogen bonds in the polymer. Thus, the polysaccharide interacts more strongly with itself than with water \[[@pone.0171469.ref021],[@pone.0171469.ref034]\]. The three evaluated curdlans also differed with respect to oil holding capacity. With the exception of the commercial curdlan, they presented greater OHC values than WHC, indicating that the samples have a higher lipolytic than hydrophilic capacity. Oil absorption values greater than three make curdlan a potentially useful ingredient in structural interactions in foods, especially in aroma retention, improved palatability and maintenance of the stability of the product during storage \[[@pone.0171469.ref011]\]. This result justifies the use of curdlan as a fat replacer or mimicker of fat in the food industry \[[@pone.0171469.ref009],[@pone.0171469.ref035]\]. Application of curdlan in yogurt {#sec027} -------------------------------- The main characteristics that define the quality of yogurts are its texture and propensity for serum separation (syneresis). Typically, polysaccharides such as xanthan, guar, gellan, pectin, carrageenan are used to give the product a firmer texture, increase its stability and hence make it more acceptable to the consumer \[[@pone.0171469.ref003],[@pone.0171469.ref020]\]. [Table 4](#pone.0171469.t004){ref-type="table"} displays the parameters of texture and syneresis after 28 days of storage of yogurt samples with and without curdlan, submitted or not to heat treatment with the aim of gelling the curdlan in the milk prior to the fermentation process. 10.1371/journal.pone.0171469.t004 ###### Texture parameters and syneresis of yogurts with and without curdlan and submitted or not to heat treatment. Values indicate mean ± standard-deviation. ![](pone.0171469.t004){#pone.0171469.t004g} Sample Firmness (× 10^−3^ N) Cohesiveness Adhesiveness (N × mm) Springiness (mm) Gumminess (N) Chewiness (N) Syneresis (%) -------- --------------------------------------------------- ----------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------------- -------------------------------------------------- A1 77 ± 2[^c^](#t004fn001){ref-type="table-fn"} 0.77 ± 0.02[^a^](#t004fn001){ref-type="table-fn"} 0.02 ± 0.01[^c^](#t004fn001){ref-type="table-fn"} 1.13 ± 0.01[^a^](#t004fn001){ref-type="table-fn"}[^b^](#t004fn001){ref-type="table-fn"} 5.9 ± 0.3[^c^](#t004fn001){ref-type="table-fn"} 6.7 ± 0.4[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"}[^d^](#t004fn001){ref-type="table-fn"} 49.3 ± 0.8[^a^](#t004fn001){ref-type="table-fn"} A2 84 ± 8[^c^](#t004fn001){ref-type="table-fn"} 0.71 ± 0.07[^a^](#t004fn001){ref-type="table-fn"}[^b^](#t004fn001){ref-type="table-fn"} 0.04 ± 0.03[^c^](#t004fn001){ref-type="table-fn"} 1.00 ± 0.04[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 5.9 ± 0.1[^c^](#t004fn001){ref-type="table-fn"} 5.9 ± 0.3[^c^](#t004fn001){ref-type="table-fn"}[^d^](#t004fn001){ref-type="table-fn"}[^e^](#t004fn001){ref-type="table-fn"} 48.7 ± 0.4[^a^](#t004fn001){ref-type="table-fn"} B1 83 ± 8[^c^](#t004fn001){ref-type="table-fn"} 0.66 ± 0.02[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 0.08 ± 0.03[^c^](#t004fn001){ref-type="table-fn"} 0.94 ± 0.04[^c^](#t004fn001){ref-type="table-fn"} 5.5 ± 0.7[^c^](#t004fn001){ref-type="table-fn"} 5.1 ± 0.7[^e^](#t004fn001){ref-type="table-fn"} 45.6 ± 0.9[^b^](#t004fn001){ref-type="table-fn"} B2 129 ± 7[^b^](#t004fn001){ref-type="table-fn"} 0.47 ± 0.04[^d^](#t004fn001){ref-type="table-fn"} 0.4 ± 0.1[^b^](#t004fn001){ref-type="table-fn"} 0.97 ± 0.05[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 6.1 ± 0.2[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 5.9 ± 0.4[^c^](#t004fn001){ref-type="table-fn"}[^d^](#t004fn001){ref-type="table-fn"}[^e^](#t004fn001){ref-type="table-fn"} 38.2 ± 0.3[^c^](#t004fn001){ref-type="table-fn"} C1 94 ± 9[^c^](#t004fn001){ref-type="table-fn"} 0.65 ± 0.02[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 0.16 ± 0.04[^c^](#t004fn001){ref-type="table-fn"} 1.2 ± 0.1[^a^](#t004fn001){ref-type="table-fn"} 6.0 ± 0.6[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 7.1 ± 0.3[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 26.4 ± 0.8[^d^](#t004fn001){ref-type="table-fn"} C2 125.1 ± 0.6[^b^](#t004fn001){ref-type="table-fn"} 0.58 ± 0.02[^c^](#t004fn001){ref-type="table-fn"} 0.58 ± 0.02[^a^](#t004fn001){ref-type="table-fn"}[^b^](#t004fn001){ref-type="table-fn"} 1.01 ± 0.01[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 7.2 ± 0.2[^b^](#t004fn001){ref-type="table-fn"} 7.3 ± 0.2[^b^](#t004fn001){ref-type="table-fn"} 29.2 ± 0.7[^d^](#t004fn001){ref-type="table-fn"} D1 128 ± 6[^b^](#t004fn001){ref-type="table-fn"} 0.46 ± 0.02[^d^](#t004fn001){ref-type="table-fn"} 0.68 ± 0.04[^a^](#t004fn001){ref-type="table-fn"} 0.96 ± 0.03[^c^](#t004fn001){ref-type="table-fn"} 5.8 ± 0.3[^c^](#t004fn001){ref-type="table-fn"} 5.6 ± 0.5[^d^](#t004fn001){ref-type="table-fn"}[^e^](#t004fn001){ref-type="table-fn"} 28.8 ± 0.2[^d^](#t004fn001){ref-type="table-fn"} D2 160 ± 7[^a^](#t004fn001){ref-type="table-fn"} 0.59 ± 0.01[^c^](#t004fn001){ref-type="table-fn"} 0.5 ± 0.1[^a^](#t004fn001){ref-type="table-fn"}[^b^](#t004fn001){ref-type="table-fn"} 0.99 ± 0.06[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 9.5 ± 0.6[^a^](#t004fn001){ref-type="table-fn"} 9.3 ± 0.3[^a^](#t004fn001){ref-type="table-fn"} 35.8 ± 0.9[^c^](#t004fn001){ref-type="table-fn"} ^a--e^ Means within the same column with different letters are significantly different (p\< 0.05). Yogurts: A1) without curdlan or heat treatment; A2) without curdlan and with heat treatment; B1) with commercial curdlan and without heat treatment; B2) with commercial curdlan and heat treatment; C1) with pre-gelled commercial curdlan and without heat treatment; C2) with pre-gelled commercial curdlan and heat treatment; D1) with curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) without heat treatment; D2) with curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) with heat treatment. It can be seen that the use of curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) produced significant alterations in the parameters of firmness, cohesiveness and adhesiveness of the yogurt, even without being subjected to heat treatment. This is due to the thickening potential of the material. Heat treatment caused major changes in these parameters, both for the yogurts with commercial curdlan and pre-gelled commercial curdlan, although the highest values for firmness, stickiness, gumminess and chewiness were obtained for the yogurt with curdlan produced by the microorganism. Heat treatment promotes the formation of a firm gel due to the intermolecular hydrophobic interactions that structure the system, making the yogurt more consistent, resulting in greater difficulty to separate in the mouth and making it denser during chewing, which therefore takes longer. It was notable that the pre-gelled commercial curdlan formed a stronger gel in water, but had less effect on the texture parameters than that produced by Agrobacterium sp. IFO 13140 (pre-gelation method) when applied to the yogurt, which was not expected. Thus, it is likely that the latter has a greater ability to interact with water and the other components of milk, especially proteins, more efficiently stiffening the protein network formed after the fermentation process. As described before, the curdlan produced by the microorganism has the capacity to absorb 64% more water and 98% more oil than the pre-gelled commercial curdlan, with these parameters being important for products such as yogurts prepared with whole milk (3% fat). The use of commercial curdlan without heat treatment did not produce many changes in the texture parameters or in the syneresis of the yogurts. This is because it was not effectively homogenized in the milk as it dispersed with greater difficulty, interfering little in the formation of the protein network during fermentation. It was found that syneresis was reduced for the yogurt samples with curdlan, particularly the pre-gelled commercial variety and the curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method). The latter displayed higher syneresis after the first 28 days of storage because its gel is less stable than the pre-gelled commercial curdlan. As such, the pre-gelled commercial curdlan demonstrated a greater ability to preserve the structure of the yogurt, avoiding rearrangements in the casein network due to the time of storage and avoiding whey expulsion. Martinez et al. \[[@pone.0171469.ref003]\] produced yogurts with added curdlan and achieved a significant reduction in the syneresis of the products compared to the control (without curdlan). The authors found that the high stability obtained by the yogurt with curdlan is attributed to interactions between the curdlan molecules, either between one another or with the proteins, which cause the formation of a more compact and continuous three-dimensional protein network, which effectively entraps the protein molecules and water in its structure. Importantly, there was no difference between any of the texture parameters and the syneresis of yogurts without curdlan submitted or not to heat treatment. This was expected as the aim of treatment was to gel the curdlan and, therefore, the A2 sample was not submitted to rheological analysis. [Fig 4](#pone.0171469.g004){ref-type="fig"} shows the flow curves ([Fig 4A and 4B](#pone.0171469.g004){ref-type="fig"}) and the viscosity curves ([Fig 4C and 4D](#pone.0171469.g004){ref-type="fig"}) for the yogurts produced with curdlan with and without heat treatment of the milk prior to fermentation. ![Flow curves of yogurt samples without heat treatment (A) with heat treatment (B); viscosity curves of yogurt samples without heat treatment (C) and with heat treatment (D).\ Yogurt without curdlan (--), with commercial curdlan (Δ), with pre-gelled commercial curdlan (○) and with curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) (■). The direction of the gray arrows indicates the ascendant and descendant curves.](pone.0171469.g004){#pone.0171469.g004} All the yogurts behaved as non-Newtonian pseudo-plastic fluids, as their viscosity decreased due to the shear rate applied. It can be seen that the samples subjected to heat treatment of the milk had higher viscosity values across the entire shear rate analyzed due to the gelling of the curdlan. However, the decrease in viscosity versus the shear rate in the return data was much more significant for these samples. All the samples also exhibited thixotropic characteristics, due to the difference of tension and viscosity between the ascending and descending shear rate curves. The hysteresis observed is due to the breakdown of the gel structure formed by coagulation of the protein during fermentation, in the presence and absence of curdlan. Hysteresis is measured as the area between the ascending and descending curves, where the greater the area (when positive), the greater the thixotropic effect. Hysteresis, together with the other rheological parameters, are described in [Table 5](#pone.0171469.t005){ref-type="table"}. 10.1371/journal.pone.0171469.t005 ###### Rheological parameters of yogurt samples with and without curdlan and submitted or not to heat treatment. Hysteresis, consistency index (K), flow behavior index (n) and yield stress (τ~0~). Values indicate mean ± standard-deviation. ![](pone.0171469.t005){#pone.0171469.t005g} Sample K (mPa s^n^) n τ~0~ (Pa) Hysteresis (Pa s^-1^) ---------- ------------------------------------------------- ----------------------------------------------------- -------------------------------------------------- ------------------------------------------------- A 98 ± 3[^d^](#t005fn001){ref-type="table-fn"} 0.997 ± 0.006[^a^](#t005fn001){ref-type="table-fn"} 2 ± 1[^e^](#t005fn001){ref-type="table-fn"} 362 ± 43[^e^](#t005fn001){ref-type="table-fn"} B1 223 ± 26[^d^](#t005fn001){ref-type="table-fn"} 0.938 ± 0.001[^b^](#t005fn001){ref-type="table-fn"} 12 ± 1[^d^](#t005fn001){ref-type="table-fn"} 1088 ± 75[^d^](#t005fn001){ref-type="table-fn"} B2 1685 ± 18[^a^](#t005fn001){ref-type="table-fn"} 0.662 ± 0.004[^e^](#t005fn001){ref-type="table-fn"} 30 ± 4[^c^](#t005fn001){ref-type="table-fn"} 2241 ± 39[^b^](#t005fn001){ref-type="table-fn"} C1 922 ± 35[^b^](#t005fn001){ref-type="table-fn"} 0.73 ± 0.04[^d^](#t005fn001){ref-type="table-fn"} 12 ± 1[^d^](#t005fn001){ref-type="table-fn"} 1134 ± 92[^d^](#t005fn001){ref-type="table-fn"} C2 ---[\](#t005fn002){ref-type="table-fn"} ---[\](#t005fn002){ref-type="table-fn"} 36 ± 2[^b^](#t005fn001){ref-type="table-fn"} 2200 ± 49[^b^](#t005fn001){ref-type="table-fn"} D1 467 ± 56[^c^](#t005fn001){ref-type="table-fn"} 0.87 ± 0.02[^c^](#t005fn001){ref-type="table-fn"} 10.9 ± 0.5[^d^](#t005fn001){ref-type="table-fn"} 1623 ± 36[^c^](#t005fn001){ref-type="table-fn"} D2 ---[\](#t005fn002){ref-type="table-fn"} ---[\](#t005fn002){ref-type="table-fn"} 75 ± 3[^a^](#t005fn001){ref-type="table-fn"} 3461 ± 10[^a^](#t005fn001){ref-type="table-fn"} ^a--e^ Means within the same column with different letters are significantly different (p\< 0.05). \It was not possible to calculate the parameters by the Oswald de Waele model. In [Fig 4B](#pone.0171469.g004){ref-type="fig"}, the heat-treated samples of yogurt with pre-gelled commercial curdlan and curdlan produced by the microorganism displayed a peak in shear strain at the beginning of the growth of the deformation rate. This peak is related to the lack of homogeneity of the product, because during the gelling process of curdlan small lumps are formed, which are not necessarily noticeable to the palate, but which alter rheology. With the presence of these lumps, the tension needed for a small deformation is high, and as a result, it was not possible to calculate the K and n parameters for samples C2 and D2 by the Oswald de Waele model. From the data of [Table 5](#pone.0171469.t005){ref-type="table"} it is clear that the higher viscosity samples in [Fig 4C and 4D](#pone.0171469.g004){ref-type="fig"}, which are those subjected to heat treatment, had more evident thixotropic characteristics (higher hysteresis), as they underwent a major reduction in apparent viscosity with time, in a rate constant with shearing. This is due to the breakdown of the organized yogurt structure when submitted to a determined strain. After remained stationary, the samples returned to their original state more quickly, but with lower viscosity during the return, being this recovery dependent on time. Therefore, the gel formed with heat treatment presented low stability, in particular the gel of the curdlan produced by the microorganism, corroborating with the syneresis data of the yogurt. Curiously, it was observed that of the samples without heat treatment, that which had the greater viscosity in the increasing shear rate curve did not present higher K and n values, or in other words, the sample with pre-gelled commercial curdlan was greater than the curdlan produced by Agrobacterium* sp. IFO 13140 (pre-gelation method) in the consistency index and flow behavior. This was due to the sharp drop in viscosity in the decreasing curve of the latter, revealing again the lower stability of the gel formed. The results of the yield strength parameter (τ~0~) were consistent with the firmness of the yogurt, as the yogurt with curdlan produced by Agrobacterium sp. IFO 13140 with thermic treatment withstood the greatest tension before suffering deformation, followed by the yogurts with pre-gelled commercial curdlan and commercial curdlan with heat treatment. It is important to note that yield stress refers to the maximum stress that the material can bear before yielding (in the elastic deformation regime) and that the greatest τ~0~ values are those of the firmer samples. These samples displayed a minimum stress for deformation that was more difficult to break due to the increased organization and stiffening of the protein network formed, originating from the intermolecular hydrophobic interactions of the curdlan with heating. Conclusions {#sec028} =========== The characteristics of dispersion, gelation and the rheological properties of curdlan depended greatly on the recovery methods employed after its production, as the curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) and the commercial pre-gelled curdlan dispersed better in water, acted as thickeners and formed more homogeneous gels. The use of the pre-gelation method exerts a major influence on the size of the particles obtained and in the presence of NaCl, which contributes significantly to the dispersion and gelation characteristics described. However, the recovery method employed did not influence the structure of the polysaccharide. Even the pre-gelled commercial curdlans and those produced by the microorganism through the pre-gelation method had significantly different technological properties. The curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) showed a greater thickening and water and oil holding capacity than the pre-gelled commercial curdlan, while the latter demonstrated a considerably greater gelling capacity, that is related to degree of polymerization of the polysaccharide. As a consequence, although both types of curdlan increased firmness, viscosity and reduced the syneresis of yogurts, the curdlan produced by Agrobacterium sp. IFO 13140 (pre-gelation method) caused a greater increase in the parameters, but also showed a less stable gel formation than the pre-gelled commercial curdlan. The commercial curdlan did not influence the firmness and syneresis of yogurts tested. Nevertheless, the curdlans recovered by the pre-gelation method provided rigid gels with a stable structure, allowing improvements in the texture of a range of products, and therefore, has great potential for application in the food industry. Supporting information {#sec029} ====================== ###### Curdlan interaction with aniline blue dye. (DOC) ###### Click here for additional data file. The authors are thankful to the Brazilian Agencies CAPES, CNPq and Fundação Araucária for their financial support of this work. [^1]: Competing Interests:The authors have declared that no competing interests exist. [^2]: Conceptualization: CSM GM.Data curation: CSM TTS VCF LNK SBSF.Formal analysis: CSM VCF MLB.Funding acquisition: GM MLB.Investigation: CSM GM.Methodology: CSM MLB GM.Project administration: GM.Resources: MLB GM.Supervision: CSM GM.Validation: CSM GM.Visualization: CSM GM.Writing -- original draft: CSM GM.Writing -- review & editing: MLB GM.	{ "pile_set_name": "PubMed Central" }
The content published in Cureus is the result of clinical experience and/or research by independent individuals or organizations. Cureus is not responsible for the scientific accuracy or reliability of data or conclusions published herein. All content published within Cureus is intended only for educational, research and reference purposes. Additionally, articles published within Cureus should not be deemed a suitable substitute for the advice of a qualified health care professional. Do not disregard or avoid professional medical advice due to content published within Cureus. Introduction ============ Enchondroma is a benign cartilaginous lesion commonly found in small bones of the hands and feet \[[@REF1]\]. It is common and affects all age groups. Generally being asymptomatic, it is discovered incidentally during an unrelated radiographic examination or in case of a pathological fracture. Proximal locations such as proximal humerus and scapula are extremely rare sites \[[@REF2]\]. Radiographs show a lytic lesion with intralesional calcification. Asymptomatic lesions can be managed conservatively with follow-up and serial radiographs. If the lesion grows or if it becomes symptomatic, extended curettage usually is curative \[[@REF3]\]. We report the case of a six-year-old male child with enchondroma of the acromion, which, to the best of our knowledge, has never been published. Case presentation ================= A six-year-old male child presented to our orthopaedics department with the complaints of pain and swelling over the right shoulder for six months. Both pain and swelling were of insidious onset and had gradually progressed. Pain was of dull-aching type. This was not associated with radiation of the pain or any history of trauma, fever or night pains. On clinical examination, there was a globular, tender swelling, about 7x7 cm in size with firm consistency, present over the posterolateral corner of the right shoulder. Overlying skin was normal and there were no dilated veins, scar marks or sinuses. There was no local rise in temperature, lymphadenopathy, distal neurovascular deficit or any other swelling elsewhere in the body. Blood investigations showed normal counts and kidney and liver functions. Serum calcium, erythrocyte sedimentation rate, C-reactive protein, alkaline phosphatase and parathyroid hormone were normal. X-ray of the right shoulder showed an expansile lytic lesion in the acromion process of scapula with no periosteal reaction or soft tissue component. CT scan showed an expansile lytic lesion with incomplete septations and a cortical breach. MRI showed the lesion to be hypointense in T1 images and hyperintense in T2 images (Figure [1](#FIG1){ref-type="fig"}). A skeletal survey did not reveal any other lesion. ![Preoperative radiographic assessment\ (A) X-ray showing an expansile lytic lesion in the acromion; (B) CT scan showing cortical breach and incomplete septations; (C) MRI showing the lesion to be hypointense on T1 and hyperintense on T2.](cureus-0012-00000007630-i01){#FIG1} The differential diagnoses of simple bone cyst, aneurysmal bone cyst and chondroblastoma were kept in mind. A core biopsy from the posterolateral corner of the shoulder was done. Straw coloured fluid within the bony cavity was found. Biopsy specimen was sent for histopathological examination, which revealed lobules of hypocellular cartilage separated by fibroconnective stroma, features suggestive of enchondroma. Due to persistent pain in the subsequent follow-ups, an extended curettage with hydrogen peroxide and fibular bone grafting of the lesion was done (Figure [2](#FIG2){ref-type="fig"}). The patient was discharged on the third day after wound inspection. Sutures were removed on day 14, and passive range of motion exercises of the shoulder were started as tolerated by the patient. ![Extended curettage and fibular grafting\ (A) Posterolateral surgical approach; (B) postoperative X-ray.](cureus-0012-00000007630-i02){#FIG2} At one-year follow-up, the patient remained symptom-free, having complete and painless range of motion of the shoulder joint (Figure [3](#FIG3){ref-type="fig"}). The patient was explained about the need of a long-term regular follow-up. ![Clinical assessment at one-year follow-up\ (A) Well-healed surgical scar; (B) full elevation at the right shoulder.](cureus-0012-00000007630-i03){#FIG3} Discussion ========== Chondroma is a benign cartilaginous lesion, and is called an enchondroma when arising from the medullary canal. It is called a perisosteal or juxtacortical chondroma when it arises from the surface of the bone \[[@REF4]\]. Enchondroma is a benign tumour of the bone containing hyaline cartilage. Multiple enchondromatosis is known as Ollier's disease, and when associated with hemangioma of overlying soft tissues, it referred to as Maffuci's syndrome \[[@REF5]\]. It is the most common tumour of the hands and feet. Proximal femur, pelvis, proximal humerus and scapula are rare sites. Proximal sites are more prone to recurrence and malignant change (chondrosarcoma), and hence should be monitored regularly \[[@REF6]\]. Various lesions arising from the acromion, such as osteochondroma, simple bone cyst (SBC), aneurysmal bone cyst (ABC), giant cell tumour with secondary ABC, chondroblastoma, metastasis and rarely multiple myeloma, have been described in the literature. Our six-year-old patient with expansile lytic lesion of the acromion had differential diagnoses of SBC, ABC, and chondroblastoma. Both SBCs and ABCs present as lytic lesions, but ABCs are blood-filled cavities with septations and fluid-fluid levels on MRI scan \[[@REF7]\]. Chondroblastoma is a rare, benign tumour with predilection for proximal humerus, but has also been reported in the acromion \[[@REF8]\]. Our case was proved to be a solitary enchondroma on biopsy. It could be managed conservatively with serial radiographs, but an extended curettage and fibular bone grafting was done due to persistent pain and its proximal location which made it prone to a malignant change. Various other options of treatment for an acromial tumour such as en bloc excision with or without stabilisation have been described \[[@REF9]\]. Conclusions =========== Enchondroma represents a rare differential diagnosis of a lytic lesion of the acromion and must be kept in mind. Extended curettage and bone grafting in a symptomatic patient yields excellent clinical outcomes. A long-term and regular follow-up is required to monitor for recurrence or malignant transformation. The authors have declared that no competing interests exist. Consent was obtained by all participants in this study	{ "pile_set_name": "PubMed Central" }
INTRODUCTION ============ Standard of care for suspicious microcalcifications in the breast is stereotactic-guided percutaneous biopsy or mammography (MG)-guided wire localization and surgical excision ([@B1]). Stereotactic biopsy, however, requires a sufficient breast volume to guarantee a safe procedure limiting its universal application. Ultrasound (US)-guided biopsy has many advantages over the MG-guided procedure including no ionizing radiation, greater patient comfort, no need for breast compression, real-time confirmation of needle location, lower cost, and universally available sonography equipment ([@B1]-[@B4]). Improvements in the resolution of sonographic equipment have led to the use of US-guided biopsy for histologic assessment of suspicious microcalcifications in the breast those are apparent on sonography ([@B1], [@B5]-[@B12]). Due to its higher sensitivity and lower rate of common sampling errors, US-guided vacuum-assisted biopsy (VAB) has been preferred to 14-gauge (G) automated core needle biopsy (CNB) ([@B5], [@B6], [@B10]-[@B12]). 14-G semi-automated CNB is a coaxial system similar to VAB in terms of the single skin puncture and repeated needle passes. Using an introducer, radiologists can precisely obtain contiguous tissue specimens and reduce the procedure time. Additionally, 14-G semi-automated CNB device is less costly than the VAB device. The purpose of the present study was to evaluate the rate of successful retrieval of microcalcifications and diagnostic accuracy of US-guided 14-G semi-automated CNB for suspicious microcalcifications in the breast those were apparent on sonography. MATERIALS AND METHODS ===================== Between January 2008 and September 2011, 138 lesions containing isolated suspicious microcalcifications without an associated mass or asymmetry were detected on contact MG. Based on findings from magnified images, each lesion was categorized as suspicious (category 4) or highly suggestive of malignancy (category 5) according to the American College of Radiology Breast Imaging-Reporting and Data System (BI-RADS) ([@B13]). Category 4 lesions were subcategorized as low (A), intermediate (B), or moderate (C) suspicious lesions. To select between US-guided 14-G semi-automated CNB and MG-guided wire localization and surgical excision, targeted US was performed for all suspicious calcifications. We prospectively determined the biopsy method and excluded lesions that were not apparent on US. In addition, lesions located in the posterior portion of the mammary zone of very thin patients who had obliterated retromammary fat layer were excluded to avoid complications such as pneumothorax. Finally, 33 consecutive lesions were selected for US-guided 14-G semi-automated CNB and were enrolled into the study. Written informed consent for US-guided 14-G semi-automated CNB was obtained from all patients and institutional review board approval was obtained from our institute. Targeted US (iU22®; Philips, Bothell, WA, USA) examination was performed using a linear array transducer (L12-5) to identify a correlating lesion containing microcalcifications. Unilateral orthogonal mammographic views were obtained with a skin marker only when there was uncertainty about the mammography-sonography correlation. The biopsy was performed using a STERICUT® 14-G semi-automated CNB set (TSK Laboratory, Tochigi, Japan). One breast radiologist with 10 years of experience performed both breast US and US-guided 14-G semi-automated CNB in all cases. For biopsy, the skin was punctured using an introducer with a stylet and advanced forward near the margin of the lesion. The stylet was removed and a biopsy gun was inserted through the introducer. The biopsy needle was pushed manually into the lesion. The location of the needle within the lesion was confirmed on transverse and longitudinal scans. The biopsy gun was fired, causing it to follow the biopsy needle and remove tissue from the lesion. The biopsy gun was then withdrawn through the introducer and the needle was pushed manually. The specimen could then be observed in the notch. Insertion of the biopsy gun through the introducer was repeated after changing the angle of insertion. Several cores of more than 1 cm length were obtained for each lesion. A 1- or 2-cm cutting edge excursion could be used; we usually selected a 2-cm excursion. After the biopsy, the retrieved cores were placed on wet paper or a plastic weigh boat and specimen radiography (Senograph DS®; GE, Buc, France) was performed to identify the presence of microcalcifications. The parameters of exposure for specimen radiography were 22 kVp, 8 mAs, and 1.8 × magnification. A core containing any calcification on specimen radiography was defined as a successfully retrieved core. The total number of retrieved cores, the number of successfully retrieved cores containing any calcification, and the total number of calcifications on specimen radiography were enumerated. Concordance between the MG findings of microcalcifications and the results of US-guided 14-G semi-automated CNB was reviewed. Radiology-pathology concordance was considered when a BI-RADS category 4 or higher lesion showed a malignant result on 14-G semiautomated CNB. There were two exceptions for category 4A lesions. In low suspicious lesions, 14-G semi-automated CNB results showing a specific benign diagnosis, such as fibroadenoma or fat necrosis, were considered to be concordant. Also, nonspecific benign results, such as fibrocystic changes (FCC), were considered to be concordant in low suspicious lesions. In contrast, for category 4B or higher lesions, nonspecific benign results were considered to be discordant because of the possible sampling error. Additionally, we reviewed the accuracy of US-guided 14-G semi-automated CNB of microcalcifications by comparing the results with surgical pathology. When both results were the same, the 14-G semi-automated CNB was considered accurate. If the 14-G semi-automated CNB result of ductal carcinoma in situ (DCIS) yielded a diagnosis of invasive ductal carcinoma (IDC) in a subsequent surgery, it was deemed a DCIS underestimate. If the 14-G semi-automated CNB result of atypical ductal hyperplasia (ADH) or ductal intraepithelial neoplasia (DIN) 1B yielded a diagnosis of DCIS in a subsequent surgical excision, it was deemed an ADH underestimate. Mean lesion size, morphology, and distribution of microcalcifications on magnified images, and associated features of calcifications, such as mass or non-mass lesions on US, were analyzed. These factors were compared between the successful and failed retrieval groups. Biopsy-related complications including infection, hemorrhage, and pneumothorax were checked by radiology reports and chart review. The chi-square test and Fisher\'s exact test were performed for nonparametric variables and Student\'s t test was performed for parametric interference. Statistical analyses were performed using SPSS 17.0 software (SPSS Inc., Chicago, IL, USA). p values \< 0.05 were considered to indicate statistical significance. RESULTS ======= Thirty-three (23.9%, 33/138) lesions in 32 patients were subjected to US-guided 14-G semi-automated CNB. One patient had suspicious microcalcifications in both breasts. Other suspicious microcalcifications underwent MG-guided wire localization excision (76.1%, 105/138). Mean patient age was 49 years (range, 24-67 years). Thirty patients were asymptomatic and were detected on screening MG. One patient had bloody nipple discharge and the other had palpable cancer in the upper inner quadrant and incidental suspicious microcalcifications in the lower inner quadrant of the same breast. Magnified images identified four low suspicious lesions (4A; 15.2%), five intermediate suspicious lesions (4B; 12.1%), eight moderate suspicious lesions (4C; 24.2%), and 16 highly suggestive of malignancy lesions (5; 48.5%) ([Table 1](#T1){ref-type="table"}). Calcifications were successfully retrieved in 30/33 lesions (90.9%) and confirmed on specimen radiography. No complications related to 14-G semi-automated CNB were observed. Thirty cases of successfully retrieved microcalcifications were assigned to the successful retrieval group and the remaining three lesions were assigned to the failed retrieval group. Results of US-guided 14-G semi-automated CNB identified 26 (78.8%) malignant and seven (21.2%) benign lesions ([Fig. 1](#F1){ref-type="fig"}). The successful retrieval group comprised 25 malignant (15 DCIS and 10 IDC) and five benign lesions. The five benign lesions were one ADH (or DIN 1B), one columnar cell hyperplasia (CCH), one FCC, one fibroadenoma, and one fat necrosis. In the failed retrieval group, one DCIS and two FCC were identified. After 14-G semi-automated CNB, 30 lesions underwent cancer surgery (n = 26) or repeated biopsy after MG-guided wire localization (n = 4), and final pathology results were provided ([Fig. 1](#F1){ref-type="fig"}). Of the 16 DCIS on 14-G semi-automated CNB, 14 were confirmed to be DCIS on final pathology ([Fig. 2](#F2){ref-type="fig"}). The remaining two DCIS, which belonged to the successful retrieval group, were upgraded to IDC. The rate of DCIS underestimates was 12.5% (2/16) and their BIRADS assessments were category 4C and 5, respectively. All 10 IDC on 14-G semi-automated CNB were confirmed to be IDC on final pathology. Of the seven benign lesions, four underwent MG-guided wire localization and excision, and were finally diagnosed as three FCC and one DCIS. The latter was the only false-negative case of 14-G semi-automated CNB and belonged to the failed retrieval group. The other three benign lesions underwent mammographic follow-up for 2 years and no missed cancers were found. There was no ADH underestimate. According to the BI-RADS categories, three of four category 4A calcifications were successfully retrieved on US-guided 14-G semi-automated CNB. The results were all benign (one fibroadenoma, one fat necrosis, and one FCC) and concordant ([Table 1](#T1){ref-type="table"}). None of the successfully retrieved benign lesions showed a missed cancer after 2 years of mammographic follow-up. One low suspicious lesion diagnosed as DCIS on biopsy, despite failed retrieval of microcalcifications, was finally confirmed to be DCIS after surgery, confirming the 14-G semi-automated CNB results to be accurate. Four of five category 4B calcifications were successfully retrieved and 14-G semi-automated CNB results were all malignant (one DCIS and three IDC) and concordant ([Table 1](#T1){ref-type="table"}). All lesions were finally diagnosed as one DCIS and three IDC after breast cancer surgery, confirming the 14-G semi-automated CNB results to be accurate. Microcalcification retrieval failed in one intermediate suspicion lesion. Although the 14-G semi-automated CNB result was FCC and considered to be discordant, the lesion was finally diagnosed as FCC after MG-guided wire localization excision, confirming the 14-G semi-automated CNB result to be accurate. All of the category 4C lesions were successfully retrieved and the 14-G semi-automated CNB results were one ADH (or DIN 1B), four DCIS, and three IDC. The ADH case was considered discordant, but was finally confirmed to be FCC after MG-guided wire localization excision, confirming the 14-G semi-automated CNB result to be accurate. All of the moderate suspicious lesions except one were finally confirmed to be malignant after breast cancer surgery. In one case among them, however, the final pathology result was upgraded to IDC, confirming the 14-G semi-automated CNB result to be a DCIS underestimate. Of the 16 category 5 calcifications, 15 were successfully retrieved ([Table 1](#T1){ref-type="table"}). The 14-G semi-automated CNB results in the successful retrieval group were 10 DCIS, four IDC, and one CCH. Of the category 5 lesions, 14 were finally diagnosed as malignancy after surgery. In one case, however, the final pathology result was upgraded to IDC, confirming the 14-G semi-automated CNB result to be a DCIS underestimate. The CCH case was considered to be discordant, but was finally confirmed to be FCC after MG-guided wire localization excision, confirming the 14-G semi-automated CNB result to be accurate ([Fig. 3](#F3){ref-type="fig"}). Retrieval of calcification failed in one highly suspicious lesion and its 14-G semi-automated CNB result was FCC, which was considered to be discordant. The case was finally diagnosed as DCIS after MG-guided wire localization and surgical excision, confirming the 14-G semi-automated CNB result to be a false-negative. Of the 30 lesions in the successful retrieval group, two (6.7%) showed a radiology-pathology discordance, whereas two of the three (66.7%) lesions in the failed retrieval group showed discordance. The discordance rate was higher in the failed retrieval group (p = 0.035). Of the 30 lesions in the successful retrieval group, 28 (93.3%) lesions were accurately diagnosed, and two of the three (66.7%) lesions in the failed retrieval group were accurately diagnosed. The accuracy of 14-G semi-automated CNB was not significantly different (p = 0.256). The overall accuracy of US-guided 14-G semi-automated CNB was 90.9% (30/33) because of two DCIS underestimates and one false-negative case. We compared the characteristics of specimen radiography according to the retrieval of calcifications ([Table 2](#T2){ref-type="table"}). The total number of retrieved cores did not differ between groups: 6.5 (range, 4-9) in the successful retrieval group and 7.3 (range, 6-10) in the failed retrieval group (p = 0.61). In the successful retrieval group, the mean number of cores containing calcification was 4.8 (range, 1-8) and the mean total number of calcifications in the retrieved cores was 24 (range, 2-101). We reviewed imaging findings of microcalcifications, including the extent of the calcification, morphology, and distribution on MG, and associated features on US ([Table 3](#T3){ref-type="table"}). Based on the MG findings, the extent of calcification did not differ between two groups (16.7 mm \[range, 3-70 mm\] in the successful retrieval group and 18.0 mm \[range, 9-30 mm\] in the failed retrieval group, p = 0.093). Punctate (66.7%, 2/3) calcifications were more common in the failed retrieval group than in the successful retrieval group (p = 0.033). Linear/branching (33.3%, 10/30) and coarse heterogeneous (16.7%, 5/30) calcifications were more common in the successful retrieval group. The proportions of linear/segmental and clustered distributions were similar in both groups (36.7% vs. 33.3%; 60.0% vs. 66.7%; p = 0.095). On US, microcalcifications were associated with mass (63.3%; 19/30) or non-mass (30.0%; 9/30) lesions in the successful retrieval group, whereas they were associated with mass (100%, 3/3) in the failed retrieval group. DISCUSSION ========== For suspicious microcalcifications in the breast those are either not palpable or not associated with a mass on MG, standard of care is stereotactic-guided VAB or MG-guided wire localization and surgical excision ([@B1], [@B2], [@B9], [@B14]). Stereotactic equipment, however, is not universally applicable. Further, an adequate breast thickness is essential for safe procedures ([@B15]). As high-resolution sonography equipment is universally available, US-guided VAB can be an effective alternative to stereotactic-guided VAB for suspicious calcifications in the breast ([@B10]-[@B12]). VAB is superior to automated CNB for the retrieval of microcalcifications because the former provides larger tissue specimens and contiguous sampling ([@B1]-[@B3], [@B5], [@B6], [@B9]-[@B12], [@B14]). On the other hand, using a 14-G semi-automated CNB device, radiologists can precisely obtain multiple cores of calcifications because the introducer reduces reverberation artifacts and the possibility of missing the target ([@B16]). The cost of 14-G semi-automated CNB is lower than that of VAB. Using an introducer in the semi-automated gun does not significantly add to the cost. Thus, US-guided 14-G semi-automated CNB may be an acceptable and cost-effective procedure for suspicious microcalcifications compared with US-guided VAB. In the present study, the retrieval rate of microcalcifications using the semi-automated gun was 91%. The retrieval rates of microcalcifications in VAB are reported to be 91-100% under stereotactic guidance ([@B1], [@B2]) and 71-97% under US guidance ([@B10]-[@B12]). Thus, in the retrieval of microcalcifications, US-guided 14-G semi-automated CNB is comparable to VAB, particularly considering the smaller tissue specimens and lower cost. The retrieval rate of microcalcifications using an automated gun is 86-94% ([@B1], [@B2]) and the results of the present study are comparable to US-guided automated CNB. Interestingly, the retrieval rate in the present study was superior to that in a report of US-guided VAB (91% vs. 71%) ([@B11]). This difference may be related to the exclusion criteria used in the present study. We excluded calcifications that were located in the posterior portion of the mammary zone in very thin patients who had obliterated retromammary fat layer to avoid pneumothorax. Cho et al. ([@B11]) reported that microcalcifications located in the anterior or middle part of the breast were retrieved more frequently than those in the posterior part of the breast. The lower retrieval of the latter likely reflected the difficulty in inserting a biopsy needle to a sufficiently deep position because of fear of an iatrogenic pneumothorax. So, the retrieval rate of posteriorly located calcifications was lower than that of other calcifications. The accuracy of 14-G semi-automated CNB was 91%, due to the two DCIS underestimates and one false-negative diagnosis. The accuracies of US-guided VAB were 94-96% ([Table 4](#T4){ref-type="table"}) ([@B10]-[@B12]). These excellent results, however, came from a mean of 10-17 VAB cores, which represent larger tissue specimens compared with stereotactic-guided VAB (mean, 11-13 cores) ([@B1], [@B2]) and a mean of seven cores in the present study. Histologic underestimation by 14-G semi-automated CNB is mostly an issue in calcified lesions rather than in mass lesions, and DCIS underestimates are more common in 14-G semi-automated CNB than that of VAB due to sampling error ([@B2]). In stereotactic-guided biopsy of microcalcifications, the rate of DCIS underestimates is reported to be 16-37% with 14-G semi-automated CNB and 0-23% with VAB ([@B2], [@B9], [@B14]). In US-guided biopsy of microcalcifications, the rate of DCIS underestimates ranges from 7-19% with VAB ([Table 4](#T4){ref-type="table"}) ([@B10]-[@B12]). Hahn et al. ([@B12]) reported a rate of DCIS underestimates of 0%, but an ADH (or DIN 1B) underestimation of 33% (1/3). In the present study, the rate of DCIS underestimates of 14-G semi-automated CNB was 13% (2/16), which is comparable to that of US-guided VAB and there was no ADH underestimate. One false-negative diagnosis was identified. The MG finding in that case was segmental fine pleomorphic calcifications and was assessed as a BI-RADS category 5. On sonography, a few calcifications with subtle non-mass lesions were identified. US-guided VAB or MG-guided wire localization excision would be a more desirable procedure for the particular case. Fortunately, the discordant 14-G semi-automated CNB result led the patient to act quickly on the diagnosis of breast cancer. The false-negative rate of the present study was 3.7% (1/27), which is comparable to those of US-guided automated CNB (mean, 2.8%; range, 0.3-8.2%) and MG-guided wire localization excision (mean, 2%; range, 0-8%) ([@B1]). Breast microcalcifications are observed on sonography as distinct echogenic dots without posterior acoustic shadowing and are more readily identified when associated with a hypoechoic mass or duct-like structures than when no associated sonographic abnormality is present ([@B1], [@B7]). In the present study, 31 of 33 microcalcifications visible on sonography were associated with a mass or nonmass lesions on sonography. The remaining two lesions of isolated calcifications on sonography were retrieved successfully and accurately diagnosed. Although the number of cases was too small to deduce a conclusive inference, it could be suggested that isolated calcifications without associated lesions could be successfully retrieved using 14-G semi-automated CNB in the era of high-resolution sonography. This study has some limitations. First, the number of enrolled cases in this study was small, which may have limited a conclusive inference. Second, all procedures were performed by one experienced breast radiologist who had used 14-G semi-automated CNB for many years. The proficiency of 14-G semi-automated CNB can affect the retrieval rate of microcalcifications. Third, US-guided 14-G semi-automated CNB was not applicable to all suspicious microcalcifications; only 24% of microcalcifications were considered amenable in the current study, similar to the findings of other studies ([@B1], [@B5], [@B9]). Nevertheless, in selected cases, US-guided semi-automated 14-G CNB can provide a reliable and cost-effective diagnosis of suspicious microcalcifications. Last, we did not deploy marker clips after US-guided biopsy because they are not generally used in Korea. In conclusion, US-guided 14-G semi-automated CNB provided a retrieval rate and diagnostic accuracy of 90.9% with a DCIS underestimate rate of 13%. Thus, this could be useful procedure for suspicious isolated microcalcifications apparent on sonography. This work was supported in part by the Soonchunhyang University Research Fund. ![US-guided 14-G semi-automated core needle biopsy results and pathologic diagnosis of microcalcifications.\ ^\^Magnification mammography, ^†^Microcalcifications. DCIS = ductal carcinoma in situ, IDC = invasive ductal carcinoma, US = ultrasound, ADH = atypical ductal hyperplasia, DIN = ductal intraepithelial neoplasia, CCH = columnar cell hyperplasia, FCC = fibrocystic changes](kjr-15-12-g001){#F1} ![Details from case of 54-year-old asymptomatic woman.\ A.* Mammography shows clustered fine pleomorphic calcifications (arrows). BI-RADS assessment is category 5. B. Targeted US shows multiple echogenic dots (arrows) within irregular indistinct isoechoic mass. C. Specimen radiography after US-guided 14-G semi-automated CNB shows multiple microcalcifications (arrows) in five of six cores retrieved. Pathology results of both CNB and final surgery were DCIS. BI-RADS = Breast Imaging-Reporting and Data System, US = ultrasound, CNB = core needle biopsy, DCIS = ductal carcinoma in situ](kjr-15-12-g002){#F2} ![Details from case of 42-year-old asymptomatic woman.\ A. Magnification mammography shows compactly clustered fine linear and fine pleomorphic calcifications (arrows). BI-RADS assessment is category 5. B. Targeted US shows multiple microcalcifications (arrows) within oval partially indistinct hypoechoic mass. C. Specimen radiography shows 69 microcalcifications in five (arrows) of six cores retrieved. Pathology result of CNB was columnar cell hyperplasia with microcalcifications and that of mammography-guided localization and surgical excision was fibrocystic change with microcalcifications. BI-RADS = Breast Imaging-Reporting and Data System, US = ultrasound, CNB = core needle biopsy](kjr-15-12-g003){#F3} ###### Results of US-Guided 14-G Semi-Automated CNB According to BI-RADS Categories ![](kjr-15-12-i001) Note.-US = ultrasound, MG = mammography, CNB = core needle biopsy, BI-RADS = Breast Imaging-Reporting and Data System, 4A = low suspicious lesions, 4B = intermediate suspicious lesions, 4C = moderate suspicious lesions ###### Characteristics of Specimen Radiography According to Retrieval of Calcifications ![](kjr-15-12-i002) Note.-^\^Microcalcifications. No = number ###### Imaging Findings of Microcalcifications According to Retrieval ![](kjr-15-12-i003) Note.-BI-RADS = Breast Imaging-Reporting and Data System, US = ultrasound ###### Comparison of Performance in US-Guided Biopsies for Suspicious Microcalcifications ![](kjr-15-12-i004) Note.-US = ultrasound, VAB = vacuum-assisted biopsy, DCIS U-E = ductal carcinoma in situ* underestimation, NA = not applicable, A-CNB = automated core needle biopsy, SA-CNB = semi-automated core needle biopsy	{ "pile_set_name": "PubMed Central" }
Ju L, Shan L, Yin B, Song Y. δ‐Catenin regulates proliferation and apoptosis in renal cell carcinoma via promoting β‐catenin nuclear localization and activating its downstream target genes. Cancer Med. 2020;9:2201--2212. 10.1002/cam4.2857 1. INTRODUCTION {#cam42857-sec-0001} =============== Renal cell carcinoma (RCC) is the ninth most common malignancy, with 403 262 new cases diagnosed and 175 098 cancer deaths worldwide in 2018.[1](#cam42857-bib-0001){ref-type="ref"}, [2](#cam42857-bib-0002){ref-type="ref"} In the management of localized or locally advanced RCC, surgical resection including partial and radical nephrectomy is a common treatment strategy.[3](#cam42857-bib-0003){ref-type="ref"}, [4](#cam42857-bib-0004){ref-type="ref"} However, up to 40% risk of developing recurrence after nephrectomy.[4](#cam42857-bib-0004){ref-type="ref"}, [5](#cam42857-bib-0005){ref-type="ref"} For patients with metastatic RCC, cytoreductive nephrectomy followed by systemic drugs is an established treatment approach.[6](#cam42857-bib-0006){ref-type="ref"}, [7](#cam42857-bib-0007){ref-type="ref"} Nevertheless, the disease still progresses after therapy and metastasis is the major cause of their death.[1](#cam42857-bib-0001){ref-type="ref"} Therefore, treatments for the recurrence and metastasis of RCC are clearly needed to improve current therapies. δ‐Catenin, alternatively known as CTNND2 or NPRAP (neural plakophilin‐related armadillo protein), is a member of the catenin family. It is the adhesive junction‐associated protein that was originally identified to be enriched in brain tissues.[8](#cam42857-bib-0008){ref-type="ref"}, [9](#cam42857-bib-0009){ref-type="ref"}, [10](#cam42857-bib-0010){ref-type="ref"}, [11](#cam42857-bib-0011){ref-type="ref"} Recent researches have heightened that δ‐catenin is also implicated in some cancers. It is evidenced that δ‐catenin facilitates the malignant phenotype of non‐small‐cell lung cancer through binding to the juxtamembrane domain (JMD) of E‐cadherin.[12](#cam42857-bib-0012){ref-type="ref"} Beyond this, two research articles indicated that ectopic overexpression of δ‐catenin is related to prostate cancer progression. One reported δ‐catenin causes the alteration of cell cycle and survival gene profiles. The other presented the mechanism that δ‐catenin induces E‐cadherin processing and activates β‐catenin‐mediated oncogenic signals.[13](#cam42857-bib-0013){ref-type="ref"}, [14](#cam42857-bib-0014){ref-type="ref"} To date, the precise role of δ‐catenin in tumorigenesis and development of RCC is still poorly understood, neither is the correlation between its expression and clinical pathological parameters. All these deserve to be explored. β‐Catenin, a crucial member of the adherent junctions,[15](#cam42857-bib-0015){ref-type="ref"} is located at adhesion complexes at the cytoplasmic side of membrane in normal cells.[16](#cam42857-bib-0016){ref-type="ref"} β‐Catenin is phosphorylated in the cytoplasm by a multiprotein destruction complex, which leads to proteasome‐mediated degradation.[17](#cam42857-bib-0017){ref-type="ref"} In pathological conditions, due to mutations on β‐catenin sequence, blocking the activity of the destruction complex or preventing β‐catenin from being phosphorylated by GSK3‐3β, β‐catenin may stabilize and escape from its degradation fate.[18](#cam42857-bib-0018){ref-type="ref"} This results in elevation of cytoplasmic β‐catenin and its translocation to the nucleus. In the nucleus, β‐catenin interacts with transcription factors from the T‐cell factor (TCF)/Lymphoid enhancer‐binding factor (LEF) family,[19](#cam42857-bib-0019){ref-type="ref"}, [20](#cam42857-bib-0020){ref-type="ref"}, [21](#cam42857-bib-0021){ref-type="ref"} which triggers transcription of Wnt/β‐catenin target genes.[22](#cam42857-bib-0022){ref-type="ref"}, [23](#cam42857-bib-0023){ref-type="ref"} The aberrant activation of these genes involved in proliferation, apoptosis and invasion contributes to cancer progression.[24](#cam42857-bib-0024){ref-type="ref"}, [25](#cam42857-bib-0025){ref-type="ref"}, [26](#cam42857-bib-0026){ref-type="ref"} Moreover, it has been demonstrated that β‐catenin can be activated in RCC development,[27](#cam42857-bib-0027){ref-type="ref"}, [28](#cam42857-bib-0028){ref-type="ref"} suggesting a possible dysregulation of this protein. However, the regulation of β‐catenin by δ‐catenin and the corresponding mechanism have not been studied yet. In the present study, we detected the expression of δ‐catenin in RCC specimens by Western blot analysis and immunohistochemistry staining, and then analyzed the correlation between its expression and clinicopathological factors. To gain insights into the role of δ‐catenin in RCC, we explored whether it was implicated in proliferation and apoptosis of RCC cells. In addition, the regulation of δ‐catenin on β‐catenin and β‐catenin‐mediated oncogenic signals were also investigated. Moreover, we determined the influence of δ‐catenin on tumor growth in the xenograft mouse model. Our results illuminated that elevated δ‐catenin expression in RCC caused the activation of β‐catenin and its target genes, thereby affecting proliferation and apoptosis. This cognition of δ‐catenin may help to establish a novel targeted therapy and improve current therapies of RCC. 2. MATERIALS AND METHODS {#cam42857-sec-0002} ======================== 2.1. Cell lines and cell culture {#cam42857-sec-0003} -------------------------------- Normal human renal epithelial cells (HK‐2) and human renal cancer cell lines (A498, ACHN and 786‐O) were obtained from Procell Life Science & Technology (Wuhan, China) and cultured in MEM medium or RPMI‐1640 medium supplemented with 10% fetal bovine serum (BI). All the cells were maintained in a humidified incubator with 5% CO2 at 37°C. 2.2. Nuclear and cytoplasm protein extraction {#cam42857-sec-0004} --------------------------------------------- The cells were lysed in 200 μL cytoplasmic lysis buffer A supplemented with PMSF on ice for 10‐15 minutes, and then added with 10 μL cytoplasmic lysis buffer B. The mixture was vortexed for 5 seconds and centrifuged at 12 000 × g for 5 minutes at 4°C. The supernatant was a cytoplasmic extract. The pellet was resuspended with 50 μL nuclei lysis buffer (Beyotime, P0028) and kept in ice for 30 minutes (vortexed 15‐30 seconds at a high speed every 1‐2 minutes). Nuclear proteins were extracted by centrifugation at 12 000 × g for 10 minutes at 4°C. The supernatant was the nuclear extract. The protein concentration was assessed by Enhanced BCA Protein Assay Kit (Beyotime, P0009). 2.3. Western blot analysis {#cam42857-sec-0005} -------------------------- Cells were collected and lysed with RIPA lysis buffer (Beyotime, P0013B). The protein concentrations were evaluated via Enhanced BCA Protein Assay Kit (Beyotime). Protein lysates were separated by electrophoresis on a sodium dodecyl sulfate‐polyacrylamide gel (8%‐15%), transferred onto PVDF membranes, and incubated with primary antibodies at 4°C overnight followed by goat anti‐rabbit or goat anti‐mouse IgG (Proteintech) at 37°C for 40 minutes. The protein bands were visualized by enhanced chemiluminescence (ECL). Primary antibodies used in this study were as follows: anti‐δ‐Catenin antibody (Bioss, bs‐22251R), anti‐Cyclin D1 antibody (CST, \#2922), anti‐CDK6 antibody (Proteintech, 19117‐1‐AP), anti‐CDK4 antibody (CST, \#12790), anti‐c‐myc antibody (Proteintech, 10828‐1‐AP), anti‐Bcl2L1 antibody (Proteintech, 10783‐1‐AP), anti‐cleaved‐caspase‐3 antibody (CST, \#9661), anti‐survivin antibody (Proteintech, 10508‐1‐AP), anti‐β‐catenin antibody (CST, \#8480), anti‐β‐actin antibody (Proteintech, 60008‐1‐Ig), anti‐Histone H3 antibody (Proteintech, 17168‐1‐AP). 2.4. Plasmid and siRNA transfection {#cam42857-sec-0006} ----------------------------------- A498 and ACHN cells were transfected by δ‐Catenin‐specific siRNA (5ʹ‐CCCUAGCAGUACUGACCAATT‐3ʹ and 5ʹ‐UUGGUCAGUACUGCUAGGGTT‐3ʹ) or siCtrl (5ʹ‐UUCUCCGAACGUGUCACGUTT‐3ʹ and 5ʹ ‐ACGUGACACGUUCGGAGAATT‐3ʹ) encapsulated with Lipofectamine 2000 (Invitrogen, 11668‐019) following its transfection protocol. After 48 hours of transfection, subsequent assays were performed as described. ACHN cells were transfected by shRNA plasmids or shCtrl plasmids as the negative control with Lipofectamine 2000 for silencing δ‐catenin. After 24 hours, the cells were selected in media supplemented with 300 μg/mL G418 to obtain the stably transduced cells. 2.5. RNA extraction and quantitative real‐time PCR (qRT‐PCR) {#cam42857-sec-0007} ------------------------------------------------------------ Total RNA was isolated from RCC cells using total RNA rapid extraction kit (BioTeke, RP1201) according to the manufacturer\'s protocol. The obtained RNA was reverse transcribed into cDNA by M‐MLV (Takara, 2641A). The cDNA was used to perform real‐time PCR with Taq HS Perfect Mix (Takara, R300A) and SYBR Green (BioTeke, EP1602). The data were determined by Exicycler™ 96 (Bioneer, Daejeon, Korea), calculated by 2^−∆∆Ct^ parameters. Sequences of real‐time PCR primers were as follows: δ‐Catenin: forward: 5ʹ‐TCTGAGAAACCTGGTGTATGG‐3ʹ, reverse: 5ʹ‐CAGTCGTCTTGCGGAGTAA‐3ʹ, β‐actin: forward: 5ʹ‐CCACTGCCGCATCCTCTT‐3ʹ, reverse: 5ʹ‐GGTCTTTACGGATGTCAACG‐3ʹ. 2.6. CCK‐8 assay {#cam42857-sec-0008} ---------------- Cell counting kit‐8 (CCK‐8) assay was employed to detect cell viability. RCC cells were seeded into 96‐well plates with 4 × 10^3^ per well beforehand. After being adhered, the cells were transfected with siCtrl and δ‐catenin siRNA; and CCK‐8 reagent (Sigma) was added into the 96‐well plates with 10 µL per well and incubated for 1 hour. The optical density (OD) value was measured at 450 nm with a microplate reader (BioTek). The OD450 values at 0, 24, 48, 72, and 96 hours were tested, respectively. 2.7. Immunofluorescence {#cam42857-sec-0009} ----------------------- The slides of cells were fixed for 15 minutes with 4% paraformaldehyde and permeabilized with 0.1% tritonX‐100 for 30 minutes. The samples were blocked by goat serum for 15 minutes. Primary antibody (diluted to 1:200 with PBS) such us Ki67 and β‐catenin was added to the cells and incubated overnight at 4°C. After PBS washing, cells were incubated with Cy3‐labeled goat anti‐rabbit IgG (Beyotime, A0516) diluted in PBS (1:200) for 1 hours at room temperature. The nuclei were counterstained by DAPI. Finally, the samples were sealed by anti‐fade reagent and observed using the OLUMPUS fluorescence microscope (at 400 × magnification). 2.8. Flow cytometry {#cam42857-sec-0010} ------------------- Flow cytometry was performed to detect the cell cycle and apoptosis of RCC cells. After transfection for 48 hours, cells were collected, treated by Cell Cycle Analysis kit (Beyotime) following the manufacturer\'s protocol, and detected using NovoCyte flow cytometer (ACEA Biosciencs). For the apoptosis analysis, the cells were treated with Annexin V‐FITC Apoptosis Detection Kit (Beyotime) and accessed through flow cytometer. 2.9. TOP/FOP luciferase reporter assay {#cam42857-sec-0011} -------------------------------------- ACHN cells transfected with δ‐catenin siRNA or siCtrl were seeded in 6‐well plates. After incubation for 24 hours, the cells were transfected with 2 μg β‐catenin responsive firefly luciferase reporter plasmid TOPflash or negative control FOPflash and 0.5 μg constitutively active vector encoding Renilla luciferase using lipofectamine 2000. At 24 hours after transfection, the cells were lysed and then firefly and Renilla luciferase activities were determined using the dual luciferase reporter assay system (Promega) according to the manufacturer\'s instructions. 2.10. Animal studies {#cam42857-sec-0012} -------------------- All animal experiments in this study were approved by the Institutional Animal Ethics Committee of China Medical University, and were carried according to the Guideline for the Care and Use of Laboratory Animals. Male Balb/c nude mice at the age of 6 weeks were divided into two groups randomly (6 mice per group). ACHN cells (1 × 10^7^) with stable knockdown of δ‐catenin with shRNA plasmids or ACHN cells transfected with shCtrl plasmids were injected subcutaneously into the right flank of the nude mice. The tumor size was measured every four days, and the tumor volume was calculated according to the following: V = (width^2^ × length)/2. The mice were sacrificed 27 days after injection. Tumors were excised and the tumor weight was recorded. 2.11. Immunohistochemistry staining assay {#cam42857-sec-0013} ----------------------------------------- The slides were heated at 60°C for 2 hours, deparaffinized in xylene, and rehydrated by graded ethanol. After antigen retrieval, the tissue sections were blocked with goat serum. The sections were then exposed to antibodies against δ‐catenin or Ki67 (Abcam, ab15580) overnight at 4°C. The slides were washed with PBS and incubated with HRP‐labeled goat anti‐rabbit antibody (ThermoFisher, \#31460) for 1 hour. After washing, the slides were added with diaminobenzidine (DAB) and counterstained with hematoxylin. In the end, the sections were dehydrated by graded ethanol, sealed by neutral balsam, and visualized with microscope (at 400 × magnification). 2.12. Statistical analysis {#cam42857-sec-0014} -------------------------- The data were presented as mean ± SD in the figures. Statistical analysis was performed with two‐tailed unpaired t‐test, one‐way ANOVA or two‐way ANOVA as indicated. P value \< .05 was considered statistically significant. 3. RESULTS {#cam42857-sec-0015} ========== 3.1. δ‐Catenin was highly expressed in human renal cancer tissues and correlated with poor prognosis of RCC patients {#cam42857-sec-0016} -------------------------------------------------------------------------------------------------------------------- We firstly determined the expression of δ‐catenin in human renal cancer tissues. Western blot analysis was performed to examine δ‐catenin expression from 28 cases of patients with RCC. The results showed that δ‐catenin was highly expressed in tumor tissues compared with adjacent normal tissues (Figure [1](#cam42857-fig-0001){ref-type="fig"}A and Figure [S1](#cam42857-sup-0001){ref-type="supplementary-material"}A). Furthermore, we detected the expression of δ‐catenin by immunohistochemistry in 58 cases of RCC, and evaluated the correlation between δ‐catenin expression and clinicopathological factors. The representative images of low and high expression of δ‐catenin from patients were presented in Figure [1](#cam42857-fig-0001){ref-type="fig"}B. As shown in Table [1](#cam42857-tbl-0001){ref-type="table"}, high expression of δ‐catenin was correlated with pTNM stage, tumor stage, and lymph node metastasis (P \< .05), but not correlated with patient\'s gender, age, tumor size, and distant metastasis (P \> .05). These results together demonstrated that the expression of δ‐catenin was enhanced in RCC tissues, which was associated with poor prognosis in RCC patients. ![δ‐Catenin was highly expressed in human renal cancer tissues. A, The representative bands of δ‐catenin detected by Western blot analysis from human renal cancer tissues and adjacent normal tissues (16 pairs). B, The representative images of low and high expression of δ‐catenin in human renal cancer tissues visualized by immunohistochemical staining. Scale bar, 50 μm](CAM4-9-2201-g001){#cam42857-fig-0001} ###### Relationship between δ‐catenin and clinical characteristics in RCC patients Parameters Group δ‐catenin expression P value ------------------------ ---------- ---------------------- ----------- ------------ Gender Male 17 13 .80628 Female 14 14 Age \<60 20 14 .47797 ≥60 11 13 Tumor size (cm) \<4 cm 1 3 .50751 ≥4 cm 30 24 pTNM stage I/II 21 27 .00378 III/IV 10 0 Tumor stage T1‐T2 23 27 .01384 T3‐T4 8 0 Lymph nodes metastasis Negative 25 27 .04747 Positive 6 0 Distant metastasis Negative 29 27 .53404 Positive 2 0 A P value of less than .05 was considered to be statistically significant (shown in bold). John Wiley & Sons, Ltd 3.2. δ‐Catenin was highly expressed in human renal cancer cells {#cam42857-sec-0017} --------------------------------------------------------------- Next, we determined the level of δ‐catenin in normal human renal epithelial cells (HK‐2) and three different RCC cell lines (A498, ACHN and 786‐O). As expected, δ‐catenin was highly expressed in RCC cells compared with HK‐2 cells (Figure [2](#cam42857-fig-0002){ref-type="fig"}A). We employed specific siRNA to knock down δ‐catenin in A498 and ACHN cells showing its relatively higher expression. qRT‐PCR measurement of A498 and ACHN cells revealed the mRNA level of δ‐catenin was reduced after δ‐catenin knockdown (Figure [2](#cam42857-fig-0002){ref-type="fig"}B,C). The remarkable reduction of δ‐catenin protein level was also observed in RCC cells after δ‐catenin silencing (Figure [2](#cam42857-fig-0002){ref-type="fig"}D,E). All the data suggested that δ‐catenin expression was elevated in RCC cells. ![δ‐Catenin was highly expressed in human renal cancer cells. A, The expression of δ‐catenin in renal cancer cell lines tested by Western blot analysis. Renal cancer cells A498 and ACHN were treated with δ‐catenin‐specific siRNA. After 48 h transfection, B, C, δ‐catenin mRNA levels were analyzed by qRT‐PCR (n = 3). D, E, The protein levels of δ‐catenin were detected by Western blot. Data were presented as mean ± SD (n = 3) with statistic significance calculated from one‐way ANOVA for multiple comparisons (\\P \< .01, \\\*P \< .001, \\\\P \< .0001)](CAM4-9-2201-g002){#cam42857-fig-0002} 3.3. δ‐Catenin promoted renal cancer cell proliferation {#cam42857-sec-0018} ------------------------------------------------------- Considering the correlation between high expression of δ‐catenin and poor prognosis, we investigated the biological role of δ‐catenin in RCC development. We performed CCK‐8 assay to evaluate the effect of δ‐catenin knockdown on RCC cell proliferation. As shown in Figure [3](#cam42857-fig-0003){ref-type="fig"}A,B, δ‐catenin knockdown resulted in decreased cell viability in A498 and ACHN cells. The immunofluorescence analysis of Ki67 also illustrated reduced proliferative capacity after δ‐catenin silencing (Figure [3](#cam42857-fig-0003){ref-type="fig"}C,D). We next explored whether δ‐catenin participated in RCC cell cycle regulation. We examined cell numbers in each phase of mitosis in RCC cells following δ‐catenin knockdown by flow cytometry. The results indicated that much higher percentages (53.87% in A498, 57.53% in ACHN) of RCC cells were arrested in G1 phase upon δ‐catenin knockdown (Figure [3](#cam42857-fig-0003){ref-type="fig"}E,F). We also determined the expression of cell cycle‐related proteins in RCC cells following δ‐catenin silencing. The expression of cyclin D1, CDK4, and CDK6 was significantly reduced after downregulating endogenous δ‐catenin (Figure [3](#cam42857-fig-0003){ref-type="fig"}G,H). These results strongly indicated that δ‐catenin knockdown has an inhibitory effect on proliferation of RCC cells. ![δ‐Catenin promoted renal cancer cell proliferation. δ‐Catenin‐specific siRNA was employed to silence δ‐catenin in A498 and ACHN cells. A, B, After 24, 48, 72, and 96 h of transfection, CCK‐8 assay was performed to observe cell viability. C, D, The immunofluorescence analysis of the expression of Ki67 in renal cancer cells transfected with siRNA. Scale bar, 50 μm. E, F, Cell cycle was analyzed in A498 and ACHN cells by flow cytometry. G, H, The expression level of cell cycle‐associated markers was determined by Western blot. Data in panel A, B, E, F, G, and H were shown as the mean ± SD from three separate experiments, with two‐way ANOVA for A, B, E, F, and one‐way ANOVA for G and H](CAM4-9-2201-g003){#cam42857-fig-0003} 3.4. δ‐Catenin inhibited apoptosis of renal cancer cells {#cam42857-sec-0019} -------------------------------------------------------- Furthermore, we identified the effect of δ‐catenin on apoptosis of RCC cells. A498 and ACHN cells transfected with siRNA were stained with Annexin V‐FITC/PI and subjected to flow cytometric analyses. The percentage of apoptotic cells was increased in δ‐catenin‐knocked down RCC cells (Figure [4](#cam42857-fig-0004){ref-type="fig"}A,B). We also examined protein expression of apoptosis‐related markers by Western blot analysis. As shown in Figure [4](#cam42857-fig-0004){ref-type="fig"}C,D, silence of δ‐catenin led to the downregulation of Bcl2L1 (anti‐apoptotic marker) but upregulation of cleaved‐caspase 3. All these data substantiated δ‐catenin restrained the induction of apoptosis in RCC cells. ![δ‐Catenin inhibited apoptosis of renal cancer cells. A498 and ACHN were transfected with δ‐catenin‐specific siRNA for 48 h. A, B, The cells were analyzed by FACS using Annexin V‐FITC/PI staining, and the percentage of apoptotic cells was calculated. C, D, The protein levels of apoptosis‐related markers in renal cancer cells following δ‐catenin knockdown were detected by Western blot. The percentage of apoptotic cells and quantitative analysis for apoptosis‐related proteins were presented as mean ± SD with one‐way ANOVA with Tukey\'s test for multiple comparisons. \\P \< .01, \\\*P \< .001, \\\\P \< .0001](CAM4-9-2201-g004){#cam42857-fig-0004} 3.5. δ‐Catenin affected β‐catenin nuclear localization and its downstream gene expression in renal cancer cells {#cam42857-sec-0020} --------------------------------------------------------------------------------------------------------------- Considering δ‐catenin could activate β‐catenin and its downstream tumor‐associated signaling pathways to participate in the development of prostate cancer,[14](#cam42857-bib-0014){ref-type="ref"} we investigated the effect of δ‐catenin on this pathway in RCC. Immunofluorescence analysis was employed to visualize the change in nuclear localization of β‐catenin under the condition of δ‐catenin silence. As shown in Figure [5](#cam42857-fig-0005){ref-type="fig"}A, δ‐catenin silencing in ACHN cells led to a remarkable decrease in nuclear β‐catenin, which was further confirmed by Western blot analysis (Figure [5](#cam42857-fig-0005){ref-type="fig"}C). To further investigate whether the altered β‐catenin distribution interacts with TCF/LEF to activate downstream gene expression, TCF/LEF transcriptional reporter activity was measured in ACHN cells. Knockdown of δ‐catenin exhibited a lower specific TCF/LEF transcriptional reporter activity (Figure [5](#cam42857-fig-0005){ref-type="fig"}B), implying an alteration in the transcription co‐activator activity of β‐catenin, which caused the subsequent downregulation of downstream target gene, such as c‐myc and survivin (Figure [5](#cam42857-fig-0005){ref-type="fig"}C). Together, these results indicated that δ‐catenin upregulated downstream gene expression of β‐catenin by promoting β‐catenin translocation to the nucleus in RCC cells. ![δ‐Catenin affected β‐catenin nuclear localization and its downstream gene expression in renal cancer cells. ACHN cells were treated with δ‐catenin specific siRNA for 48 h. A, The expression and localization of β‐catenin were observed by immunofluorescence analysis. Scale bar, 50 μm. B, The transcriptional activity of β‐catenin following δ‐catenin knockdown was analyzed by the TOP/FOP luciferase reporter assay. The relative luciferase activity was normalized and shown (n = 3). C, The expressions of β‐catenin in the whole cell lysate, cytoplasm, and nucleus as well as the protein levels of c‐myc and survivin were tested by Western blot. Data in panel B and C were shown as the mean ± SD based on one‐way ANOVA with Tukey\'s test (\*P \< .05, \\P \< .01, \\\*P \< .001, \\\\P \< .0001)](CAM4-9-2201-g005){#cam42857-fig-0005} 3.6. δ‐Catenin promoted tumor growth in ACHN xenograft mouse model {#cam42857-sec-0021} ------------------------------------------------------------------ To further detect the regulation of δ‐catenin on tumorigenesis in vivo, we examined the effect of δ‐catenin knockdown on tumor growth of ACHN cell‐derived xenograft in nude mice. The ACHN cells transfected with shCtrl plasmids and δ‐catenin shRNA plasmids were subcutaneously injected into BALB/c nude mice, which were divided into two groups (n = 6 for each group). Tumor volume was recorded over a period of 27 days and the solid tumors were photographed and weighed after the mice were sacrificed. As shown in Figures [6](#cam42857-fig-0006){ref-type="fig"}A‐C, knockdown of δ‐catenin markedly suppressed tumor growth in mice, compared with the shCtrl group. Moreover, immunohistochemical analysis of the tumor tissues also revealed δ‐catenin knockdown suppressed tumor growth, as evidenced by reduced Ki67 expression (Figure [6](#cam42857-fig-0006){ref-type="fig"}D). Furthermore, δ‐catenin knockdown restrained translocation of β‐catenin to the nucleus, as well as regulated the expression of c‐myc, surviving, cyclinD1, Bcl2L1, and cleaved‐caspase‐3 (Figure [6](#cam42857-fig-0006){ref-type="fig"}E). All these results validated that δ‐catenin facilitated tumor growth in vivo in RCC. ![δ‐Catenin promoted tumor growth in ACHN xenograft mouse model. A, The images of tumor harvested at 27 d after transplantation of ACHN cells. B, Tumor diameters were measured at a regular interval of 4 d for up to 27 d and the tumor volume was calculated. C, Tumor weight from each group was measured after the mice were sacrificed. D, The expression of δ‐catenin and Ki67 in tumor tissue sections was tested by immunohistochemistry staining. Scale bar, 50 μm. E, The protein expression of δ‐catenin, β‐catenin (in the whole cell lysate, cytoplasm, and nucleus), c‐myc, survivin, cyclinD1, Bcl2L1, and cleaved‐caspase‐3 in tumor tissues was detected by Western blot. Data in panel B were shown as the mean ± SD based on two‐way ANOVA (n = 6). Data in C and E were presented as the mean ± SD (n = 6, two‐tailed unpaired t test). \\\*P \< .001, \\\\P \< .0001](CAM4-9-2201-g006){#cam42857-fig-0006} 4. DISCUSSION {#cam42857-sec-0022} ============= It has been proven that δ‐catenin has ectopic overexpression in lung cancer and prostate cancer and promotes tumor progression,[12](#cam42857-bib-0012){ref-type="ref"}, [13](#cam42857-bib-0013){ref-type="ref"}, [14](#cam42857-bib-0014){ref-type="ref"} but little is known about its biological role and the associated signaling pathways in RCC. Our study set out to determine whether δ‐catenin was implicated in the development of RCC. Based on the results of Western blot and immunohistochemistry, we found that the level of δ‐catenin is elevated in clinical tissue samples with RCC. Therefore, we reasoned that δ‐catenin also played a role in promoting the development of RCC. In subsequent exploration, we found that δ‐catenin affected the translocation of β‐catenin, which further modulated the expression of its downstream apoptosis‐related protein Bcl2L1, cell cycle‐associated marker cyclin D1 as well as tumorigenic gene c‐myc and survivin, thereby facilitated proliferation and suppressed apoptosis in RCC. Moreover, the RCC cell xenograft mouse model revealed knockdown of δ‐catenin restrained tumor growth in vivo, which was concordant with our in vitro findings. Researches conducted by Zhang et al[12](#cam42857-bib-0012){ref-type="ref"} have assessed the clinical significance of δ‐catenin and its correlation with clinicopathological factors in a cohort of 115 specimens with non‐small‐cell lung cancer (including 65 cases with follow‐up records and 50 cases with paired lymph node metastases lesions). The results demonstrated that high expression of δ‐catenin could be observed in adenocarcinoma. Positive expression of δ‐catenin was enhanced in lymph node metastasis lesions, compared with corresponding primary tumors. The elevated δ‐catenin expression was significantly correlated with higher pTNM stage and lymph node metastasis. Furthermore, patients harbouring tumors with δ‐catenin positive expression had significantly shorter postoperative survival period than patients harboring tumors with negative expression. Besides, it has been reported that increased δ‐catenin expression in estimated 85% of prostate cancer based on tissue samples (composed of 90 cases with prostate cancer and 90 cases with benign prostate). Meanwhile, the enhanced δ‐catenin expression was concerned with increased Gleason scores.[29](#cam42857-bib-0029){ref-type="ref"} Consistent with these reports, a significant finding of our study is that RCC patients showed high expression of δ‐catenin and it is correlated with some clinicopathological factors, such as pTNM stage, tumor stage and lymph nodes metastasis. It gave evidence that positive expression of δ‐catenin was relevant to poor prognosis in RCC. δ‐Catenin might be established as a potential prognostic factor and a therapeutic target of RCC. In this study, we found δ‐catenin promoted RCC cell proliferation by altering G1 to S phase transition in cell cycle. Furthermore, δ‐catenin restrained apoptosis of RCC cells through upregulating Bcl2L1 (a long isoform of Bcl2) that protects cells from undergoing apoptosis.[30](#cam42857-bib-0030){ref-type="ref"} In the past decades, a large and growing body of literature has shown that cyclin D1 was induced by hyperactivated β‐catenin, and thereby led to G1 phase progression and proliferation of tumor cell.[31](#cam42857-bib-0031){ref-type="ref"}, [32](#cam42857-bib-0032){ref-type="ref"}, [33](#cam42857-bib-0033){ref-type="ref"}, [34](#cam42857-bib-0034){ref-type="ref"} On the other hand, Rosenbluh et al stated that β‐catenin forms a complex with YAP1 and transcription factor TBX5 which binds to the promoter of BCL2L1, thereby directly regulating BCL2L1 expression at transcriptional level.[35](#cam42857-bib-0035){ref-type="ref"} Taken together, aberrant cyclin D1 and BCL2L1 upregulation take place when the Wnt/β‐catenin pathway is activated. Combined with our investigation about the effects of δ‐catenin on proliferation and apoptosis in RCC cells, we deduced that δ‐catenin may exert these functions by regulating β‐catenin in RCC development. As expected, our data showed that δ‐catenin knockdown suppressed β‐catenin nuclear localization and subsequent alteration of Bcl2L1 and cyclin D1 level in both ACHN cells and ACHN‐derived xenograft tumor tissue. Given that δ‐catenin affected subcellular β‐catenin distribution and activated the downstream effectors of β‐catenin in prostate cancer,[14](#cam42857-bib-0014){ref-type="ref"} we explored whether δ‐catenin played the same role in RCC progression. Figure [5](#cam42857-fig-0005){ref-type="fig"} elucidated δ‐catenin realized the tumor‐promoting function in renal tumorigenesis by altering β‐catenin translocation. Notably, silencing of δ‐catenin prevented β‐catenin from translocating to the nucleus, which resulted in the repression of β‐catenin‐TCF/LEF‐mediated downstream effector activation. Knockdown of δ‐catenin in RCC cells revealed a markedly decreased expression in β‐catenin target genes such as Bcl2L1, cyclin D1, c‐myc, and survivin. In our study, we used RNA interference (RNAi) including transient siRNA and stable shRNA for loss‐of‐function studies to investigate the role of δ‐catenin in RCC development. One major drawback of these means is the limited utility, as evidenced by δ‐catenin‐silenced cells still could grow and preserve tumorigenesis capacity. In future research, the inducible system (such as the tetracycline‐controlled operator system) should be combined with RNAi, which could reveal the direct effects caused by the loss of δ‐catenin via comparing noninduced and induced cells, thereby accurately evaluating the crucial role of δ‐catenin in RCC cell growth and apoptosis. In summary, our study is the first to report δ‐catenin was overexpressed in tissues of RCC patients and its correlation with pTNM stage, tumor stage, and lymph nodes metastasis. In addition, knockdown of δ‐catenin can obviously decrease the proliferation capacity and induce apoptosis of RCC cells by affecting β‐catenin nuclear localization and its downstream target gene expression. Therefore, our results illustrated that δ‐catenin might be a prognostic marker and a promising therapeutic target in RCC. CONFLICT OF INTEREST {#cam42857-sec-0024} ==================== The authors declare that they have no competing interests. AUTHOR CONTRIBUTIONS {#cam42857-sec-0025} ==================== Yongsheng Song conceived and designed the study. Ju Lincheng performed the experiments and analyzed the data. Shan Liping wrote the paper. Yinbo obtained funding. All authors have read and approved the final manuscript. Supporting information ====================== ###### ###### Click here for additional data file. This study was supported by a grant from the Guide Project for the Key Research and Development Project of Liaoning Province (No. 2017225038) and 345 Talent Project of Shengjing Hospital of China Medical University. The pathologic sections were provided by the Biobank from Shengjing Hospital of China Medical University. DATA AVAILABILITY STATEMENT {#cam42857-sec-0027} =========================== All data generated or analyzed during this study are included in this published article.	{ "pile_set_name": "PubMed Central" }
Introduction {#s1} ============ Over the last two decades, optical traps (or optical tweezers) have become a standard tool in the physical and biological sciences, allowing the measurement of sub-nanometer displacements of optically trapped microparticles, as well as the exertion of piconewton-level, controlled forces on these particles [@pone.0057383-Svoboda1], [@pone.0057383-Bustamante1], [@pone.0057383-Ashkin1]. Taking advantage of this versatility, studies on the mechanical and biochemical properties of biomolecules at the single-molecule level using optical tweezers are now widespread. In these assays, individual molecules, attached to trapped particles, can be pulled on or stretched (or even twisted) using the laser trap, and the molecular displacement and force (or torque) responses can be measured with high spatial (∼1 nm) and temporal (∼100 kHz) resolution [@pone.0057383-Fazal1]. These experiments provide unique information on molecular mechanisms that complement traditional biochemical studies. Typically, an optical tweezers arrangement consists of a single laser beam tightly focused by a microscope objective of high numerical aperture (NA) (1.0--1.4), which traps microparticles near its focal point [@pone.0057383-Ashkin2], [@pone.0057383-Neuman1]. Position detection of the trapped particle relative to the laser beam axis is achieved through back focal plane (BFP) detection, where the trapping beam (or a secondary beam of low-power) is directed to a position-sensitive detector (PSD) located in a plane conjugate to the condenser BFP [@pone.0057383-Pralle1]. In many trapping setups, these requirements are implemented by substantial modifications of a commercial inverted light microscope in order to accommodate holders, mounts, and a stable stage platform fitted with a piezoelectric stage (for fine sample movement). Alternatively, an optical trap can be built entirely from individual optical components, providing increased flexibility in the design, often reducing costs, and facilitating the choice of parts. This last option is the one we have followed in this work. Successful development of an optical tweezers apparatus often requires appropriate means to visualize small objects (∼20--200 nm). Indeed, a variety of cellular and single-molecule assays require the localization and subsequent tracking of nanoparticles [@pone.0057383-Sako1], or the imaging of individual slender polymer filaments such as microtubules (MTs, with diameters of only ∼25 nm) [@pone.0057383-Block1]. Because small objects scatter light weakly, their visualization has traditionally not been performed using bright field microscopy. Instead, localization is achieved by employing a number of specialized imaging techniques, in the case of MTs: fluorescence [@pone.0057383-Sammak1], dark field [@pone.0057383-Kuriyama1], polarization [@pone.0057383-Oldenbourg1], phase contrast [@pone.0057383-Kuriyama1], or Nomarski differential interference-contrast (DIC) microscopy [@pone.0057383-Schnapp1]. However, some of these visualization methods exhibit inherent limitations in the context of optical trapping. For example, it is well-known that the Wollaston prisms required in DIC optics can introduce significant asymmetries in the optical trap [@pone.0057383-Lang1]. Likewise, because the Wollaston prisms generally produce or recombine two displaced beams with orthogonal polarizations, DIC microscopy prohibits the operation of an optical tweezers setup where polarization of the trapping beam requires high purity and independent adjustment (as in an "optical torque wrench" [@pone.0057383-GutirrezMedina1]). Introducing optical elements such as prisms, apertures, phase masks, or polarizers in the laser beam path unavoidably constrains the possibilities for optical trapping, especially when control of phase or polarization effects is necessary, as in holographic [@pone.0057383-Grier1] or interferometric [@pone.0057383-MacDonald1] optical tweezers. Recently, it has been shown that unlabeled, individual MTs can be visualized by computer-enhanced bright field microscopy (CEBFM) [@pone.0057383-GutirrezMedina2], a method that does not require the addition of specific optical elements or processing instrumentation to effectively remove the background, reduce noise, and enhance contrast of images. CEBFM provides an alternative of remarkable simplicity to visualize small objects, and is therefore ideally suited to complement optical trapping. In bright field microscopy, ample access to optical paths in the setup is ensured, and it is possible to independently optimize Koehler illumination and the BFP detection stage of the optical trap, allowing excellent imaging and trapping. By using CEBFM, a basic microscope could be fit to perform standard single-molecule "gliding filament" or optical trapping "bead" assays without resorting to advanced imaging modalities. Furthermore, CEBFM may prove useful in emerging microscopy techniques that use image contrast to perform biophysical measurements and that incorporate optical tweezers as a tool for microparticle manipulation, such as "defocusing microscopy" [@pone.0057383-Agero1] (see Section "Digital image processing"). In this work, we describe the design, construction, and testing of an elementary optical tweezers instrument that allows trapping and visualization of weakly scattering objects, integrating a straightforward approach for the study of biomaterials and individual biomolecules. Our design is based on CEBFM together with a single laser beam coupled into the microscope for optical trapping and BFP detection. This arrangement keeps the number of mechanical and optical components to a minimum, resulting in ease of construction and cost-effectiveness that do not compromise final stability or resolution. To test the capabilities of our system, we performed two single-molecule assays and compared our results with previous reports. We have measured the persistence length of individual double-stranded DNA (dsDNA) molecules at relatively high ionic strength (\[Na^+^\] = 150 mM), and have observed the stepping of single kinesin motor proteins as they advance along clearly imaged, individual MTs. To our knowledge, this is the first time that the classic cytoskeletal motor protein bead-assay has been demonstrated using plain bright field illumination to localize filament substrates. Results and Discussion {#s2} ====================== Optical and Mechanical Layout {#s2a} ----------------------------- Our optical tweezers system (shown in [Figure 1](#pone-0057383-g001){ref-type="fig"}) was built entirely from individual components, mounted on a vibration-isolation optical table (I-2000, Newport). Trapping light is provided by a CW, Yb-doped, fiber laser (YLR-10-1064-LP, 10 W, λ = 1064 nm, IPG Photonics) featuring excellent pointing stability (deviations \<1 µrad/s), and good power stability (fluctuations ∼2%). A 10 m-long optical fiber coupled to the laser allows installation of the laser head (together with the noisy power electronics) in a separate room. The laser is nominally operated at ∼4.2 W of output power, above the minimum ∼3.0 W, that according to the manufacturer accounts for the stability figures cited above. ![Schematic of the optical tweezers instrument.\ A fiber-coupled laser provides the trapping light (red lines), the intensity of which is controlled with a combination of a λ/2-waveplate and a polarizing beam splitter (PBS). An acousto-optic modulator (AOM) is part of a feedback loop that stabilizes laser intensity. The trapping beam is coupled into and out of a custom-built inverted microscope (black dotted line box) using two dichroic mirrors (DM). Illumination (blue lines) for the inverted microscope is provided by a blue LED. The green and blue arrows indicate the positions of the two sets of optically conjugated planes that satisfy the conditions for Koehler illumination and back focal plane detection. See text for details. BS: beam sampler, PD: photodiode, M: mirror, L: lens, D: diaphragm iris, PSD: position-sensitive detector.](pone.0057383.g001){#pone-0057383-g001} We now describe our setup in detail. The end terminal of the laser fiber is secured onto the optical table, and a combination of a half-wave plate and a polarizing cube beamsplitter (PBS) control the laser power going to the optical trap. To provide means of automatically steering the laser trap on the sample plane, an x-y acousto-optic modulator (AOM) (DTD-274HD6M, IntraAction) is placed immediately after the PBS, and an iris diaphragm (D1) selects the (+1,+1) diffracted beam that emerges from the AOM. Next, two lenses (L1 and L2), forming a Keplerian telescope, expand the beam to a final waist size, w, of ∼3 mm, corresponding to a beam diameter that roughly matches the diameter of the objective back aperture (∼6 mm). Two additional lenses (L3 and L4) then form a 1∶1 telescope, with L3 (see [Figure 1](#pone-0057383-g001){ref-type="fig"}) mounted to an x-y-z translation stage to provide fine manual control to steer and collimate the laser beam going into the objective. A stage-mounted lens is a useful feature during the initial alignment of the trap. To transform beam deflection by either the AOM or lens L3 into beam translation in the specimen plane and minimize beam clipping, the AOM and the lens L3 are optically conjugate to the back aperture of the objective [@pone.0057383-Neuman1] (see [Figure 1](#pone-0057383-g001){ref-type="fig"}). The AOM is useful not only to steer the trap but also to control the laser power going into the objective. Our instrument is placed in a room that does not have temperature control, and we have found that the output laser power may drift during the course of the day by as much as ∼5%, presumably due to room temperature changes. To correct for this problem, a laser power feedback loop was implemented using the AOM. A beam sampler placed after the AOM picks up part of the main beam and sends it to a photodiode, whose signal is amplified and fed into a data-acquisition board (U12, LabJack) connected to a computer. A proportional-integral-derivative (PID) feedback routine is implemented in software, using LabView 8.5 (National Instruments), and its corrected amplitude signal is communicated to the AOM driver (AFG3022B, Tektronix) via a USB port. Finally, the x and y outputs of the AOM driver are each amplified by an RF amplifier (ZHL-3A, Minicircuits), and connected to the AOM, closing the feedback loop. The resulting feedback control has a characteristic time response of ∼1 s, and maintains the power constant (to a level of ∼0.2%) for hours. Alternatives to this feedback scheme using an AOM driver are available [@pone.0057383-Mack1]. To afford imaging of the specimen, trapping chamber manipulation, and laser focusing by the objective, an inverted microscope was built. Because one of the main purposes of this work is to attain visualization of weak light-scattering objects by means of bright field microscopy (without the aid of additional optical parts or special types of illumination), the components of the optical train of our microscope are arranged to satisfy Koehler illumination. Among other advantages of Koehler illumination, the field is homogeneously bright, image contrast is optimum, and maximal lateral and axial resolutions are achieved [@pone.0057383-Inou1]. In our microscopy system (see [Figure 1](#pone-0057383-g001){ref-type="fig"}), the illuminator is composed of a high-power light-emitting diode (LED) (LXHL-LR5C, 700 mW, peak λ = 455 nm, Luxeon) as the light source, together with a collector lens (L6) and a field iris diaphragm (D3) placed immediately adjacent to L6. The LED is driven by an inexpensive analog current controller, originally designed for laser diodes, that provides extremely stable current (∼10^−4^ rms variations) up to 500 mA (design available at: <http://george.ph.utexas.edu/~meyrath/informal/laser%20diode.pdf>). This intensity stability is necessary to perform appropriate background subtraction during image processing (as described in the next section). The collector lens collimates incoming light from the LED, and an additional lens (L7) focuses the image of the light source onto the condenser iris diaphragm (D4). To maximize contrast during imaging, we close this aperture almost entirely (∼85%) to effectively produce a point-like illumination source. Relay lens (L8) images the condenser iris onto the condenser BFP. The condenser lens (1.4NA, Zeiss) is mounted on an x-y-z translation stage (461-XYZ-M, Newport), and aligned to focus an image of the field diaphragm onto the sample plane. This ensures that the specimen and the images of the light source are in different (reciprocally related) planes, and that the specimen is illuminated with a set of plane waves. Following the optical train, the objective lens captures the illuminating plane waves and creates an image of the light source onto the rear focal plane of the objective (exit pupil). We use an infinity-corrected, oil-immersion objective of high numerical aperture (100X/1.3NA/PlanFluorite, part RMS100X-PFO, Olympus). Finally, a tube lens (L9) captures illuminating and scattered light from the specimen and images the sample onto an inexpensive CCD camera (STC-TB33USB-B, Sentech) that is connected to the computer via a USB interface. With the condition for Kohler illumination satisfied in this optical arrangement, two mutually reciprocal sets of optically conjugated planes are formed. The first set includes the field diaphragm, the sample plane, and the camera sensor plane. The second set includes the light source, the condenser iris, the rear focal plane of the condenser, and the back focal plane of the objective. Coupling of the trapping laser into and out of the microscope is achieved using two dichroic mirrors (Z1064RDC-SP, Chroma). To perform BFP detection, a single lens (L5) images the rear focal plane of the condenser onto a PSD (DL100-7-PCBA2, Pacific Silicon Sensor). A diaphragm (D2) placed directly in front of the PSD modulates the NA of the detection optics, optimizing the PSD detection signal [@pone.0057383-Rohrbach1]. The output voltage signals from the PSD (fitted with a built-in preamplifier) are further amplified using a custom-made electronic circuit (200 kHz bandwidth) and sent to a data-acquisition board (PCI-6052E, National Instruments) fitted in a computer, where data are processed by custom software written in LabView. A laser power of 90 mW applied to the trap (measured before entering the objective rear pupil) was typically used for our experiments. To block the trapping laser whenever necessary, an inexpensive, custom-made mechanical shutter was used (design available at: <http://george.ph.utexas.edu/~meyrath/informal/shutter.pdf>). An advantage of our optical trapping and microscopy setup over previous designs is that BFP detection and Koehler illumination are effectively decoupled and can be optimized independently. In typical microscopes working under Koehler illumination, the condenser iris is located at the back focal plane of the condenser lens (entrance pupil). This arrangement is not ideal for an optical tweezers arrangement, as any aperture placed in the proximity of the condenser compromises BFP detection. Conversely, although standard optical trapping systems often do not use the condenser iris, here it is necessary to maximize contrast during visualization of weak light-scattering specimens. We resolved this potential conflict by using relay lens L8 to image the condenser iris onto the condenser BFP, leaving the condenser lens entrance pupil unobstructed. Lens L8 thus allows us to adjust the condenser iris aperture without affecting the NA of the BFP detection optics. The mechanical design of the microscope is completed with a piezoelectric stage (Nano-LP100, Mad City Labs) that is used for fine sample movement, and a supporting crossed-roller-bearing mechanical stage (Manual MicroStage, Mad City Labs) that provides coarse positioning. The x-y-z piezoelectric stage has nanometer-level step resolution over 100×100×100 µm. The stacked stages are supported by four 8″-tall, 1.5″-diameter stainless steel posts that provide mechanical stability to the stage platform. All of the condenser, detection, and illumination optics are mounted on a breadboard that is suspended vertically on a structure formed by two large vertical construction rails (XT95, Thorlabs), cross-linked at the top by a third rail, and further supported by one additional beam joining the top rail to the optical table. Digital Image Processing {#s2b} ------------------------ Images are digitally processed to remove background, perform frame averaging, and enhance contrast. We implemented computer-based image processing essentially as described in Ref. [@pone.0057383-GutirrezMedina2], with appropriate improvements for our work. All of the procedures described here are implemented in software using LabView 8.5 (add-on package Vision, National Instruments). First, using the piezoelectric stage, the surface of the coverslip is brought into focus and subsequently defocused by a few micrometers. At this point, the sample is oscillated in the x or y direction (about 5 micrometers in amplitude, 2 seconds in period) using the piezoelectric stage, while background images (typically 250) are simultaneously acquired. An average of the background images provides a reference image that is subsequently subtracted from all incoming frames. As background frames are acquired this way, images belonging to any relatively strong scattering objects that may be stuck to the coverslip surface and present in the field of view are defocused and blurred in the final reference background image, thereby minimizing their influence in subsequent processing. This allows for changes in the field of view without the need to acquire a new background reference. After completing this operation, focus is readjusted to visualize the sample. Next, background-free images are averaged using a simple exponential-averaging (or exponential-smoothing) procedure, during which incoming time series data are averaged using exponentially decreasing weights [@pone.0057383-Salmon1]. One advantage of exponential over boxcar averaging is that it is easier to implement, as only two data (the last, smoothed datum and the incoming datum) need to remain in memory. In addition, the weight assigned to recent or older data can be readily adjusted. Thus, let F~i~ represent the i ^th^ original, background-free image acquired sequentially in time and S~i~ the corresponding final, smoothed image. Exponential averaging is implemented according to the following rule [@pone.0057383-NISTSEMATECH1]:where parameter α is the smoothing constant (see [Figure 2A](#pone-0057383-g002){ref-type="fig"}). Full or zero weight is assigned to the most recent datum in the time series by setting α = 1 or α = 0, respectively. We typically use α = 0.3 in our experiments. ![Individual microtubules are visualized using bright field microscopy together with digital image processing.\ (A) Flow diagram of digital image processing, depicting operations for background acquisition (dashed area) and subtraction, exponential averaging, and contrast enhancement. (B-E) Frames show MTs immobilized on coverslips, as imaged during the various stages of processing marked in (A): (B) raw image; (C) the same as (B) with the background removed; (D) the same as (C) after multiple-frame exponential averaging; (E) the same as (D) after a convolution routine (shown with the symbol in the diagram) that enhances contrast in the vertical direction, and a fast-Fourier-transform routine (shown with the symbol in the diagram) that smoothes out rapidly-varying spatial features. Field of view is 24 µm × 21 µm.](pone.0057383.g002){#pone-0057383-g002} Background-free, exponentially-averaged images can be further processed using two additional routines that enhance contrast and reduce noise: (i) a spatial convolution routine, which involves a point-by-point multiplication of the pixel values of the original image multiplied by the corresponding pixel values of a filtering convolution mask [@pone.0057383-GutirrezMedina2], and (ii) a fast Fourier transform smoothing routine, which removes spatial features within a given size range. To test our system, we visualized MTs, polymer filaments that are involved in cellular structure and organization [@pone.0057383-WatermanStorer1], previously immobilized on coverslips (see [Materials and Methods](#s3){ref-type="sec"}). [Figures 2B--2E](#pone-0057383-g002){ref-type="fig"} show that individual MTs are clearly distinguished, with an apparent thickness determined by the diffraction limit of the microscope (∼250 nm). We noticed that the contrast of MT images changes as the coverslip surface is brought in and out of focus, reaching an optimum when the coverslip surface is slightly (∼300 nm) defocused (see [Figure 3](#pone-0057383-g003){ref-type="fig"}). This fact is consistent with the early observation by Fritz Zernike that when a bright field microscope is focused precisely, the image of a thin, transparent specimen disappears due to interference of light diffracted by the specimen and of background illumination, and contrast is maximized only by slightly defocusing the microscope [@pone.0057383-Inou1]. Recently, this effect has been used to develop "defocusing microscopy," a quantitative visualization method where image contrast (shown proportional to the amount of defocusing and to the curvature of the specimen) is measured, from where it is possible to learn about the material properties of the sample [@pone.0057383-Agero1]. In our experiment, the MT filaments are transparent under visible light and present enough shape curvature, therefore they act as effective phase objects despite being hollow and having a diameter of only λ/20. ![Microtubule image contrast is optimized by microscope defocusing.\ Images of a single MT (top row) displaying the same field of view when the focus is placed 300 nm below (left), precisely at (center), and 300 nm above (right) the coverslip surface. The bottom row shows the corresponding pixel count profiles, obtained for each image by averaging all the pixel count values along a given pixel column. Images were digitally-processed as described in the text to remove background and perform exponential averaging (α = 0.3). No convolution or fast-Fourier-transform routines were applied in this case. Field of view is 3.4 µm × 4.8 µm.](pone.0057383.g003){#pone-0057383-g003} The previous MT imaging result demonstrates that our setup achieves excellent sample visualization at standard video rates (∼30 Hz), offering several advantages over a recent system that also imaged MTs using bright field microscopy [@pone.0057383-GutirrezMedina2], among which are: improved background subtraction (by using a light source with excellent intensity stability and by moving the piezoelectric stage during background acquisition), real-time specimen visualization (by implementing exponential averaging of images), and the possibility of using the full NA of the condenser (by using relay lens L8 to image the condenser iris onto the condenser BFP). Most importantly, the instrument presented here has the full capabilities of optical trapping for measuring molecular-level displacements and forces. Calibration of the Optical Trap {#s2c} ------------------------------- We perform position and force calibrations using well-established procedures for trapping spherical beads [@pone.0057383-Neuman1]. First, the relationship between the PSD voltage in a given direction, V~x~, and the displacement of the bead from the equilibrium position, x, is determined by scanning a bead affixed to the coverslip surface across the laser beam. As shown in [Figure 4A](#pone-0057383-g004){ref-type="fig"}, the resulting profile V(x) is well-described by the derivative of a Gaussian function, and the slope of the central, linear part of the profile is the conversion factor from nanometers to volts, ξ~x~ (units of V/nm), which is subsequently used to convert all PSD signals along x from volts to nanometers. Next, the stiffness of the optical trap, κ, along the lateral directions is established by using three separate methods, which are based on measurements of particle position variance, power spectrum, and Stokes' drag, respectively. The stiffness along the axial direction is established using the particle position variance and power spectrum methods only. ![Calibration of the optical trap.\ (A) Response of the position-sensitive detector (PSD) as a bead immobilized on the coverslip is scanned across the laser beam along the x axis (red circles). The data are well-fit by the derivative of a Gaussian (red line), from which the nanometer to volt conversion factor is determined, ξ~x~ = 0.013 V·nm^−1^. (B) The position of a trapped bead along the x axis is monitored and the respective histogram is well-fit by a Gaussian (thick red line). The position variance is = 51 nm^2^ (or, the raw PSD signal variance is = 0.0086 V^2^). Using the principle of equipartition of energy, the trap stiffness is found: κ~x~ = 0.080 pN·nm^−1^. (C) The power spectrum of the position signal along the x (red line), y (blue line), and z (green line, inset) directions is shown, together with fittings to a Lorentzian function for x, y, and z (red, blue, and green thin lines, respectively). Measurement of the roll-off frequency of the fit yields the trap stiffness: κ~x~ = 0.063 pN·nm^−1^ and κ~y~ = 0.055 pN·nm^−1^. (D) In the Stokes drag calibration method, a trapped bead is subject to fluid drag by moving the stage at velocity v, resulting in the bead moving to a new equilibrium position that is measured (inset). Bead displacement vs. stage velocity data are shown for the x (red squares) and y (blue circles) axes. Linear fits to the data (red and blue lines) yield the slopes: s~x~ = 0.090 nm·µm^−1^·s, and s~y~ = 0.099 nm·µm^−1^·s (or the corresponding raw-signal slopes s~Vx~ = 0.0012 V·µm^−1^·s, and s~Vy~ = 0.0013 V·µm^−1^·s), which yield trap stiffness along each direction: κ~x~ = 0.076 pN·nm^−1^, and κ~y~ = 0.063 pN·nm^−1^. As discussed in the text, ξ and κ can alternatively be found using: ξ = (γ/s~V~), and κ = γξ/s~V~. This method yields ξ~x~ = 0.012 V·nm^−1^, and κ~x~ = 0.07 pN·nm^−1^ for the example presented here. The average value of κ~x~ using all four of these methods provides the best estimate of trap stiffness along x: κ~x~ = 0.072±0.007 (mean ± SD). Calibrations were performed using 0.54-µm diameter polystyrene beads.](pone.0057383.g004){#pone-0057383-g004} Briefly, for the first method, the variance of the position signal of a trapped bead () is measured and, using the principle of equipartition of energy, the stiffness κ~x~ is given by κ~x~ = , where k ~B~ is Boltzmann's constant and T is temperature (see [Figure 4B](#pone-0057383-g004){ref-type="fig"}). For the other two procedures, the drag coefficient (γ) of a bead of radius r immersed in a fluid of viscosity η, γ = 6πηr, is assumed to be known. In the second method, the power spectrum of the position signal of a trapped bead is computed and fitted by a Lorentzian profile (see [Figure 4C](#pone-0057383-g004){ref-type="fig"}). From a measurement of the roll-off frequency of the fit, f ~0~, the stiffness can be derived as κ = 2πγf ~0~. Finally, in the third method, a steady drag force is exerted on a trapped bead by moving the sample with the piezoelectric stage at constant speed (v~x~). The resultant viscous drag places the bead at a new equilibrium position, x = γv~x~/κ~x~ (from Stokes' law), that is measured. This procedure is then repeated for various v~x~, and the stiffness is inversely proportional to the slope of the linear fit of the x vs. v~x~ data (see [Figure 4D](#pone-0057383-g004){ref-type="fig"}). Alternatively, the variance and the Stokes' drag methods can be combined to provide simultaneous position (ξ) and force (κ) calibrations along the lateral directions, using the raw PSD signal, V~x,y~. The advantage of this approach over the previous independent methods is that it does not require an explicit position calibration (based on coverslip-stuck beads), and can therefore be applied to free beads. In this case, we measure the variance of the position signal, = k ~B~ Tξ~x,y~ ^2^/κ~x,y~ (units of V^2^), and the slope of the V~x,y~ vs. v~x,y~ linear fit in the Stokes' drag method, s~x,y~ = γξ~x,y~/κ~x,y~ (units of V·µm^−1^·s). These two relationships are combined to yield ξ~x,y~ = (γ/s~x,y~), and κ~x,y~ = γξ~x,y~/s~x,y~. Like the power spectrum and Stokes' drag methods, with the previous procedure it is assumed that the drag coefficient of the bead is known. For a sphere near a surface, the relationship γ = 6πηr for the hydrodynamic drag no longer holds, due to surface proximity effects [@pone.0057383-Neuman1]. Accordingly, to determine the drag coefficient, we take into account Faxen's law in all our measurements. Very good agreement is obtained between all calibration methods presented here, resulting in variations of ∼10% in values of κ~x,y~. The Persistence Length of dsDNA {#s2d} ------------------------------- A number of studies have measured the persistence length of dsDNA under a variety of buffer conditions [@pone.0057383-Smith1], [@pone.0057383-Wang1], [@pone.0057383-Seol1], [@pone.0057383-Baumann1]. We tested our optical trapping instrument by pulling on individual dsDNA molecules, using a standard surface-based assay [@pone.0057383-Seol1], at a relatively high ionic strength (\[Na^+^\] = 150 mM). Following established protocols, a ∼3.1 kb dsDNA template was attached by the 5′ end of one of its strands to a 0.7 µm-diameter polystyrene bead, and affixed by the 5′ end of the complementary strand to the coverslip surface (see [Figure 5A](#pone-0057383-g005){ref-type="fig"} and [Materials and Methods](#s3){ref-type="sec"}). The procedure to measure force-extension curves for dsDNA is described in Ref. [@pone.0057383-Wang1]. ![Stretching single dsDNA molecules.\ (A) Schematic of the surface-based assay, not to scale. (B) A typical force-extension record (black circles) together with a worm-like chain (WLC) polymer model fit (black line). (C) Histogram of contour length values. Fit to a Gaussian (black line) yields: 1116±2 nm (μ ± σ~μ~). (D) Histogram of persistence length values. Fit to a Gaussian (black line) yields: 47.0±0.2 nm (μ ± σ~μ~). Legends display data mean ± SEM.](pone.0057383.g005){#pone-0057383-g005} Briefly, surface-tethered beads were held by the laser tweezers and centered with respect to the surface attachment point, along the x and y directions, by adjusting the piezoelectric stage position such that stage motions in the ±x or ±y directions produced symmetric bead displacement responses (as measured by the PSD). Next, beads were placed at a specified distance (20 nm) above the coverslip surface, taking into account the effective focal shift (∼0.8) of the optical trap [@pone.0057383-Neuman1]. Then, the DNA molecule was stretched by moving the coverslip horizontally (using the piezoelectric stage) in steps of 10 nm in distance, 10 ms in duration. For each step, x,y,z bead displacements were recorded at a rate of 10 kHz and averaged in each direction. Finally, the extension of the elongated molecule (L) was determined from the known stage movement and the measured bead position with respect to the trap, and the force (F) acting on the molecule along the stretching direction was determined from the position of the bead with respect to the trap and the trap stiffness [@pone.0057383-Wang1]. Because the method used here to determine bead-surface distances is precise to 20 nm [@pone.0057383-Lang1], the distance between the bead and the surface was set to 0 nm to obtain force-extension curves. All DNA molecules were stretched in the ±x and ±y directions. To discriminate single from multiple tethers, only ±x and ±y symmetrical force-extension curves were considered for analysis. We report results from stretching along the y direction only. For each molecule, data corresponding to the +y scan direction were superimposed on the --y scan direction, and treated as a single y record. Care was taken to perform measurements within the respective linear regions of position and force calibrations. [Figure 5B](#pone-0057383-g005){ref-type="fig"} shows a typical single-molecule force-extension record. Maximal stretching forces of 4 pN were applied to all molecules. In this low-force regime, the elasticity of DNA is essentially entropic, and the corresponding F vs. L relationship is well-described by the worm-like chain (WLC) polymer model [@pone.0057383-Bouchiat1]:where L ~0~ and L ~p~ are the contour length and the persistence length of the polymer, respectively. Force-extension records were well-fit using the WLC model (see [Figure 5B](#pone-0057383-g005){ref-type="fig"}), from which values for L ~0~ and L ~p~ were inferred. Statistical analysis of data for L ~0~ and L ~p~ (see [Figs. 5C and 5D](#pone-0057383-g005){ref-type="fig"}) yielded measured values: nm and nm (mean ± SEM), which are in excellent agreement with the expected value for the contour length L ~0~ ∼1053 nm (see [Materials and Methods](#s3){ref-type="sec"}) and with previous reports of L ~p~ at comparable ionic strengths [@pone.0057383-Baumann1]. The Stepping of Kinesin along MTs {#s2e} --------------------------------- Kinesin is a homodimeric motor protein that is involved in intracellular transport, using the energy of ATP hydrolysis to ferry cargo along MTs [@pone.0057383-Asbury1]. Previous studies using optical trapping, among other single-molecule techniques, have determined that kinesin advances in 8.2-nm increments and can sustain retarding loads as large as ∼7 pN before stalling [@pone.0057383-Block2]. In traditional optical-tweezers based bead assays of kinesin, DIC microscopy is used to localize MTs [@pone.0057383-Block1]. Here, we demonstrate the versatility of our instrument by observing kinesin motors stepping along MTs, where we simultaneously use both optical trapping and the excellent visualization features available from our experimental setup. To record kinesin motion, a standard motility bead assay was carried out (see [Figure 6A](#pone-0057383-g006){ref-type="fig"}) [@pone.0057383-Svoboda2]. Single recombinant kinesin motors were attached to 0.54-µm polystyrene beads (see [Materials and Methods](#s3){ref-type="sec"}). Kinesin-carrying beads were optically trapped and placed directly above individual MTs that were immobilized on the coverslip surface, previously identified using the microscope's digitally-enhanced imaging (see [Figure 6B](#pone-0057383-g006){ref-type="fig"}). When kinesin binds to the MT in the presence of ATP in the buffer solution, it advances unidirectionally towards the 'plus' end of the MT. In our experiment, the (fixed) optical trap was used both to measure the stepping of kinesin, and to exert a retarding force on motors that increases as beads are transported away from the center of the laser beam. For MTs aligned along the x direction, the retarding force on kinesin is determined through Hookés law (F~x~ = -κ~x~·x). [Figure 6C](#pone-0057383-g006){ref-type="fig"} presents examples of kinesin stepping records obtained in this work, showing that expected steps of 8.2 nm are clearly resolved. Additionally, we observed maximal forces sustained by motors before stalling or detachment from the microtubule of ∼7 pN, consistent with previous, detailed studies of kinesin motility [@pone.0057383-Block2]. ![Kinesin stepping along MTs.\ (A) Schematic of the kinesin bead assay, not to scale. (B) A kinesin-carrying trapped bead is placed directly above an individual MT, visualized using digital image processing. Field of view is 27 µm × 6 µm. (B) Representative records of kinesin-driven bead motion (gray line: unfiltered trace acquired at 20 kHz, black line: trace after a 20-point median filtering), display steps at 8.2-nm intervals (dashed horizontal lines) and the development of forces up to ∼7 pN. \[ATP\] = 100 µM.](pone.0057383.g006){#pone-0057383-g006} Conclusions {#s2f} ----------- An optical trapping system has been developed with exceptional visualization features at minimal cost and complexity. The technique of CEBFM provides an alternative to elaborate microscopy techniques for high-contrast visualization, avoiding the inclusion of optical elements in the instrument (such as prisms, apertures or polarizers) that could compromise operation of optical trapping. Using our instrument, it is possible to achieve independent control of BFP detection (in optical trapping) and Koehler illumination (in sample visualization). The approach presented here is uniquely suited for single-molecule experiments and studies of biomaterials at the micro- and nano-scales. Unconventional optical trapping techniques, such as an optical torque wrench (for rotating microparticles), or holographic and interferometric optical tweezers (for creating multiple-trap arrangements) can be incorporated immediately without sacrificing imaging capabilities. The setup can also be readily expanded to incorporate additional features, such as a second trapping beam (to perform off the surface, multiple-bead assays) [@pone.0057383-Abbondanzieri1], or strategies to maintain a constant force during experiments (a 'force clamp') [@pone.0057383-Lang1], [@pone.0057383-Greenleaf1]. Materials and Methods {#s3} ===================== Unless specified, all reagents were purchased from Sigma. Microscope slides and coverslips were cleaned prior to use for 5 min in a plasma cleaner (Harrick Plasma) at 1 Torr (ambient air). Flow channels were made using double-sided tape as described [@pone.0057383-Appleyard1]. All buffers were filtered before use through 0.22 µm filters (Millipore). Preparation of dsDNA {#s3a} -------------------- A 3039-bp dsDNA coding for a fragment of the blr-1 gene of Trichoderma atroviride [@pone.0057383-CasasFlores1] was amplified by a polymerase chain reaction (PCR) (Applied Biosystems). To allow attachment of the dsDNA to a polystyrene bead and to the coverslip, two functionalized PCR primers (Integrated DNA Technologies) were used: a forward primer (biotin-5′-GGGCTTCTACCAGACAAACCA-3′), and a reverse primer (digoxigenin-5′-CGCTCTTCTCGTATTGAAGCC-3′). The reaction tube contained: 5.0 µL of the 5XPCR buffer (Promega), 2.5 µL of 25-mM MgCl~2~ (Promega), 0.5 µL of 100-mM dNTPs, 0.5 µL of 10-µM reverse primer, 0.5 µL of 10-µM forward primer, 1.0 µL of cDNA from the blr-1 gene (generous gift of Sergio Casas Flores, IPICYT, Mexico), 0.5 µL of Taq Polimerase (GoTaq, Promega), diluted in 15 µL of Milli-Q water. The PCR ran for 25 cycles with an alignment temperature of 58°C and 1 min for extension. The amplified fragment was purified using a QIAQuick PCR purification kit (Qiagen). The expected length of the dsDNA molecule is ∼1053 nm, computed by taking into account a 3039-bp chain with 0.34 nm rise per bp [@pone.0057383-Watson1], together with ∼20 nm for the biotin-streptavidin and digoxigenin-anti-digoxigenin linkages. Recombinant Kinesin {#s3b} ------------------- We expressed the homodimeric, recombinant kinesin construct DmK401, a His-tagged derivative of Drosophila melanogaster kinesin heavy chain, that includes the first 401 N-terminal residues (previously described [@pone.0057383-Asbury2]). Briefly, BL21(DE3) cells transformed with plasmid pCA1 (generous gift of Steven Block, Stanford University) were grown to logarithmic phase in Luria Broth medium (10 g/L tryotone, 5 g/L yeast extract, 10 g/L NaCl) supplemented with 0.1 mg/mL ampicillin (GIBCO). Kinesin expression was chemically induced with 1 mM IPTG (Invitrogen) at 27°C during 12 h. Cells were lysed by sonication in extraction buffer (200 mM Na~2~HPO~4~, 50 mM NaCl, 2 mM imidazole, 20 uM ATP, 2 mM MgCl~2~, 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride (PMSF), and protease inhibitor cocktail (P8465, Sigma); 1% Tween was added after sonication). Cellular lysate was clarified by centrifugation (30 min, 15,000 rpm, 4°C). Clarified lysate was stored at −20°C in 10% glycerol, and used in motility assays. Single-molecule Assays {#s3c} ---------------------- ### dsDNA flexibility {#s3c1} A solution of 0.05 mg/mL of antidigoxigenin (3210--0488, Spherotech) in phosphate buffer saline (137 mM NaCl, 2.7 mM KCl, 10 mM Na~2~HPO~4~, 2 mM KH~2~PO~4~) was introduced in a flow channel, and incubated for 10 min. After washing with 200 µL of washing buffer (5 mg/mL BSA, 77.4 mM Na~2~HPO~4~, 0.1% Tween), the channel was filled with a sample of 0.25 nM dsDNA diluted in phosphate buffer (77.4 mM Na~2~HPO~4~), and incubated for 10 min. Unbound DNA was removed by flowing 200 µL of washing buffer through the channel. Finally, 30 µL of 730-nm diameter, avidin-coated beads (generous gift of Steven Block, Stanford University), diluted in phosphate buffer to a final concentration of ∼1 pM, were introduced into the channel, and the flow cell was sealed using nail polish. ### Kinesin motility assay {#s3c2} The motility, bead assay follows previous experiments. Briefly, 10 µL of the stock solution of 540-nm diameter, streptavidin-coated beads (Spherotech, SVP-05--10) were diluted in 70 µL of PEMBSA buffer (4 mg/mL BSA, 80 mM PIPES, 1 mM EDTA, 4 mM MgCl~2~, pH 6.9), and sonicated for 10 min., after which 20 µL of penta-His biotin conjugate antibody (34440, Qiagen) were added. After incubating for 1 hour at room temperature, beads were washed 5 times by centrifugation, and stored at 4°C. To bind kinesin to the beads, antibody-coated beads were diluted in assay buffer (3 mg/mL BSA, 0.05 M potassium acetate, 100 µM ATP, 1 µM DTT, 80 mM PIPES, 1 mM EDTA, 4 mM MgCl~2~, 0.02 mM Taxol (TXD01, Cytoskeleton), pH 6.9), sonicated, and incubated for 12 h at 4°C with clarified lysate diluted in assay buffer at various concentrations. To immobilize MTs on coverslips, flow channels were prepared using coverslips coated with poly-L-lysine. A rack of plasma-cleaned coverslips was submerged in a solution of 600 µL of poly-L-lysine diluted with 300 mL of ethanol, incubated for 15 min, dried in an oven at 40°C, and stored in a closed container. Tubulin (TL238-C, Cytoskeleton) was polymerized to produce MTs as described [@pone.0057383-GutirrezMedina2]. Stabilized MTs were diluted in PEMTAX buffer (0.02 mM Taxol, 80 mM PIPES, 1 mM EDTA, 4 mM MgCl~2~, pH 6.9), introduced into the flow channel and incubated for 10 min. Unbound MTs were removed by washing the channel with 40 µL of PEMTAX. This procedure was followed to obtain samples used to tests the imaging quality of our microscope. To complete the kinesin motility assay, the channel was washed first with 60 µL of 20-mg/mL BSA diluted in PEMTAX (to minimize the sticking of beads to the coverslip surface), and then with 100 µL of assay buffer. Finally, 40 µL of kinesin-bead complexes were introduced into the channel and the flow cell was sealed. To minimize the presence of reactive oxygen species, final samples were protected with an oxygen scavenger system (0.25 mg/mL glucose oxidase, 0.03 mg/mL catalase, 4.7 mg/mL beta-D-glucose (MP Biomedicals)). To ensure motility in the single-molecule regime, kinesin dilutions were used in which only ∼50% of the tested beads in a given sample displayed movement. Data processing and analysis were performed using Igor Pro 5.0 (Wavemetrics). We thank J.S. Casas Flores, M.G. Cervantes Badillo, and C. García-García for continuous support and advice; F.J. Cárdenas Flores for a generous loan of the AOM driver; and P.E. Riley, C. García-García, N.R. Guydosh, and anonymous reviewers for comments on the manuscript. [^1]: Competing Interests:The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: CNHC STM BGM. Performed the experiments: CNHC STM BGM. Analyzed the data: CNHC STM BGM. Contributed reagents/materials/analysis tools: BGM. Wrote the paper: CNHC BGM. [^3]: Current address: Program in Biophysics, University of California, Berkeley, California, United States of America	{ "pile_set_name": "PubMed Central" }
1.. Introduction {#sec1} ================== Recent progress in direct electron detector technology, the provision of intense and coherent electron beams using field emission guns (FEGs and X-FEGs) in cryo-electron microscopes and advances in image-processing algorithms and sample preparation have led to the so-called 'resolution revolution' (Kühlbrandt, 2014[@bb15]; Mitra, 2019[@bb17]) in cryo-electron microscopy (cryo-EM). Indeed, two thirds of the more than 6000 entries currently in the Electron Microscopy Data Bank (EMDB; Patwardhan, 2017[@bb20]) have been deposited within the last five years. However, an average resolution for single-particle cryo-EM structures of between 5 and 6 Å for the 3DEM density maps deposited suggests that there is still some way to go before cryo-EM becomes a routine technique for the production of high-resolution structural information. Nevertheless, with this aim in mind, and coupled with significant advances in sample preparation and data processing, many national and international centres for cryo-EM have opened during the last few years (see, for example, Alewijnse et al., 2017[@bb2], and references therein; Stuart et al., 2016[@bb27]; Clare et al., 2017[@bb6]). In 2015, the European Synchrotron Radiation Facility (ESRF) took the decision to complement its portfolio of cutting-edge X-ray-based facilities for structural biology (Pernot et al., 2013[@bb21]; von Stetten et al., 2015[@bb26]; Linden et al., 2014[@bb16]; Mueller-Dieckmann et al., 2015[@bb19]) by acquiring, installing and commissioning a cryo-EM facility (CM01; <http://www.esrf.eu/home/UsersAndScience/Experiments/MX/About_our_beamlines/CM01.html>) based around a Titan Krios microscope and making it available to its structural biology user community. CM01 is managed as is any other public ESRF beamline, with user access via either peer-reviewed applications or a paid-for proprietary mechanism. CM01 forms part of the Partnership for Structural Biology (PSB; <http://www.psb-grenoble.eu>) and its operation and development is assured by staff from a consortium of institutes co-located on the European Photon and Neutron Campus (EPN) in Grenoble. Experiments at CM01 are currently restricted to pre-characterized projects for which the sample-freezing conditions have been optimized and initial cryo-EM analyses indicate a high potential for obtaining information to the secondary-structure level. In the intermediate future it is planned that this restriction will be removed and that access to CM01 will be based on scientific merit alone. Here, preliminary sample screening will be carried out using the PSB cryo-EM platform (<http://www.ibs.fr/research/research-groups/methods-and-electron-microscopy-group/electron-microscopy-platform/>) microscopes located at the Grenoble Institut de Biologie Structurale (IBS). Once this mode of operation is fully functional, it will provide a unique opportunity for scientists of the ESRF's international community with limited or no access to EM facilities to access the technique for scientifically important projects. 2.. General setup and infrastructure {#sec2} ====================================== CM01 is sited on the 'golden slab' of the Belledonne extension of the ESRF experimental hall (Fig. 1[▸](#fig1){ref-type="fig"} a), which was constructed as part of Phase 1 of the ESRF Upgrade. The location was chosen following preliminary measurements confirming that it would provide a stable, vibration-free environment coupled with minimal variations in magnetic field (Table 1[▸](#table1){ref-type="table"}) that might otherwise be associated with installing a cryo-electron microscope in the experimental hall of a synchrotron source. The infrastructure (Figs. 1[▸](#fig1){ref-type="fig"} b--1[▸](#fig1){ref-type="fig"} e) in which CM01 is hosted comprises a grey room (ISO8 level) air-conditioned with a temporal thermal stability of 20 ± 0.1°C within 30 min and ±0.8°C per 24 h, a technical cabinet and an experimental control cabin. This room also contains the high-voltage generator, electronics and fluidics cabinets as well as the computer and processor units for the detector. These are screened from the microscope itself (Fig. 1[▸](#fig1){ref-type="fig"} c). All walls and the ceiling are acoustically damped to avoid any external disturbances, while the water-chilling unit that is needed to cool the electro-magnetic constant power lenses of the microscope is located in a technical cabinet (Fig. 1[▸](#fig1){ref-type="fig"} d) outside the experimental hutch, again for acoustic and vibrational insulation reasons. The microscope around which CM01 is built is a Titan Krios G3 (ThermoFisher Scientific; <https://www.fei.com>) currently operated at its maximal voltage of 300 kV and equipped with a GIF Quantum LS energy filter coupled to a Gatan K2 summit direct electron-counting camera. The microscope also features a Volta phase plate (VPP), which increases the contrast and as such enables the high-resolution imaging of smaller macromolecules of less than 150 kDa (Danev & Baumeister, 2016[@bb8]). CM01's spacious control cabin is equipped with the computing infrastructure required to control both the microscope and the detector (Fig. 1[▸](#fig1){ref-type="fig"} e) and also includes infrastructure for data processing and backup. Before each experiment, the microscope is set up and aligned to reduce the astigmatism and coma values to less than 0.5 µm (AutoCTF) by the facility personnel, who also clip the grids if necessary, load them into the cassette and then into the microscope. The microscope is operated in EFTEM mode with a nanoprobe for data acquisition. Currently, automatic data collection for single-particle reconstruction is carried out using the manufacturer's control software package EPU v.1.11 (ThermoFisher). A typical user session involves 4 h for grid screening (up to six grids) followed by the selection of the best grid squares and hole positions for image acquisition. Many experiments are carried out at a nominal magnification of 130 000 (pixel size of 1.056 Å) using a detector flux density of 5 e^−^ per pixel per second (usually denoted the 'dose rate' in EM) and exposure times of 8 s per movie (40 frames per movie), resulting in an accumulated fluence of 40 e^−^ Å^−2^ (usually denoted as the 'dose' in EM). Although the setting up and launching of image acquisition is carried out by the CM01 staff, the data-collection strategies are discussed directly with the users, who are either present on-site or access the facility remotely (see below). Using the current setup, the throughput is about 75--80 movies per hour for collection in counting mode for grids with a 2 µm hole size (five movies per hole). 3.. Computing infrastructure, data pre-processing and presentation of data using EXI/ISPyB {#sec3} ============================================================================================ The cryo-EM facility greatly benefits from the computing and network infrastructure available at the ESRF for experiments on its X-ray beamlines (Fig. 2[▸](#fig2){ref-type="fig"} a). Images or movies recorded by the detector are directly transferred to the ESRF central storage system (NICE, Networked Interactive Computing Environment; <https://www.esrf.eu/Infrastructure/Computing/NICE>) using a dedicated fibre-optic connection (10 Gb s^−1^). This facilitates not only fast data transfer but also allows both on-the-fly data backup onto an external hard drive and image pre-processing. In general, raw data are maintained on disk for up to 50 days from the experiment date and are then archived on tape and stored without time limit. Data backup to external hard disk drives is usually carried out by the users using the same easy-to-use interface that is available on all ESRF beamlines for structural biology. Alternatively, users can also download and transfer data to their home storage system via an rsync command. CM01 implements the ESRF data policy (<http://www.esrf.fr/datapolicy>). Parallelized image pre-processing (Fig. 2[▸](#fig2){ref-type="fig"} b) using several GPU-equipped computing clusters is automatically launched on the collected movies using the SCIPION wrapper (Conesa Mingo et al., 2018[@bb7]; Gómez-Blanco et al., 2018[@bb11]). The imported movies are drift-corrected using MotionCor2 (Zheng et al., 2017[@bb30]) and contrast-transfer function (CTF) parameters are estimated using Gctf (Zhang, 2016[@bb29]). With the current GPU infrastructure, which is subject to constant upgrades, and for an incoming rate of up to 80 movies per hour, pre-processing results are calculated with almost no delay. The output of MotionCor2 and Gctf are uploaded into the ISPyB database (Delagenière et al., 2011[@bb9]) and are displayed, along with other metadata, as snapshots in EXI (Fig. 3[▸](#fig3){ref-type="fig"}; <https://exi.esrf.fr>). Several relevant parameters, including maximum resolution and astigmatism, are constantly uploaded and plotted both in EXI and on a separate screen inside the control cabin, allowing on-the-fly quality control of incoming data. In the imminent future, extension of the existing data-processing pipeline to include automated or semi-automated particle picking and the generation of 2D classes are foreseen. 4.. Sample-preparation laboratory and storage facility {#sec4} ======================================================== As noted above, experiments at CM01 are currently restricted to pre-characterized samples for which frozen grids have been prepared. However, and in preparation for the operation phase of CM01, in which preliminary sample screening and sample optimization will be carried out using the PSB cryo-EM platform, a fully equipped sample-preparation laboratory is associated with the facility (Fig. 4[▸](#fig4){ref-type="fig"}). The equipment in this laboratory comprises a ThermoFisher Scientific Mark IV Vitrobot and all of the tools necessary for grid preparation, including a glow-discharge apparatus (Ted Pella easyGlow). Users currently applying for microscope time to analyse grids based on pre-characterization can ask for access to this laboratory during the application procedure (see below). External users of CM01 have the option of either coming to the ESRF to oversee their experiments or sending samples for remote and mail-in data collection. In both cases, cryo-EM grids can be sent to the ESRF prior to the experiment date and stored using individually barcoded grid-boxes provided by the ESRF (<https://www.mitegen.com/product/mitegen-cryo-em-pucks-generation-2-0/>) in a dedicated storage area. As for experiments on the ESRF's X-ray-based structural biology beamlines, samples must be registered using the ISPyB sample-tracking module (Delagenière et al., 2011[@bb9]). In the case that more data need to be collected from a set of grids, samples can be stored for up to six months, a period that can be prolonged upon request. Again, as for the ESRF's X-ray-based structural biology beamlines, if users do not come to the ESRF for their experiment then the costs of sample transport to and from Grenoble are met by the ESRF (see <https://www.esrf.eu/UsersAndScience/Experiments/MX/How_to_use_our_beamlines/Prepare_Your_Experiment/Dewar_sending>). 5.. How to apply for access {#sec5} ============================= Currently, CM01 delivers ∼600 8 h shifts per year for external user experiments. Experiments are scheduled on a three-day, two-day or one-day basis according to the beamtime demanded in the proposal, with the vast majority of experiments using three days. 10% of the beamtime given is dedicated to in-house research. One day per week is reserved for microscope maintenance and the remaining time is used for commissioning (including service interventions and upgrades) and developments on the microscope. As for all of the ESRF's facilities for structural biology, access to CM01 can be on either a public or a proprietary basis. For the former, applications, peer-reviewed by the ESRF's MX Beam-time Allocation Panel (MX-BTAP) comprising several cryo-EM specialists, can currently be submitted at any time via the ESRF's Rolling Access Mechanism (<http://www.esrf.eu/UsersAndScience/UserGuide/Applying/MXApplications>). For accepted proposals, the ESRF provides financial support for either user travel to Grenoble (including accommodation and subsistence) or for sample transport to/from Grenoble. Aside from this core mission of providing user service, hands-on workshops on sample preparation (see, for example, <http://www.esrf.eu/cryo-em2019-1-workshop.html>) as well as on theoretical principles of cryo-EM are held on a regular basis. The normal waiting time from proposal submission until experiment is currently around two months. Proprietary access to CM01 is via the ESRF's Business Development Office (BDO; <http://www.esrf.fr/Industry/contact-industrial-services>). 6.. Microscope benchmarking {#sec6} ============================= In order to benchmark the microscope installed on CM01, we used Tobacco mosaic virus (TMV) as a test sample for single-particle reconstruction. TMV grids were prepared using the Vitrobot by pipetting 3.5 µl of 3 mg ml^−1^ TMV onto Quantifoil 1.2/1.3 400 mesh grids (<https://www.quantifoil.com/>); excess liquid was blotted off for 2 s. The K2 detector was used in counting mode for automatic data collection using the EPU software. Movies were collected as gain-normalized and unpacked MRC files at a magnification of 130 000, yielding a calibrated pixel size of 1.067 Å. Drift correction was carried out using MotionCor2 (Zheng et al., 2017[@bb30]) and CTF estimation was performed by Gctf (Zhang, 2016[@bb29]). Further image-processing steps were performed using RELION 2.1 (Kimanius et al., 2016[@bb14]; Scheres, 2012[@bb24]). Segments were automatically picked and extracted with about 90% overlap between segments. 2D reference-free classification was performed and class averages showing high-resolution features were selected and included for 3D refinement. As a final step, local refinement of the CTF parameters was performed using Gctf (Zhang, 2016[@bb29]). The map was further post-processed by automatically applying a B factor of −83 Å^2^, yielding a gold-standard FSC (Henderson et al., 2012[@bb13]) resolution of 2.3 Å (Fig. 5[▸](#fig5){ref-type="fig"}), one of the highest resolutions for the TMV structure to be published (Weis et al., 2019[@bb31]). Even though the resolution of the TMV structure presented here is not the highest to be reported (Song et al., 2019[@bb25]), the result obtained clearly shows the very high quality of data that can be obtained using CM01. The final helical parameters are 1.408 Å for the helical rise and 22.03° for the helical twist. One TMV protomer (PDB entry [4udv](http://scripts.iucr.org/cgi-bin/cr.cgi?rm=pdb&pdbId=4udv); Fromm et al., 2015[@bb10]) was rigid-body fitted into the density map using UCSF Chimera (Pettersen et al., 2004[@bb22]) and the atomic positions were further refined using PHENIX real-space refinement (Afonine et al., 2018[@bb1]). The final model fits well into the density with very good statistics (Table 2[▸](#table2){ref-type="table"}). The TMV protomer modelled consists of 153 residues (residues 65 and 98--102 were not visible). Most of the side-chain densities are well defined (Fig. 5[▸](#fig5){ref-type="fig"}) and only 17% of residues exhibit poorly defined side-chain density. Of these, about 40% are negatively charged amino acids such as aspartic and glutamic acid residues, which are prone to radiation damage (Hattne et al., 2018[@bb12]). 7.. The first year of operation {#sec7} ================================= The vast majority of users are from academic institutions located in the 13 ESRF member states (Fig. 6[▸](#fig6){ref-type="fig"}). At the time of writing, seven publications based, in whole or in part, on data collected at CM01 have already been published. These include different states of the 5HT3 receptor, which allow a description of its activation cycle upon binding to its serotonin substrate (Polovinkin et al., 2018[@bb23]), the structure of the membrane complex (MC) of a type VI secretion system (T6SS), providing insights into the mechanism of action of the MC complex during T6SS assembly (Cherrak et al., 2018[@bb5]), the natural tetrameric structure of human butyrylcholinesterase (BChE; Boyko et al., 2019[@bb4]), the structure of Hantaan virus (Arragain et al., 2019[@bb3]), the structure of the human ferritin--transferrin receptor 1 complex (Montemiglio et al., 2019[@bb18]) and the structure of adenovirus type 3 fibre with its receptor desmoglein 2 (Vassal-Stermann et al., 2019[@bb28]). However, given that most of the experiments carried out are rather challenging, no reliable statistics on publications per year can yet be given. 8.. Perspectives {#sec8} ================== With the rapid growth of the cryo-EM community in the last few years, the demand for data collection at high-end microscopes such as the Titan Krios installed at CM01 has skyrocketed. Since its installation, CM01 has proved to be a highly successful facility, acting in concert with the ESRF's X-ray-based beamlines for structural biology as a source of primary or complementary structural information in a wide range of projects. One of the major driving forces for the creation of CM01 was to provide a facility that can be accessed by all of the ESRF's international structural biology user community. However, a lack of infrastructure for sample preparation and screening remains a bottleneck, particularly for researchers in smaller laboratories. To address this and to broaden the accessibility of cryo-EM to researchers working in ESRF member states, the PSB Cryo-EM platform will, for projects approved by the ESRF MX-BTAP, soon provide additional services to users based on one of the following models: (i) the screening of frozen grids and a preliminary data collection if negative-stain characterization has already been carried out and (ii) sample characterization using negative-staining methods followed by cryo-EM sample preparation and screening. If the above steps are successful, data collection using the Titan Krios at CM01 will follow. The ESRF will also continue its training workshops for external users on both theoretical principles and the practical aspects of sample preparation. We would like to thank all of the ESRF support groups, such as the technical infrastructure division, safety, communication and data-analysis units and computing and network services, for their invaluable support during the construction, setup and commissioning of the microscope. The support of Francesco Sette (Director General of the ESRF), Harald Reichert (Director of Research of the ESRF), Winfried Weissenhorn (Director of the IBS) and Stephen Cusack (Head of EMBL Outstation Grenoble) was crucial for this project. We would also like to mention FEI, Gatan and particularly their application support teams for continuous training and support, specifically Pierre-Yves Mille, Remi Boucher, Bruno Barlassina, Ana Alarcon, Wim Woorhut, Eric Bathelt, Felix de Haas, Thomas Hoffmann and Evgeniya Pechnikova. ![CM01 infrastructure. (a) Location of CM01 in the Belledonne extension of the experimental hall of the ESRF. (b) Schematic top view of CM01 showing the experimental hutch (EH) in blue, the technical gallery in yellow and the control cabin (CC) in violet. (c) The CM01 experimental hutch (grey room) showing the Titan Krios microscope (1), the duct sock porous mesh for uniform air flow (2), SF6 extraction (3) and the acoustic damping sheets (4) behind which the high-voltage generator is located. (d) Technical gallery, where the chiller unit (indicated) is placed. (e) The large CM01 control cabin.](d-75-00528-fig1){#fig1} ![(a) Schematic of the IT infrastructure at CM01 for data collection and transfer connections between computers (pc TF, pc GATAN and the microscope are connected via ethernet). (b) The single-particle cryo-EM image pre-processing workflow currently carried out at CM01. The collected movies are motion-corrected and CTF values are estimated. The obtained metadata are then accessed and uploaded to the EXI web interface.](d-75-00528-fig2){#fig2} ![The EXI web interface (<https://exi.esrf.fr>) provides a summary of the data-collection parameters and pre-processing results on the fly. (a) Snapshot of the pre-processing results showing the unaligned average (first column), motion-corrected average (second column) and CTF estimation (third column) images and associated values. (b) Overview of data quality. The first three rows show the image series plots and histograms for the information limit in the micrographs, the astigmatism angle and defocus U and V (astigmatism in the middle frame), respectively. The last plot indicates the average motion per frame in micrometres from the alignment of the acquired movies.](d-75-00528-fig3){#fig3} ![Laboratory access is provided to users for cryo-EM sample preparation. The laboratory includes a Vitrobot positioned in the fume hood, as can be seen on the left side.](d-75-00528-fig4){#fig4} ![3D reconstruction of TMV particles. (a) 3D reconstruction of TMV. Alternate subunits are coloured in yellow and green. The scale bar represents 5 nm. (b) Masked-corrected Fourier shell correlation (FSC) curve for resolution. (c) An asymmetric unit modelled in the electron density. (d) Some examples representing the quality of the EM density and the model.](d-75-00528-fig5){#fig5} ![Breakdown of the shifts delivered in 2018 by proposal country of origin for experiments at CM01.](d-75-00528-fig6){#fig6} ###### Vibrational and electromagnetic measurements inside the experimental hutch of CM01 The mechanical vibrations are measured in units of g (the acceleration owing to gravity in m s^−2^) in three directions approximately at the position of the column at the floor level. Both the AC and DC electromagnetic fields (EMF) are measured in nanotesla p/p in three directions to the column position for three different heights \[the GIF approximately 0.5 m from the floor, the compustage approximately 1.5 m from the floor and the gun (source) approximately 2.5 m from the floor\]. Front to back Left to right Vertical ---------------------- --------------------- --------------------- -------------------- Mechanical vibration 1.1 (1.56 µm s^−1^) 1.1 (1.56 µm s^−1^) 2.3 (3.2 µm s^−1^) EMF AC GIF level 42 (50 nT) 21 (50 nT) 29 (75 nT) Stage level 43 (80 nT) 20 (80 nT) 33 (80 nT) Source level 42 (80 nT) 21 (80 nT) 39 (80 nT) Filtered GIF level 19 (50 nT) 18 (20 nT) 23 (75 nT) Stage level --- --- --- Source level --- --- --- Near DC GIF level 14 (50 nT) 7 (50 nT) 19 (75 nT) Stage level 6 (80 nT) 11 (80 nT) 19 (80 nT) Source level 3 (80 nT) 12 (80 nT) 58 (80 nT) ###### Data-collection and refinement statistics for the single-particle reconstruction of TMV -------------------------------- -------------- Data collection Accelerating voltage (kV) 300 Nominal magnification 130 000 Calibrated pixel size (Å) 1.067 Total No. of movies collected 357 Total fluence (e^−^ Å^−2^) 40 No. of frames 40 Defocus range (µm) −1.0 to −3.0 Data processing Final segments 109763 Helical parameters Twist (°) 22.04 Rise (Å) 1.41 Gold-standard resolution (Å) 2.3 MolProbity statistics All-atom clashscore 2.41 Ramachandran plot Outliers (%) 0 Allowed (%) 3.97 Favoured (%) 96.03 Rotamer outliers (%) 0 C^β^ deviations (%) 0 Deviations from ideal values Bond lengths (µm) 0.007 Bond angles (°) 1.030 Chirality (Å^3^) 0.060 Planarity (°) 0.009 Dihedral (°) 8.534 Map--model CC Main chain (760 atoms) 0.68 Side chain (513 atoms) 0.63 EMDB ID EMD-4928 -------------------------------- --------------	{ "pile_set_name": "PubMed Central" }
1. Introduction =============== Suture anchors are increasingly used to fix soft tissue to bone.^\[[@R1]\]^ The modified Brostrom operation (MBO) is the most widely used procedure for the correction of ankle instability.^\[[@R2]\]^ Favorable outcomes of the MBO using suture anchors have been reported.^\[[@R1],[@R3]\]^ The infection rate after such surgery is low: 0% to 4.5%.^\[[@R1],[@R2],[@R4]\]^ However, a suture anchor is not free of complications because a suture anchor acts as an implant. After shoulder surgery, suture anchor complications include pull-out,^\[[@R5],[@R6]\]^ arthritis,^\[[@R5]\]^ and osteolysis.^\[[@R7]\]^ However, complications of suture anchors used in ankle surgery are rare. We encountered a chronic infection of the sinus tract after the MBO using a suture anchor. The wound healed completely after removal of the sinus tract and infected suture anchor. 2. Case description =================== This case report was approved by the Institutional Review Board of Eulji Medical Center, and the patient gave written informed consent. The patient was a healthy 18-year-old male who underwent anterior talofibular ligament (ATFL) and calcaneofibular ligament (CFL) repair using a suture anchor at a local clinic 1 month after being injured while playing soccer. There were no other injuries. The ATFL and CFL were repaired using a JuggerKnot 1.45-mm LongSoft Anchor (Zimmer Biomet, Warsaw, IN). The ATFL was repaired at the talar side and the CFL at the fibular side. According to the medical record of the first hospital, an infection featuring a purulent exudate developed in the surgical site 1 week after the operation. Irrigation and debridement were performed in the operating room every day for 10 days. The wound discharge and swelling decreased, and a simple suture was placed. There was no evidence of wound infection, and the suture was removed and the patient discharged. However, abrupt wound discharge and dehiscence developed after suture removal. He did not visit that hospital again, for personal reasons, and personally dressed the chronic wound at home. He visited our outpatient clinic 1 year later. A clinical examination revealed a 1.5 × 2-cm-sized ulcerative wound accompanied by a red-black discharge on the anterior border of the right lateral malleolus (Fig. [1](#F1){ref-type="fig"}). We could approximate the wound to the joint capsule level by applying forceps at the wound site. A simple standing anteroposterior radiograph of the right ankle revealed mild lateral soft-tissue swelling. Magnetic resonance imaging revealed a sinus tract running from the wound to the talar body across the anterior joint capsule, and bone marrow edema surrounding the talus (Fig. [2](#F2){ref-type="fig"}). The American Orthopedic Foot and Ankle Society (AOFAS) Ankle-Hind Foot score was 45. Intravenous vancomycin was commenced on the day of surgery. ![A 1.5 × 2-cm-sized ulcerative wound accompanied by a red-black discharge on the anterior border of the right lateral malleolus.](medi-97-e13738-g001){#F1} ![A T2-weighted axial image (A) and a proton density coronal image (B) showing the sinus tract (arrow) running from the wound to the talar body (crossing the anterior joint capsule), and bone marrow edema (arrowhead) surrounding the talus.](medi-97-e13738-g002){#F2} In the operation theater, we excised the sinus tract, sent material for culture, and performed irrigation and debridement; our surgical approach was via the chronic ulcerative wound. Surgery was performed with the patient right side up in the lateral position after placement of a pneumatic tourniquet under general anesthesia. A rectangular skin incision encompassing the ulcerative wound was created, followed by dissection along the chronically infected sinus tract (Fig. [3](#F3){ref-type="fig"}). Intraoperatively, the sinus tract was seen to lie on top of the suture anchor (Fig. [4](#F4){ref-type="fig"}). The suture anchor was loose and infected and was removed (Fig. [5](#F5){ref-type="fig"}). The chronically infected sinus tract was excised (Fig. [6](#F6){ref-type="fig"}) and a pathological examination was performed (Fig. [7](#F7){ref-type="fig"}). A hole through the sinus tarsi was evident near the wound; we performed wound culture (Fig. [8](#F8){ref-type="fig"}). As the surgical site was clean and no discharge was evident, we decided to perform immediate repair. As a rectangular wound cannot be easily sutured, we created rectangular superior and inferior skin flaps (Fig. [9](#F9){ref-type="fig"}). The anterior flap was lowered and the posterior flap raised to create a zigzag-shaped wound (Fig. [10](#F10){ref-type="fig"}). After simple suturing, a short leg cast was applied and the wound dressed through the cast window (Fig. [11](#F11){ref-type="fig"}). The patient received intravenous vancomycin during 14 days of hospitalization. He was discharged on postoperative day 14 with a 1-week prescription of oral amoxicillin/clavulanic acid. At 4 weeks postoperatively, we removed the suture and cast; no wound dehiscence was evident. No bacterium grew on wound culture; the pathological diagnosis was chronic inflammation and abscess formation. He visited our outpatient clinic for scheduled follow-up at 12 months after surgery. The wound had completely healed (Fig. [12](#F12){ref-type="fig"}). He reported no symptoms of infection and regained full function of the right ankle. The anterior drawer test was negative. The AOFAS Ankle-Hind Foot score improved to 90. ![A rectangular skin incisions encompassing the ulcerative wound was created, followed by dissection along the chronically infected sinus tract.](medi-97-e13738-g003){#F3} ![Intraoperatively, the sinus tract was found to lie over the suture anchor. The anchor was loose and infected, and was removed.](medi-97-e13738-g004){#F4} ![The removed suture anchor.](medi-97-e13738-g005){#F5} ![The chronically infected sinus tract was excised.](medi-97-e13738-g006){#F6} ![The sinus tract; pathological examination.](medi-97-e13738-g007){#F7} ![A hole through the sinus tarsi was evident at the wound site.](medi-97-e13738-g008){#F8} ![Rectangular skin flaps were created superior and inferior to the wound.](medi-97-e13738-g009){#F9} ![The anterior flap was lowered and the posterior flap raised to create a zigzag-shaped wound.](medi-97-e13738-g010){#F10} ![After simple suturing, a short leg cast was applied and the wound dressed through the cast window.](medi-97-e13738-g011){#F11} ![The wound exhibited no complication or recurrence of infection at the 1-year follow-up exam.](medi-97-e13738-g012){#F12} 3. Discussion ============= Chronic lateral ankle instability (CLAI) is defined as an instability more than 6 months in duration, manifesting as repeated episodes of "giving out", persistent pain, and an inability to resume the pre-injury activity level.^\[[@R1]\]^ About 50 surgical procedures are available to treat CLAI. Anatomical repair of the ATFL and CFL can restore normal ankle and subtalar joint mechanics. The MBO is the most popular technique for anatomical reconstruction of lateral ankle ligaments; the success rate is 85% to 90%.^\[[@R8]--[@R10]\]^ Suture anchors were developed to facilitate soft tissue fixation to bone, and have been placed during both open and arthroscopic MBOs.^\[[@R1]--[@R3],[@R11],[@R12]\]^ Such placement is technically simple and allows accurate ligament re-attachment to the normal anatomical origins.^\[[@R1]\]^ However, suture anchors, being implants, can cause problems.^\[[@R5],[@R13],[@R14]\]^ Reported complications associated with suture anchor used in shoulder surgery include pull-out,^\[[@R5],[@R15]\]^ arthritis,^\[[@R5]--[@R7]\]^ and osteolysis.^\[[@R7],[@R16]\]^ To the best of our knowledge, complications associated with suture anchor use during ankle surgery have been rarely reported. Suture anchors implanted during orthopedic surgery can cause infections, both perioperatively and postoperatively. In our present case, the exact cause of infection was unknown, but given that a chronically infected sinus tract lay over the suture anchor, the anchor is likely to have caused the infection. The removal of orthopedic implants is important to control infection. We removed the infected suture anchor and the chronically infected sinus tract; the wound healed after repair without infection recurrence. A few reports have described the incidence and management of culture-negative infections developing after orthopedic surgery. Such infections arise after prior antibiotic treatment to counter infections with unusual organisms such as fungi, mycobacteria, or bacteria that form biofilms. Two case reports have appeared.^\[[@R17],[@R18]\]^ Bacterial cultures were negative, but Mycobacterium was found in the wounds. This pathogen must be considered if, postoperatively, wounds present with chronically discharging sinuses, even in healthy non-immunocompromised patients. In our case, the sinus tract was chronically infected, but Gram staining and bacterial cultures were negative. We did not stain for acid-fast bacilli or seek to culture Mycobacterium; these tests should be considered in cases featuring the chronic infection of a sinus tract. In conclusion, the MBO with placement of a suture anchor affords good results, but the anchor may become a source of infection. Removal of the chronically infected sinus tract and the infected suture anchor are required to heal the wound. Acknowledgments =============== The authors would like to thank the Soonchunhyang University Research Fund for support. Author contributions ==================== Conceptualization: Woo Jong Kim. Investigation: Ki Won Young. Software: Sung Hun Won, Kyu Hwan Bae. Supervision: Hong Seop Lee. Writing -- original draft: Hong Seop Lee. Writing -- review & editing: Chang Hyun Kim, Hong Seop Lee. Hong Seop Lee orcid: 0000-0002-3589-314X. Abbreviations: AOFAS = American Orthopedic Foot and Ankle Society, ATFL = anterior talofibular ligament, CFL = calcaneofibular ligament, CLAI = chronic lateral ankle instability, MBO = modified Brostrom operation. WJK and KWY contributed equally to this study and should be considered co-first authors. No benefit in any form has been or will be received from a commercial party related directly or indirectly to the subject of this article. The authors have no potential conflicts of interest to disclose.	{ "pile_set_name": "PubMed Central" }
The face is perhaps the most powerful communication channel for human social interaction, enabling the observer to quickly infer information about the sender[@b1][@b2]. Faces allow not only recognizing a unique identity, but also communicating basic emotional states. Dating back to Darwin[@b3], faces and facial expressions have been extensively studied in psychology and neuroscience, resulting in elaborate cognitive[@b4] and neuroscientific theories (e.g. see refs [@b5] and [@b6]). Some theories also had a big impact on other disciplines. Ekman's theory of universal basic emotions[@b7][@b8] and the resulting Facial Action Coding System (FACS) became very popular in computer graphics and are a fundamental concept in current facial animation systems. By blending between different FACS expressions a complex facial expression is created for a virtual character. In this regard, advancements in face perception research are also of high interest for many commercial productions in the game or visual effects industry. Convincing characters highly contribute to the story and visual quality of a movie binding a lot of time and money at production time. In the movie "Maleficent" for example, expensive hardware, custom software, and months of work by highly skilled professionals and researchers was required to create the digital faces of the flower pixies[@b9]. However, little is known about how stylized faces are processed by the brain and about which facial details are important in particular. Even less is known about how a presented emotional expression interacts with stylization of a face. Concerns about the influence of stylization have recently also become an issue for face perception research in psychology, where computer generated characters start to replace photographs of real people due to higher flexibility and better experimental stimulus control[@b10]. Regarding the effect of character realism on human perception, so far, the uncanny valley theory[@b11] dominates the discourse in research and practice. According to Mori, humanoids are in general perceived as more appealing if they are more realistic. However, humanoids that are more realistic than cartoon characters, but do not fully achieve human-likeness, may cause an "eerie feeling" and fall into the "uncanny valley". This theory has been developed on the basis of Mori's personal observation in the 1970s and has been subject of some empirical research in the last decade. Initial research used morphs of different characters[@b12] or videos of different robots at different stylization levels[@b13]. Also image editing algorithms[@b14] or different rendering algorithms[@b15] have been applied to create facial stimuli at different stylization levels. Other work investigated the threshold beyond which characters are perceived as real, by morphing between photographs and puppet faces[@b16] or photographs and virtual faces[@b17]. Despite its intuitive appeal, the uncanny valley concept still lacks empirical support[@b18] and no final agreement exists about the measurement scales[@b19] or potentially underlying perceptual mechanisms. However, facial features such as skin impurities[@b20] or strong deviations of facial proportions like too big eyes[@b21] have been identified as reliable characteristics of unappealing characters. Event-related potentials (ERPs) are an excellent way to analyse face perception, as the measured brain responses can distinguish between highly automatic and more controlled stages of processing and do not require overt responses that may be distorted by experimental demand characteristics. Neuroscientific research has shown that face processing strongly activates dedicated areas in the visual cortex, in particular the occipital face area[@b22][@b23] and the fusiform face area[@b24]. The fusiform face area is also involved in generating the face-sensitive N170 component in ERP studies[@b25]. Previous work showed that the N170 component peaks selectively for faces[@b26]. Real faces elicit a stronger N170 compared to abstract sketches of faces, but the difference is not statistically significant compared to schematic faces[@b27], suggesting a gradual increase. Larger N170 amplitudes are found for human faces compared to those of other species[@b28]. Attractiveness also seems to induce small modulations in the N170 for characters of the same stylization level[@b29][@b30]. Further, it has been found that baby faces cause a higher N170 amplitude than adult faces, most likely due to neotenic features, like proportionally larger eyes[@b31]. Finally, a recent meta-analysis indicated that larger N170 amplitudes are observed for angry, fearful, and happy than for neutral facial expressions[@b32]. Moreover, emotion effects at the N170 can be found for faces of medium stylization[@b33] and even for robots with rather schematic faces[@b34]. Enhanced processing of emotional faces is also reflected in an enhanced Early Posterior Negativity (EPN), as well a larger Late Positive Potential (LPP)[@b35][@b36]. The EPN indicates early attention mechanisms, whereas LPP is viewed as reflecting higher-order evaluation and episodic memory encoding (for an overview see ref. [@b37]). Both EPN and LPP amplitudes are sensitive to voluntary attention deployment to a stimulus and can be increased by explicit attention instructions[@b38][@b39][@b40]. Both ERP components are also sensitive to emotional content of various stimulus types, including faces[@b41][@b42][@b43][@b44][@b45]. This emotion sensitivity is thought to reflect intrinsically motivated attention to emotional stimuli and often co-varies with stimulus intensity or arousal[@b35][@b46][@b47][@b48]. As, in the absence of relevant social context, participants typically report less subjective arousal for picture of emotional faces than for pictures of emotional scenes, differences in arousal-level might account for generally smaller ERP emotion effects in response to faces compared to scenes[@b49]. Apart from emotional expressions, larger EPN[@b50] and LPP amplitudes[@b30][@b50] are reported for attractive faces compared to unattractive faces, also applying to moderate stylizations[@b29]. Interestingly, in a recent study contrasting real neutral faces with neutral faces of puppets, no differences at the N170 level were observed, while from 400 ms onwards a larger LPP was found for real faces[@b51]. This was attributed to the salience and unique identity of a real face and to mentalizing about the depicted individual[@b51], as generally computer generated faces are harder to remember[@b52][@b53]. In this vein, manipulating perceived uniqueness or distinctiveness via shape or reflectance manipulations of initially non-distinctive real faces has been found to result in a larger late positivity as well as a better memory performance[@b54][@b55]. Further, biographical information facilitates the processing of real faces at late ERP components[@b56][@b57], which has recently also been shown for realistic computer generated faces[@b58]. The combined evidence suggests that a unique identity, either pre-existing or ascribed, enhances processing particularly at late stages. Against the above background, the present research systematically addresses how brain responses known to reflect distinct stages of face processing vary across different levels of character stylization and for different emotional expressions. While previous work, in general, compared a real photograph of a person to a single stylization level, no previous study systematically manipulated face stylization of the same identity and studied how this influences brain responses towards emotional and neutral expressions. Our stimulus set exhibits six levels of stylization for two identities, depicting happy, neutral and angry expressions[@b20]. These cover a wide range of stylization, were professionally created, and match quality and style of current movies. All stimuli are rendered under the same conditions, thus reducing inconsistencies to a minimum. We expected to find strongest responses for real faces at the N170, EPN and LPP[@b26]. For very abstract faces the proportionally larger eyes and uniformly coloured textures might also cause a strong N170[@b31]. For angry and happy faces, we expected to find larger N170, EPN and LPP amplitudes. Interactions between level of stylization and emotional expression were explored on all components. Moreover, source analyses were employed to uncover the cortical generators of stylization effects, addressing the possibility of differential recruitment of visual areas such as fusiform or occipital face areas. Results ======= ERP results ----------- ### N170 For the N170 component significant modulations of stylization (F~(5,160)~ = 15.93, p \< 0.001, partial η~P~^2^ = 0.332; see [Fig. 1](#f1){ref-type="fig"}), emotion (F~(2,64)~ = 50.33, p \< 0.001, partial η~P~^2^ = 0.611; see [Fig. 2](#f2){ref-type="fig"}), as well as an interaction of stylization with emotion (F~(10,320)~ = 2.44, p = 0.008, partial η~P~^2^ = 0.071) were found over the two large symmetrical occipital sensor groups (see [Fig. 1](#f1){ref-type="fig"}). For stylization levels, we tested a linear (F~(1,32)~ = 0.09, p = 0.765, partial η~P~^2^ = 0.003), compared to a u-shaped, quadratic, contrast (F~(1,32)~ = 68.50, p \< 0.001, partial η~P~^2^ = 0.682), and found a u-shaped form for the face-specific N170 component. Although these u-shaped forms are visible and highly significant for all emotions, we identified with increased realism less intense reactions towards happy expressions and more intense reactions towards angry faces (see [Fig. 1](#f1){ref-type="fig"}, upper panel). Moreover, emotional modulations resulted in a stronger quadratic (F~(1,32)~ = 59.81, p \< 0.001, partial η~P~^2^ = 0.651), compared to a linear contrast (F~(1,32)~ = 41.64, p \< 0.001, partial η~P~^2^ = 0.565), showing the most pronounced N170 for angry faces, smallest for neutral faces and in between happy expressions (all ps \< 0.001). Additionally, there was a trend for a main effect of channel group (F~(1,32)~ = 3.37, p = 0.075, partial η~P~^2^ = 0.095) and an interaction of stylization with channel group (F~(5,160)~ = 3.30, p = 0.025, partial η~P~^2^ = 0.087), showing in tendency a u-shaped distribution with two maxima over the left and a u-shaped distribution with one maximum over the right sensor cluster. The interaction of emotion with channel group (F~(2,64)~ = 0.58, p = 0.532, partial η~P~^2^ = 0.018) and triple interaction were both insignificant (F~(10,320)~ = 1.03, p = 0.471, partial η~P~^2^ = 0.031). ### EPN In the EPN time range, over the same cluster, main effects of emotional content (F~(2,64)~ = 19.89, p \< 0.001, partial η~P~^2^ = 0.383) and laterality were observed (F~(1,32)~ = 33.83, p \< 0.001, partial η~P~^2^ = 0.514; see [Fig. 2](#f2){ref-type="fig"}). The EPN effect was somewhat better explained by a u-shaped form (F~(1,32)~ = 22.73, p \< 0.001, partial η~P~^2^ = 0.415), compared to a linear contrast (F~(1,32)~ = 17.59, p \< 0.001, partial η~P~^2^ = 0.355), where the strongest responses were found for angry faces, least responses to neutral faces and happy faces in between. The effect of laterality showed a larger EPN on the right compared to the left sensor group. There was no effect of stylization (F~(5,160)~ = 1.56, p = 0.215, partial η~P~^2^ = 0.046), and no interaction of stylization with emotion (F~(10,320)~ = 1.04, p = 0.412, partial η~P~^2^ = 0.031), or channel group (F~(5,160)~ = 0.80, p = 0.555, partial η~P~^2^ = 0.024). Further, there was no interaction of emotion with channel group (F~(2,64)~ = 0.76, p = 0.474, partial η~P~^2^ = 0.023) and no triple interaction (F~(10,320)~ = 1.59, p = 0.108, partial η~P~^2^ = 0.047). ### LPP In the LPP time window, over a large parietal sensor group, we found only a main effect of stylization (F~(5,160)~ = 12.62, p \< 0.001, partial η~P~^2^ = 0.283). Here, a linear increase in the LPP amplitude was observed (see [Fig. 3](#f3){ref-type="fig"}). The linear contrast (F~(1,32)~ = 31.28, p \< 0.001, partial η~P~^2^ = 0.494) accounted for substantially more variance explained than did a u-shaped contrast (F~(1,32)~ = 11.39, p = 0.002, partial η~P~^2^ = 0.262). There was no effect of emotion (F~(2,64)~ = 0.04, p = 0.965, partial η~P~^2^ = 0.001), and no interaction of stylization with emotion (F~(10,320)~ = 1.00, p = 0.444, partial η~P~^2^ = 0.030). Source reconstruction --------------------- Source reconstructions were calculated for significant effects of stylization level. All presented faces elicited strong visual responses (see [Figs 4](#f4){ref-type="fig"} and [5](#f5){ref-type="fig"} upper panel). However, the extreme poles (photographs compared to cartoon faces) showed a distinct processing already in the N170: While real faces led to larger inferior and middle occipital activations, highly stylized faces caused stronger responses in the right cuneus/lingual gyrus (see [Fig. 4](#f4){ref-type="fig"} and [Table 1](#t1){ref-type="table"}). Thus, despite similar N170 peaks for the extreme poles, the cortical generators differ. For neither of the extreme poles did we find statistical differences in source localization compared to moderately stylized characters. Later, in the LPP stronger superior occipital and superior parietal activations are observed for real compared to cartoon faces (see [Fig. 5](#f5){ref-type="fig"} and [Table 2](#t2){ref-type="table"}). These source estimations mirror the linearly increasing LPP for higher realism. With increasing realism of the faces, the differences to real faces become smaller and finally disappear in middle occipital areas. For the reverse contrasts, no differences were found. Discussion ========== This study investigated the cerebral processing of stylized faces across six levels of realism and three levels of emotional expression (happy, angry, and neutral), as reflected by the N170, EPN and LPP components. The results demonstrate that character stylization affects both the N170 and the LPP amplitudes, albeit in a qualitatively different manner. For the N170, a u-shaped modulation was observed, while continuous amplitude increases with increasing realism occurred for the LPP. For the N170 both highly stylized faces and real faces elicited strongest responses. Further, at the N170 level, a differential effect of face stylization on emotional expression was found: For cartoon characters, happy expressions caused similar N170 amplitudes as did angry faces, while for realistic faces (levels 5 and 6) only angry expressions were selectively processed. The EPN component was modulated solely by emotional expression, angry faces eliciting largest amplitudes. The results are striking in that they reveal a dissociation of stylization effects on the N170 and LPP: Taking into account that (i) the N170 amplitude is larger for faces than for objects[@b24][@b25][@b31], (ii) larger for real compared to schematic faces[@b27] and (iii) cuteness and baby-like features have been associated with a larger N170[@b31], we suggest neoteny and perceived face realism to drive the u-shaped N170 modulation. Analysis of stimulus properties (see [Fig. 6](#f6){ref-type="fig"}) indicates that neotenic features, such as eye size, decrease non-linearly and very quickly for stylization levels 1 to 3. On the other hand, perceived realism increases linearly for the tested characters. N170 generators were found to differ between very abstract and realistic faces: Although all faces activated extended visual regions, including the right fusiform gyrus, cartoon faces elicited stronger early visual cortex activations (cuneus, lingual gyrus, inferior occipital gyrus), while for real faces, stronger activations were found in middle occipital regions. Results suggest that processing of highly stylized faces relies more on structural analysis, associated with the so-called occipital face area, whereas realistic faces activate to a greater extent holistic processing, associated with the fusiform face area[@b6]. Fusiform responses have been found for a variety of face stimuli[@b59]. However, within computer generated characters, stronger fusiform responses were found when these looked and acted naturally and meaningfully[@b60]. The present U-shaped modulation could result from an interaction of perceived realism activating fusiform-dependent holistic processing and neoteny features activating feature-based processing in more occipital face areas. Emotion effects on the N170 and EPN are in line with previous work[@b32][@b33][@b35]. They indicate that emotional expressions modulate the N170 and EPN responses across stylization levels, while, also in line with the literature, in real faces, angry expressions had the largest impact on ERPs[@b32]. It is remarkable that stylization and emotional expressions interact on the N170, indicating an early interplay of structural analysis and emotional classification, rather than dual processing routes for identity and expression. The present data suggest that with increasing realism more resources are captured by cues signalling threat[@b61]. Accordingly, more realistic angry faces were rated more intense compared to happy faces see [Figs 6](#f6){ref-type="fig"} and [7](#f7){ref-type="fig"} and[@b20]. On the other hand, for very stylized faces, a relatively stronger processing of happy expressions was observed. Neotenic features may selectively enhance the processing of happy expressions. With the exception of the forehead, all neotenic features decrease or remain nearly unchanged with increasing realism ([Fig. 6b](#f6){ref-type="fig"}). As big eyes and a small nose contribute highly to a cute perception of the character, possibly amplifying processing of positive expressions. Unlike the N170 amplitude, LPP amplitude increase parallels perceived face realism across the tested stylization levels. This might be due to the uniqueness of a real face, prompting a multitude of ad hoc social inferences[@b62]. Bruce and Young's influential model suggests that after initial perceptual and structural analyses, the observed stimuli are compared with face representations stored in memory and if there is a match, person-specific knowledge is retrieved[@b63]. Indeed, person-related semantic information enhances LPP amplitudes[@b56]. Recently, it has been further shown that biographical information can increase the LPP to computer-generated faces[@b58]. Although we did not provide participants with explicit biographical information, as a result of social inferences, the more realistic faces might be perceived as having a unique biographical identity. Behavioural evidence showed that computer generated faces are harder to remember, possibly because they are not encoded as a unique person[@b52][@b53]. The noticeable discontinuity between levels 1--4 and 5--6 could also imply a categorical change between realistic and non-realistic characters as shown by classification tasks at a similar stylization level[@b16][@b17]. Typically, distinctiveness is achieved by exaggerating certain spatial differences between an individual and an average face[@b64]. Distinctiveness by shape or reflectance manipulations has been found to result in a larger EPN and LPP as well as a steeper learning curve and better memory trace for initially non-distinctive faces[@b54][@b55][@b65]. However, naturally distinctive faces lead to the largest LPP and are remembered even better[@b55]. Regarding the creation of the currently used face stimuli, spatial differences were not overexaggerated in comparison to an average face. However, rated face-realism and distinctiveness might not be uncorrelated, as more realistic faces for example exhibit a more detailed texture. Other studies have related the enhanced LPP for real compared to doll faces to the unique identity of the real face, generating an impression of personal social presence[@b51]. Future studies should aim to disentangle effects of face-distinctiveness from face-realism. Other factors can also modulate the LPP, but are unlikely to play a role in the current experiment: For instance, LPP responses increase with higher perceived familiarity[@b66]. However, Zell et al.[@b20] report equal familiarity across the present stylization levels. Similarly, facial attractiveness enhances the LPP[@b30][@b50]. For our stimuli, appeal, which is conceptually similar to attractiveness, was rated highest for medium-stylized faces (see [Fig. 6](#f6){ref-type="fig"} and Zell et al.[@b20]). Therefore, it is unlikely that attractiveness is responsible for LPP modulations in our experiment. In source space, the linear modulation of the LPP was reflected in larger and broader activations in occipito-parietal areas. The localization of this increase is in line with both enhanced perceptual processing of more realistic faces and, in particular, also memory-related processes. Although higher LPP amplitudes have been reported for emotional than for neutral stimuli e.g. refs [@b36], [@b41] and [@b44], we found no differences between emotional and neutral expressions on this component. In general, during passive viewing, emotion effects are smaller for faces compared to complex scenes and participants typically report less subjective arousal for faces[@b49]. Nevertheless, large emotion effects were present for the N170 and EPN time window. Similarly, Thom and colleagues[@b49] found emotion effects for the N170, while for the EPN emotion effects were only descriptively visible and no differences were found for the LPP. This suggests that without an explicit task a highly automatic response towards emotional facial expressions modulates early components without affecting late stages of processing. In this vein, it has been shown that at late stages emotion effects benefit more from attention to the emotional category than do early responses[@b39][@b40]. Overall, we demonstrated that stylized characters elicit neural effects that are different from the ones elicited by real faces. Importantly, the pattern changes qualitatively across different processing stages, although the measured facial features changed continuously across similar stylization levels. For face perception experiments, which use computer generated stimuli, this means that, unless a high level of realism is achieved, results cannot be transferred directly to real humans. Thus, computer generated stimuli may be suitable to test initial hypotheses, but require final validation with real photographs. So far, it is unclear why or when exactly realism is beneficial in practical applications like games or perceptual studies, but the present study, in demonstrating that realism affects different processing stages in a distinct manner, may offer some clues: Both highly stylized faces, with their neotenic features and very realistic faces influence early stage processing and are equally efficient in transient attention capture. On the other hand, only more realistic faces induce the kind of post-processing necessary to build an individual identity representation and likely facilitating identification with the character. Beyond gaming, these findings have implications for the design of virtual reality therapy settings, for instance of social phobia. They underscore that depending on the overall goal optimal character design will differ. If so, the uncanny valley phenomenon may also arise from a perceived mismatch between situational expectations resulting from a given virtual scenario and character appearance. In character design, the main problem for artists is that, for adult characters, neotenic and realistic features often exclude each other. For instance, skin smoothness is a neotenic feature, but detailed pores and skin-impurities are required to achieve full realism for a virtual character. Similarly, big eyes are considered as cute, but realistic characters that have unnaturally big eyes are perceived as creepy[@b21], inverting the intended effect. This dichotomy between realistic characters on one side and rather cute characters on the other, could also explain the plausibility of the "uncanny valley" concept and the present data indeed reveal a neural dissociation that might support it. In practice, considering two independent scales--one for realism and one for neotenic features--seems to be a promising future direction to predict whether a stylized character will be perceived positively or negatively when used in game or as an interactive agent. Moreover, these parameters can be controlled more easily than appeal or attractiveness, which depend on many aspects and are more subjective. As a limitation of the present work, it has to be noted that only two different identities were used. Therefore, our results might not generalize across all conceivable characters. In order to increase validity of the obtained results, our characters have been stylized based on popular 3D characters. Furthermore, the stimuli have been designed to match the quality of current animation movies as much as possible. To our best knowledge, our results are unique in that they provide many different stylization levels for the same characters. Conclusion ========== We measured EEG responses elicited by carefully manipulated faces with six different degrees of stylization and three emotional expressions. We tested a stimulus set for which a linear modulation of realism was achieved for the same identity. Our results indicate that face realism has a strong, but qualitatively different, influence on the N170 as well as the LPP component. While perceived realism influenced the N170 component in a u-shaped manner and interacted with emotional expression, the LPP component was only influenced by perceived realism, increasing continuously with face realism. For the N170, main generators differed between highly stylized and very realistic faces, suggesting to distinct contributing processes. The increased LPP was based on enhanced activity in broad occipito-parietal areas, in line with enhanced perceptual processing and memory encoding of more realistic faces. Methods ======= Participants ------------ Thirty-three participants were recruited at Bielefeld University. They gave written informed consent and received 11 Euros or course credit for participation. The study was approved by the Bielefeld University ethics committee (EUB number 2016-112). All methods were performed in accordance with the guidelines and regulation at Bielefeld University. The participants (22 females) were 23.30 years on average (SD = 3.68), all of them right-handed and had normal or corrected-to normal vision. Upon structured interview, no participant reported a current neurological or psychiatric disorder or relevant medication intake. Stimuli ------- Two face characters (one male, one female) varying across six stylization levels, with three emotional expressions (happy, angry, neutral) per stylization level, were used as stimuli. Creating continuous stylization of the same character is still an unsolved problem. While certain stylization effects can be achieved by image editing, using non-photorealistic rendering algorithms or character generators, all these approaches either do not modify the shape at all or address only a single stylization level. So far no automatic approach exists that accomplishes character stylization comparable to trained artists. Professional stylization is in general very time consuming and expensive. To circumvent this problem, most previous work used unrelated characters at different stylization levels. In contrast, Zell et al.[@b20] published a set of stimuli that contains six stylization levels of the same person matching the state of the art in computer graphics. For completeness, we shortly report the stimuli creation process. Level 6 are real photographs. Models for level 5 have been created from high resolution 3D scans of the photographed people. These 3D scans have been post-processed to remove visible artefacts or to add hair. The remaining stylization levels have been created by professional artists targeting popular looks of animation movies. All emotions have been created so that specific features remain consistent (e.g., teeth shown consistently for happy and angry expressions). Levels 2, 3 and 5 were created first and evaluated in a study according to perceived realism and familiarity. All characters had very similar familiarity ratings as none of the characters is known to the public. In addition, for the initial characters, discontinuities were detected for rated face-realism. In order to achieve a stepwise increase in rated realism and sample the stylization scale more uniformly, level 1 and 4 have been created and added afterwards. The whole stimulus set was finally rated according perceived realism, appeal and expression intensity ([Fig. 6](#f6){ref-type="fig"}). For the real photographs, only the neutral expression was rated in perceived realism and appeal. In addition, facial neotenic cues have been computed by measuring the relative size of facial parts[@b67]. We refer the reader to Zell et al.[@b20] for a full description of the stimuli creation process, including all technical details and initial evaluation. For the current study, camera view and aspect ratio have been adjusted, such that all faces are nearly of the same size and the eyes are located at similar positions. Background planes with a 50% grey were inserted in the 3D scenes before rendering. For stimuli examples see the depicted faces in [Figs 6](#f6){ref-type="fig"} and [7](#f7){ref-type="fig"}. Procedure --------- The faces were randomly presented for 600 ms, followed by a fixation cross with variable latencies (randomly between 400 and 500 ms). All faces were repeated fifteen times, for a total number of 540 presented faces, while it was prevented that the same stimulus was presented twice in a row. Faces were presented on a 15.4-inch screen (Dell Latitude D830) with a 1600 × 1200 pixel resolution (image width: 800; height: 1142). The background colour was 50% grey (RGB color values: 128; 128; 128). The stimulus presentation lasted for about 10 minutes, while the whole session took approximately 50 minutes. Participants had no task but were instructed to attend to the presented faces, while moving as little as possible. Participants were encouraged to reduce their eye-movements by focusing on the fixation cross. EEG recording and analyses -------------------------- EEG was recorded from 128 BioSemi active electrodes ([www.biosemi.com](http://www.biosemi.com)). Recorded sampling rate was 2048 Hz. During recording Cz was used as reference electrode. Biosemi uses two separate electrodes as ground electrodes: A Common Mode Sense active electrode (CMS) and a Driven Right Leg passive electrode (DLR). The two electrodes form a feedback-loop which enables to measure the average potential close to the reference in the AD-box (see <http://www.biosemi.com/faq/cms&drl.htm>, where also information about extra functions of the CMS/DRL loop can be retrieved). Four additional electrodes (EOG) measured horizontal and vertical eye-movement. These were placed at the outer canthi of the eyes and below the eyes. Pre-processing and statistical analyses were done using BESA^®^ ([www.besa.de](http://www.besa.de)), EMEGS[@b68] and SPM8 for EEG data (<http://www.fil.ion.ucl.ac.uk/spm/>). Offline, data were re-referenced to the average reference and then filtered with a forward 0.16 Hz high-pass and a zero-phase 30 Hz low-pass filter. Filtered data were segmented from 100 ms before stimulus onset until 600 ms after stimulus presentation. The 100 ms before stimulus onset were used for baseline correction. Eye-movements were corrected using the automatic eye-artefact correction method implemented in BESA[@b69]. Additionally, trials exceeding a threshold of 120 μV were rejected. Overall, 4.04 percent of all electrode measurements were interpolated. Per participant, on average 5.41 percent of all trials were rejected, leaving 28.38 trials per cell, leading to 85 trials per realism condition and 170 trials per emotion condition. Cortical source reconstructions of significant ERP differences were generated and statistically assessed with SPM8 for EEG[@b70], following recommended procedures. First, a realistic boundary element head model (BEM) was derived from SPM's template head model based on the Montreal Neurological Institute (MNI) brain. Electrode positions were then transformed to match the template head, which is thought to generate reasonable results even when individual subjects' heads differ from the template[@b71]. Average electrode positions as provided by BioSemi were co-registered with the cortical mesh template for source reconstruction. This cortical mesh was used to calculate the forward solution. For the inverse solution, the group inversion algorithm was used[@b70] and the solution was calculated from 100 ms pre-baseline to 600 ms after stimulus onset. Statistical analyses -------------------- EEG scalp-data were statistically analyzed with EMEGS. Six (stylization: level 1, level 2, level 3, level 4, level 5, level 6) by three (emotional display: angry, neutral, happy) repeated measure ANOVAs were set-up to investigate main effects of the communicative sender, emotion and their interaction in time windows and electrode clusters of interest. Partial eta-squared (η~P~^2^) was estimated to describe effect sizes, where η~P~^2^ = 0.02 describes a small, η~P~^2^ = 0.13 a medium and η~P~^2^ = 0.26 a large effect[@b72]. When Mauchly's Test detected a violation of sphericity, degrees of freedom were corrected according to Greenhouse-Geisser. For readability, the original degrees of freedom but corrected p-values and effect sizes are reported. For significant main effects linear compared to U-shaped contrasts were calculated. Time windows were segmented from 150 to 190 ms to investigate the N170, from 250 to 400 ms to investigate the EPN, and from 400 to 600 ms to the LPP component, after collapsing all conditions and visual inspection of the ERP components. For the N170 and EPN time windows, two large symmetrical temporo-occipital clusters of thirteen electrodes each were examined (left: I1, OI1, O1, POO3, PO9, PO9h, PO7, PO7h, P9, P9h, P7, TP9h, TP7; right: I2, OI2, PO10, POO4, PO10, PO10h, PO8, PO8h, P10, P10h, P8, TP10h, TP8). For the LPP time windows a large parietal cluster was investigated (twenty-six electrodes: CCPz, CP5, CP5h, CP3, CP1, CPz, CP2, CP4, CP6, CPP5h, CPP3, CPPz, CPP4, P3, P1, Pz, P2, P4, PPO3, PPO1, PPOz, PPO2, PPO4, PO3, POz, PO4). Results did not change qualitatively, when selecting different literature-based electrode clusters for the N170, EPN or LPP[@b34][@b36][@b54]. Source reconstructions were performed for the main effects of face-stylization. For each analyzed time window in scalp space, three-dimensional source reconstructions were generated as NIFTI images (voxel size = 2 mm · 2 mm · 2 mm). These images were smoothed with a Gaussian kernel using an 8 mm full-width half-maximum. The statistical comparisons used in source space were based on significant differences on the scalp. In line with previous studies[@b39][@b42][@b73], we describe statistical differences in source activity of voxels differing at least at an uncorrected threshold of p \< 0.005 and a minimum of twenty-five significant voxels per cluster. Additionally, in all tables results are shown applying a family-wise error corrected threshold of p \< 0.05. The identification of activated brain regions was performed using the LONI atlas[@b74]. Additional Information ====================== How to cite this article: Schindler, S. et al. Differential effects of face-realism and emotion on event-related brain potentials and their implications for the uncanny valley theory. Sci. Rep. 7, 45003; doi: 10.1038/srep45003 (2017). Publisher\'s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. We thank Liane Stritz, Dominik Neu and Maria Vogt for help with data acquisition and all participants contributing to this study. All authors were supported by the Cluster of Excellence Cognitive Interaction Technology "CITEC" (EXC 277) at Bielefeld University, funded by the German Research Foundation (DFG). We acknowledge support of the publication fee by Deutsche Forschungsgemeinschaft and the Open Access Publication Funds of Bielefeld University. The authors declare no competing financial interests. Author Contributions E.Z. and S.S. contributed to the study design. S.S. measured participants and analysed the data under supervision of J.K. E.Z. and S.S. drafted and revised the manuscript under supervision of M.B. and J.K. All authors read and approved the final version. ![N170 ERP results.\ The upper panel displays the N170 at electrode PO8. From this panel, the interaction between emotion and realism can be observed. The lower panel shows the mean N170 over the occipital sensor cluster. Error bars are +/− one standard error of the mean. Note that, while negative-going, the N170 peak is still in the positive range (see top panel). Therefore, smaller bars represent higher N170 amplitudes.](srep45003-f1){#f1} ![Emotion main effects for the N170 and the EPN.\ Left: Displayed are the difference topographies for the main effects of emotion across all stylization levels. Blue colours indicate a relatively larger negativity and red colours a larger positivity. Right: Displayed is the time course at electrode PO8.](srep45003-f2){#f2} ![LPP ERP results.\ The upper panel shows the time course for electrode CPPz. Here an increase of LPP amplitudes can be seen for increasing realism. The lower panel displays the mean LPP over the parietal sensor cluster. Error bars are +/− one standard error of the mean.](srep45003-f3){#f3} ![Source estimations for main effects of stylization level for the N170.\ Upper panel: Displayed are the family-wise error corrected (p \< 0.05) changes in source activity for each stylization level. For all faces, significant generators can be observed for the N170 in bilateral superior, middle and inferior occipital gyri, as well as in the bilateral fusiform gyri. Lower panel: Displayed are the differences in source activity between stylization levels (post-hoc contrasts, uncorrected p \< 0.005). In the N170 the real faces lead to larger middle and inferior occipital activations, while the most stylized faces are processed more intensely in the right inferior occipital gyrus/cuneus/lingual gyrus.](srep45003-f4){#f4} ![Source estimations for main effects of stylization for the LPP time window.\ Upper panel: Displayed are the family-wise error corrected (p \< 0.05) changes in source activity for each realism level. For all faces, significant generators can be observed for the LPP in bilateral superior, middle and inferior occipital gyri, as well as in the bilateral fusiform gyri and bilateral superior parietal areas. Lower panel: Displayed are the differences between realism levels (post-hoc contrasts, uncorrected p \< 0.005). In the LPP, real faces are processed more intensely in bilateral middle and superior occipital and superior parietal areas. However, with increased realism, these differences become smaller and finally disappear in middle occipital regions.](srep45003-f5){#f5} ![(a) Realism, appeal and intensity ratings from Zell et al.[@b20], which tested the same characters but from slightly different point of view. Icons along the x-axis represent the stylization level. (b) Length and width of different facial parts, proportionally to the length and width of the face--averaged across both characters. For more stylized characters, the bigger eyes and smaller nose follow neotenic characteristics. However, this is not the case for vertical eye placement or forehead height. Error Bars denote 95% confidence levels.](srep45003-f6){#f6} ![Example for emotional and neutral expressions for all six realism levels.\ The stimulus set consists of a male and female character. Here, the male character is shown for illustration purpose. Please refer to [Fig. 6](#f6){ref-type="fig"} for illustration of the female character.](srep45003-f7){#f7} ###### Source estimations for stylization level main effects for the N170 (150--190 ms). Cluster-level peak-level MNI coordinates LONI ---------------------- ------------ ----------------- ------ ----- ----- ------------------- Level 6 \> Level 1 408 3.35 \<0.001 28 −90 2 Mid Occipital G R 86 2.70 \<0.005 −34 −90 8 Mid Occipital G L Level 1 \> Level 6 169 3.31 \<0.001 18 −98 −14 Inf Occipital G R Results show differences only between very stylized faces (Level 1) and real faces (Level 6). Real faces elicited more activity in middle occipital regions, while cartoon faces elicited more activity in the right inferior frontal gyrus/cuneus/lingual gyrus. Notes. No. of sig. voxel = the number of voxel which differ significantly between both conditions. Peak p-unc = uncorrected p value. For each significant peak, respective coordinates (x, y and z) are displayed in MNI space. If a cluster exhibited more than one peak, only the largest peak is reported. Area = peak-level brain region as identified by the LONI atlas. R/L = laterality right or left. G = Gyrus; Mid = middle. ###### Source estimations for stylization level main effects for the LPP (400--600 ms). Cluster-level Peak-level MNI coordinates LONI ---------------------- ------------ ----------------- ------ ----- ---- ------------------- Level 6 \> Level 1 1331 (707^a^) 7.47 \<0.001 −36 −86 18 Mid Occipital G L 700 (483^a^) 7.09 \<0.001 28 −90 2 Mid Occipital G R 673 (209^a^) 4.74 \<0.001 20 −82 30 Sup Occipital G R Level 6 \> Level 2 1192 (639^a^) 6.51 \<0.001 -38 −84 20 Mid Occipital G L 617 (356^a^) 5.51 \<0.001 34 −88 16 Mid Occipital G R 685 (301^a^) 4.94 \<0.001 20 −82 30 Sup Occipital G R Level 6 \> Level 3 1215 (610^a^) 6.06 \<0.001 −36 −86 20 Mid Occipital G L 684 (345^a^) 5.13 \<0.001 20 −82 30 Sup Occipital G R 749 (53^a^) 4.98 \<0.001 34 −92 2 Mid Occipital G R Level 6 \> Level 4 1167 (348^a^) 5.63 \<0.001 −34 −88 20 Mid Occipital G L 705 (369^a^) 5.24 \<0.001 20 −82 30 Sup Occipital G R 565 4.10 \<0.001 36 −86 0 Mid Occipital G R Level 6 \> Level 5 243 3.18 \<0.001 −38 −82 22 Mid Occipital G L 329 2.91 \<0.005 20 −82 30 Sup Occipital G R Results show enhanced activity for real faces compared to stylized faces. Real faces (Level 6) elicited more activity in middle and superior occipital regions. Differences become smaller with increasing realism of the stylized faces. Notes. ^a^Resulting cluster size with FWE-corrected threshold of p \< 0.05 (≥25 significant voxels). No. of sig. voxel = the number of voxel which differ significantly between both conditions. Peak p-unc = uncorrected p Value. For each significant peak, respective coordinates (x, y and z) are displayed in MNI space. If a cluster exhibited more than one peak, only the largest peak is reported. Area = peak-level brain region as identified by the LONI atlas. R/L = laterality right or left. G = Gyrus; Inf = inferior, Mid = middle, Sup = superior. [^1]: These authors contributed equally to this work.	{ "pile_set_name": "PubMed Central" }
Introduction {#s1} ============ Genome size, defined as the haploid DNA content of a cell, has increased as eukaryotes evolved from single-cell species to more complex, multicellular organisms. Within the same evolutionary timeframe, nuclear volume has also increased but at a slower rate than genome size expansion ([@bib45]; [@bib49]). While the ratio of nuclear to cell size has remained essentially constant in eukaryotes ([@bib8]), the disproportional increase in genome size relative to the nuclear volume has required organisms with larger genomes to compact their chromatin to greater extents than organisms with small sized genomes. Indeed there is a positive correlation between genome size and native chromatin compaction as measured by dye incorporation into chromatin ([@bib66]). In most eukaryotes, the genome is organized into chromatin by the repeating nucleosomal structure ([@bib42]). The nucleosomes stack and fold into higher order structures, serving to systematically compact the genome ([@bib38]; [@bib17]) and to regulate molecular processes that are based on DNA ([@bib9]; [@bib67]; [@bib24]; [@bib35]; [@bib26]). The surface of the histone octamer has 14 DNA interaction sites. Each interaction is mediated by an arginine residue that intercalates into the minor groove of the DNA to stabilize the nucleosomal structure ([@bib42]; [@bib71]). Arginine is the most commonly used amino acid for interaction with DNA due to its positive charge and the lower energetic cost compared to lysine for displacing water when intercalating into the minor groove ([@bib54]). Nucleosomes mediate chromatin compaction through a variety of mechanisms. For instance, nucleosomes form higher order structures through inter-nucleosomal contacts between the histone H4 N-terminal domain (NTD) and the acidic patch of H2A and between two H2B C-terminal domains (CTD) ([@bib42]; [@bib15], [@bib16]; [@bib27]; [@bib56]). Histone variants, such as H2A.Z or H2A.Bbd, as well as post-translational modifications of histones, such as H4K16ac, can further regulate the degree of compaction ([@bib63]; [@bib1]; [@bib61]; [@bib73]; [@bib10]; [@bib34]; [@bib23]). Polycomb complexes compact large domains of chromatin ([@bib21]) and are important for proper development. Histones of the H1 family promote additional compaction by binding between nucleosomes to linker DNA near the DNA entry/exit site on nucleosomes and stabilize the intrinsic ability of nucleosomal arrays to fold in vitro ([@bib6]; [@bib53]; [@bib64]). Linker histones may affect chromatin compaction globally ([@bib22]), at specific stages of the cell cycle such as mitosis ([@bib44]) or at specific regions of the genome ([@bib37]). In contrast to canonical histones, the linker histones are much less conserved ([@bib7]), and ectopic expression of human linker histones in the budding yeast even at low levels is lethal for the cell ([@bib46]). Finally, structural proteins such as condensin also contribute to chromatin condensation ([@bib65]). Many of these modulatory mechanisms are dynamic in nature ([@bib41]) and may help explain why multicellular organisms can compact chromatin to different degrees in different cell types. However, despite the existence of these mechanisms for genome compaction in higher eukaryotes, it has not been known whether the canonical histones themselves have evolved sequence features that also contribute to the generally increased chromatin compaction observed in organisms with larger genomes. In this study, we provide evidence from analysis of 160 fully-sequenced eukaryotic genomes that arginine (R) residues at specific positions in the N-terminal tail of histone H2A---which protrudes from the nucleosome on the opposite side of DNA entry/exit site---have co-evolved with increasing genome size with a concomitant decrease in serines/threonines. Although increases in genome size are associated with phylogenetic evolution from protozoa to fungi to more complex plants and animals, we present genetic and molecular evidence from the budding yeast and human cells as well as in vitro biochemical data to demonstrate that the evolutionary changes in H2A directly regulate chromatin compaction in vivo and in vitro with consequences for the nuclear volume. The evolutionary changes in H2A regulate chromatin compaction in yeast and human cells, revealing a surprising flexibility in the dynamics of the chromatin fiber that has been conserved across distant eukaryotes. This previously unrecognized structural feature of the nucleosome has evolved to enable greater chromatin compaction when genome size is disproportionately larger than the nuclear volume. Our findings also suggest that the reported mutations in the histone H2A NTD may contribute to the altered chromatin compaction that is commonly observed in cancer cells ([@bib74]). Results {#s2} ======= H2A acquires specifically positioned arginines as genome size increases {#s2-1} ----------------------------------------------------------------------- To determine whether specific residues in the four core histones have co-evolved with increasing genome size, we performed residue composition analysis of canonical histone protein sequences from 160 fully sequenced eukaryotes with genome sizes ranging from 8 to 5600 Mbp encompassing protozoa, fungi, plants, and animals. The canonical histone proteins for each organism were defined based on at least 90% overlap and 35% identity with the histone fold domain of the corresponding human sequence ('Materials and methods'). Each organism was categorized as having a small (\<100 Mbp), medium (100--1000 Mbp), or large (\>1000 Mbp) genome ([Figure 1---figure supplement 1A](#fig1s1){ref-type="fig"}). Of the canonical histones, the H2A NTD showed the most statistically significant variability in amino acid residues, where the number of arginines increased with increasing genome size ([Figure 1A](#fig1){ref-type="fig"}), while the number of serines (S) and threonines (T) decreased ([Figure 1B](#fig1){ref-type="fig"}). Other amino acid residues in the H2A NTD, including lysines (K), did not correlate with genome size ([Figure 1---figure supplement 1B](#fig1s1){ref-type="fig"}).10.7554/eLife.02792.003Figure 1.Histone H2A N-terminal sequence has co-evolved with genome size.Violin plots of the number of (A) arginines or (B) serines/threonines in the H2A NTD for species with small, medium, and large genomes. Plot widths correspond to species frequency within each group. (C) H2A NTD sequences for S. cerevisiae and H. sapiens. (D) Heat map of H2A NTD residue composition at the indicated positions ordered by genome size. Example species are shown with kingdom and genome size information. (E) Protein sequence motifs surrounding the four H2A NTD arginine residues. (F) Positioning of evolutionarily variable residues relative to the H2A N-terminus (left) or histone fold (right). See also [Figure 1---figure supplements 1 and 2](#fig1s1 fig1s2){ref-type="fig"}.DOI: [http://dx.doi.org/10.7554/eLife.02792.003](10.7554/eLife.02792.003)10.7554/eLife.02792.004Figure 1---source data 1.H2A multiple sequence alignments, heat map data, and canonical H2A isoforms.DOI: [http://dx.doi.org/10.7554/eLife.02792.004](10.7554/eLife.02792.004)10.7554/eLife.02792.005Figure 1---figure supplement 1.Phylogenetic distribution of species analyzed in this paper.(A) The bar graph indicates the number and proportion of organisms in our data set that belong to the indicated phylogenetic kingdoms for each genome size category. (B) Violin plots of the number of lysines in the H2A NTD grouped by genome size as in [Figure 1](#fig1){ref-type="fig"}. (C) Boxplot of genome sizes of the organisms which have an H2A without (left) or with (right) the indicated residue. p-values (Mann--Whitney U test) of the difference in means between the absence or presence of the indicated residue are indicated.DOI: [http://dx.doi.org/10.7554/eLife.02792.005](10.7554/eLife.02792.005)10.7554/eLife.02792.006Figure 1---figure supplement 2.H2A arginines 3 and 11 are situated adjacent to DNA within the nucleosome.Structure of a di-nucleosome obtained from the crystal lattice of a mono-nucleosome structure is shown from (A) the side or (B) close up highlighting potential intra- and inter-nucleosomal interactions between arginines and DNA backbones. Green is H2A, yellow is H2B, cyan is H3, and salmon is H4. The red and blue spheres are R3 and R11, respectively, both of which are in chain C ([@bib12]).DOI: [http://dx.doi.org/10.7554/eLife.02792.006](10.7554/eLife.02792.006) The acquisition of arginines and loss of serines/threonines in the H2A NTD with increasing genome size occur at specific positions and in sequential order. For instance, the human H2A NTD contains arginine residues at positions 3 and 11 that are absent in yeast and at position 20 which is correspondingly a lysine in yeast ([Figure 1C](#fig1){ref-type="fig"}). In contrast, the human sequence lacks S10 and S15 that are observed in the yeast H2A NTD ([Figure 1C](#fig1){ref-type="fig"}). Alignment of all H2A NTD sequences also revealed similar trends across all eukaryotes studied here. The heat map in [Figure 1D](#fig1){ref-type="fig"} shows the occurrence of arginines and serines/threonines in the H2A NTD as a function of genome size (see [Figure 1---source data 1](#SD1-data){ref-type="supplementary-material"} for raw data and [Figure 1---figure supplement 1C](#fig1s1){ref-type="fig"} for statistical analysis). At position 3, an arginine (R3) is predominantly present in medium and large species but is lacking in small species. At position 11, a lysine (K11) is observed in species with medium genomes that evolves to an arginine (R11) mainly in organisms with large genomes. R17 is present in most organisms examined, suggesting a very conserved function for this residue ([@bib72]). At position 20, small genomes contain predominantly a lysine residue, which converts to arginine in medium and large genomes. In contrast, serines/threonines at positions 10 and 15 are found primarily in organisms with small genomes and much less so in organisms with medium and large genomes ([Figure 1D](#fig1){ref-type="fig"}). Additionally, each of the four H2A NTD arginines is surrounded by a conserved motif ([Figure 1E](#fig1){ref-type="fig"}). The residues surrounding R3 and R17 are mainly glycine and serine, respectively. At position 11, the motif varies based on genome size. Species with medium-sized genomes contain VKG and those with large genomes contain ARA. The same is true of position 20, where AKA is present in organisms with small genomes and (S/T)RA in larger genome species ([Figure 1E](#fig1){ref-type="fig"}). Interestingly, except for R3, the positions of all the other evolutionarily varying residues in the H2A NTD are strongly conserved relative to the histone fold domain and not the N-terminus ([Figure 1F](#fig1){ref-type="fig"}). When counting conventionally from the N-terminus, amino acids R11, R17, and R20---which are numbered based on the human sequence---were not observed consistently at the same positions in other organisms. However, these residues are respectively 12, 6, and 3 amino acids away from the histone fold in most species (note the vertical axes in [Figure 1F](#fig1){ref-type="fig"}). S10 and S15---which are numbered based on the yeast sequence---also show more uniform positioning when counted from the histone fold. Altogether, as genome size increases, arginines appear in conserved positions within the H2A NTD relative to the histone fold, and serines and threonines are lost. Arginines 3 and 11 in the H2A NTD increase chromatin compaction {#s2-2} --------------------------------------------------------------- To determine whether arginines and serines/threonines of the H2A NTD affect chromatin compaction in vivo, we took advantage of a strain of Saccharomyces cerevisiae, that has both chromosomal copies of H2A deleted and carries a single copy of H2A on a plasmid (TSY107), to construct mutant strains containing single or multiple insertions of arginines into their conserved motifs, deletions of serines, or combinations thereof (see [Table 1](#tbl1){ref-type="table"} for specific amino acid changes and [Supplementary file 1A](#SD2-data){ref-type="supplementary-material"} for a description of the mutant strains). Two mutants, R3(ΔGS10)R11 and R11ΔS15, were also designed such that the spacing between R3 and R11 or R11 and the histone fold, respectively, is the same as in the H2A NTD of organisms with large genomes ([Figure 1C,F](#fig1){ref-type="fig"}). As a control for positive charge, mutant strains with lysines inserted in the same positions as arginines were also generated.10.7554/eLife.02792.007Table 1.List of H2A mutations, sequence changes and their effects on chromatin compaction and nuclear volumeDOI: [http://dx.doi.org/10.7554/eLife.02792.007](10.7554/eLife.02792.007)FISHNuclear volumeH2A mutantH2A NTD Protein sequence% Changep-value% Changep-valueYeast WTSG--GKG--GKAGSA--AKASQSRSAKAG--1.0E+00--1.0E+00 R3SG[R]{.ul}GKG--GKAGSA--AKASQSRSAKAG−189.5E−04−54.1E−01 R11SG--GKG--GKAGSA[R]{.ul}AKASQSRSAKAG−158.6E−04−205.9E−05 R3R11SG[R]{.ul}GKG--GKAGSA[R]{.ul}AKASQSRSAKAG−228.2E−06−163.0E−03 R3(ΔGS10)R11SG[R]{.ul}GKG--GKA··A[R]{.ul}AKASQSRSAKAG−302.1E−06+63.7E−01 R11ΔS15SG--GKG--GKAGSA[R]{.ul}AKA·QSRSAKAG−413.9E−08−9[\](#tblfn1){ref-type="table-fn"}4.7E−04 K3SG[K]{.ul}GKG--GKAGSA--AKASQSRSAKAG+98.6E−01+139.4E−03 K11SG--GKG--GKAGSA[K]{.ul}AKASQSRSAKAG+163.1E−01+32.6E−01 K3K11SG[K]{.ul}GKG--GKAGSA[K]{.ul}AKASQSRSAKAG+68.3E−01+315.4E−08 K11ΔS15SG--GKG--GKAGSA[K]{.ul}AKA·QSRSAKAG−79.2E−02+26.6E−01 ΔGS10SG--GKG--GKA··A--AKASQSRSAKAG−63.2E−02+103.0E−02 ΔS15SG--GKG--GKAGSA--AKA·QSRSAKAG+39.4E−02+99.7E−03 R6SG--GKG[R]{.ul}GKAGSA--AKASQSRSAKAG−55.6E−02+101.0E−03 K20RSG--GKG--GKAGSA--AKASQSRSA[R]{.ul}AG−33.0E−01+71.5E−02 R17KSG--GKG--GKAGSA--AKASQS[K]{.ul}SAKAG−17.9E−0102.7E−01Human---HA Tag WTSGRGKQGGKTRAKAKSRSSRAG--1.0E+00--1.0E+00 ΔR3SG·GKQGGKTRAKAKSRSSRAG+398.3E−03+421.3E−08 R11KSGRGKQGGKT[K]{.ul}AKAKSRSSRAG+202.3E−02+141.2E−03 R11ASGRGKQGGKT[A]{.ul}AKAKSRSSRAG+431.0E−05+215.7E−07 ΔR3R11ASG·GKQGGKT[A]{.ul}AKAKSRSSRAG+353.5E−03+185.9E−04Human---FLAG Tag WTSGRGKQGGKARAKAKSRSSRAG--1.0E+00--1.0E+00 Δ1--12············KAKSRSSRAG+474.9E−03+182.8E−04[^2][^3][^4] To test the effects of H2A NTD changes on chromatin compaction, the physical distance between two probes on chromosome XVI spaced 275 kb apart was assessed in each of the H2A mutants using fluorescent in situ hybridization (FISH) ([Figure 2A](#fig2){ref-type="fig"}; [@bib30]; [@bib4]). The probes were differentially labeled and visualized by confocal microscopy. The distance between the probes was measured in a single plane in which both probes were present within each nucleus ([@bib4]). When compared to the isogenic wild type (WT), addition of a single arginine at position 3 (R3) or 11 (R11) to the H2A NTD was sufficient to significantly decrease the average interprobe distance by 18% and 15%, respectively ([Figure 2B,C](#fig2){ref-type="fig"}; [Table 1](#tbl1){ref-type="table"}). The average interprobe distance was further decreased by 22% when both arginines were present (R3R11) and even more so (30%) in R3(ΔGS10)R11. Deleting G9S10 (ΔGS10) alone caused slightly increased compaction with low statistical significance ([Table 1](#tbl1){ref-type="table"}). The largest decrease in interprobe distance (41%) was observed in the R11ΔS15 mutant, which places R11 12 amino acids from the histone fold, the same position as in organisms with large genomes. Removal of S15 (ΔS15) alone had no effect. The effect of arginines was not simply due to increasing the positive charge of the H2A NTD, as insertions of lysines at positions 3 and 11 did not significantly affect the interprobe distances ([Figure 2C](#fig2){ref-type="fig"} and [Supplementary file 2](#SD3-data){ref-type="supplementary-material"}). Although lysines are found at these positions in certain species ([Figure 1D](#fig1){ref-type="fig"}), the lack of potential compaction by lysines may be due to the absence of other evolutionary changes in yeast histones (see [Figure 1E](#fig1){ref-type="fig"}). Additionally, R17K or K20R mutations did not affect compaction, nor did a randomly inserted arginine at position 6 (R6) ([Figure 2C](#fig2){ref-type="fig"} and [Supplementary file 2](#SD3-data){ref-type="supplementary-material"}), suggesting that not every arginine in the H2A NTD contributes to chromatin compaction.10.7554/eLife.02792.008Figure 2.Ectopic expression of H2A NTD arginines causes compaction in yeast.(A) Schematic position of probes on chromosome XVI that were used for FISH. The letters correspond to the probe sets. (B) FISH images and (C) boxplot of the distributions of interprobe distances for probe set A in the indicated strains. (D) The mean interprobe distances for the indicated yeast strains for probe sets A, B, C, and D are plotted as a function of genomic distance. Solid lines are best fit equations. (E) Boxplot of the distributions of interprobe distances for probe set A in the indicated strains. Dashed lines mark the median value for the WT strain. The boxplot whiskers contain 90% of the data. All scale bars are 1 µm. Boxes are colored if the mean of the indicated strain is significantly different from WT. Red stars denote level of significance: \p\<0.01; \\p\<0.001; \\\p\<0.0001 (For exact values, see [Supplementary file 2](#SD3-data){ref-type="supplementary-material"}). (F) Agarose gel electrophoresis of MNase-digested chromatin in the indicated strains including the densitometric profiles comparing the WT to each of the mutant H2A strains for a given amount of enzyme. See also [Figure 2---figure supplement 1](#fig2s1){ref-type="fig"}.DOI:* [http://dx.doi.org/10.7554/eLife.02792.008](10.7554/eLife.02792.008)10.7554/eLife.02792.009Figure 2---figure supplement 1.Ectopic expression of H2A NTD arginines causes compaction in yeast.(A--C). Boxplot of the distributions of interprobe distances in the indicated H2A mutant strains for probe sets B, C, and D as shown in [Figure 3A](#fig3){ref-type="fig"}. Boxes are colored if the mean of the indicated strain is significantly different from WT. Red stars denote level of significance: \p\<0.01; \\\p\<0.0001 ([Supplementary file 2](#SD3-data){ref-type="supplementary-material"}). Cell cycle analysis of yeast strains in the TSY107 (D) or FY406 (E) background. 1C and 2C refer to G1 and G2 DNA content, respectively. Note that the WT strains (TSY107, FY406) carry one copy of the H2A gene on a plasmid with the two chromosomal copies deleted ([@bib59]; [@bib33]). Dosage alterations of the H2A protein cause G2/M arrest ([@bib62]) which is evident in our WT strains. Nonetheless, all the mutants display highly similar cell cycle profiles. (F) Agarose gel electrophoresis of MNase-digested chromatin in the indicated strains. The amount of enzyme used to digest chromatin is indicated. Also shown are the densitometric profiles of the agarose gel that compares WT to each of the indicated mutant H2A strains for a given amount of enzyme.DOI: [http://dx.doi.org/10.7554/eLife.02792.009](10.7554/eLife.02792.009) We further confirmed the effects of R11 on chromatin compaction using three additional probe sets ([Figure 2A](#fig2){ref-type="fig"}). The level of compaction seen in our WT strain is similar to what has been previously reported in yeast using a different strain background ([@bib4]). The interprobe distances for all probe sets were significantly decreased in R11 compared to WT and even more so in R11ΔS15 but not ΔS15 alone ([Figure 2---figure supplement 1A--C](#fig2s1){ref-type="fig"} and [Supplementary file 2](#SD3-data){ref-type="supplementary-material"}). Plotting the physical vs genomic distances for all probe sets revealed uniform compaction across large genomic distances ([Figure 2D](#fig2){ref-type="fig"}). The effect of R11 on chromatin compaction was not strain-specific as H2A R11 and R11ΔS15, but not ΔS15, caused chromatin compaction in a different strain background ([Figure 2E](#fig2){ref-type="fig"}; [Supplementary file 2](#SD3-data){ref-type="supplementary-material"}). We therefore conclude that H2A arginines at positions 3 and 11, especially when R11 is placed at the evolutionarily-conserved position relative to the histone fold, increase the degree of chromatin compaction. Chromatin is differentially compacted at different cell cycle stages ([@bib30]). Cell cycle profile analysis showed little difference between the strain harboring WT H2A and any of the mutant strains ([Figure 2---figure supplement 1D--E](#fig2s1){ref-type="fig"}), indicating that the observed differences in chromatin compaction are not due to altered cell cycle profiles. Chromatin compaction may also be influenced by nucleosomal spacing; indeed the linker DNA length is larger in human cells than in yeast ([@bib28]). We find that there are essentially no differences in nucleosomal density in H2A arginine mutants using Micrococcal nuclease (MNase) digestion ([Figure 2F](#fig2){ref-type="fig"}, [Figure 2---figure supplement 1F](#fig2s1){ref-type="fig"}), indicating that the average nucleosomal spacing is not affected by these mutations. But the more compact mutants displayed decreased accessibility to MNase as indicated by the delayed appearance of the nucleosomal digestion pattern ([Figure 2F](#fig2){ref-type="fig"}, [Figure 2---figure supplement 1F](#fig2s1){ref-type="fig"}). H2A arginines and serines affect nuclear volume in yeast {#s2-3} -------------------------------------------------------- Since chromatin structure may influence the volume of the nucleus ([@bib8]), we asked whether nuclear volume was affected by H2A-mediated chromatin compaction. We tagged a nuclear pore protein, Nup49, in its chromosomal locus with GFP to visualize the nuclear membrane and used confocal microscopy to capture three-dimensional images of the nucleus to quantify volumes of ≥150 cells per H2A mutant ([Figure 3A,B](#fig3){ref-type="fig"}; [Table 1](#tbl1){ref-type="table"}; [Supplementary file 3](#SD4-data){ref-type="supplementary-material"}, see 'Materials and methods' for volume calculations). As compared to WT cells, H2A mutants containing R11 or R3R11, both of which contain more compact chromatin, displayed significantly decreased nuclear volumes. The average nuclear volume in the R3 mutant was also less than WT but did not reach statistical significance. Interestingly, H2A mutants from which serines 10 and 15 were removed displayed larger nuclear volumes. Simultaneous insertions of arginines into these strains (R3(ΔGS10)R11 and R11ΔS15) decreased their nuclear volume (R11ΔS15 p\<0.001 compared to ΔS15), restoring them to levels similar to WT. The control strains with either lysines or R6 had nuclear volumes similar to or larger than WT. Neither arginines nor serines had any effect on total cell size as measured by concanavalin A staining ([Figure 3---figure supplement 1A,B](#fig3s1){ref-type="fig"}; [Supplementary file 3](#SD4-data){ref-type="supplementary-material"}). In the FY406 strain background, ΔS15 did not cause an increase in nuclear volume; and thus both R11 and R11ΔS15 strains exhibited smaller nuclear volumes than isogenic WT ([Figure 3C](#fig3){ref-type="fig"}; [Supplementary file 3](#SD4-data){ref-type="supplementary-material"}). These data suggest that modulation of chromatin compaction through the H2A NTD, especially in the presence of R11, affects the nuclear volume but this effect may be indirect (see human data below).10.7554/eLife.02792.010Figure 3.Ectopic expression of H2A NTD arginines decreases nuclear volume in yeast.(A) Images of the nuclear envelope, as visualized by Nup49p-GFP, and boxplot of the distributions of nuclear volumes in the indicated strains in the TSY107 background (B) or the FY406 background (C). Dashed lines mark the median value for the WT strain. The boxplot whiskers contain 90% of the data. All scale bars are 1 µm. Boxes are colored if the mean of the indicated strain is significantly different from WT. Red stars denote level of significance: \p\<0.01; \\p\<0.001; \\\p\<0.0001 ([Supplementary file 3](#SD4-data){ref-type="supplementary-material"}). Red dagger (†) indicates that mean nuclear volume of R11ΔS15 is significantly smaller than its isogenic WT strain (ΔS15; p\<0.001). See also [Figure 3---figure supplement 1](#fig3s1){ref-type="fig"}.DOI: [http://dx.doi.org/10.7554/eLife.02792.010](10.7554/eLife.02792.010)10.7554/eLife.02792.011Figure 3---figure supplement 1.H2A arginines do not affect cell size.(A) Images of the cell wall, as visualized by concanavalin A staining, and (B) boxplot of the distributions of cellular volumes in the indicated yeast strains. The scale bar is 5 µm.DOI: [http://dx.doi.org/10.7554/eLife.02792.011](10.7554/eLife.02792.011) Loss of H2A arginines causes de-compaction of chromatin in human cells {#s2-4} ---------------------------------------------------------------------- Since the H2A NTD in large genomes contains both R3 and R11, we expected that their removal would cause de-compaction of chromatin. To test this prediction, we ectopically expressed WT or mutant H2A in several human cell lines and measured the distances between probes 0.49 Mbp apart on chromosome 1 by FISH, as well as the largest nuclear cross-sectional areas ('Materials and methods'). The H2A gene was HA-tagged and mutated to remove R3 (ΔR3), to replace R11 with alanine (R11A) or lysine (R11K), or to combine two mutations (ΔR3R11A). The H2A constructs were overexpressed using the strong CMV promoter in the normal human IMR90 fibroblasts, the breast cancer cell line MDA-MB-453, or the HEK293 cells. Cells overexpressing ΔR3, R11A, or ΔR3R11A H2A mutants had increased interprobe distances, indicating de-compaction of chromatin. Expression of H2A R11K had modest effects on chromatin de-compaction with marginal statistical significance ([Figure 4A,B](#fig4){ref-type="fig"}, [Figure 4---figure supplement 1A,B](#fig4s1){ref-type="fig"}; [Supplementary file 4](#SD5-data){ref-type="supplementary-material"}). Cells expressing any of the H2A mutants displayed larger nuclear areas, suggesting that nuclear size is increased ([Figure 4C,D](#fig4){ref-type="fig"}, [Figure 4---figure supplement 1C--F](#fig4s1){ref-type="fig"}; [Supplementary file 5](#SD6-data){ref-type="supplementary-material"}). Equal degrees of overexpression were confirmed by immunofluorescence analysis with an anti-HA antibody and detection of HA-H2A by western blotting ([Figure 4D](#fig4){ref-type="fig"}, [Figure 4---figure supplement 1G](#fig4s1){ref-type="fig"}). Ectopic expression of a C-terminally FLAG-tagged H2A mutant missing residues 1--12 (Δ1--12) also caused significant de-compaction of chromatin and increased nuclear area despite being expressed at a lower level than WT ([Figure 4E,F](#fig4){ref-type="fig"}). These data demonstrate that, consistent with our predictions, the H2A NTD, especially arginines 3 and 11, function to compact chromatin in human cells.10.7554/eLife.02792.012Figure 4.Loss of H2A NTD arginines decreases chromatin compaction in human cells.(A) FISH images of probes on chromosome 1 in normal primary IMR90 fibroblasts with HA-tagged WT or mutant H2A overexpressed as indicated. (B) Boxplot of the distributions of inter-probe distances. Note that R11K was only marginally significant at p=0.023. (C) Immunofluorescence images of IMR90 cells overexpressing HA-tagged WT or mutant H2A as indicated. (D) Top: boxplot of the distributions of largest nuclear cross-sectional areas in the indicated H2A overexpressing cells. Bottom: boxplot of the distributions of α-HA fluorescence intensities. (E) Left: FISH images, as in (A), of IMR90 cells expressing a C-terminal FLAG-tagged WT or tailless (Δ1--12) H2A. Right: boxplot of the distributions of inter-probe distances. (F) Top: immunofluorescence images of IMR90 cells overexpressing FLAG-tagged WT or tailless H2A. Bottom: boxplot of nuclear areas and fluorescence intensities, as indicated. Dashed lines mark the median value for the WT strain. All scale bars are 10 µm. Boxes are colored if the mean of the indicated strain is significantly different from WT. Red stars denote level of significance: \p\<0.01; \\p\<0.001; \\\p\<0.0001 ([Supplementary files 4 and 5](#SD5-data SD6-data){ref-type="supplementary-material"}). See also [Figure 4---figure supplement 1](#fig4s1){ref-type="fig"}.DOI: [http://dx.doi.org/10.7554/eLife.02792.012](10.7554/eLife.02792.012)10.7554/eLife.02792.013Figure 4---figure supplement 1.Loss of H2A NTD arginines decreases chromatin compaction in human cells.(A) FISH images of probes on chromosome 1 in MDA-MB-453 cells with either WT or mutant HA-tagged H2A overexpressed. (B) Boxplot of the distributions of interprobe distances. Immunofluorescence images of (C) MDA-MB-453 or (E) HEK293 cells overexpressing WT or mutant HA-tagged H2A. Boxplot of the distributions of the largest nuclear cross-sectional areas in (D) MDA-MB-453 or (F) HEK293 for the indicated H2A over-expressing cells. (G) Western blot of lysates from HEK293 cells overexpressing the indicated WT or mutant HA-tagged H2A. All scale bars are 10 µm. Dashed lines mark the median value for the WT strain. Boxes are colored if the mean of the indicated strain is significantly different from WT. Red stars denote level of significance: \p\<0.01; \\\p\<0.0001.DOI: [http://dx.doi.org/10.7554/eLife.02792.013](10.7554/eLife.02792.013) H2A R11 regulates compaction of nucleosomal arrays in vitro {#s2-5} ----------------------------------------------------------- Because R11 compacts chromatin in vivo, we investigated whether this effect is directly on the chromatin fiber. We used step-wise salt dialysis to assemble nucleosomal arrays with a DNA template containing 12 copies of the 177 bp '601' nucleosome positioning sequence (601-177-12) and recombinant Xenopus laevis histone octamers that contain either WT H2A or one with R11 deleted (ΔR11). We assembled nucleosomal arrays at different octamer-to-template ratios (0.9, 1, and 1.1 octamer to 1 template) and monitored the quality of the arrays by MgCl~2~ precipitation and restriction digest analysis using ScaI. We found that a 1:1 octamer-to-template ratio gave the best results as the ScaI digest demonstrated well-assembled arrays compared to the 5% free DNA loaded as a comparison ([Figure 5A](#fig5){ref-type="fig"}). We used analytical ultracentrifugation to determine the sedimentation velocity combined with van Holde--Weischet analysis ([@bib70]) to ascertain the distribution of sedimentation coefficients (S) for each nucleosomal array in the absence or presence of 0.8 mM MgCl~2~, a concentration of the divalent cation that promotes intra-molecular folding of nucleosomal arrays ([@bib60]). In the absence of Mg^2+^, arrays containing WT H2A sedimented with a coefficient of 33.1, which is a value that has been previously shown for similar arrays ([@bib15]; [@bib61]; [@bib73]). In contrast, arrays missing R11 adopted a more extended conformation with a smaller sedimentation coefficient of 31.0 ([Figure 5B](#fig5){ref-type="fig"}). Addition of Mg^2+^ increased compaction of both arrays and shifted the sedimentation coefficients to 39.3 and 37.4 for WT and ΔR11 H2A, respectively ([Figure 5B](#fig5){ref-type="fig"}). A second independent chromatin assembly and ultracentrifuge analysis confirmed these results ([Figure 5---figure supplement 1](#fig5s1){ref-type="fig"}). Thus, in the absence of R11 in the H2A NTD, nucleosomal arrays adopt a less compact conformation even in the presence of divalent cations, showing that R11 directly increases chromatin compaction.10.7554/eLife.02792.014Figure 5.H2A NTD R11 directly modulates the compaction of chromatin fibers in vitro.(A) Polyacrylamide gel electrophoresis (PAGE) of ScaI-digested 601-177-12 DNA template assembled with octamers containing recombinant WT or ΔR11 H2A. As a control, 5% of the 601-177-12 DNA without octamers was also digested. (B) The distribution of sedimentation coefficients determined by van Holde-Weischet analysis plotted against the percent boundary fraction in the absence or presence of 0.8 mM MgCl~2~ as indicated. S~20°C,W~ is the sedimentation coefficient corrected to water at 20°C. See also [Figure 5---figure supplement 1](#fig5s1){ref-type="fig"}.DOI: [http://dx.doi.org/10.7554/eLife.02792.014](10.7554/eLife.02792.014)10.7554/eLife.02792.015Figure 5---figure supplement 1.H2A NTD R11 directly modulates the compaction of chromatin fibers in vitro.The distribution of sedimentation coefficients determined by van Holde-Weischet analysis plotted against the percent boundary fraction in the absence or presence of 0.6 mM MgCl~2~ as indicated. S~20°C,W~ is the sedimentation coefficient corrected to water at 20°C.DOI: [http://dx.doi.org/10.7554/eLife.02792.015](10.7554/eLife.02792.015) Compaction of chromatin by H2A NTD arginines does not alter global gene expression in yeast {#s2-6} ------------------------------------------------------------------------------------------- To determine whether chromatin compaction through H2A arginines interferes with transcription regulation, we examined gene expression patterns in the H2A yeast mutants. Remarkably, there was a high level of correlation (≥0.99) between all strains examined ([Figure 6A](#fig6){ref-type="fig"}), and no specific gene ontology was found among the genes that were differentially expressed by twofold or more. The expression levels of the histone genes were similar, indicating that altered levels of histone genes expression do not account for the changes in chromatin compaction. These data indicate that compaction of chromatin by H2A does not significantly alter global gene expression in exponentially growing cells.10.7554/eLife.02792.016Figure 6.Mutations to H2A NTD decrease the fitness of yeast.(A) Pearson correlations between the global gene expressions of the indicated strains grown in YPD. Correlations are calculated from an average of at least two experiments. (B) Growth curves of the indicated H2A yeast strains over 10 hr in YPD. (C) Spot tests with 10-fold serial dilutions for the indicated strains in the presence of different drugs. (D) The proportion of yeast cells in a co-culture of WT and the indicated mutant H2A carrying Pgk1 gene fusion to GFP (green) or RFP (red) as indicated by color.DOI: [http://dx.doi.org/10.7554/eLife.02792.016](10.7554/eLife.02792.016) All strains also showed similar growth rates in rich media ([Figure 6B](#fig6){ref-type="fig"}) and no significant differences in sensitivity to hydroxyurea, methyl methanesulfonate (MMS), bleomycin, 4-nitroquinoline 1-oxide (4NQO), cycloheximide, and rapamycin, indicating no major defects with DNA replication or repair, protein synthesis, or the TOR signaling pathways ([Figure 6C](#fig6){ref-type="fig"}). But in competition growth assays in which equal amounts of WT and H2A mutant cells harboring the PGK1 gene fused to either GFP or RFP were co-cultured, the H2A mutants regardless of any effect on chromatin compaction, were outcompeted ([Figure 6D](#fig6){ref-type="fig"}). This suggests that changes in the H2A NTD sequence can affect the overall fitness of the cell. H2A NTD arginines and their surrounding residues are mutated in cancer and affect chromatin compaction {#s2-7} ------------------------------------------------------------------------------------------------------ Deregulated chromatin compaction is often a pathological hallmark of cancer cells ([@bib18]), although the underlying mechanisms are not well-understood. A survey of the COSMIC database ([@bib25]), as of the time of writing, revealed 41 documented missense mutations within the H2A NTD with 29 (71%) affecting a residue within one of the four arginine motifs ([Figure 7A](#fig7){ref-type="fig"}). R11, which had the strongest effect of any single arginine residue on chromatin compaction, is the most commonly mutated residue in the H2A NTD. We tested the effects of three of these mutations, R11C, H, and P and found that ectopic expression of each in normal human fibroblasts decreases chromatin compaction significantly with R11P having the strongest effect ([Figure 7B,C](#fig7){ref-type="fig"}). These cancer mutations have little effect on increasing nuclear area, however ([Figure 7D,E](#fig7){ref-type="fig"}), in contrast to R11A ([Figure 4D](#fig4){ref-type="fig"}). It is unclear to what extent the H2A mutants have to be expressed in cancer cells relative to the 17 canonical H2A genes in the human genome to affect chromatin compaction. But our data suggest that over-expression of an H2A mutant has the potential to disrupt chromatin compaction in cancer.10.7554/eLife.02792.017Figure 7.Mutations of H2A NTD found in cancers decreases chromatin compaction in human cells.(A) Schematic of the H2A NTD showing only the mutations within the arginine motifs found in various cancers as indicated by the colored shapes ([@bib25]). The letter within each shape represents the mutated amino acid. (B) FISH images of probes on chromosome 1 in normal primary IMR90 fibroblasts with HA-tagged WT or mutant H2A overexpressed as indicated. (C) Boxplot of the distributions of inter-probe distances. (D) Immunofluorescence images of IMR90 cells overexpressing HA-tagged WT or mutant H2A as indicated. Anti-HA primary and Alexa Fluor 647-conjugated secondary antibodies were used to determine expression in FISH images and for measurement of nuclear areas. (E) Top: boxplot of the distributions of largest nuclear cross-sectional areas in the indicated H2A overexpressing cells. Bottom: boxplot of the distributions of α-HA fluorescence intensities. Dashed lines mark the median value for the WT strain. All scale bars are 10 µm. Boxes are colored if the mean of the indicated strain is significantly different from WT. Red stars denote level of significance: \p\<0.01; \\\p\<0.0001 ([Supplementary files 4 and 5](#SD5-data SD6-data){ref-type="supplementary-material"}).DOI: [http://dx.doi.org/10.7554/eLife.02792.017](10.7554/eLife.02792.017) Discussion {#s3} ========== In this study, we describe evolutionary adaptations of the histone H2A whereby single arginines in the NTD function to dramatically affect the degree of genome compaction. This mechanism is distinct from several other known chromatin compaction mechanisms in higher eukaryotes ([@bib2]; [@bib15]; [@bib61]; [@bib73]; [@bib23]) in that it involves the histone proteins themselves. The H2A-mediated chromatin compaction thus provides a novel but potentially complementary mechanism for genome compaction. Organisms with small genomes but also very small cell size may face similar physical constraints as those with larger genomes, and may therefore use arginine-containing H2A as a means for chromatin compaction. For instance, Ostreococcus tauri which possesses an R3-containing H2A, is a free-living unicellular algae that has a very small genome of 12.6 Mbp but a cell diameter of 0.8 µm ([@bib50]). S. cerevisiae, which does not contain an H2A with R3, has a similarly sized genome but has a cell diameter that is approximately five times larger. Furthermore, certain organisms such as Oikopleura dioica, which has one of the smallest genomes in animals, have distinctive life cycles and possess H2A genes with and without arginines, which may enable them to dynamically regulate genome compaction at different stages of their life cycles ([@bib47]) (for species with H2A isoforms, see attached spreadsheet). So, the H2A arginines may have evolved in circumstances when the genome size became disproportionately large compared to nuclear volume. Interestingly, the toad, Bufo gargarizans, which has a genome size that is twice as large as the human genome, possesses an H2A gene with not only R3 and R11 but also glutamine 6 replaced with an arginine, suggesting that additional arginines in the H2A tail may enable further compaction in organisms with even larger genomes. To better understand the three dimensional positions of the H2A NTD arginines, we examined a crystal structure of the mono-nucleosome in which R3, R11, R17, and R20 were all simultaneously crystalized, and visualized interactions between nucleosomes within the crystal lattice ([@bib12]). Interestingly, while R17 and R20 are more buried within the octamer, R3 and R11 are situated close to the DNA backbone. R3 is at 2.87 Å from the DNA and could potentially bind the DNA gyre as the DNA wraps around the histone octamer. R11 forms close contacts with the DNA phosphate backbone of self and neighboring nucleosomes (4.09, 2.90 Å, respectively). Although these interactions may have helped form the crystal lattice, they also suggest a possible mechanism for tighter nucleosomal stacking in vivo through shielding of the DNA negative charge ([Figure 1---figure supplement 2A,B](#fig1s2){ref-type="fig"}; [@bib12]). Thus, the evolutionary appearance of arginines in the H2A NTD sequence at positions 3 and 11 corresponds to strategic positioning of R3 and R11 within the nucleosome structure that may enable interactions with the DNA, leading to more compact chromatin. Our in vivo data in yeast cells demonstrate that interprobe distances shorten in the presence of the H2A arginines R3 and R11. While the mechanism of this shortening is still unknown, the two most likely explanations are due to linear chromosomal compaction or increased chromatin looping ([@bib3]). However, our data are more consistent with increased linear compaction due to several reasons. First, our analysis of multiple probes along Chr XVI in yeast demonstrates a uniform compaction between all probe pairs examined. Second, our in vivo data with human cell lines shows de-compaction of chromatin in the absence of R11. If chromatin looping was the mechanism, loops would have to be disassembled in human nuclei independent of factors such as CTCF and condensin. Third, our in vitro data show that R11 alone affects chromatin compaction even in absence of divalent cations. Because the in vitro experiments were performed with unmodified histones in arrays with equal linker lengths, this strongly points to a direct effect of H2A NTD arginines on chromatin compaction that occurs in short arrays. Both R3 and R11 are at contact distances from the DNA, and R11 may also contact the DNA backbone of the neighboring nucleosome ([Figure 1---figure supplement 2](#fig1s2){ref-type="fig"}). These intra- and inter-nucleosomal interactions with the arginines and the DNA may serve to neutralize the negative charge of the DNA backbone, leading to enhanced stacking of nucleosomes and hence increased compaction. Consistent with this model, the other two arginine residues in the H2A NTD, R17 and R20, which are more buried from the surface, do not affect compaction by themselves. However, their functions may be to modulate the effects of the surrounding residues. Although all canonical H2A genes contain R3 and R11 in humans, the cell may still be able to dynamically regulate chromatin compaction by these arginines. For instance, arginines may be subject to posttranslational modifications, such a methylation which makes the arginine residue bulkier, or citrullination which removes the positive charge ([@bib69]; [@bib31]; [@bib14]; [@bib68]). Interestingly, the H2A NTD is situated in close proximity to the H2B CTD which when ubiquitylated, disrupts chromatin compaction in vitro ([@bib23]), lending support to the ability of this region of the nucleosome to modulate chromatin compaction. The inability of lysines, especially at position 11, to increase chromatin compaction suggests exquisite structural constraints for H2A-mediated chromatin compaction. Although lysines and arginines both are positively charged, the positive charge of arginine is due to the presence of a guanidinium group that is structurally different from the positive charge of an amino group of a lysine residue. In this regard, it is interesting to note that only arginines contact DNA as it wraps around the nucleosome core ([@bib42]); and arginines preferentially bind the minor groove of DNA compared to lysines ([@bib54]). The context in which lysines appear in evolution may be important as well. We did not observe a lysine at position 3 in our list of organisms, and K11 was present in organisms with medium-sized genomes and surrounded mainly by the motif VKG ([Figure 1D,E](#fig1){ref-type="fig"}). When tested in our S. cerevisiae strains, K11 was in the context of AKA. So, it is conceivable that additional amino acid changes would be required for lysines in the H2A NTD to increase genome compaction. The nucleoskeletal theory proposes that chromatin structure influences the shape of the nucleus, and thus is a major determinant of nuclear volume ([@bib8]), although the amount of DNA per se does not affect nuclear volume ([@bib48]). Non-chromatin components such as nuclear import factors from the cytoplasm may also modulate nuclear size ([@bib36]). Our data suggest that in particular cases, the effects of H2A NTD mutations on chromatin compaction are linked to nuclear volume, although not in a straightforward relationship. While arginines at positions 3 and 11 increase chromatin compaction and reduce nuclear volume, lysines at the same positions have no effect on chromatin compaction yet increase nuclear volume. Removal of serines at positions 10 or 15 has little effect on compaction but also increase nuclear volume. In human cells, expression of all R11 mutants (R11A, C, H, and P) decreased compaction but only R11A also affected nuclear area. Although we do not observe a clear-cut relationship between nuclear volume and chromatin compaction, our data identify a region of the nucleosome that is directly or indirectly linked to nuclear volume control mechanisms. Since alterations in chromatin structure often cause changes in transcription ([@bib51]), we were surprised that mutant H2A-containing yeast had very similar gene expression profiles as WT cells, grew at similar rates, did not have altered cell cycle profiles, and were not sensitive to DNA damaging drugs or environmental challenges. These data raise the possibility that H2A-mediated compaction of chromatin may have evolved as a mechanism to enable regulation of chromatin compaction without having to make compensatory changes to all other processes that are also based on DNA such as transcription. Nevertheless, it remains to be determined what molecular or cellular processes govern the optimal level of chromatin compaction and nuclear volume for an organism. The stable alterations of chromatin compaction in eukaryotic model organisms through genetic manipulation of H2A should facilitate further experiments to uncover these processes. Materials and methods {#s4} ===================== Strains and media {#s4-1} ----------------- The yeast strains used in this study are listed in [Supplementary file 1A](#SD2-data){ref-type="supplementary-material"}. Yeast cells were grown in YPD at 30°C unless otherwise noted. C-terminal tagging of yeast proteins was performed as described previously ([@bib39]). Mammalian cell lines were maintained at 37°C and 5% CO~2~ and cultured with 10% fetal bovine serum and DMEM (Life Technologies, Grand Island, NY). Histone sequence database construction and analysis {#s4-2} --------------------------------------------------- Sequences were initially extracted from the Entrez database using a keyword search for 'histone', and removing non-histone sequences by using keyword searches such as 'histone-like', 'ubiquitin', and 'acetyl', yielding 54,646 results. Blast 2.0 ([@bib5]) was used to align the sequences against the highly conserved histone fold region of the four core histones from Homo sapiens. Thresholds for true hits were set at \>35% identity match and \>90% overlap match with the histone fold globular domain region. All duplicate sequences were removed, and further sequence comparisons were made for histone H3 and H2A sequences to filter variants within them. The canonical sequence data sets comprised 672 sequences for histone H3, 357 sequences for histone H4, 518 sequences for histone H2B, and 435 sequences for histone H2A. To further select one canonical sequence for a species among isotypes and variants when annotation was missing, the sequences were compared to the canonical H. sapiens and S. cerevisiae sequence, and the sequence with the highest similarity was selected. Using only completely sequenced species, the final histone sequence data set included canonical sequences for 160 species from plants, fungi, protozoa, and animals, with genome sizes ranging from 8 to 5600 Mbp. Sequences for the four core histones were subsequently split into the N-terminal tail, globular domain, and C-terminal tail (in the case of H2A and H2B) sub-sequences. For discovery of patterns of residue changes according to genome size, each of the sub-sequences was further sub-grouped into small (\<100 Mbp), medium (100--1000 Mbp), and large (\>1000 Mbp) genome sizes. The frequency of the amino acid residues in each sequence in the sub-groups was determined, and a p-value for the comparison between sub groups was obtained using a Mann--Whitney U Test. Multiple sequence alignment profiles were created using the Muscle sequence comparison tool from Embl-EBI ([@bib19], [@bib20]). Weblogo3 ([@bib58]; [@bib11]) was used for motif discovery. Heat maps for residue positions were constructed using Cluster 3.0 ([@bib13]) and Java Treeview ([@bib55]). Yeast H2A mutagenesis {#s4-3} --------------------- Site directed mutagenesis was performed using the QuickChange Lightning kit (Agilent Technologies, Santa Clara, CA) on the pFL142 plasmid. [Supplementary file 1B](#SD2-data){ref-type="supplementary-material"} contains all the plasmids that were used and constructed in this study. The sequences of primers are listed in [Supplementary file 1C](#SD2-data){ref-type="supplementary-material"}. The correct mutation was verified by sequencing. Measurement of yeast nuclear volume {#s4-4} ----------------------------------- Yeast strains were generated that contained a C-terminally tagged Nup49p-GFP fusion. Cells were grown in a rich medium to 0.6--0.8 × 10^7^ cells/ml, fixed in a growth medium with 4% paraformaldehyde for 15 min at room temperature, washed twice in PBS, and mounted on a poly-L-lysine-coated slide with mounting medium (Vector Laboratories, Burlingame, CA). Z-stacks were obtained as described in the microscopy imaging section, and GFP excitation was achieved at 488 nm. Resulting z-stack images were de-convolved using a constrained iterative algorithm from SlideBook 5.0 software and nuclear volumes were measured by masking the inside of each nucleus, which were delineated by the GFP signal. The resulting mask was used to calculate volumes through the SlideBook software. Statistical analysis was performed using the Student\'s t test. Measurement of yeast cellular volume {#s4-5} ------------------------------------ Yeast strains were grown in rich medium to 0.6--0.8 × 10^7^ cells/ml, fixed in growth medium with 4% paraformaldehyde for 15 min at room temperature, washed twice in PBS, and stained with a 1:50 dilution of concanavalin A conjugated with tetramethylrhodamine (Life Technologies) for 15 min at room temperature. Cells were washed twice in PBS, once in water, and mounted on a poly-L-lysine-coated slide with mounting medium. Z-stacks were obtained as described in the microscopy imaging section with mRFP excitation. Cell volume was measured by masking the inside of the RFP signal as described in the measurement of yeast nuclear volume. FISH probes {#s4-6} ----------- For yeast FISH analysis, DNA templates for probes 1, 3, and 4 came from cosmids 71042, 70912, and 70982 (American Type Culture Collection, Manassas, VA) as described elsewhere ([@bib29]). DNA templates for Probe 2 were obtained by PCR amplification of a 10-kb region starting at coordinate 364647 of chromosome 16 using three primer pairs (Probe2_P1, Probe2_P2, Probe2_P3, [Supplementary file 1C](#SD2-data){ref-type="supplementary-material"}). All DNA templates were digested to smaller fragments using Sau3a (New England BioLabs, Ipswich, MA). Fragments were directly labeled using BioPrime labeling kit (Life Technologies) with either ChromaTide Alexa Fluor 488-5-dUTP or ChromaTide Alexa Fluor 568-5-dUTP (Life Technologies). For human cell FISH analysis, DNA templates for probes came from BACS RP11-252L24 and RP11-195J4 spaced 0.488 Mb apart on chromosome 1. Each BAC was digested into smaller fragments using Sau3a and fragments were directly labeled using BioPrime labeling kit with either ChromaTide Alexa Fluor 488-5-dUTP or ChromaTide Alexa Fluor 568-5-dUTP, as described above. Fluorescent in situ hybridization analysis in yeast {#s4-7} --------------------------------------------------- Yeast strains were grown in rich medium to 0.6--0.8 × 10^7^ cells/ml and fixed in a growth medium with 4% paraformaldehyde for 15 min at room temperature. Cells were then washed twice in the growth medium and re-suspended in 2 ml of EDTA-KOH (0.1 M, pH 8.0) and 10 mM DTT and incubated for 10 min with shaking at 30°C. Cells were spun down and re-suspended in 2 ml of YPD + 1.2 M sorbitol with 50 µg/ml of Zymolyase 100-T (Sunrise Science Products, San Diego, CA) and 400 U/ml of lyticase (Sigma-Aldrich, St. Louis, MO) and incubated at 30°C for 16 min with shaking. Spheroplasts were then washed twice in YPD + 1.2 M sorbitol and transferred to a poly-L-lysine-coated slide. After settling for 5 min, excess liquid was aspirated away and the slides were allowed to air dry for 5 additional min. The slides were washed in methanol for 10 min and then acetone for 30 s before air drying. Cells were then dehydrated in a series of cold ethanol washes (70%, 80%, 90%, 100%, 1 min each) and allowed to air dry. Denaturing solution (70% deionized formamide, 2 × SSC) was added to the slide, and cells were denatured at 75°C for 7--10 min. The slides were immediately put through another cold ethanol dehydration series and allowed to air dry. Hybridization solution (50% deionized formamide, 2 × SSC, 10% dextran sulfate, 100 ng/µl salmon sperm DNA) containing fresh probes was added to the slide, and the probes were hybridized for 40--48 hr at 37°C. The slides were then washed in two 5 min washes in 0.05 × SSC at 48°C and washed twice in BT Buffer (0.15 M NaHCO3 pH 7.5, 0.1% Tween) for 5 min at room temperature. Mounting medium containing DAPI (Vector Laboratories) was added to the slides, and a coverslip was sealed with nail polish. Inter-probe distances were measured in single projections as described elsewhere ([@bib4]) by finding the pixel distance between weighted centers of the green signal and red signal and converted to nm by the appropriate factor. Microscopy imaging {#s4-8} ------------------ A 3i Marianas SDC confocal microscope equipped with a Zeiss AxioObserver Z1 with a 100 × /1.45 NA objective and Yokogawa CSU-22 confocal head was used. Images were captured by a Hamamatsu EMCCD C9100-13 camera controlled by Slidebook 5.0/5.5 (Intelligent Imaging Innovations, Denver, CO). DAPI, GFP, mRFP, and Far-red images were acquired by excitation at 360 nm, 488 nm, 561 nm, and 640 nm from a high-speed AOTF laser launch line. A step size of 0.3 (yeast) or 0.5 (human) µm was used for z-stack acquisition. Micrococcal nuclease digestion {#s4-9} ------------------------------ Micrococcal nuclease (MNase) digestions were performed on exponentially growing yeast cells as described previously, except that the enzyme was obtained from Sigma-Aldrich (Sigma-Aldrich) ([@bib52]). RNA expression analysis {#s4-10} ----------------------- RNA was extracted from exponentially growing yeast as described previously ([@bib57]). PolyA-RNA was prepared, labeled, and hybridized to Affymetrix Gene ChIP Yeast Genome 2.0 array by the UCLA clinical microarray core facility and data normalized according to manufacturer\'s indications. The data are accessible at Gene Expression Omnibus with accession number GSE50440. DNA template and histone preparation for in vitro studies {#s4-11} --------------------------------------------------------- A plasmid containing 12 tandem 177 bp repeats of the high affinity 601 sequence was obtained from Craig L Peterson\'s laboratory ([@bib61]). DNA arrays were prepared as described previously ([@bib43]). After excision with EcoRV, the arrays were gel purified. QuikChange Lightning Site-Directed Mutagenesis (Agilent Technologies) was used to create H2A ΔR11 using primers as listed in [Supplementary file 1C](#SD2-data){ref-type="supplementary-material"}. Recombinant X. laevis histones were expressed in bacteria and purified as described previously ([@bib43]). Equimolar amounts of all histones were co-folded to form octamers. Intact octamers were purified from aggregates and free H2A-H2B dimers using Pharmacia Superdex 200 gel filtration column (GE Healthcare Bio-Sciences, Pittsburgh, PA). Nucleosome array assembly {#s4-12} ------------------------- Recombinant histone octamers and the 601-177-12 DNA template ([@bib40]) were combined in stoichiometric amounts where 1.0 equivalent of histone octamers and 1.0 equivalents of DNA template were mixed in 2.0 M NaCl. Nucleosome arrays were assembled by step-wise salt dialysis in decreasing NaCl concentration: 1.6 M, 1.2 M, 1.0 M, 0.6 M, 0.4 M, 0.1 M, and 0.025 M (in 10 mM Tris pH 8.0, 0.25 mM EDTA), followed by exchanges with 2.5 mM NaCl and 10 mM Tris pH 8.0 without EDTA. Each dialysis step was performed at 4°C for 4 hr to overnight. Partially assembled chromatin was eliminated by precipitation in 4.0 mM MgCl~2~ ([@bib15]). The extent of array saturation was assessed by ScaI digestion (200 ng total DNA/chromatin, 3 units ScaI, 50 mM NaCl, 50 mM Tris pH 7.4, 0.5 mM MgCl~2~), performed for 16 hr at room temperature followed by 1 hr at 37°C, and subsequent analysis using a 5% native polyacrylamide gel ([@bib43]). Analytical ultracentrifugation {#s4-13} ------------------------------ Nucleosome arrays were allowed to equilibrate at room temperature in buffer (2.5 mM NaCl, 10 mM Tris--HCl pH 8.0) containing either 0.1 mM EDTA or 0.6 and 0.8 mM MgCl~2~. Samples were centrifuged at 20,000 RPM on a Beckman Optima XL-I analytical ultracentrifuge using an An60 Ti rotor after a 1 hr equilibration at 20°C under vacuum. Time-dependent sedimentation was monitored at 260 nm. Boundaries were analyzed by the method of van Holde and Weischet ([@bib70]; [@bib32]). Combined immunofluorescence and fluorescent in situ hybridization in human cells {#s4-14} -------------------------------------------------------------------------------- N-terminally HA-tagged WT H2A of X. laevis was cloned by PCR into mammalian expression vector pCMV-HA (Clontech Laboratories, Mountain View, CA) between EcoRI and NotI sites. C-terminally Myc-FLAG-tagged human H2A, in a mammalian expression vector, was obtained from OriGene (RC200688, Origene Technologies, Rockville, MD). Site directed mutagenesis was performed using the QuickChange Lightning kit (Agilent Technologies) on these expression plasmids. Human cells (HEK293, IMR90 and MDA-MB-453) were grown on glass coverslips in 24-well plates in DMEM containing 10% fetal bovine serum and transfected with the indicated H2A expression plasmids using BioT transfection reagent (Bioland Scientific, Paramount, CA) or Lipofectamine LTX with Plus reagent (Life Technologies). Cells were grown for 48 hr post-transfection. For immunofluorescence only, transfected cells were fixed with ice-cold methanol for 15 min at −20°C followed by washing with PBS-T. For combined immunofluorescence and FISH, transfected cells were fixed with 4% paraformaldehyde in PBS for 10 min at room temperature followed by washing with PBS. Cells were then permeabilized in 0.5% Triton X-100 in PBS for 10 min at room temperature followed by washing with PBS. Cells were blocked in 5% BSA and incubated with anti-HA antibody (ab9110; 1:250 dilution, Abcam, Cambridge, MA) or anti-FLAG antibody (F1804; 1:1000 dilution, Sigma-Aldrich). Cells were washed and incubated with secondary antibody (A11008; 1:500 Alexa Fluor 488 goat anti-rabbit, A21245; 1:250 Alexa Fluor 647 goat anti-rabbit, A11001; 1:500 Alexa Fluor 488 goat anti-mouse, or A21235; 1:100 Alexa Fluor 647 goat anti-mouse, Life Technologies). For immunofluorescence, cells were washed and then incubated with Hoechst stain (0.001 mg/ml in PBS). After final washes, cover slips were mounted and imaged. Fluorescence was visualized as above except with the use of 63X magnification. For FISH, cells were washed, following secondary antibody incubation, in CSK buffer (100 mM NaCl, 300 mM sucrose, 3 mM MgCl~2~, 10 mM PIPES pH 6.8) and permeabilized in CSKT buffer (CSK+0.5% Triton X-100) before being fixed for 10 min in 4% paraformaldehyde in PBS at room temperature. Cells were immediately put through a cold ethanol dehydration series (5 min each at 85%, 95%, and 100%) and allowed to air dry. Cells were rehydrated in 2 × SSC for 5 min and then RNase-treated for 30 min at 37°C in a humid chamber. Cells were washed with 2 × SSC and denatured at 80°C for 15--20 min with 70% deionized formamide and 2 × SSC. They were immediately cooled with cold 2 × SSC and put through another cold ethanol dehydration series. Probes were added to cells and allowed to hybridize for 48 hr. After hybridization, cells were washed with 50% formamide in 2 × SSC, 2 × SSC, and 1 × SSC containing DAPI. Slides were mounted, imaged, and analyzed as described above. Nuclear staining, in H2A-expressing cells, was used to measure lengths of the long and short orthogonal nuclear axes. Estimated nuclear cross-sectional area was calculated using the following formula: Area = (D~1~/2) \* (D~2~/2)\π, where D~1~ and D~2~ are long and short axis lengths, respectively. Competition assays {#s4-15} ------------------ Two sets of yeast strains were generated in which Pgk1p was C-terminally fused with either GFP or RFP ([Supplementary file 1A](#SD2-data){ref-type="supplementary-material"}). GFP-labeled WT H2A strains were co-cultured with RFP-labeled mutant H2A strains at a 1:1 ratio and at an optical density of ∼0.4. Corresponding co-cultures with switched fluorescent labels were also made. Cultures were incubated at 30°C for 72 hr and were diluted every 6--12 hr to maintain cells in exponential growth phase. Samples were collected every 12 hr for analysis by flow cytometry. Collected cells were fixed in 70% ethanol, washed, and re-suspended in 50 mM sodium citrate, pH 7.0, and mildly sonicated to disrupt aggregates. GFP- and RFP- labeled cells were counted using a Becton Dickinson FACScan cytometer, and the proportion of each in the population was calculated. Cell cycle analysis {#s4-16} ------------------- Cell cycle analysis of exponentially growing cells was performed essentially as described previously ([@bib75]), except that cells were stained with 1 μM SYTOX Green (Life Technologies). Spot tests {#s4-17} ---------- Approximately 1.0 × 10^7^ exponentially growing yeast cells were collected and re-suspended in 100 μl of H~2~O and 10-fold serially diluted. Subsequently, 5 μl was spotted on agar plates containing media and drugs as indicated in the figures and incubated at 30°C for 2--6 days. Funding Information =================== This paper was supported by the following grants: - National Institutes of Health [FundRef identification ID:](http://www.crossref.org/fundref/)http://dx.doi.org/10.13039/100000002 Director\'s Innovator Award, 1 DP2 OD006515 to Siavash K Kurdistani. - Howard Hughes Medical Institute [FundRef identification ID:](http://www.crossref.org/fundref/)http://dx.doi.org/10.13039/100000011 Medical Student Fellowship to Benjamin R Macadangdang. - University of California [FundRef identification ID:](http://www.crossref.org/fundref/)http://dx.doi.org/10.13039/100005595 Philip Whitcome Pre-doctoral Training Program to Benjamin R Macadangdang. - University of California [FundRef identification ID:](http://www.crossref.org/fundref/)http://dx.doi.org/10.13039/100005595 UCLA Cancer Cell Biology Postdoctoral Training Program to Tanya Spektor. We acknowledge the support of Martin Phillips and the UCLA-Department of Energy Biochemistry Instrumentation Facility for analytical ultracentrifuge experiments, the Jonsson Comprehensive Cancer Center flow cytometry core facility (supported by grants P30 CA016042 and 5P30 AI028697) and the UCLA Broad Stem Cell Center Sequencing Core. We thank Michael Grunstein and Fred Winston for providing yeast strains TSY107 and FY406, respectively. BM was supported partially by a Howard Hughes Medical Institute Medical Student Fellowship and a Philip Whitcome Pre-doctoral Training Program grant. TS was supported partially by a UCLA Cancer Cell Biology Postdoctoral Training grant. This work was funded by an NIH Director\'s Innovator Award to SKK. Additional information ====================== The authors declare that no competing interests exist. BRM, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. AO, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. TS, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. OAC, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. MV, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. FS, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. MFC, Conception and design, Analysis and interpretation of data, Drafting or revising the article. SKK, Conception and design, Analysis and interpretation of data, Drafting or revising the article. Additional files ================ 10.7554/eLife.02792.018 ###### Tables of yeast strains, plasmids, and primers. DOI:* [http://dx.doi.org/10.7554/eLife.02792.018](10.7554/eLife.02792.018) 10.7554/eLife.02792.019 ###### Table of yeast FISH results. DOI: [http://dx.doi.org/10.7554/eLife.02792.019](10.7554/eLife.02792.019) 10.7554/eLife.02792.020 ###### Table of yeast nuclear and cellular volumes. DOI: [http://dx.doi.org/10.7554/eLife.02792.020](10.7554/eLife.02792.020) 10.7554/eLife.02792.021 ###### Table of human FISH results. DOI: [http://dx.doi.org/10.7554/eLife.02792.021](10.7554/eLife.02792.021) 10.7554/eLife.02792.022 ###### Table of human nuclear area results. DOI: [http://dx.doi.org/10.7554/eLife.02792.022](10.7554/eLife.02792.022) Major dataset ------------- The following dataset was generated: SiavashKurdistani, 2014,Expression data from Saccharomyces cerevisiae histone H2A mutants,<http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE50440>,Publicly Available at NCBI Gene Expression Omnibus. 10.7554/eLife.02792.023 Decision letter Kadonaga James T Reviewing editor University of California, San Diego , United States eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see [review process](http://elifesciences.org/review-process)). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers. Thank you for sending your work entitled "Evolution of Histone 2A for Chromatin Compaction in Eukaryotes" for consideration at eLife. Your article has been favorably evaluated by James Manley (Senior editor) and 3 reviewers, one of whom is a member of our Board of Reviewing Editors. The Reviewing editor and the other reviewers discussed their comments before we reached this decision, and the Reviewing editor has assembled the following comments to help you prepare a revised submission. This paper explores a novel and interesting hypothesis, which is that the inclusion of extra arginine residues at position 3 and 11 in the N-terminal tail of histone H2A allows a greater degree of compaction of the chromatin. The authors found a strong correlation between the number of arginine residues in the N-terminal domain (NTD) of histone H2A and genome size. This observation led them to postulate that the NTD arginines promote chromatin compaction and a consequent reduction in the nuclear volume. A variety of substitution, insertion, and deletion mutations were constructed in the NTDs of yeast and human H2A. A good correlation between NTD arginines and chromatin compaction was observed. There was not, however, a strong correlation between chromatin compaction and nuclear volume. This study was based on an innovative idea and contains information that will be of interest to chromatin biologists. This work could be appropriate for publication in eLife if the following points are suitably addressed. 1\) [Figure 2](#fig2){ref-type="fig"} is flawed and should be deleted as a main figure. [Figure 2](#fig2){ref-type="fig"} shows the locations of R3 and R11 on an image of two nucleosomes stacked against each other, to infer that these amino acids are involved in interactions between nucleosomes to promote chromatin compaction. However, the structure used (Davey et al., 2002) was a mononucleosome. This paper itself did not show anything more than a mononucleosome, and any potential nucleosome-nucleosome interactions that have been inferred from the coordinates of the mononucleosome may be crystal-packing artifacts. Furthermore, this modeling should be done using instead Tim Richmond\'s tetranucleosome structure, albeit that the nucleosome interactions in that structure have not been validated experimentally. As such, [Figure 2](#fig2){ref-type="fig"} does not stand on its own and should be deleted. The models should be remade with two nucleosomes from the tetranucleosome structure and presented as a supplement to [Figure 1](#fig1){ref-type="fig"}. If the authors wish to keep the Davey et al. models, the text also needs to acknowledge that the structure that was used to generate the models was a mononucleosome and that the interactions shown have not been validated and may be crystal packing artifacts. 2\) The authors should consider and discuss the possibility that the changes in inter-locus distances (i.e., apparent chromatin compaction) may be due to chromatin looping rather than chromatin compaction. This chromatin looping hypothesis could be tested with a 3C-like method, such as Hi-C, but we do not request that such extensive experiments be performed for this paper. This issue will, however, ultimately need to be addressed by the authors. 3\) Changes in nuclear volume do not always correlate with chromatin compaction (as summarized in Table 1). Hence, there is not a straightforward relationship between chromatin compaction and nuclear volume. This point needs to be made clearly in the text. 4\) There are two issues with [Figure 6](#fig6){ref-type="fig"}: a\) The authors reconstituted oligonucleosome arrays with either wild-type H2A or deltaR11. The sedimentation coefficients of the mutant arrays were decreased, suggesting loss of inter-fiber compaction. It should be noted, however that results from the lab of Jeff Hansen (Gordon et al., 2005) showed that reconstituted nucleosome arrays missing only the H2A N-terminal region have a sedimentation profile that is indistinguishable from that of wild-type arrays. Hence, the results from the Hansen lab argue against an essential role of the H2A N-terminal region in chromatin compaction and appear to be inconsistent with the data in [Figure 6](#fig6){ref-type="fig"}. b\) The biochemical experiments provide an excellent opportunity to examine the effect of charge on chromatin compaction in the absence of potential histone acetylation. Thus, it could be informative to test the relative effect of R vs. K in these experiments. However, given the ambiguity in the sedimentation data from different labs (as discussed in point 4a, above), it is not certain how useful such experiments would be. 5\) [Figure 7E](#fig7){ref-type="fig"} shows some mutations of H2A that have been found in cancer. The authors make the argument that the cancer mutations in a single copy of an H2A gene have the potential to affect chromatin compaction, as supported by their overexpression analysis in tissue culture cells. However, in their overexpression study, they show that the mutant histones are expressed to the same level as the endogenous histones. This is the equivalent of a heterozygous mutation in a single copy gene. However, there are 17 canonical H2A genes in humans; hence, the authors are not making a fair argument, as a mutant histone in humans would only be present at 1/34th of the level of the non-mutant histones (assuming equal expression of all histone H2A genes). This fact needs to be made clear in the Discussion. 10.7554/eLife.02792.024 Author response 1) [Figure 2](#fig2){ref-type="fig"} is flawed and should be deleted as a main figure. [Figure 2](#fig2){ref-type="fig"} shows the locations of R3 and R11 on an image of two nucleosomes stacked against each other, to infer that these amino acids are involved in interactions between nucleosomes to promote chromatin compaction. However, the structure used (Davey et al., 2002) was a mononucleosome. This paper itself did not show anything more than a mononucleosome, and any potential nucleosome-nucleosome interactions that have been inferred from the coordinates of the mononucleosome may be crystal-packing artifacts. Furthermore, this modeling should be done using instead Tim Richmond\'s tetranucleosome structure, albeit that the nucleosome interactions in that structure have not been validated experimentally. As such, [Figure 2](#fig2){ref-type="fig"} does not stand on its own and should be deleted. The models should be remade with two nucleosomes from the tetranucleosome structure and presented as a supplement to [Figure 1](#fig1){ref-type="fig"}. If the authors wish to keep the Davey et al. models, the text also needs to acknowledge that the structure that was used to generate the models was a mononucleosome and that the interactions shown have not been validated and may be crystal packing artifacts. We agree with the comments and have moved [Figure 2](#fig2){ref-type="fig"} to the supplementary material (as a supplement to [Figure 1](#fig1){ref-type="fig"}). We have also edited the text to reflect that the internucleosomal contacts are indeed inferred from crystal lattice packing. The reason we performed this analysis in the first place was because the interaction between H4 K16 and the H2A/H2B acidic patch was described initially from crystal lattice packing (Luger et al., 1997). 2) The authors should consider and discuss the possibility that the changes in inter-locus distances (i.e., apparent chromatin compaction) may be due to chromatin looping rather than chromatin compaction. This chromatin looping hypothesis could be tested with a 3C-like method, such as Hi-C, but we do not request that such extensive experiments be performed for this paper. This issue will, however, ultimately need to be addressed by the authors. This is a valid point that we have now explicitly discussed in the paper (see Discussion section). However, we think that this is a less likely possibility because examination of inter-probe distances for four probe-pairs on the same chromosome suggests linear compaction ([Figure 2D](#fig2){ref-type="fig"}). 3) Changes in nuclear volume do not always correlate with chromatin compaction (as summarized in Table 1). Hence, there is not a straightforward relationship between chromatin compaction and nuclear volume. This point needs to be made clearly in the text. We agree with the reviewers and in fact had attempted to make this point in the discussion of the original submission. We have now further amended our discussion on nuclear volume changes to better emphasize this point. 4) There are two issues with [Figure 6](#fig6){ref-type="fig"}: a) The authors reconstituted oligonucleosome arrays with either wild-type H2A or deltaR11. The sedimentation coefficients of the mutant arrays were decreased, suggesting loss of inter-fiber compaction. It should be noted, however that results from the lab of Jeff Hansen (Gordon et al. 2005) showed that reconstituted nucleosome arrays missing only the H2A N-terminal region have a sedimentation profile that is indistinguishable from that of wild-type arrays. Hence, the results from the Hansen lab argue against an essential role of the H2A N-terminal region in chromatin compaction and appear to be inconsistent with the data in [Figure 6](#fig6){ref-type="fig"}. Although we do not have a satisfactory explanation for this discrepancy, we have two counterpoints. 1\) We have performed additional in vivo experiments in which we ectopically expressed either WT H2A or an H2A gene missing amino acids 1-12 ( H2A-NTD) in IMR90 cells. Both constructs were FLAG tagged at the C-terminus. As would be predicted from our data, expression of H2A-NTD which is missing R3 and R11 decreased chromatin compaction (increased inter-probe distances) ([Figure 4E](#fig4){ref-type="fig"}) and also caused an increase in nuclear area ([Figure 4F](#fig4){ref-type="fig"}). We therefore conclude that deletion of the H2A NTD does indeed affect chromatin compaction in vivo. 2\) In Gordon et al., nucleosomal arrays missing the H4 NTD sedimented at a value almost identical to WT octamers: 23.1 S for ΔH4 NTD vs 23.2 S for WT. (Arrays missing H2A sedimented slightly slower than WT arrays with a sedimentation coefficient of 22.7 S). However, other studies have demonstrated that H4 NTD or acetylation of H4K16 do affect chromatin compaction in vitro (Dorigo et al., 2003; Shogren-Knaak et al., 2006). Some of these discrepancies may be attributable to the nature of assembled chromatin (DNA sequence, method of assembly, etc.) and inherent intricacy of such experiments. But we note that none of the referenced studies have validated their findings in vivo. Our in vitro data are supported by additional in vivo evidence for both ectopic addition of the H2A arginines as well as their removal in yeast and human cell line model systems, respectively. b) The biochemical experiments provide an excellent opportunity to examine the effect of charge on chromatin compaction in the absence of potential histone acetylation. Thus, it could be informative to test the relative effect of R vs. K in these experiments. However, given the ambiguity in the sedimentation data from different labs (as discussed in point 4a, above), it is not certain how useful such experiments would be. We share the same point of view and in fact did attempt to assemble chromatin in vitro with H2A R11K mutation. But for unknown reasons we were not able to obtain satisfactorily assembled chromatin for sedimentation velocity analysis. However, in addition to the original data on K11 insertion in yeast H2A ([Figure 2B-C](#fig2){ref-type="fig"}), we ectopically expressed an H2A gene with R11K mutation in normal human fibroblasts and found no statistically significant effects on chromatin compaction ([Figure 4A-B](#fig4){ref-type="fig"}), albeit the average values were slightly larger than WT and trending toward significance (p value=0.02 for FISH distances). Also, R11K did increase the average nuclear area. These data suggest that R11K may have subtle effects on chromatin compaction that would be more apparent if more than one region of the genome was analyzed. Together with the findings from yeast, the data suggest that R11 may be somewhat better than K11 in compacting chromatin and has a greater effect on nuclear area, explaining the evolution of R11 in large genome species. 5) [Figure 7E](#fig7){ref-type="fig"} shows some mutations of H2A that have been found in cancer. The authors make the argument that the cancer mutations in a single copy of an H2A gene have the potential to affect chromatin compaction, as supported by their overexpression analysis in tissue culture cells. However, in their overexpression study, they show that the mutant histones are expressed to the same level as the endogenous histones. This is the equivalent of a heterozygous mutation in a single copy gene. However, there are 17 canonical H2A genes in humans; hence, the authors are not making a fair argument, as a mutant histone in humans would only be present at 1/34th of the level of the non-mutant histones (assuming equal expression of all histone H2A genes). This fact needs to be made clear in the Discussion. This is a valid observation and we have now modified the text to better explain this point. We have also performed additional experiments showing that ectopic expression of H2A that bears cancer mutations also affect chromatin compaction ([Figure 7](#fig7){ref-type="fig"}). [^1]: These authors contributed equally to this work. [^2]: The -marks indicate spacing for sequence alignment purposes. The inserted residues are bold typed and underlined. Deletions are indicated by ·. [^3]: Percent (%) change refers to the difference in median values relative to WT unless otherwise indicated; the statistically significant differences are bold typed. p-values were calculated using the t-test (yeast) and Mann--Whitney U test (human). [^4]: percent change was calculated relative to isogenic WT control (ΔS15).	{ "pile_set_name": "PubMed Central" }
INTRODUCTION ============ Alternative splicing (AS) is considered as the major mechanism that contributes to the increased proteomic diversity in multicellular eukaryotes (Blencowe, [@b7]; Maniatis & Tasic, [@b27]; Nilsen & Graveley, [@b33]). Through AS, one pre-mRNA could produce multiple mRNA isoforms that might be under different post-transcriptional regulation and/or encode proteins with different functions. Recent transcriptome analysis by massive parallel RNA sequencing (RNA-seq) indicated that more than 90% human genes underwent AS (Pan et al, [@b35]; Wang et al, [@b42]). Very often, the pattern of AS was tissue and developmental stage specific, thought to be under precise regulation modulated by cooperative interplays between trans-acting RNA binding proteins (RBPs) and cis-regulatory elements in nascent transcripts (Barash et al, [@b3]; Black, [@b6]; Chen & Manley, [@b10]; Witten & Ule, [@b48]). Mutations in splicing regulators (Padgett, [@b34]; Yoshida et al, [@b51]) and abnormal splicing of RNA targets have been associated with many human diseases (Cooper et al, [@b11]; Garcia-Blanco et al, [@b12]; Wang & Cooper, [@b43]). Nevertheless, the exact molecular mechanisms controlling the AS process in physiological and pathological conditions are not well-understood to date. Several splicing regulating RBPs have recently been found to modulate hundreds even thousands of functional targets (Lebedeva et al, [@b24]; Licatalosi et al, [@b25]; Mukherjee et al, [@b31]; Ule et al, [@b40]; Wang et al, [@b41]; Xue et al, [@b49]; Yeo et al, [@b50]). Therefore, elucidating the regulatory roles of splicing related RBPs requires comprehensive identification of the RBP--RNA interactions and global quantification of the splicing outcomes induced by RBPs. RBM10 encodes a 930 amino acid protein containing two RNA recognition motifs (RRM), two zinc fingers and one G patch motif. These motifs were often found in RNA-binding proteins involved in pre-mRNA splicing, such as heterogeneous nuclear ribonucleoproteins (hnRNPs) and protein components of small nuclear ribonucleoproteins (snRNPs; Glisovic et al, [@b13]; Keene, [@b22]). Through mass spectrometric analysis, RBM10 has been reported to associate with purified splicing complex (Rappsilber et al, [@b37]), and was further identified as a component of U2 snRNPs (Makarov et al, [@b26]), spliceosomal A (or prespliceosomal) (Agafonov et al, [@b1]; Behzadnia et al, [@b4]) and B complexes (Agafonov et al, [@b1]; Bessonov et al, [@b5]). Most recently, based on yeast two hybridization method, a study on interactions between more than 200 proteins previously known to be present in spliceosome could demonstrate the physical interaction between RBM10 and multiple spliceosomal components (Hegele et al, [@b17]). Moreover, its closest paralogue RBM5, a putative tumour suppressor of lung and other cancers (Sutherland et al, [@b39]), has been shown to regulate AS of apoptosis related genes, Fas receptor and c-FLIP, resulting in isoforms with antagonistic functions in controlling programmed cell death (Bonnal et al, [@b8]). Although all these observations would suggest the potential role of RBM10 in pre-mRNA splicing regulation, it remains unclear whether and how RBM10 could regulate splicing. Nonsense and frame shift mutations in RBM10 have been identified to be causative for TARP syndrome (Talipes equinovarus, atrial septal defect, Robin sequence and persistent left superior vena cava, MIM \#311900), an X-linked inherited disorder leading to multiple organ malformation in affected males (Gripp et al, [@b14]; Johnston et al, [@b21]). More recently, multiple truncating and missense somatic mutations were detected in lung adenocarcinomas (Imielinski et al, [@b20]). These findings implicated the important role of RBM10, but whether its potential function in splicing regulation is involved in these different pathological contexts has not been explored. In this study, we explored the AS regulated by RBM10. Here, we combined photoactivatable-ribonucleoside-enhanced crosslinking and immunoprecipitation (PAR-CLIP) with massive parallel sequencing to identify RNA binding sites for RBM10 in human embryonic kidney (HEK) 293 cells, which turned out to be significantly enriched in the vicinity of both 5′ and 3′ splice sites. Using RNA-seq, we identified 304 and 244 significant exon splicing changes following RBM10 depletion or overexpression (OE) in HEK293 cells, respectively. Among these changes, more than 74% were RBM10 enhanced exon skipping events and they were correlated with strong RBM10 binding near 5′ and 3′ splicing sites of both upstream and downstream introns. Furthermore, in a patient suffering from TARP syndrome, we identified an in-frame deletion in RBM10 and demonstrated that the splicing defects in the lymphoblastoid cells derived from the patient were largely due to the loss of nuclear function of RBM10. Overall, our data provides direct experimental evidence supporting the role of RBM10 in splicing regulation. Our transcriptome-wide analysis of binding pattern and RBM10 splicing profile allows the illustration of the molecular mechanism underlying RBM10 regulated AS. RESULTS ======= Transcriptome-wide binding sites of RBM10 in HEK293 cells --------------------------------------------------------- To identify in vivo binding sites of RBM10, we performed PAR-CLIP sequencing (Hafner et al, [@b15]; Lebedeva et al, [@b24]) in HEK293 cells that expressed epitope (FLAG/HA)-tagged RBM10 (Materials and Methods Section). 4-Thiouridine (4SU) labelled and crosslinked cells were immunoprecipitated with monoclonal anti-FLAG antibody. The bound RNAs was then partially digested and radioactively labelled. Protein--RNA complexes were resolved on a denaturing gel. The band corresponding to RBM10--RNA complexes was excised (Supporting Information [Fig S1](#SD1){ref-type="supplementary-material"}A). The RNA was recovered, converted into cDNA and sequenced on an Illumina platform. In total, we performed two biological replicate experiments. The sequencing reads were processed and clustered as described in Materials and Methods Section. A total of 20.6 million sequencing reads could be mapped to the human genome with at most one mismatch (Supporting Information [Table S1](#SD1){ref-type="supplementary-material"}). Compared with all other mutations in the mappable sequence reads, T to C transitions were significantly enriched (Supporting Information [Fig S1](#SD1){ref-type="supplementary-material"}B), manifesting efficient crosslinking of 4SU labelled RNA (Hafner et al, [@b15]). We identified 240,712 and 218,281 RBM10 sequence clusters (putative binding sites) in the two replicates, respectively (Supporting Information [Table S1](#SD1){ref-type="supplementary-material"}, [Fig S1](#SD1){ref-type="supplementary-material"}C and D for the length distribution of binding clusters as well as the number of PAR-CLIP reads within each cluster). Of these, 87,957 sequence clusters had their preferred crosslinking sites, i.e. the position with the highest number of T to C transitions within a site, to be within the binding site identified in the other replicate. We defined these clusters as consensus binding clusters. Comparison of the binding scores of these consensus binding clusters between the two replicates revealed a high correlation (R^2^ = 0.619) (Supporting Information [Fig S1](#SD1){ref-type="supplementary-material"}E). RBM10 binding in the vicinity of intronic splicing sites -------------------------------------------------------- Ninety-one percent of the consensus binding sites could be assigned to 6396 RBM10 target genes. According to Refseq annotation, 39 and 52% of them fell into exonic and intronic regions, whereas 9% mapped to intergenic regions, which might harbour previously unannotated transcripts ([Fig 1](#fig01){ref-type="fig"}A). Given the possible involvement of RBM10 in splicing process, we examined the distribution of binding sites relative to splice sites. Intriguingly, we found that they were significantly enriched in exons and in the vicinity of both 5′ and 3′ splice sites of the introns ([Fig 1](#fig01){ref-type="fig"}B). Notably, the binding sites were more enriched at the vicinity of (∼70 nt upstream) of 3′ splice site than at 5′ splice site. Interestingly, we also observed the specific binding of RBM10 at U2 snRNA, which is known to pair with 3′ branch site (Supporting Information). Together, these binding patterns were consistent with the previous findings of RBM10 in pre-spliceosomal A and B complex (Agafonov et al, [@b1]; Behzadnia et al, [@b4]; Bessonov et al, [@b5]) and indicated that RBM10 very likely involves in splice site recognition and/or pairing, as well as further intron removal processes via coordinated interactions with snRNPs and the pre-mRNA substrates. ![RBM10 RNA binding sites identified by PAR-CLIP\ The genomic distribution of 87,957 RBM10 consensus binding clusters.Distribution of RBM10 consensus binding clusters (red line), PAR-CLIP reads (green line) and PAR-CLIP reads containing at least one T to C change (converted reads, blue line) along exon--intron and intron--exon boundaries. The density of RBM10 converted PAR-CLIP reads in the vicinity of intronic splice sites were significantly higher than that of Ago2 converted PAR-CLIP reads (grey line).](emmm0005-1431-f1){#fig01} Alternative splicing regulated by RBM10 --------------------------------------- The RNA binding patterns presented above suggested that RBM10 might function as a splicing regulator. To test this possibility, we performed RNA-seq and quantified changes in gene expression as well as AS in HEK293 cells upon RBM10 knockdown (KD) or overexpression (OE) respectively (Materials and Methods Section). The efficiency of KD and OE were assessed both at mRNA level by qPCR and at protein level by Western blot (Supporting Information [Fig S2](#SD1){ref-type="supplementary-material"}A and B). In total, we performed two and four biological replicate experiments for OE and KD, respectively. Sixty-one to 185 millions 100 nt sequencing reads were generated for each sample, of which 92--96% could be mapped to the genome reference (UCSC genome browser hg19) or a reference set of exon--exon junction sequences (see Materials and Methods Section and Supporting Information [Table S2](#SD1){ref-type="supplementary-material"}). The gene expression level was estimated based on RPKM value (reads per kilobase of exon per million mapped sequence reads, (Mortazavi et al, [@b30]), Materials and Methods Section). At false discovery rate (fdr) \<0.05, 171 and 105 genes were found to be significantly upregulated and downregulated by at least 1.5-fold upon RBM10 KD (Supporting Information [Fig S2](#SD1){ref-type="supplementary-material"}C and [Table S3](#SD1){ref-type="supplementary-material"}), whereas 19 and 49 genes were upregulated and downregulated to the same level (fdr \< 0.05, fold change ≥1.5) in response to RBM10 OE, respectively (Supporting Information [Fig S2](#SD1){ref-type="supplementary-material"}D and [Table S3](#SD1){ref-type="supplementary-material"}). Overall, the expression changes induced by KD and OE were not inversely correlated (Supporting Information [Fig S2](#SD1){ref-type="supplementary-material"}E). We then sought to characterize the splicing changes induced by RBM10 OE/KD. Based on RNA-seq data, we defined the inclusion ratio (PSI: percentage splicing in) of each exon in Refseq transcripts as the number of reads exclusively supporting inclusion divided by total number of reads supporting inclusion and exclusion of the specific exon (Supporting Information [Fig S2](#SD1){ref-type="supplementary-material"}F; Polymenidou et al, [@b36]; Wang et al, [@b42]). We then compared the inclusion ratio between KD and control, OE and control, respectively. The changes were transformed into Z-value (Supporting Information [Fig S2](#SD1){ref-type="supplementary-material"}H) and the results from replicate experiments were combined to evaluate statistical significance using the rank product method (Materials and Methods Section). At a stringent cutoff (fdr \< 0.05, \\|ΔPSI\\| ≥ 10%), we identified 256 induced cassette exon inclusion and 48 exclusion events upon RBM10 KD ([Fig 2](#fig02){ref-type="fig"}A and Supporting Information [Table S4](#SD1){ref-type="supplementary-material"}). In comparison, 27 exon inclusion and 217 exclusion events were observed upon RBM10 OE ([Fig 2](#fig02){ref-type="fig"}A and Supporting Information [Table S4](#SD1){ref-type="supplementary-material"}). ![Exon splicing changes induced by RBM10 perturbation and their association with RBM10 binding profile\ Exon splicing changes (ΔPSI, percentage splicing in) induced by RBM10 OE (X axis) were plotted against those induced by RBM10 KD (Y axis). A total of 412 cassette exons were found to be differentially spliced after RBM10 OE or KD (FDR ≤ 5%, \\|ΔPSI\\| ≥ 0.1).The splicing changes (left panel, ΔPSO, percentage splicing out; right panel, ΔPSI) of 21 exons (orange dots in A) quantified by RNA-seq (X axis) are highly correlated with those measure by qPCR (Y axis).Representative examples of RBM10 dependent exon skipping events. The density of RNA-seq reads obtained in control, RBM10 OE and KD experiments together with RBM10 binding sites were shown for gene UBN1 and PUF60. Reads supporting the inclusion or exclusion of the cassette exon, or spanning exon--intron junctions were shown in blue, orange or green, respectively. The remaining reads were shown in grey. The size of the triangle marking RBM10 binding sites reflected the number of PAR-CLIP reads.The density of RBM10 binding clusters close to 5′ and 3′ splicing sites of introns flanking all cassette exons (black) is significantly lower than those flanking the exons more excluded after RBM10 OE (orange, Z~PSI~ ≤−2 and green, Z~PSI~ ≤−1).Cumulative distribution functions of splicing changes upon RBM10 OE for different groups of cassette exons with RBM10 binding close to none or one of the four splicing sites (upper panel), or to different number of the four splicing sites (lower panel). The numbers of exons within different groups were printed in parenthesis.](emmm0005-1431-f2){#fig02} We then selected 21 candidate transcripts for which we had detected splicing changes with different Z values for validation by qPCR using junction specific primers ([Fig 2](#fig02){ref-type="fig"}A and Supporting Information [Fig S3](#SD1){ref-type="supplementary-material"}). The abundance of transcript isoforms including or excluding the cassette exons was normalized based on that of constitutive exons. We could validate splicing changes in all the 21 cases (Supporting Information [Fig S3](#SD1){ref-type="supplementary-material"}). The splicing changes detected by qPCR were quantitatively correlated with that determined by RNA-seq ([Fig 2](#fig02){ref-type="fig"}B). Comparison of splicing changes induced by OE and that by KD revealed a clear inverse correlation ([Fig 2](#fig02){ref-type="fig"}A). The majority (74%) of the splicing changes observed upon RBM10 OE and KD are RBM10-enhanced exon exclusion events ([Fig 2](#fig02){ref-type="fig"}A), indicating that RBM10 primarily mediated the skipping of cassette exons. An RNA splicing map integrating RBM10 binding profile and induced splicing changes ---------------------------------------------------------------------------------- We then took a close look at functional annotations of the genes that changed the expression level and/or the splicing pattern as a result of RBM10 perturbation. Interestingly, 14 RNA-binding proteins and five known splicing regulators were found with significant expression changes (Supporting Information [Table S3](#SD1){ref-type="supplementary-material"}), and even more genes with splicing changes (22 and eight) were found to be involved in RNA-binding or splicing regulation (Supporting Information [Table S4](#SD1){ref-type="supplementary-material"}). Therefore, the overall splicing changes described above represented not only the direct RBM10-targeted splice events, but also the secondary effects resulting from the expression and/or splicing changes in those splicing regulators. In order to understand the mechanisms for AS directly under RBM10 regulation, we correlated the RBM10 RNA binding pattern with the splicing changes upon RBM10 OE and KD. Among the RBM10-enhanced exon skipping events, we often observed RBM10 binding(s) close to 5′ and/or 3′ splicing sites at upstream and/or downstream introns. Two representative examples were depicted in [Fig 2](#fig02){ref-type="fig"}C. To examine general mechanism for RBM10 enhanced exon skipping, we integrated the PAR-CLIP data and splicing profiles into an RNA splicing map. As shown in [Fig 2](#fig02){ref-type="fig"}D, the map revealed increased density of RBM10 binding clusters close to the splice sites of both introns flanking the skipped cassette exons, with the most prominent enrichment at 3′ splice site of downstream intron. The RNA splicing map of RBM10 suggested that RBM10 binding close to the splice sites of neighbouring introns are enriched for skipped exons. To assess whether such binding pattern could predict exon exclusion events, we searched our PAR-CLIP data for non-constitutive exons with RBM10 binding close (i.e. up to 150 nt) to the splice sites of adjacent introns. In total, 5262 such exons were found. Among these exons, 147 showed significant splicing changes (fdr \< 0.05, \\|ΔPSI\\| ≥ 10%) upon RBM10 OE and/or KD, accounting for 30.8% (147/412) of all the exons with significant splicing changes upon RBM10 perturbation. As shown in [Fig 2](#fig02){ref-type="fig"}E, the exons with RBM10 binding close to one of the four splice sites were more likely excluded upon RBM10 OE, and those with binding close to 3′ splice sites of upstream introns exhibiting the weakest skipping propensity. Intriguingly, exons with binding close to more of the four splice sites showed progressively stronger skipping tendency upon RBM10 OE ([Fig 2](#fig02){ref-type="fig"}E). Similarly, exons with binding close to the same four splicing junctions showed progressively stronger inclusion tendency upon RBM10 KD (Supporting Information [Fig S4](#SD1){ref-type="supplementary-material"}). Mechanistic study of RBM10 enhanced exon skipping using minigenes ----------------------------------------------------------------- We demonstrated that RBM10 binding near splice sites of flanking introns would enhance the skipping of cassette exons. In order to test the direct effect of the RBM10 binding on pre-mRNA splicing, we fused RBM10 with a modified pumilio domain, PUF3-2, which specifically recognizes an eight nucleotide sequence 'UGUAUGUA' with high affinity ([Fig 3](#fig03){ref-type="fig"}A; Wang et al, [@b44]). By co-transfecting the RBM10-PUF fusion protein with the splicing reporter pZW2C-A6G that contains PUF cognate sequence, we could tether RBM10 to an intronic region 18-nt downstream of the middle exon ([Fig 3](#fig03){ref-type="fig"}A; Wang et al, [@b44], [@b45]). Whereas the expression of PUF domain alone hardly induced any splicing changes, the expression of RBM10-PUF fusion protein exhibited strong exon skipping effects ([Fig 3](#fig03){ref-type="fig"}B and Supporting Information [Fig S5](#SD1){ref-type="supplementary-material"}), providing unequivocal support to our hypothesis that RBM10 intronic binding in the vicinity of splicing sites would facilitate the skipping of cassette exons. ![Validation of RBM10-enhanced exon exclusion with minigene experiments\ Schematic representation of the minigene experiment, in which RBM10 was tethered to an intronic region 18-nt downstream of the middle cassette exon via fusing to the PUF domain that recognizes its cognate sequence (blue rectangle).Splicing changes of the cassette exons upon the OE of RBM10-PUF fusion protein or PUF alone, detected in three biological replicates. RBM10 tethered in the vicinity of intronic splice site could significantly promote exon skipping (two-tailed paired t-test p-value: 0.001832).](emmm0005-1431-f3){#fig03} Mechanistic model underlying RBM10 regulated alternative splicing ----------------------------------------------------------------- RBM10 binding close to splice sites might interfere with splicing sites recognition and/or splice sites pairing. Our observations strongly supported the model that RBM10 binding in the vicinity of splicing sites might repress the splicing of introns and delay the splicing choice; thereby facilitate the skipping of cassette exons flanked with relatively weaker splicing sites ([Fig 4](#fig04){ref-type="fig"}). Indeed, we found several lines of evidences supporting such a working model. First, a clear positive correlation between exon skipping and the retention of flanking introns could be observed based on RNA-seq results ([Fig 5](#fig05){ref-type="fig"}A). Second, the intron retention appeared to be also enhanced by RBM10 binding in the vicinity of its splice sites, especially the binding near 5′ splice sites ([Fig 5](#fig05){ref-type="fig"}B). Third, we proposed that the exon skipping was largely due to the weaker flanking splice sites. Once the splicing of flanking intron was repressed, the use of stronger distal splicing sites would be enhanced. As shown in [Fig 5](#fig05){ref-type="fig"}C, the strength of splicing sites distal to the cassette exons was generally stronger than that of those immediately flanking the exons. Finally, RBM10 binding close to downstream 3′ splice sites might also facilitate their paring with upstream 5′ splicing sites at later steps of splicing process, an effect proposed previously by Bonnal et al for RBM5 (Bonnal et al, [@b8]). Although it would be difficult to formally disentangle such effect from its repression of intron splicing, the observation that RBM10 binding at downstream 3′ splicing sites exerts in general stronger effects than that at the immediately flanking ones ([Fig 2](#fig02){ref-type="fig"}E) implicated a possible dual function of RBM10 binding at 3′ splicing sites, i.e. repression of intron splicing and facilitation of the splicing site pairing. ![Mechanistic model for RBM10-mediated splicing regulation\ RBM10 binding in the vicinity of splicing sites might repress the splicing of flanking introns and delay the splicing choice, thereby facilitate the skipping of cassette exons flanked with relatively weaker splice sites.](emmm0005-1431-f4){#fig04} ![Correlation between RBM10 binding, exon splicing changes, intron splicing changes and splicing site strengths\ Cumulative distribution functions (CDF) of splicing change of introns (Z~PIR~, percentage intron retention) upstream (left) or downstream (right) to the cassette exons that were differentially spliced upon RBM10 OE. Exon exclusion upon RBM10 OE (orange: Z~PSI~ ≤−2; green: Z~PSI~ ≤−1, PSI: percentage splicing in) is associated with higher retention of both flanking introns than background (black) and exon inclusion (blue: Z~PSI~ ≥1). The numbers of exons were indicated in parenthesis.Cumulative distribution functions of Z~PIR~ for different groups of introns with or without RBM10 binding at the 5′ and/or 3′ splice sites. The numbers of introns within different groups were printed in parenthesis.Box plots of the strengths of splicing splice sites at upstream and downstream introns flanking all cassette exons (white) or those with higher exclusion after upon RBM10 OE (orange: Z~PSI~ ≤−2; green: Z~PSI~ ≤−1).](emmm0005-1431-f5){#fig05} Functional investigation of an in-frame deletion of RBM10 identified in a patient with TARP syndrome -------------------------------------------------------------------------------------------------------- Nonsense and frame-shift mutations in RBM10 have been identified to be causative for TARP syndrome (Johnston et al, [@b21]). More recently, during a screen of \>400 index patients from families with X-linked intellectual disability (ID; Kalscheuer et al., manuscript in preparation), a deletion of 1292 nt (ChrX: 46929367--46930658 bp, UCSC genome browser hg18) in RBM10 were found in a German family ([Fig 6](#fig06){ref-type="fig"}A and Supporting Information [Fig S6](#SD1){ref-type="supplementary-material"}A). Apart from severe ID, the two patients also suffered from multiple congenital malformations and their leading pathologic phenotypes overlapped with TARP syndrome (see Supporting Information and [Table S7](#SD1){ref-type="supplementary-material"} for detailed reports of clinical findings). Based on the annotated gene structure, the deletion spanned six exons, but appeared not to disrupt the open reading frame. We performed RT-PCR and Western blot on the lymphoblast cells deriving from the patient, and demonstrated that the in-frame deletion did not induce nonsense mediated decay, which would otherwise degrade the truncated RNA and reduce protein levels (Supporting Information [Fig S6](#SD1){ref-type="supplementary-material"}B and C). The deletion removed 239 amino acids (651--889 amino acid at NP_005667), including the second zinc finger domain and a portion of the G patch domain ([Fig 6](#fig06){ref-type="fig"}B). ![In-frame deletion of RBM10 identified in a family afflicted with multi-organ malformation\ Pedigree of the family.RBM10 protein domain structure, the deletion removed the second zinc finger domain, a portion of the G patch domain and a nuclear localization signal (NLS).Splicing difference (ΔPSI) between lymphoblastoid cell lines (LCLs) derived from the patient and those from healthy controls were correlated with splicing changes induced by RBM10 KD in HEK293.Subcellular localization of wild type and mutant RBM10 detected by antibody against RBM10 (a-RBM10) or that against HA tag. Co-staining with DAPI showed that in contrast to the nuclear localization of wild type, RBM10 mutant predominantly localized in the cytoplasm. Scale bar: 10 µm.Splicing changes of the two cassette exons in HEK293 cells upon overexpression (OE) of RBM10 wild type or mutant were measured by qPCR. The exclusion and inclusion levels were normalized based on constitutive exon expression.](emmm0005-1431-f6){#fig06} To understand how this in-frame deletion in RBM10 could contribute to disease phenotype, we examined in more detail the sequence of the deleted fragment and identified a potential nuclear localization signal (NLS) within the deleted sequence (821--837 aa; [Fig 6](#fig06){ref-type="fig"}B and see also Supporting Information Supplementary Methods). To determine whether this finding was functionally relevant, we compared the subcellular localization of wild type RBM10 with that of the mutant. In contrast to the nuclear localization of wild type, RBM10 mutant predominantly localized in the cytoplasm ([Fig 6](#fig06){ref-type="fig"}D), consistent with the loss of NLS. This finding suggests that the deletion might result in loss of nuclear functions of RBM10. To investigate the impact of this deletion on gene expression especially splicing pattern, we performed RNA-seq on lymphoblastoid cell lines (LCLs), derived from the patient carrying the mutation and from healthy controls, respectively. We determined the changes in gene expression and splicing pattern using the same strategy applied on HEK 293 cells (Supporting Information [Table S3](#SD1){ref-type="supplementary-material"} and S4). In total, we identified 206 and 102 exons showing enhanced inclusion and exclusion (\\|Z~PSI~\\| ≥ 3, \\|ΔPSI\\| ≥ 10%) in the patient derived LCLs, respectively. Intriguingly, the splicing changes observed here correlated well with changes induced by RBM10 KD in HEK293 ([Fig 6](#fig06){ref-type="fig"}C). This observation indicated that RBM10 mutant lost its function in splicing regulation, in accordance with the observed change in subcellular localization. To further validate the functional impact of the mutant, we overexpressed RBM10 in HEK 293 cells. As shown in [Fig 6](#fig06){ref-type="fig"}E, for the two cassette exons, the OE of the mutant could not induce the same splicing changes as that of the wild type. DISCUSSION ========== RBM10 has been characterized in vitro as an RNA-binding protein and identified as a component of spliceosome complex. However, its putative role in splicing regulation has not been established. In this study, our transcriptome-wide analysis of RBM10 binding profile as well as changes in splicing pattern induced by RBM10 perturbation provided experimental evidence supporting its role as a novel splicing regulator. Using PAR-CLIP, we identified thousands of consensus binding sites, 51% of which are located in the introns, with a significant enrichment in the vicinity of splicing sites. Using RNA-seq, we identified hundreds of exons, the splicing pattern of which was significant changed upon increasing or decreasing cellular RBM10 abundance. An RNA splicing map that associated RBM10-binding profiles with those observed splicing changes yielded a mechanistic model underlying RBM10 mediated splicing regulation. Nonsense and frame-shift mutations in RBM10 have been identified to cause TARP syndrome (Johnston et al, [@b21]). In this study, in the two male cousins with congenital multi-organ malformation, we identified an in-frame deletion in RBM10, which removed the NLS of the protein and thereby largely disrupted its nuclear function. Whole-mount in situ expression analysis of the murine Rbm10 has shown that the gene was expressed during embryonic development in a pattern consistent with the human malformations observed in TARP syndrome (Johnston et al, [@b21]). Therefore, due to the loss of function of RBM10, our patients as well as the previously reported TARP patients would suffer from molecular defects in those tissues expressing critical amount of RBM10 during development. Given that RBM10 regulates many genes and possibly different sets of target genes in different tissues, it is likely that multiple targets would contribute to the phenotype. Indeed, among the genes with splicing pattern regulated by RBM10, some have been implicated in the TARP syndrome associated anomalies (Supporting Information [Table S4](#SD1){ref-type="supplementary-material"}). For example, DNML1 has found to be mutated in the patients with microcephaly and optic atrophy, overlapping features of TARP syndrome (Waterham et al, [@b46]). Mutations in CEP290 could cause Joubert syndrome, a heterogenous ciliopathy characterized by cerebellar vermis hypoplasia and severe ID, two clinical findings also typical for TARP syndrome (Sayer et al, [@b38]). Another interesting gene, CASK, once mutated could lead to brain anomalies similar to the patients reported here (Najm et al, [@b32]). Finally, great phenotypic overlap is also noted to individuals with mutations in the PIGN gene, which lead to a syndromic entity characterized by hypotonia, seizures, neonatal hypotonia, lack of psychomotor development and dysmorphic features, associated with cardiac, urinary and gastrointestinal malformations (Maydan et al, [@b28]). Among the differentially expressed genes upon RBM10 perturbation, there are also candidate genes causing entities with overlapping features. For example, ECEL1 is a gene responsible for distal arthrogryposis type 5D with similar limb anomalies observed in our patients (McMillin et al, [@b29]). Mutation in LHB gene could causes hypogonadism, which is also manifested in individuals with TARP syndrome (Weiss et al, [@b47]). On one hand, the leading pathologic phenotypes observed in our patients largely overlapped with TARP syndrome (see Supporting Information for the discussion of clinical findings), indicating many, if not most of the TARP associated malformations resulted from loss of RBM10 nuclear function, i.e. regulation of exon skipping. On the other hand, comparing with typical TARP patients, our patients are relatively milder affected. Indeed they are the eldest patients reported so far. Given that only nonsense and frame-shift mutations in RBM10 have been reported in TARP patients, it is tempting to speculate that the mutant RBM10 in our patients might retain either some residue nuclear function or other unknown functions of the protein. Notably, a large number of RBM10-RNA interactions, especially those in exons, appeared not to be directly associated with splicing regulation. Whether such interactions hold other regulatory roles awaits further investigation. More recently, in a large sequence analysis of lung adenocarcinomas, RBM10 was found to be frequently mutated and subject to recurrent nonsense, frame-shift or splice-site mutations (Imielinski et al, [@b20]). Interestingly, among the genes differentially spliced upon RBM10 perturbation in HEK293 cells, several were known factors associated with cancers (Supporting Information [Table S4](#SD1){ref-type="supplementary-material"}). Given the frequent observation of splicing deregulation in different types of cancers, it is plausible that RBM10 might also play an important role in cancers other than lung adenocarcinomas. Taken together, our study established RBM10 as an important regulator of AS, yielded a mechanistic model for RBM10-mediated splicing regulation and provided a starting point for the future functional characterization of RBM10 in different biological systems. The paper explained ------------------- PROBLEM: TARP syndrome is an X-linked inherited disorder leading to multiple organ malformation in affected males. Nonsense and frame shift mutations in RBM10, a gene encoding an RBPs, have been identified to cause TARP syndrome. Although the protein has been reported to associate with spliceosome complex, the exact molecular function of RBM10 is not clear. RESULTS: We combined photoactivatable-ribonucleoside-enhanced crosslinking and immunoprecipitation (PAR-CLIP) with massive parallel sequencing to identify RNA binding sites for RBM10 and observed significant RBM10-RNA interactions in the vicinity of splice sites. Using RNA-seq, we identified hundreds of splicing changes following perturbation of cellular RBM10 abundance. Integrative analyses of binding sites as well as splicing profile suggested a mechanistic model underlying RBM10-mediated splicing regulation, which could be subsequently validated by minigene experiments. Furthermore, we demonstrated the splicing defects in a TARP patient carrying an in-frame deletion in RBM10, which could be explained by disrupted function of RBM10 in splicing regulation. IMPACT: Our study for the first time established RBM10 as an important regulator of AS, presented a mechanistic model for RBM10-mediated splicing regulation and provided a molecular link to understanding a human congenital disorder. MATERIALS AND METHODS ===================== Cell lines ---------- Stable HEK 293 T-REx Flp-In cell lines inducibly expressing FLAG/HA-tagged wild type and mutant RBM10, respectively were generated and maintained as described previously (Landthaler et al, [@b23]) with minor modifications (Supporting Information Supplementary Methods). Expression of FLAG-HA-tagged RBM10 was induced with 10 ng/ml doxycycline for 16 h. PAR-CLIP -------- The cells were labelled with 100 µM 4-thiouridine (4SU) and induced with 10 ng/ml doxycycline for 16 h. PAR-CLIP was performed as described previously (Hafner et al, [@b15]) with the following modifications. Cells were lysed in high salt lysis buffer (50 mM Tris--HCl pH 7.2, 500 mM NaCl, 1% NP40, 1 mM DTT, complete protease inhibitor (Roche)). For the second RNase T1 digestion, 10 U/µl RNase T1 and 5 min incubation was used. PAR-CLIP libraries were sequenced 1 × 50 cycles using Illumina HiSeq following the standard protocol. Detailed procedures see Supporting Information Supplementary Methods. RBM10 knockdown --------------- siRNA (Applied Biosystems, s15747) against RBM10 was reverse transfected at a final concentration 20 nM with lipofectamine RNAiMAX (Invitrogen) in HEK293 T-REx Flp-In cells. Controls were treated with only transfection reagents. Cells were harvested 48 h after transfection, respectively. Total RNA was extracted using Trizol (Invitrogen) and the quality was assessed by Agilent Bioanalyser according to the manufacturer\'s instructions. The KD efficiency was assessed by qPCR and Western blot. RBM10 overexpression -------------------- Stable HEK293 T-REx Flp-In cells inducibly expressing FLAG-HA-tagged RBM10 was induced with 10 ng/ml doxycycline for 16 h. Control was treated with equal amount of medium. Total RNA was extracted using Trizol (Invitrogen) and the quality was assessed by Agilent Bioanalyser according to the manufacturer\'s instructions. The OE efficiency was assessed by qPCR and Western blot. mRNA sequencing --------------- mRNA sequencing was performed using 1 µg total RNA. Briefly, poly (A) RNA was isolated by two rounds of oligo (dT)~25~ Dynabeads (Invitrogen) purification. Purified poly (A) RNA was fragmented at 94°C for 3.5 min using 5× fragmentation buffer (200 mM Tris--acetate, pH 8.1, 500 mM KOAc, 150 mM MgOA). The fragmented RNA was precipitated and converted to first strand cDNA using random hexmer primer and Supescript II (Invitrogen), followed by second strand cDNA synthesis with Eschericcia coli DNA pol I (Invitrogen) and RNAse H (Invitrogen). Then the paired-end sequencing library was prepared and sequenced on Illumina HiSeq for 2 × 100 cycles following the standard protocol. qRT-PCR ------- Total RNA was treated with TURBO DNase (Ambion) following the manufacturer\'s protocol. Reverse transcription was performed using 1 µg of DNase treated total RNA, random hexamer and Superscript III reverse transcriptase (Invitrogen) according to manufacturer\'s protocol. First stranded cDNA was diluted 1:20 and 2 µl was used as template in a 20 µl qPCR reaction system. qPCR was carried out using SYBRGreen Masrermix I (Roche) on LightCycler 480 (Roche) according to manufacturer\'s instructions. All assays were performed in triplicates. For expression quantification, the average fold change was calculated by normalization to GAPDH. For exon inclusion or exclusion quantification, the relative ratios were calculated by normalization to corresponding constitutive exons. The sequences of all PCR primers were listed in Supporting Information [Table S5](#SD1){ref-type="supplementary-material"}. Western blot ------------ Western blot was performed as described in Supporting Information Supplementary Methods. The following antibodies were used: rabbit polyclonal anti-RBM10 (Abcam, ab26046, 1:2000), mouse monoclonal anti-HA (Covance, MMS-101P, 1:4000), mouse monoclonal anti-FLAG (Sigma, F1804, 1:4000) and rabbit polyclonal anti-GAPDH (Santa Cruz, sc-25778, 1:2000). Secondary HRP-conjugated goat anti-mouse or human IgG (Santa Cruz) was detected with SuperSignal Kit (Thermo). Minigene experiments -------------------- To validate the direct effects of RBM10 intronic binding on exon skipping, RBM10 protein was fused upstream of a modified PUF domain (PUF3-2) with specific and high affinity RNA recognition sequence (UGUAUGUA, i.e. A6G) as previously reported (Wang et al, [@b44]). The plasmid expressing the RBM10-PUF fusion protein was generated by overlapping PCR as previously described (Heckman & Pease, [@b16]) and then inserted into pcDNA3.1D/V5-His-TOPO expression vector (Invitrogen). Five hundred nanograms of splicing reporter (pZW2C-A6G) containing the PUF3-2 recognition sequence (Wang et al, [@b44]) was transfected alone, or cotransfected with 100 ng of RBM10-PUF or PUF expression vector, respectively into HEK293T cells in 12-well plate. After RT-PCR, the expression level of the two isoforms including or excluding the cassette exons was measured by Bioanalyser DNA 1000 chip (Agilent). All the PCR primers were listed in Supporting Information [Table S5](#SD1){ref-type="supplementary-material"}. PAR-CLIP sequencing data analysis --------------------------------- The PAR-CLIP reads were processed as described before (Lebedeva et al, [@b24]). In brief, the reads were aligned to the human genome (UCSC Genome Browser, hg19) allowing at most one mismatch, or indel of one nucleotide. Uniquely mapped reads were overlapped to define binding clusters. For each cluster, the preferred crosslinking position was defined as the site with the highest number of T to C conversions. Based on the binding clusters identified in the two biological replicates, a consensus binding cluster was defined as a pair of clusters from the two replicates, if preferred crosslinking site of one cluster from one replicate was located within the other cluster from the other replicate and vice versa. The preferred crosslinking site in the first replicate was used as that of the consensus cluster. RNA-seq data analysis --------------------- The RNA-seq reads were mapped with at most two mismatches to the human genome reference (UCSC genome browser hg19) and a set of sequences consisting of all possible junctions between the exons of each Refseq gene. The expression level of a gene was calculated as RPKM values by dividing the number of reads which could be mapped to the exons or exon--exon junctions of this gene by its cumulative exon length (in kb) and the total number of mappable reads (in million). For each of the internal exon E, we computed percent splicing in value, PSI = e~in~/(e~in~ + e~out~), where e~in~ denotes the number of reads which could only be mapped to E or exon junctions containing E with an overlap of at least 6 bp, and e~out~ denotes the number of reads which could be mapped only to exon junctions skipping E and overlap with both exons by at least 6 bp. The internal exons with its PSI value between 0.02 and 0.98 in at least one data set were defined as cassette exons. For each intron I, we computed percent intron retention value, PIR = i~in~/(i~in~ + i~out~), where i~in~ denotes the number of reads which could only be mapped to junctions between I and adjacent exons, and overlap with I by at least 6 bp, and i~out~ denotes the number of reads which could be mapped only to exon--exon junctions skipping I and overlap with both exons by at least 6 bp. To estimate the significance of the change in gene expression level, PSI or PIR upon RBM10 KD and OE, we applied a Z-value transformation, i.e. divided Δlog~2~ RPKM/ΔPSI/ΔPIR by a local standard deviation which we computed using a sliding window approach as following. After sorting the exons according to the total number of reads used for computing the RPKM/PSI/PIR values (e.g. in the case of PSI, it is the sum of e~in~ and e~out~ from both OE/KD and control), we calculated for each data point the standard deviation of the respective values inside a window consisting 1% genes/exons/introns. The local standard deviations were then smoothed using loess regression before we used them for calculating Z values. We then estimated false discovery rates using the rank product method (Breitling et al, [@b9]). For each independent replicate, the genes/exons/introns were ranked according to the respective Z values and the ranks obtained in the replicates were multiplied for each gene/exon/intron. The number of genes/exons/introns expected to have a given rank product by chance was estimated using random permutations of the rank lists. The enrichment and annotation of functional categories in the set of genes with significant splicing changes or differentially expressed upon RBM10 perturbation was computed using the Database for Annotation, Visualization and Integrated Discovery (Huang et al, [@b18], [@b19]). This work was supported by the Federal Ministry for Education and Research (BMBF) and the Senate of Berlin by funds aimed at establishing the Berlin Institute of Medical Systems Biology (BIMSB) (315362A, 315362C). YW was funded by China Scholarship Council (CSC). HH and VMK were funded by project GENCODYS (241995) in the European Union Framework Programme 7. We thank Mirjam Feldkamp, Claudia Langnick from BIMSB, Melanie Bienek and Susanne Freier from MPIMG for their excellent technical assistance. We thank Gabriele Gillessen-Kaesbach from University of Luebeck for providing clinical information of the patient. We thank Zefeng Wang from UNC for help with minigene constructs. Supporting Information is available at EMBO Molecular Medicine online. The authors declare that they have no conflict of interest. Author contributions ==================== YH and WC conceived and supervised the project. YW performed the experiments. AGD analysed the data. HH and VK performed mutation screening. SF, JM, MJ performed PAR-CLIP data analysis. YM, CQ and YaW helped with minigene experiments. YaM and ML help with PAR-CLIP experiments. DW provided the clinical expertise. YW, AGD, DW, YH contributed to part of the manuscript. WC wrote the paper. For more information ==================== The raw sequencing data has been deposited in NCBI GEO database (Accession number: GSE44976). The PAR-CLIP data could also be accessed at doRiNA database (Anders et al, [@b2]): <http://dorina.mdc-berlin.de/rbp_browser/dorina.html> Supplementary material ====================== As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. [^1]: These authors contributed equally to this work.	{ "pile_set_name": "PubMed Central" }
The authors confirm that all data underlying the findings are fully available without restriction. All relevant data are within the paper and its Supporting Information files. Introduction {#s1} ============ The thymus plays a crucial role in the development and maturation of T lymphocytes. Inside the organ, thymocytes (T cells in the thymus) are provided with a wealthy microenvironment consisting of cytokines, chemokines and cell-cell interactions that ultimately leads to proliferation, T-cell receptor gene rearrangements and thymocyte differentiation into mature T cells [@pone.0110739-Gameiro1]. However, the thymus undergoes weight and cellular loss caused by ageing and environmental disturbances, such as infections and malnutrition [@pone.0110739-Savino1]. We have previously shown that infection with Plasmodium berghei NK65 infection, the non-cerebral malaria pathological agent, renders the thymus atrophic through the enhanced thymocyte death by apoptosis and premature egress of CD4^+^CD8^+^ (Double-positive, DP) T cells to the periphery [@pone.0110739-Andrade1]--[@pone.0110739-Gameiro2]. It is already known that some viral and bacterial infections can promote the development of autoimmunity by inducing the breakdown of T cell tolerance and development of effector T cells reactive with the self-antigens or by the phenomenon called molecular mimicry, where a foreign antigen shares sequence or structural similarities with self-antigens [@pone.0110739-Gilden1], [@pone.0110739-Herrmann1]. For instance, acute rheumatic fever, where antibodies attack the heart, can occur after the body makes immune responses against Group A β-hemolytic streptococci [@pone.0110739-Kawakita1], [@pone.0110739-Read1]. In addition, it has been proposed that the prematurely egressed DP-T cells observed during Trypanosoma cruzi infection play an important role in the autoimmune cardio-inflammation [@pone.0110739-MendesdaCruz1]. Experimental Autoimmune Encephalomyelitis is a T cell-driven inflammation of the Central Nervous System (CNS) that presents similar characteristics to human Multiple Sclerosis [@pone.0110739-Mix1]. In this model, following an inflammatory stimulus containing neuro-peptides T cells migrate from the peripheral immune system towards the CNS where they promote inflammation through the release of inflammatory mediators such as cytokines and chemokines [@pone.0110739-Dittel1], [@pone.0110739-Fletcher1]. Cells from the Th1 and Th17 subsets are important for disease establishment, as evidenced by previous reports [@pone.0110739-Chu1]--[@pone.0110739-Peron1]. Both in the human and animal diseases T cells play a major role. Therefore, changes in the subpopulations of T cells influence the outcome and susceptibility to autoimmune development. In this context, we aimed to evaluate whether the previous infection with Plasmodium berghei NK65 would interfere with the clinical course of Experimental Autoimmune Encephalomyelitis, a mouse model for human Multiple Sclerosis (MS). We observed that EAE-susceptible mice cured from malaria developed an aggravated form of EAE, with increased infiltration of DP-T cells in the Central Nervous System (CNS). Further analyses showed that thymic-prematurely egressed DP-T cells were important for the enhanced clinical manifestation of the disease. To our knowledge, this is the first study to demonstrate the possible integration between malaria and EAE through the contribution of the thymus. Materials and Methods {#s2} ===================== Animals {#s2a} ------- Six- to eight-week-old female C57BL/6 mice from the Multidisciplinary Center for Biological Research, University of Campinas, were used in this study. Mice were kept in specific-pathogen free conditions, in a controlled temperature and photoperiod environment, with free access to autoclaved food and water throughout the experiment. All protocols involving laboratory animals were approved and performed in accordance with the guidelines of the Institutional Committee on the Use and Care of Animals (CEUA, \#2687--1). Infection and treatment {#s2b} ----------------------- For these experiments, we used the NK65 strain of Plasmodium berghei, because this parasite do not accumulate in the CNS of mice to cause cerebral malaria. Mice (n = 6 mice/group) were intra-peritoneally injected with 10^6^ Plasmodium berghei-infected red blood cells (iRBCs) obtained from a source mouse. The frequency of iRBCs was assessed daily by examination with Giemsa-stained thin blood smear. Ten days after infection, animals of each group started the treatment with chloroquine (5 mg.kg^−1^, via i.p. for five consecutive days). Three days after the last dose of the drug, mice were immunized for the induction of EAE. EAE induction and evaluation {#s2c} ---------------------------- EAE was induced and evaluated in mice according to a previous published paper [@pone.0110739-Thome1]. Briefly, each mouse was subcutaneously injected with 100 µg MOG~35--55~ (MEVGWYRSPFSRVVHLYRNGK, Genemed Syn, USA) emulsified with Complete Freunds Adjuvant (CFA, Sigma-Aldrich, USA). 200 ηg Pertussis toxin (Ptx, Sigma-Aldrich, USA) was administrated via i.p. at 0 and 48 h after MOG~35--55~ inoculation. Clinical signs were followed and graded daily according to a score method, where 0: no sign, 1: flaccid tail, 2: hind limbs weakness, 3: hind limbs paralysis, 4: hind paralysis and fore limbs weakness, 5: full paralysis/dead. Disease severity was evaluated for thirty days. At the end of the experiment, results were analyzed by linear regression, thinner lines indicate 95% confidence interval. Determination of the thymic index {#s2d} --------------------------------- The calculations to determine the thymic index were previously described [@pone.0110739-Andrade1]. Briefly, at the indicated time-points, the gross weight of each mouse was recorded and then the thymuses were collected and weighed. The thymic index was calculated using the formula: organ weight (g)/body weight (g) ×100. Cell sorting and flow cytometry {#s2e} ------------------------------- Spleen cells were obtained from P.berghei-infected mice at ten days of infection. Total splenic T lymphocytes were isolated using Dynabeads following the manufactureŕs instructions (Mouse Pan T cell isolation kit, Life Technologies, USA). For DP-T cell isolation, spleen-derived single cell suspension was incubated with FITC-conjugated anti-CD4 (clone H129.19) and PE-conjugated anti-CD8a (clone 53--6.7) for 20 minutes. CD4^+^, CD8^+^ and CD4^+^CD8^+^ (double-positive) T lymphocytes from malaria-bearing mice at ten days of infection were sorted in FACSAria II cell sorter (BD Biosciences, USA). The purity of isolations was assessed and accounted for 98% pure subpopulations. For adoptive transfer experiments, 1,5×10^5^ cells were adoptively transferred (via i.v.) to each group of mice at the onset of EAE (around the tenth day after immunization). For activation experiments, sorted cells were seeded in U-bottom 96 well/plates (2×10^4^ cells/well) and incubated for 72h with lipopolysaccharide (1 ng/mL, from E. coli, Sigma-Aldrich, USA), Plasmodium berghei extracts (PbX, 50 µg/mL) or MOG~35--55~ peptide (20 µg/mL, Genscript, USA). At the end of culture period, the supernatants were collected and assayed for detection of mouse IFN-γ and IL-17 by Cytometric Bead Assay (CBA, BD Biosciences, USA). In some set of experiments, total T cells isolated from naïve and malaria-bearing mice were transferred (1x10^6^ cells/mouse) at the onset of EAE. Preparation of Plasmodium berghei extracts {#s2f} ------------------------------------------ The production of extracts from P. berghei-infected RBCs (iRBC) followed a previously published recommendation [@pone.0110739-Thome2]. Briefly, iRBC-enriched suspension was submitted to 20 cycles of freeze/thawing in liquid nitrogen and warm bath (37°C). The protein concentration was determined using the Bradford Protein Assay following the manufactureŕs instructions (Sigma-Aldrich, St Louis, MO, USA). Analysis of cellular infiltration in the CNS {#s2g} -------------------------------------------- Fourteen days after EAE induction, mice were anesthetized, perfused with ice cold PBS and half of the spinal cords were removed and stored at −80°C until use for RT-PCR assays; the remaining tissue was prepared for the enrichment of infiltrating leukocytes according to a previously described methodology and analyzed by flow cytometry [@pone.0110739-McCall1]. RT-PCR assays {#s2h} ------------- Mice were killed at the indicated time-points and frozen tissues were used for RNA extraction (Trizol reagent, Life Technologies, USA) and cDNA synthesis according to the manufactureŕs recommendations (High Capacity RNA-to-cDNA converter kit, Life Techcnologies, USA). RNA extraction from DP-T cells was carried our immediately after sorting, using an extraction kit according to the manufactureŕs instruction (RNeasy Micro kit, Qiagen USA). Expression of AIRE (Mm00477461_m1), IL-7 (Mm01295803_m1), IL-6 (Mm00446190_m1), IL-17 (Mm00439618_m1), IFNg (Mm01168134_m1), FOXP3 (Mm00475162_m1) and RAR-related orphan receptor C (RORc) (Mm01261022_m1) were analyzed in comparison to GAPDH (Mm99999915_g1, housekeeping gene) levels. RT-PCR reactions were performed using Taqman reagents according to manufactureŕs recommendations (Applied Biosystems, USA). Expression levels of genes were represented as a relative copy numbers by using the method of delta threshold (2^−ΔΔCt^). Histopathology and immunofluorescence {#s2i} ------------------------------------- At the indicated time-points mice were killed, spinal cords were removed and snap frozen; 12 µm thin slices were made in cryostat and stained with haematoxylin and eosin (H&E) for histopathological analysis. For the characterization of the infiltrating cells, the reactive sites of the slices were blocked with Phosphate-Buffer (0,1M, pH7,2)-BSA 3%. The cells were stained with FITC-conjugated rat anti-mouse CD4 (clone H129.19, at a 1:100 dilution, BD Biosciences, USA) and PE-conjugated rat anti-mouse CD8a (clone 53--6.7, at a 1:100 dilution, BD Biosciences, USA). In some set of experiments, the lumbar spinal cords were incubated with purified anti-mouse GFAP, COX-2, IL-6, NF-κB, phosphorylated iKB and iNOS. Later, secondary Cy3- or TRITC-conjugated antibodies were added. The reaction was analyzed under epifluorescence microscope (Leica, GER). In vitro re-stimulation and cytokine dosage {#s2j} --------------------------------------------- Splenic cells were aseptically collected from mice after 10 days of MOG~35--55~ immunization. Single cell suspensions were stained with Carboxyfluorescein succinimidyl ester (CFSE, 2,5 µM, Sigma-Aldrich, USA) following the manufactureŕs instructions. Cells (5×10^5^/well) were diluted in RPMI 1640 media supplemented with Fetal Calf Serum (FCS; 10% vol/vol), guaramicine (50 µg/mL), 2-Mercaptoethanol (2 mM) and myelin oligodendrocyte glycoprotein peptide (MOG~35--55~; 20 µg/mL), plated in U-bottom plates and incubated for 96 h. After the incubation period, cells were stained with PercPCy5-conjugated anti-CD3e, PE-conjugated anti-CD8a and PECy7-conjugated anti-CD4 antibodies and fixed in 1% paraformaldehyde prior to flow cytometer analysis. CFSE^low^ cells inside each population were considered proliferating T cells. Culture supernatants were collected and assayed for cytokines (IL-4, IL-6, IL-10, IL-17, IFN-γ and TNF-α) secretion using the Cytometric Bead Array (CBA, BD Biosciences, USA) according to manufactureŕs instructions. T cell co-cultures {#s2k} ------------------ For the co-cultures, total spleen T cells were isolated from malaria-bearing and EAE-inflicted mice after ten days of immunization. T cells derived from EAE mice were stained with CFSE according to the manufactureŕs instructions (2,5 µM, Sigma-Aldrich, USA) and seeded in U-bottom 96-well plates (5×10^5^ cells/well). T cells derived from naïve or P.berghei-infected mice were added afterwards at a 1∶1 proportion. MOG~35--55~ peptide was added to a final concentration of 20 µg/mL. The cultures were incubated for 96 h at 37°C. At the end of culture time, the cells were collected and stained with PercP-conjugated anti-CD3e antibodies and fixed in 1% paraformaldehyde prior to flow cytometer analysis. CFSE^low^CD3^+^ cells were considered proliferating responder T cells. Statistical analysis {#s2l} -------------------- Clinical score comparisons between control and experimental groups were done by Two-Way ANOVA and post-tested with Bonferroni. Other analyses among two and three (or more) groups were carried out with Students t test and One-Way ANOVA, respectively. Results are expressed as mean ± standard error mean (SEM) and p\<0,05 value were defined as significant. Results {#s3} ======= Exacerbation of Experimental Autoimmune Encephalomyelitis in mice cured from P.berghei infection {#s3a} -------------------------------------------------------------------------------------------------- The present study aimed to evaluate whether the previous infection with P.berghei NK65 would interfere with the clinical outcome of Experimental Autoimmune Encephalomyelitis. For that purpose, C57BL/6 mice were infected with parasitized erythrocytes and at the tenth day of infection, parasites were eliminated following chloroquine (CQ) treatment. Three days after the last dose of CQ, mice were immunized with MOG~35--55~ peptide for the induction of EAE (NK+CQ+EAE group). As a control, besides EAE-bearing mice (EAE group), we used naïve mice injected with non-parasitized erythrocytes and treated with chloroquine before EAE induction (CQ+EAE group). The results showed that mice from the NK+CQ+EAE group developed an aggravated disease course with higher clinical score when compared to EAE group ([Figure 1A](#pone-0110739-g001){ref-type="fig"}). Also noteworthy is the mild clinical scores developed by mice from CQ+EAE group. As recently published by our group, CQ reduces the severity of EAE probably by induction of regulatory T cells [@pone.0110739-Thome1]. Thus, our results also show that P.berghei infection was able to overcome the suppressive effect of chloroquine. ![Aggravation of EAE in mice cured from malaria correlates with increased cellular immune response towards myelin.\ C57BL/6 mice (n = 6 mice/group) were intraperitoneally (i.p.) infected with 1×10^6^ P.berghei-infected Red Blood Cells and treated with chloroquine (CQ, 5 mg/Kg) for five consecutive days starting at the 10^th^ day after infection. Three days after the last dose of CQ, mice were immunized with 100 µg of MOG~35--55~ peptide and Pertussis toxin was administrated (via i.p.) at 0 and 48h after peptide immunization for EAE induction. A) The clinical course of EAE was then monitored. Linear regression analyses are exposed in the side panels, thinner lines indicate 95% confidence interval. B) At the 10^th^ day after MOG-immunization, the spleens of mice were collected and dissociated. Total leukocytes (5×10^5^/well) were CFSE-stained (2,5 µM) and cultured in the presence of MOG~35--55~ (10µg/mL) peptide for 96h. At the end of culture period, the cells were surface stained with anti-CD3/CD4/CD8 antibody cocktail and events were acquired in a flow cytometer. The proliferation was analyzed inside each T cell population. C) The culture supernatants were assayed for the secreted cytokines IL-10, IL-4, IL-6, IL-17, TNF-α and IFN-γ. Data was analyzed by One-Way Anova and post-tested with Bonferroni. In all analyses, \: p\<0,05; ns: not significant. Representative data of three independent experiments.](pone.0110739.g001){#pone-0110739-g001} The aggravation of EAE correlates with DP-T lymphocytes reactivity to neuro-peptides {#s3b} ------------------------------------------------------------------------------------ Since EAE model is highly dependent on the cellular response against neuro-antigens, the cellular immune response against MOG peptide was evaluated. Spleen cells from EAE-bearing mice were harvested ten days after neuro-peptide immunization and cultivated in the presence of MOG~35--55~. The data obtained showed that NK+CQ+EAE-derived cells proliferated significantly more when compared to cells from EAE control group ([Figure 1B](#pone-0110739-g001){ref-type="fig"}). Among the proliferating cells, the supopulation of DP-T cells had the most active proliferation, while CQ+EAE mice presented no DP-T cells in the periphery. Cells from CQ+EAE group did not present significant proliferation or the presence of DP-T cells in the spleens ([Figure 1B](#pone-0110739-g001){ref-type="fig"}). The cytokine secretion was altered as well. Cell culture supernatants from NK+CQ+EAE mice presented significant higher levels of pro-inflammatory cytokines (TNF-α, IL-17 and IFN-γ) compared with the EAE group, whereas the levels of IL-10, IL-6 and IL-4 remained similar between the two groups ([Figure 1C](#pone-0110739-g001){ref-type="fig"}). Interestingly, cell cultures from CQ+EAE mice showed significantly higher levels of IL-10 and IL-4, and lower levels of IL-17, IL-6 and TNF-α than the other two groups ([Figure 1C](#pone-0110739-g001){ref-type="fig"}). Plasmodium berghei NK65 infection promotes thymic alterations and the premature egress of CD4^+^CD8^+^ double-positive lymphocytes to the periphery {#s3c} --------------------------------------------------------------------------------------------------------------------------------------------------- The results presented so far show that the exacerbated clinical score from NK+CQ+EAE mice correlates with an enhanced cellular immune response towards neuro-antigens orchestrated, partly, by DP-T cells. The presence of DP-T cells in the spleens of NK+CQ+EAE mice is interesting. It was previously shown that during P.berghei* infection, the thymus of BALB/c mice undergoes structural and phenotypical changes that together with the uncontrolled parasitemia culminates in death of the host fourteen days after infection [@pone.0110739-Andrade1]. To investigate whether the infection with P.berghei would promote similar alterations in the spleens and thymuses of a different mouse strain, C57BL/6 (B6) mice were injected with Plasmodium berghei-infected Red Blood Cells. In this strain, the infection is controlled when chloroquine (5 mg.kg^−1^) is administrated for five consecutive days starting at day 10 after infection ([Figure 2A](#pone-0110739-g002){ref-type="fig"}). Similarly as in the BALB/c model, the thymus of B6 mice is reduced with loss of the thymic index, starting at the third day of infection ([Figure 2B](#pone-0110739-g002){ref-type="fig"}). Interestingly, the expression of the Autoimmune Regulator (AIRE) gene was found up-regulated compared with control mice as well as the expression of IL-7 and IL-6 ([Figure 2C](#pone-0110739-g002){ref-type="fig"}). The gene expression within the thymus of B6 mice infected with other Plasmodium species, such as P.yoelli and P.chabaudi was also evaluated and the results showed that the gene expression of cytokines and AIRE differs between groups ([Figure S1](#pone.0110739.s001){ref-type="supplementary-material"}). ![P. berghei infection provokes thymic alterations and the premature egress of double-positive T lymphocytes in the spleens.\ C57BL/6 mice (n = 6 mice/group) were intraperitoneally (i.p.) infected with 1x10^6^ P. berghei-infected Red Blood Cells and treated with chloroquine (CQ, 5mg/Kg) for five consecutive days starting at the 10^th^ day after infection. A) Mice treated with CQ showed decrease in the parasite burden. B) The thymuses were collected at different time-points and the relative body weight was determined. C) The gene expression of AIRE, IL-7 and IL-6 was assessed in the thymuses of P. berghei NK65-infected mice at the fifteenth day of infection. Flow cytometry analysis of the frequency (in D) and absolute numbers (in E) of T lymphocytes subpopulations (CD4^+^ and CD8^+^) in the thymuses and spleens. Data was analyzed by Student\'s t test. In all analyses, \: p\<0,05 and \\: p\<0,01. Ns: not significant. \#: p\<0,05 in comparison with untreated mice. Representative data of four independent experiments.](pone.0110739.g002){#pone-0110739-g002} It is already clear nowadays that disarranged thymic structure associated with an inflamed microenvironment result in altered maturation of thymocytes [@pone.0110739-Savino2]. Accordingly, the thymocyte subpopulations were altered in the thymus of P.berghei-infected mice, with reduced numbers of CD4^+^CD8^+^ (Double-positive, DP) and increased frequency of CD4^+^ and CD8^+^ thymocytes ([Figure 2D](#pone-0110739-g002){ref-type="fig"}). In the spleen, results showed an increased frequency of DP-T lymphocytes as well as CD4^+^ and CD8^+^ T cells. Naïve mice treated with chloroquine showed no alterations regarding the frequency of DP-T cells in the spleen and in thymic T cell subpopulations in comparison to mice without treatment (data not shown). Thus, it is possible to suppose that the responding DP-T cells observed in the NK+CQ+EAE mice was elicited during P.berghei* NK65 infection and that these cells are related to the exacerbated EAE outcome. Infiltration of inflammatory cells in the central nervous system is characterized by the presence of DP-T cells and the local production of inflammatory cytokines {#s3d} ------------------------------------------------------------------------------------------------------------------------------------------------------------------ To investigate the possible association of DP-T cells with the exacerbation of EAE, the frequency of these cells in the Central Nervous System was evaluated. Fourteen days after EAE induction, mice were killed and spinal cords were prepared for histological analyses. The phenotype of the infiltrating T cells was determined by direct immunofluorescence labeling technique using FITC-conjugated anti-CD4 and PE-conjugated anti-CD8 antibodies. Results showed that in EAE group, among the infiltrating cells, a high frequency of CD4^+^ T cells was observed, while no infiltration was observed in the CNS of CQ+EAE mice ([Figure 3](#pone-0110739-g003){ref-type="fig"}). As expected, an elevated frequency of CD4^+^CD8^+^ (DP) T cells among the infiltrating cells in the CNS of NK+CQ+EAE mice was identified ([Figure 3](#pone-0110739-g003){ref-type="fig"}). These data indicate that the double-positive T cells that prematurely migrated from the thymus during P.berghei infection are antigen-responsive ([Figure 1B](#pone-0110739-g001){ref-type="fig"}) and capable to migrate to the CNS during neuro-inflammation ([Figure 3](#pone-0110739-g003){ref-type="fig"}). In addition, the CNS of NK+CQ+EAE mice showed an increased production of IL-6, NF-κB, phosphorylated iκB (iκBα-p), iNOS, COX-2 and GFAP compared with EAE group ([Figure S2](#pone.0110739.s002){ref-type="supplementary-material"}). ![Central Nervous System of malaria-cured EAE mice show increased cellular infiltration of DP-T cells.\ C57BL/6 mice (n = 6 mice/group) were intraperitoneally (i.p.) infected with 1×10^6^ P.berghei-infected Red Blood Cells and treated with chloroquine (CQ, 5 mg/Kg) for five consecutive days starting at the 10^th^ day after infection. Three days after the last dose of CQ, EAE was induced. As controls, naïve mice were treated with CQ or vehicle before EAE induction. The spinal cords of EAE-inflicted mice were collected fourteen days after MOG-immunization. Frozen thin sections (12 µm) were made and fixed in formalin. Cells were stained with FITC-conjugated anti-CD4 and PE-conjugated anti-CD8 and analyzed in epifluorescence microscope. Figures are representative of three independent experiments. Magnification: 200X.](pone.0110739.g003){#pone-0110739-g003} In the CNS, an increased gene expression of IL-17 in tissue from NK+CQ+EAE mice was observed in comparison with mice from EAE group ([Figure 4A](#pone-0110739-g004){ref-type="fig"}). In contrast, the expression of Foxp3 and IL-10 was not significantly altered in comparison with the naïve group ([Figure 4A](#pone-0110739-g004){ref-type="fig"}). IFN-γ expression was not significantly altered. To investigate whether DP-T cells found in the CNS tissue were able to produce inflammatory mediators, mice were killed at the fourteenth day of EAE and the infiltrating cells were enriched from the CNS. The intracellular staining results showed that NK+CQ+EAE mice had higher frequency of IFN-γ- and IL-17-producing DP-T cells than EAE mice. ([Figure 4B](#pone-0110739-g004){ref-type="fig"}). ![Inflammation in the CNS of NK+CQ+EAE mice correlates with an increased production of inflammatory cytokines by DP-T cells.\ Groups of mice (n = 6 mice/group) subjected to infection and EAE induction. A) At the 10^th^ day after MOG-immunization, mice were killed and spinal cords were removed to analyze the gene expression of IL-17, IFN-γ, Foxp3 and IL-10 in the lumbar spinal cords of mice. Data was analyzed by One-Way Anova and post-tested with Bonferroni. B) The infiltrating cells of the CNS were enriched and stimulated by Phorbol Myristate Acetate and Ionomycin in the presence of Brefeldin A for 4 h. The frequency of IFN-γ- and IL-17-producing cells inside CD4^+^CD8^+^ T cell gate was analyzed. In all analyses, \: p\<0,05. ns: not significant. Representative data of three independent experiments.](pone.0110739.g004){#pone-0110739-g004} Transfer of lymphocytes from P.berghei-infected mice increases the severity of EAE {#s3e} ------------------------------------------------------------------------------------ In order to better characterize the relationship among malaria-elicited DP-T lymphocytes and the exacerbation of EAE, a series of experiments to demonstrate that the mechanism of disease aggravation is T-cell dependent was conducted. First, splenic T cells isolated from P.berghei-infected or naïve mice were co-cultured with splenic T cells from EAE-bearing mice in the presence of MOG~35--55~ peptide for four days. Results showed that the cultures conducted in the presence of T cells from malaria-infected mice proliferated significantly more than cells from co-cultures with naïve T cells ([Figure 5A](#pone-0110739-g005){ref-type="fig"}). ![Exacerbation of EAE in mice transferred with total T from naïve mouse or DP-T cells from malaria-cured mouse.\ A) C57BL/6 mice (n = 6) were intraperitoneally (i.p.) infected with 1×10^6^ P.berghei-infected Red Blood Cells. At the 10^th^ day of infection, splenocytes were collected and total T cells were isolated using dynabeads (Pan T cell isolation kit). Naïve-derived splenic T cells were used as well. EAE-inflicted mice (n = 6) were killed at the 10^th^ day after immunization and the total splenic T cell were isolated with the same methodology and CFSE-stained (2,5 µM). As controls, T cells from EAE-inflicted mice were cultured without the presence of other cells. B) Total T cells isolated from naïve and malaria-bearing mice were transferred (1×10^6^ cells/mouse) at the onset of EAE (n = 6 mice/group), and clinical course of the disease was evaluated daily. Data was analyzed by Two-Way ANOVA and post-tested with Bonferroni. C) Sorted splenic CD4^+^, CD8^+^ and CD4^+^CD8^+^ T cells from malaria-bearing mice were adoptively transferred (1,5×10^5^ cells/mouse) to EAE-inflicted mice (n = 6 mice/group) at disease onset. The clinical score of the disease was evaluated daily. The side panel contains the linear regression lines with the 95% confidence interval (thinner lines). D) Malaria-bearing mice were killed at 10 days after infection and the DP-T cells were sorted and incubated with LPS (1 ng/mL), P. berghei* extracts (50 µg/mL) and MOG~35-55~ peptide (20 µg/mL) for 72 h, at the end of culture period the supernatants were removed and assayed for the detection of IFN-γ and IL-17. Analyses were conducted using Student\'s t test. \: p\<0,05. \\: p\<0,01. Representative data of two independent experiments with similar results.](pone.0110739.g005){#pone-0110739-g005} Then, it was investigated whether the adoptive transfer of splenic T cells from malaria mice would alter the course of EAE. For that purpose, B6 mice were infected with Plasmodium berghei* NK65 and the total splenic T cells were isolated ten days after infection. These cells were adoptively transferred to EAE mice at the beginning of the clinical signs (around the 10^th^ day after immunization). As expected, the clinical course of EAE was significantly aggravated in malaria T cells-transferred mice compared to naïve T cell-recipient mice ([Figure 5B](#pone-0110739-g005){ref-type="fig"}). Lastly, since the infiltrating T cells in the CNS of NK+CQ+EAE mice comprised mostly of DP-T cells, the next goal was to evaluate whether these cells play a role in the aggravation of EAE. Therefore, sorted splenic CD4^+^, CD8^+^ and DP-T cells from malaria-infected mice were intravenously transferred to EAE mice as soon as the clinical signs of EAE started to appear (around the 10^th^ day after MOG~35--55~ immunization). The results showed that mice that received DP-T cells developed a more severe disease than mice that received CD4^+^ and CD8^+^ T cells ([Figure 5C](#pone-0110739-g005){ref-type="fig"}). In addition, we aimed to evaluate whether the cellular response elicited by DP-T cells were antigen-specific or were raised in a by-stander fashion. For that purpose, we sorted DP-T cells from the spleens of naïve and P. berghei-infected mice and cultivated the cells in the presence of LPS, P. berghei extracts and MOG peptide. At the end of the culture time, the supernatants were assayed for the presence of inflammatory cytokines. As depicted in [figure 5D](#pone-0110739-g005){ref-type="fig"}, cultures conducted in the presence of LPS and PbX produced elevated levels of IFN-γ and IL-17 in comparison with DP-T cells from naïve mice. Interestingly, the presence of MOG was able to stimulate the production of IL-17, but not IFN-γ, in DP-T cell cultures. This data shows that the prematurely egressed DP-T cells from P. berghei-infected mice are able to produce inflammatory cytokines in the presence of PAMPs (LPS), specific stimulus (PbX) and auto-antigens (MOG). Discussion {#s4} ========== In this study, we show that, after infection with Plasmodium berghei NK 65, MOG-immunized mice develop an aggravated form of Experimental Autoimmune Encephalomyelitis. Further analysis demonstrated that the thymic prematurely egressed CD4^+^CD8^+^ (double-positive) T cells play an important role in the enhanced severity of the disease. The thymus is a key organ for the development of T lymphocytes and plays a major role in the elimination of auto-reactive T cells [@pone.0110739-Asano1], [@pone.0110739-Starr1]. T cell precursors enter the thymus through the high endothelial venules found in the cortico-medullary junction and lack the expression of antigen receptor and co-receptors CD4 and CD8 [@pone.0110739-Lind1]. The maturation process consists of both migration of thymocytes through the cortex towards the medulla and expression of T cell receptor and co-receptors [@pone.0110739-Anderson1]. During this process, immature T cells interact with a myriad of cytokines, chemokines and self-peptides presented in association with MHC molecules on thymic epithelial cells (TECs) and cells that recognize antigens with high avidity are deleted [@pone.0110739-Savino2]. We have previously shown that the lethal Plasmodium berghei NK65 infection promotes thymic atrophy and that this phenomenon is dependent on thymocyte death and its premature egress to the periphery [@pone.0110739-Andrade1]--[@pone.0110739-Gameiro2]. The lymphocytes subpopulations was altered in the thymus as well, with increased frequency of simple positive CD4^+^ and CD8^+^ T cells. However, due to the loss of thymic mass, the absolute numbers of cells were reduced. In the spleens, the augmentation in the numbers of CD4 and CD8 T cells may be explained by the clonal expansion observed in the course of malaria [@pone.0110739-Falanga1]--[@pone.0110739-Sardinha1]. In addition, we observed an increased expression of AIRE, IL-6 and IL-7 in thymuses from P.berghei-infected mice, which indicate that an altered repertoire selection might be occurring. An impaired T cell maturation and selection in the thymus may lead to the release of self-reactive T cells to the periphery and the sub-sequential development of autoimmune disorders [@pone.0110739-Capalbo1]--[@pone.0110739-Itoh1]. Indeed, we observed that mice, which were cured from P.berghei infection, developed a more severe form of EAE compared with uninfected animals. When we analyzed the inflammatory infiltration in the CNS, we found that most of the infiltrating cells were double-positive T cells. These DP-T cells were functional as they produced high levels of IL-17, an inflammatory cytokine related to EAE worsening [@pone.0110739-Chu1], [@pone.0110739-OConnor1], [@pone.0110739-Willenborg1]. As expected, when we adoptively transferred total T cells or sorted DP-T cells derived from malaria-bearing mice towards EAE-inflicted mice, the clinical score arose significantly. Taken together, these set of experiments demonstrated that the DP-T cells elicited in P.berghei-infected mice are inflammatory and exacerbate autoimmune neuro-inflammation. Co-cultures showed that malaria-elicited T cells stimulated the proliferation of T cells from MOG-immunized mice. Together with the proliferative response observed in DP-T cells from NK+CQ+EAE, these data suggest that DP-T cells, originated in the course of malaria, exacerbate EAE by, at least, two different mechanisms: (i) DP-T cells self-react and proliferate against neuro-antigens and (ii) DP-T cells stimulate the proliferation of encephalitogenic T cells. Of note, when cultured in the presence of different stimuli, DP-T cells in the spleens produced inflammatory cytokines IFN-γ and IL-17 at high levels. These observations indicate that DP-T cells elicited in P. berghei infection respond to Pathogen-Associated Molecular Patterns and specific antigens as well as to unrelated auto-antigens. Interestingly, infection overcame the suppressive effect of chloroquine. We previously observed that CQ treatment reduces the severity of Experimental Autoimmune Encephalomyelitis through the expansion of regulatory T cells [@pone.0110739-Thome1]. Later, our results showed that CQ is able to modulate dendritic cells in vitro directing them towards a tolerogenic profile that can be used to treat EAE [@pone.0110739-Thome3]. The mechanisms by which P. berghei surpasses the CQ suppressive effect remains to be elucidated. Still, the recently described effect of Plasmodium infection in Toll-like receptors (TLR) priming may explain, at least in part, the aggravation of inflammation [@pone.0110739-Hartgers1], [@pone.0110739-McCall1]. The data presented here might be concerning as malaria remains the worlds most prevalent infectious disease, with over 250 million people infected [@pone.0110739-Bhaumik1]. Several reports have correlated autoimmunity prevalence in countries that have eradicated the Plasmodium infection [@pone.0110739-Sotgiu1]. Although the hygiene hypothesis supports the idea of infection-induced autoimmune resistance, this theory may not fully apply to malaria. There is an increasing amount of evidence supporting or rejecting the role of malaria in autoimmune disease both in human and animal models [@pone.0110739-Butcher1]--[@pone.0110739-Adu1]. Similarly, it was previously proposed that in Trypanosoma cruzi infection, the inflammatory profile of prematurely egressed DP-T cells from the thymus may be involved in the autoimmune process observed in murine and human Chagas\' disease [@pone.0110739-MendesdaCruz1]. It was demonstrated that Plasmodium chabaudi infection increases the frequency of regulatory T cells in spleens, and these cells suppressed the development of EAE in a by-stander fashion [@pone.0110739-Farias1]. However, P. berghei infection was shown to induce the secretion of anti-DNA and anti-nuclear antibodies in a T cell-dependent manner [@pone.0110739-Poels1]. These observations indicate that different Plasmodium species induce distinct types of immune responses. Indeed, we observed that the gene expression of inflammatory cytokines and AIRE within the thymus vary depending on the type of Plasmodium species. Nonetheless, infected Red Blood Cells from P.falciparum, P.yoelli, P.berghei and P.vivax induce distinct maturation status of dendritic cells (DCs) [@pone.0110739-Thome2], [@pone.0110739-Bettiol1]--[@pone.0110739-Wykes1]. Indeed, we have shown that DCs modulated with plasmodium extracts are able to reduce the severity of EAE through the suppression of inflammatory responses in the Central Nervous System [@pone.0110739-Thome2]. Interestingly, it was recently observed that Plasmodium chabaudi-infected mice presented exacerbated response to bacterial challenge [@pone.0110739-Ataide1]. In this study, the authors showed that following infection, mice were hyper-responsive to E. coli-derived LPS in a mechanism that is dependent on NLRP12/NLRP3 inflammasome activation. The results presented by the authors are in line with others [@pone.0110739-Hartgers1], [@pone.0110739-McCall1], [@pone.0110739-Franklin1]. In our model, although we did not investigate the role of inflammasomes, we found that the T cell compartment is severely altered in P. berghei-infected mice resulting in an exacerbation of autoimmune responses. Still, further studies must be conducted in order to ascertain the possible interplay between innate cells-derived inflammasome and T cells in the hyper-responsive inflammation elicited by P. berghei infection. Based on our observations, we hypothesized that during malaria infection, thymocytes prematurely egress from the thymus towards the peripheral immune system. In the periphery, due to a trigger event (which can be genetic susceptibility, infection, or radiation, for instance) these cells proliferate and migrate to the target organ, where they produce high amounts of inflammatory cytokines leading to exacerbated inflammation ([Figure 6](#pone-0110739-g006){ref-type="fig"}). ![Hypothesis model for the exacerbation of autoimmune neuro-inflammation.\ Based in our observations, we propose a model for EAE exacerbation. Malaria infection promotes thymic atrophy and the premature egress of DP-T cells to the peripheral immune system. After an inflammatory trigger, which can be infection, genetic susceptibility or chronic inflammation, these cells proliferate and migrate to the target organ where they stimulate CNS inflammation by secreting cytokines. However, there is still much to be explored, as for example, whether these DP-T cells are able to induce leukocyte recruitment, microglia and astrocyte activation, and, Blood-Brain Barrier (BBB) destruction.](pone.0110739.g006){#pone-0110739-g006} Further studies must be conducted in order to evaluate how these DP-T cells are activated and whether they are antigen-specific or not. Although an increasing amount of evidences show the presence of extra-thymic DP-T cells in the control of inflammation [@pone.0110739-Bang1]--[@pone.0110739-Tutaj1], our data clearly shows that in malaria infection, the thymic-derived DP-T cells are inflammatory. To our knowledge, this is the first study that shows the interplay between malaria infection and autoimmune neuro-inflammation through the contribution of the thymus-derived double-positive T cells. Supporting Information {#s5} ====================== ###### Analysis of the gene expression in the thymus of mice infected with distinct *Plasmodium* species. C57BL/6 mice were intraperitoneally (i.p.) infected with 1×10^6^ infected Red Blood Cells. Mice were infected with P.berghei NK65, P.chabaudi AS and P.yoelli CL. The gene expression of AIRE, FOXP3, Rorγt, IL-17, IL-7 and IL-6 was evaluated. Results showed that the different Plasmodium species triggers distinct gene expression. Data was analyzed by One-Way ANOVA and post-tested with Bonferroni, where \: p\<0,05. Representative data of four independent experiments. (TIF) ###### Click here for additional data file. ###### Analysis of the inflammation in the CNS of malaria-cured EAE mice.* C57BL/6 mice were intraperitoneally (i.p.) infected with 1×10^6^ P.berghei-infected Red Blood Cells and treated with chloroquine (CQ, 5 mg/Kg) for five consecutive days starting at the 10^th^ day after infection. Three days after the last dose of CQ, mice were immunized with 100 µg of MOG~35--55~ peptide and Pertussis toxin was administrated (via i.p.) at 0 and 48 h after peptide immunization for EAE induction. The spinal cords of EAE mice were collected fourteen days after immunization. Frozen thin sections (12 µm) were made and fixed in formalin. Cells were stained with purified anti-mouse iNOS (in A), IL-6 (in B), COX-2 (in C), NF-κB (in D), phosphorylated iκBα (in E) and GFAP (in F). DAPI was added to stain DNA (blue color). The slices were analyzed in epifluorescence microscope. Figures are representative of three independent experiments. Magnification: 400X. (TIF) ###### Click here for additional data file. [^1]: Competing Interests:The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: RT FTMC MACH ALRO LV. Performed the experiments: RT ALB LKI CR SCPL TAC RDG ITF ALFL. Analyzed the data: RT MACH FTMC LV. Contributed reagents/materials/analysis tools: LV FTMC MACH ALRO. Wrote the paper: RT LV.	{ "pile_set_name": "PubMed Central" }
All relevant data are within the paper and its Supporting Information files. Introduction {#sec001} ============ A high body mass index is a major risk factor for the incidence of numerous non-communicable diseases (NCD), such as cardiovascular and kidney diseases, diabetes and some cancers \[[@pone.0182333.ref001]--[@pone.0182333.ref006]\]. Indeed, obesity has been recognised as a major public health challenge for the 21^st^ century \[[@pone.0182333.ref007]\], with concerns regarding the health and economic burden, which have led to the identification of global targets to halt the rise in obesity prevalence by 2025 \[[@pone.0182333.ref008], [@pone.0182333.ref009]\]. However, recent figures from the NCD Risk Factor Collaboration, which analysed the aggregated data of 19.2 million participants from 200 countries, suggest that if current, post-2000, trends continue, the probability of meeting these targets is almost zero. These findings highlight the critical need to develop and implement effective interventions for the prevention and treatment of excess body mass \[[@pone.0182333.ref010]\]. A fundamental principle in the development of such interventions is the assessment of the energy expenditure associated with activities of daily living; weight management strategies are most effective when individuals can accurately determine how much energy they have expended \[[@pone.0182333.ref011]\]. Indeed, a failure to assess energy expenditure appropriately may, at least in part, explain the inconsistent evidence regarding intervention effectiveness and sustainability. Walking represents a popular, convenient and relatively safe form of activity that can easily be incorporated into weight management programmes \[[@pone.0182333.ref012]--[@pone.0182333.ref014]\]. The energy expenditure associated with walking are reported to be either linearly or slightly exponentially related to speed \[[@pone.0182333.ref015]\]. However, the applicability of these findings is based on walking in a straight line which does not align with everyday activities. In particular, recent studies have suggested that the process of changing direction is associated with significant additional energy expenditure \[[@pone.0182333.ref016]--[@pone.0182333.ref018]\]. Wilson et al. \[[@pone.0182333.ref018]\] reported that the energy expenditure of turning is linearly related to the degree of turning angle at 6 km∙hr^-1^ while Hatamoto et al. found quadratic \[[@pone.0182333.ref017]\] or linear \[[@pone.0182333.ref016]\] functions best represented the relationship between running speed and the energy expenditure of a 180° turn. A failure to account for this additional energy expenditure of turning has significant implications for both sporting and health contexts. For example, most English Premier league footballers make more than 700 turns per match \[[@pone.0182333.ref019]\] while medical treatment effectiveness is often assessed using a six-minute walking test. While the latter is intended to be conducted over a standardised 30 m straight line distance \[[@pone.0182333.ref020]\], some studies have used 20 m or 50 m straights \[[@pone.0182333.ref021], [@pone.0182333.ref022]\], significantly influencing the number of turns completed and thus potentially confounding inter-study comparisons. Indeed, the difference in the number of turns completed, and thus overall energy expenditure, may explain studies, which utilised shorter straights reporting significantly shorter distances covered \[[@pone.0182333.ref023]--[@pone.0182333.ref025]\]. The purpose of the present study was therefore to investigate the influence of walking speed, angle, and their interaction, on energy expenditure. We hypothesised that 1) as walking speed increased, so would energy expenditure; 2) as angle increased, so would energy expenditure and that 3) angle and speed would demonstrate a synergistic effect on energy expenditure while walking. Materials and methods {#sec002} ===================== Participants {#sec003} ------------ In total, 20 healthy adults (7 female, 13 male; 28 ± 7 yrs; 20.5 ± 4.1 kg∙m^2^) were recruited for the study. The participants were all recreationally active but none were highly trained. Prior to testing, participants were informed of the protocol and risks and provided written consent. All procedures were approved by a Swansea University ethics committee and were conducted in accordance with the Declaration of Helsinki. Participants were asked to arrive at the laboratory in a rested state, at least two hours postprandial and to avoid strenuous exercise in the 24 hours preceding each testing session. Participants were also asked to refrain from caffeine and alcohol for 6 and 24 h before each test, respectively. All tests were performed at the same time of day (± 2 h). Experimental design {#sec004} ------------------- Participants were required to visit the laboratory/indoor track on three occasions, separated by at least 24 hours recovery. Participants initially completed an incremental treadmill test for determination of peak oxygen uptake ($\overset{˙}{V}\text{O}_{2}{}_{\text{peak}}$) and the gas exchange threshold (GET). On each of the two subsequent visits, participants completed the turning protocol. Incremental treadmill test {#sec005} -------------------------- Following a three-minute warm-up at 6 km·hr^-1^, the treadmill speed increased by 1km·hr^-1^ every minute until participants reached volitional exhaustion. The treadmill gradient was set at 1% throughout the test \[[@pone.0182333.ref026]\], until participants reached their maximal running speed at which point it subsequently increased by 1% every minute until volitional exhaustion. Participants were given strong verbal encouragement throughout the test. Turning protocol {#sec006} ---------------- On subsequent visits to the indoor track, each participant was asked to complete repeated three-minute bouts of walking interspersed by three minutes of rest. In a randomised order, each participant walked at four different walking velocities (2.5, 3.5, 4.5, 5.5 km·hr^-1^) in combination with four different angles (0, 45, 90, 180°). Specifically, each of the sixteen conditions involved 5 m of straight walking interspaced by prescribed turns with the speed dictated by a digital, auditory metronome, which sounded once half-way between turns and once on the turns. Each condition incorporated an equal number of left- and right-handed turns, as illustrated in [Fig 1](#pone.0182333.g001){ref-type="fig"}. ![Schematic of the turning protocol showing 5m straights interspersed by prescribed turns of a) 0° b) 45° c) 90° and d) 180° with equal left and right hand turns.](pone.0182333.g001){#pone.0182333.g001} Measurements {#sec007} ------------ Throughout all the tests, gas exchange variables (MetaMax Cortex 3B, CORTEX Biophysik GmbH, Germany) were measured on a breath-by-breath basis and displayed online. Prior to each test, the gas analysers were calibrated using gases of known concentration and the turbine volume transducer was calibrated using a 3-litre syringe (Hans Rudolph, Kansas City, MO). The delay in the capillary gas transit and analyser rise time were accounted for relative to the volume signal, thereby time-aligning the concentration and volume signals. Additionally, two combined tri-axial accelerometers and tri-axial magnetometers (SLAM Tracker, Wildbyte Technologies Ltd, Swansea, UK), measuring at 60 Hz, were worn by participants; one set was worn on the right mid-axilla line at the level of the iliac crest and one set at the middle of the lower back. Data analysis {#sec008} ------------- The peak $\overset{˙}{V}\text{O}_{2}$ was defined as the highest 10 s stationary average during the incremental treadmill test. The GET was determined by the V-slope method \[[@pone.0182333.ref027]\] as the point at which carbon dioxide production began to increase disproportionately to $\overset{˙}{V}\text{O}_{2}$, as identified using purpose written software developed using LabVIEW (National Instruments, Newbury, UK). The mean $\overset{˙}{V}\text{O}_{2}$ during each condition was taken as the first 45 seconds of the final minute of that bout. Subsequent analyses were based on the premise that the energy expenditure of turning was superimposed on the baseline energy expenditure of straight line travel. Thus, the difference in $\overset{˙}{V}\text{O}_{2}$ during straight line walking (0°) at each speed compared to the $\overset{˙}{V}\text{O}_{2}$ engendered during walking with 45, 90 or 180° turns, was attributed to the additional energy expenditure of turning. This $\overset{˙}{V}\text{O}_{2}$ was converted to gross energy expenditure in kJ using a conversion factor of 20.1 J per ml of oxygen and subsequently divided by the total number of turns per condition to provide an estimate of the energy expenditure of each angle and speed combination. The raw accelerometer data were converted to dynamic body acceleration (DBA) by first smoothing each channel to derive the static acceleration using a running mean over 2 s \[[@pone.0182333.ref028]\] and then subtracting this static acceleration from the raw data \[[@pone.0182333.ref029]\]. The resulting values for dynamic acceleration were all then converted to positive values. These values for DBA were then vectorially summed to give: $$VeDBA = \ \sqrt{\left( {A_{x}^{2} + A_{y}^{2} + A_{z}^{2}} \right)}$$ Where VeDBA is the vectorial dynamic body acceleration, A~x~, A~y~, and A~z~ are the derived dynamic accelerations at any point in time corresponding to the three orthogonal axes of the accelerometer \[[@pone.0182333.ref030]\]. Mean and summed VeDBA were derived for each individual turn and straight during the middle minute of the each condition and for the overall three minute bout. Individual turns and straights were determined using custom designed C++ software (DDMT Wildbyte Technologies Ltd) written specifically for the SLAM Tracker devices, to visualise the accelerometry and magnetometry traces and identify the point at which each trace significantly deviated from the local mean. Statistics {#sec009} ---------- Gaussian distributions in data were confirmed by Shapiro-Wilks tests. To account for the repeated and correlated nature of the data, linear mixed-effects models were used to determine the influence of, and interaction between, walking speed and turn angle on energy expenditure and VeDBA ([S1 File](#pone.0182333.s001){ref-type="supplementary-material"}). All condition combinations were placed in one model with covariates (sex, stature, peak $\overset{˙}{V}\text{O}_{2}$ and/or turning $\overset{˙}{V}\text{O}_{2}$ for VeDBA) added to subsequent adjusted models to determine their modulatory effect. Pearson product moment correlation coefficients were used to analyse the degree of association between key variables. Statistical analyses were conducted using PASW Statistics 21 (SPSS, Chicago, IL). All data are presented as means ± SD. Statistical significance was accepted when P≤0.05. Results {#sec010} ======= Descriptive characteristics of the sample population are shown in [Table 1](#pone.0182333.t001){ref-type="table"}. The male participants were significantly taller and demonstrated a higher peak $\overset{˙}{V}\text{O}_{2}$, both in absolute and relative terms (normalised per kg body mass). 10.1371/journal.pone.0182333.t001 ###### Participant characteristics. ![](pone.0182333.t001){#pone.0182333.t001g} Total Male Female -------------------------------------------------------------------------------- ------------- ------------- -------------------------------------------------- n 10 13 7 Age, yrs 28.0 ± 6.7 28.5 ± 7.5 27.2 ± 5.3 Stature, m 1.74 ± 0.09 1.78 ± 0.08 1.67 ± 0.02[\](#t001fn002){ref-type="table-fn"} Body mass, kg* 72.1 ± 16.1 74.5 ± 16.1 67.6 ± 16.4 BMI, kg∙m^2^ 20.6 ± 4.1 20.8 ± 3.9 20.3 ± 4.9 Peak $\overset{˙}{V}\mathbf{O}_{2}$, l∙min^-1^ 3.54 ± 0.75 3.92 ± 0.64 2.85 ± 0.33[\](#t001fn002){ref-type="table-fn"} Relative peak $\overset{˙}{V}\mathbf{O}_{2}$, l∙kg^-1^∙min^-1^ 49.8 ± 7.9 53.2 ± 5.5 43.5 ± 8.1[\](#t001fn002){ref-type="table-fn"} GET, l∙min^-1^ 2.31 ± 0.66 2.52 ± 0.71 1.92 ± 0.34[\](#t001fn002){ref-type="table-fn"} GET, %* 65 ± 9 64 ± 10 67 ± 7 Mean ± SD. BMI, body mass index; $\overset{˙}{V}\text{O}_{2}$, oxygen uptake, GET, gas exchange threshold. \* Significant difference between males and females P \< 0.05 As shown in [Table 2](#pone.0182333.t002){ref-type="table"}, there was a significant main effect for speed (F = 121.609, P \< 0.001) and turn angle (F = 19.186, P \< 0.001) on $\overset{˙}{V}\text{O}_{2}$ and a significant interaction between these parameters (F = 4.433, P \< 0.001). Specifically, as speed increased, $\overset{˙}{V}\text{O}_{2}$ increased, but significant increases in $\overset{˙}{V}\text{O}_{2}$ relative to straight line walking were only observed for 90° and 180° turns at the two highest speeds (4.5 and 5.5 km∙hr^-1^; [Table 2](#pone.0182333.t002){ref-type="table"}). Males demonstrated a significantly greater $\overset{˙}{V}\text{O}_{2}$ across all conditions (F = 25.322, P \< 0.001), although this difference was reversed when stature was included in the model (Sex: F = 16.77, P \< 0.001; Stature: F = 152.493, P \< 0.001). $\overset{˙}{V}\text{O}_{2}$ during the turning protocol was dependent on peak $\overset{˙}{V}\text{O}_{2}$ (F = 100.970, P \< 0.001) but once scaled to account for differences in body size, this relationship was no longer significant (F = 0.708, P \> 0.05). 10.1371/journal.pone.0182333.t002 ###### Mean energy expenditure and VeDBA during each combination of walking speed and angle. ![](pone.0182333.t002){#pone.0182333.t002g} $\overset{˙}{V}O_{2}$ (l∙min^-1^) Mean VeDBA (g) ----------------------- -------------------------------------------------- -------------------------------------------------- ----------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------- -------------------------------------------------- -------------------------------------------------- -------------------------------------------------- -------------------------------------------------- 2.5 km∙hr^-1^ 0.66 ± 0.20 0.63 ± 0.19 0.63 ± 0.20 0.67 ± 0.22 0.19 ± 0.03 0.19 ± 0.02 0.19 ± 0.04 0.20 ± 0.03 3.5 km∙hr^-1^ 0.73 ± 0.20[\](#t002fn002){ref-type="table-fn"} 0.74 ± 0.22[\](#t002fn002){ref-type="table-fn"} 0.74 ± 0.21[\](#t002fn002){ref-type="table-fn"} 0.82 ± 0.24[\](#t002fn002){ref-type="table-fn"} 0.24 ± 0.04[\](#t002fn002){ref-type="table-fn"} 0.25 ± 0.05[\](#t002fn002){ref-type="table-fn"} 0.25 ± 0.04[\](#t002fn002){ref-type="table-fn"} 0.27 ± 0.04[\](#t002fn002){ref-type="table-fn"} 4.5 km∙hr^-1^ 0.86 ± 0.26[\](#t002fn002){ref-type="table-fn"} 0.88 ± 0.26[\](#t002fn002){ref-type="table-fn"} 0.93 ± 0.24[\](#t002fn002){ref-type="table-fn"}[^\#^](#t002fn003){ref-type="table-fn"} 1.10 ± 0.33[\](#t002fn002){ref-type="table-fn"}[^\#^](#t002fn003){ref-type="table-fn"} 0.32 ± 0.03[\](#t002fn002){ref-type="table-fn"} 0.32 ± 0.04[\](#t002fn002){ref-type="table-fn"} 0.32 ± 0.05[\](#t002fn002){ref-type="table-fn"} 0.37 ± 0.05[\](#t002fn002){ref-type="table-fn"} 5.5 km∙hr^-1^ 1.00 ± 0.26[\](#t002fn002){ref-type="table-fn"} 1.00 ± 0.33[\](#t002fn002){ref-type="table-fn"} 1.14 ± 0.30[\](#t002fn002){ref-type="table-fn"}[^\#^](#t002fn003){ref-type="table-fn"} 1.54 ± 0.36[\](#t002fn002){ref-type="table-fn"}[^\#^](#t002fn003){ref-type="table-fn"} 0.44 ± 0.05[\](#t002fn002){ref-type="table-fn"} 0.44 ± 0.05[\](#t002fn002){ref-type="table-fn"} 0.46 ± 0.06[\](#t002fn002){ref-type="table-fn"} 0.53 ± 0.08[\](#t002fn002){ref-type="table-fn"} Mean ± SD. $\overset{˙}{V}\text{O}_{2}$, net oxygen uptake; VeDBA, vectorial dynamic body acceleration. \* indicates significant difference to 2.5 km∙hr^-1^ within angle (P\<0.05) ^\#^ indicates significant difference to straight walking within speed (P\<0.05) The estimated energy expenditure associated with an individual turn is represented in [Fig 2](#pone.0182333.g002){ref-type="fig"}, showing the synergistic interaction between speed and turn angle in determining the energy expenditure. ![Estimated energetic cost of individual turns at each combination of walking speed and turning angle.\ \* indicates a significant difference in the energy expenditure of turning compared to 45°. Values shown are the difference in energy expenditure between each condition and straight walking.](pone.0182333.g002){#pone.0182333.g002} In accord with $\overset{˙}{V}\text{O}_{2}$, there was a significant main effect for speed and angle on mean VeDBA during the straights (speed: F = 1058.734, P \< 0.001; angle: F = 43.416, P \< 0.001) and corners (speed: F = 996.671, P \< 0.001; angle: F = 12.534, P \< 0.001), with a significant interaction effect between speed and angle (straight: F = 4.497, P \< 0.001; corners: F = 3.053, P \< 0.01). Specifically, while both straight and corner VeDBA increased with speed, the effect of angle was only manifest at 4.5 and 5.5 km∙hr^-1^ ([Table 3](#pone.0182333.t003){ref-type="table"}). For straight sections, VeDBA was dependent on sex (F = 15.781, P \< 0.001) and peak $\overset{˙}{V}\text{O}_{2}$ (F = 7.437, P \< 0.01) whilst stature was not a significant contributor (F = 0.038, P \> 0.5). In contrast, corner VeDBA was significantly influenced by sex (F = 6.723, P \< 0.01) and stature (F = 4.507, P \< 0.05), but not peak $\overset{˙}{V}\text{O}_{2}$ (F = 0.243, P \> 0.05). 10.1371/journal.pone.0182333.t003 ###### Mean straight and corner VeDBA during each combination of walking speed and angle. ![](pone.0182333.t003){#pone.0182333.t003g} Straight Mean VeDBA (g) Corner Mean VeDBA (g) ----------------------- -------------------------------------------------- -------------------------------------------------- ----------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------- -------------------------------------------------- -------------------------------------------------- ----------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------- 2.5 km∙hr^-1^ 0.19 ± 0.03 0.18 ± 0.03 0.19 ± 0.04 0.20 ± 0.03 0.18 ± 0.04 0.18 ± 0.03 0.18 ± 0.04 0.18 ± 0.03 3.5 km∙hr^-1^ 0.24 ± 0.04[\](#t003fn002){ref-type="table-fn"} 0.25 ± 0.05[\](#t003fn002){ref-type="table-fn"} 0.25 ± 0.04[\](#t003fn002){ref-type="table-fn"} 0.28 ± 0.05[\](#t003fn002){ref-type="table-fn"} 0.23 ± 0.04[\](#t003fn002){ref-type="table-fn"} 0.25 ± 0.06[\](#t003fn002){ref-type="table-fn"} 0.24 ± 0.05[\](#t003fn002){ref-type="table-fn"} 0.24 ± 0.03[\](#t003fn002){ref-type="table-fn"} 4.5 km∙hr^-1^ 0.32 ± 0.04[\](#t003fn002){ref-type="table-fn"} 0.32 ± 0.04[\](#t003fn002){ref-type="table-fn"} 0.33 ± 0.04[\](#t003fn002){ref-type="table-fn"}[^\#^](#t003fn003){ref-type="table-fn"} 0.39 ± 0.04[\](#t003fn002){ref-type="table-fn"}[^\#^](#t003fn003){ref-type="table-fn"} 0.30 ± 0.05[\](#t003fn002){ref-type="table-fn"} 0.33 ± 0.04[\](#t003fn002){ref-type="table-fn"} 0.32 ± 0.04[\](#t003fn002){ref-type="table-fn"}[^\#^](#t003fn003){ref-type="table-fn"} 0.34 ± 0.05[\](#t003fn002){ref-type="table-fn"}[^\#^](#t003fn003){ref-type="table-fn"} 5.5 km∙hr^-1^ 0.44 ± 0.05[\](#t003fn002){ref-type="table-fn"} 0.44 ± 0.05[\](#t003fn002){ref-type="table-fn"} 0.46 ± 0.06[\](#t003fn002){ref-type="table-fn"}[^\#^](#t003fn003){ref-type="table-fn"} 0.55 ± 0.09\[^\#^](#t003fn003){ref-type="table-fn"} 0.42 ± 0.06[\](#t003fn002){ref-type="table-fn"} 0.44 ± 0.05[\](#t003fn002){ref-type="table-fn"} 0.45 ± 0.06[\](#t003fn002){ref-type="table-fn"}[^\#^](#t003fn003){ref-type="table-fn"} 0.49 ± 0.07[\](#t003fn002){ref-type="table-fn"}[^\#^](#t003fn003){ref-type="table-fn"} Mean ± SD. VeDBA, vectorial dynamic body acceleration. \* indicates significant difference to 2.5 km∙hr^-1^ within angle (P\<0.05) ^\#^ indicates significant difference to straight walking within speed (P\<0.05) Overall $\overset{˙}{V}\text{O}_{2}$ during each condition was significantly correlated with straight (r^2^ = 0.61; P \< 0.001), corner (r^2^ = 0.58; P \< 0.001) and total mean VeDBA (r^2^ = 0.61; P \< 0.001), demonstrating a weaker but statistically significant relationship with peak $\overset{˙}{V}\text{O}_{2}$ (r^2^ = 0.36; P \< 0.001) and stature (r^2^ = 0.32; P \< 0.001). Discussion {#sec011} ========== This is the first study to investigate the interaction between speed and turn angle in determining the energy expenditure associated with walking. In agreement with one hypothesis, as speed increased for any given turning angle, the associated energy expenditure similarly increased. However, whether angle comprised a significant additional energy expenditure was dependent on the degree of turn angle. Specifically, irrespective of speed, 45° turns did not significantly increase energy expenditure, whilst 180° turns were always associated with a greater energy expenditure than straight line walking. Speed and angle demonstrated a significant interaction; 90° turns were only associated with significantly increased energy expenditure relative to straight line walking at 4.5 and 5.5 km∙hr^-1^. This synergistic interaction was further supported by the exponential relationship found to best represent the relationship between speed and angle \[[@pone.0182333.ref015]\]. These findings therefore highlight the importance of accounting for the quantity and magnitude of turn completed when estimating energy expenditure, particularly at higher speed and angles. In recent years there has been increasing recognition of the physiological demands engendered by turning 180° when running. Dellal et al. \[[@pone.0182333.ref031]\] reported a greater heart rate, blood lactate and ratings of perceived exertion (RPE) during intermittent shuttle runs involving 180° turns compared to straight line running at the same average running velocities, subsequently confirmed by Buchheit et al. \[[@pone.0182333.ref032]\]. Furthermore, Bekraoui et al. \[[@pone.0182333.ref033]\] found that covering the same distance at the same average speed resulted in a significantly greater physiological response when the course was 3.5m compared to 7.0m. These earlier findings were recently extended by Hatamoto et al. \[[@pone.0182333.ref017]\] who found that, even at running speeds as low as 3 km∙hr^-1^, thirty 180° turns per minute elicited a similar metabolic demand as straight line running at 6 km∙hr^-1^. In the present study, a significant increase in total energy expenditure relative to straight line walking was not observed at 2.5 km∙hr^-1^, but was observed at 3.5 km∙hr^-1^. Whilst these findings are largely in accord with those of Hatamoto et al. \[[@pone.0182333.ref017]\], it is pertinent to note certain methodological discrepancies, such as the training status of the sample population and turning frequencies utilised, which limit inter-study comparisons. Specifically, there were considerable differences in the number of turns completed, with Hatamoto et al. \[[@pone.0182333.ref017]\] utilising up to 30 turns per minute compared to the 35 turns in 3 minutes at 3.5 km∙hr^-1^ used in the present study. The greater energy expenditure associated with turning whilst walking is likely to be primarily attributable to the deceleration and subsequent acceleration required to make a turn, both of which necessitate eccentric and concentric muscle contractions, respectively \[[@pone.0182333.ref034]\]. Acceleration has been shown to engender a greater energy expenditure than travelling at a constant speed, with the energy expenditure dictated by the rate of acceleration \[[@pone.0182333.ref035]\]. A high acceleration rate requires a high degree of horizontal propulsion \[[@pone.0182333.ref036]\], therefore the change in acceleration is greater when performing a 180° turn at higher running velocities, thereby resulting in greater energy expenditure. The extent of the angle turned has also been shown to alter the biomechanical properties in running; a 90° turn exerts a significantly higher vertical, braking and propelling force than a 45° turn \[[@pone.0182333.ref037]\]. It would therefore be postulated that greater angles would also be associated with further increases in directional forces and thus energy expenditure during walking. In accord with this hypothesis, a linear relationship has been suggested between angle and energy expenditure when walking at 6 km∙hr^-1^ \[[@pone.0182333.ref018]\]. However, the present study suggests a synergistic interaction between speed and angle, with the influence of increasing angle within a speed only evident at 4.5 km∙hr^-1^ and above. This discrepancy may be attributable to differences in the walking velocities, the specific techniques used to turn, stature or training status \[[@pone.0182333.ref032], [@pone.0182333.ref038]\]. Indeed, both stature and peak $\overset{˙}{V}\text{O}_{2}$, an indicator of aerobic fitness and training status, were significant predictors of energy expenditure in the present study. Hatamoto et al. \[[@pone.0182333.ref016]\] previously suggested that ball game players, who are likely to be mainly running rather than walking and who turn more frequently anyway, were likely to have a more efficient turning technique. However, the mean $\overset{˙}{V}\text{O}_{2}$ of an individual turn was reported to be 0.34 ± 0.13 ml∙kg^-1^ and 0.55 ± 0.09 ml∙kg^-1^ at 4.3 km∙hr^-1^ and 5.4 km∙hr^-1^, respectively \[[@pone.0182333.ref016]\]. These values are substantially more than the values observed in the present study ([Fig 2](#pone.0182333.g002){ref-type="fig"}: 4.5 km∙hr^-1^ = 0.07 ± 0.03 ml∙kg^-1^; 5.5 km∙hr^-1^ = 0.13 ± 0.07 ml∙kg^-1^), despite the less trained status of the present participants. The reason for this discrepancy and its contradiction to the postulated role of aerobic fitness and technique are presently unclear, although it is perhaps pertinent to note the different methods of calculating the energy expenditure of an individual turn and the recent findings of Zagatto et al. \[[@pone.0182333.ref039]\] who found a lower metabolic power to be associated with more frequent changes of direction. It is interesting to note the apparent dissociation between VeDBA and turning angle in the present study, whereby increasing the angle of the turn was not associated with any significant increase in VeDBA. This could be attributable to the short duration of the turns, although the high measurement resolution makes this unlikely, or measurement error associated with the use of magnetometry to isolate the turn. However, whilst the magnitude of change in the signal was decreased at lower turn angles, this is unlikely to entirely explain the present findings. Rather, this finding may largely be attributable to the complex and individual-specific interaction between the surge, heave and sway components of DBA as well as muscular effort that involves generation of high forces without the dynamism typical of straight-line travel. Indeed, recent studies using force plates to investigate turn kinetics suggest that during a turn, the surge (inline) component of DBA is accompanied by a sway (perpendicular) component (Griffiths et al., In press). Furthermore, the surge component tends to 'average' zero over the straight sections (equal deceleration and acceleration phases) but during a turn section, the surge component becomes negative on average to provide the deceleration required to enter and execute the turn. In addition, the heave component (vertical) component of DBA may increase above and beyond normative walking values but this may depend on the turning technique being employed, e.g. some participants may elect to turn using a 'stop and reverse direction' method while others may prefer a 'gradual cornering' approach. The authors are of the opinion that this is by far the most likely explanation for the lack of sensitivity of VeDBA to turns. The present findings have significant implications within both sporting and health contexts given that few sporting, fitness or functional activities occur in a strictly linear fashion \[[@pone.0182333.ref037]\]. Indeed, whilst the present study only considered walking and caution should be exercised when extrapolating the findings to speeds associated with running and team sports, it is perhaps pertinent to note the similarity between the current findings and those reported elsewhere. Specifically, Dellal et al. \[[@pone.0182333.ref031]\] reported a greater heart rate, blood lactate and ratings of perceived exertion (RPE) during intermittent shuttle runs involving 180° turns compared to straight line running at the same average running velocities, subsequently confirmed by Buchheit et al. \[[@pone.0182333.ref032]\]. Furthermore, Bekraoui et al. \[[@pone.0182333.ref033]\] found that covering the same distance at the same average speed resulted in a significantly greater physiological response when the course was 3.5m compared to 7.0m. These earlier findings were recently extended by Hatamoto et al. \[[@pone.0182333.ref017]\] who found that, even at running speeds as low as 3 km∙hr^-1^, thirty 180° turns per minute elicited a similar metabolic demand as straight line running at 6 km∙hr^-1^. In the present study, a significant increase in total energy expenditure relative to straight line walking was not observed at 2.5 km∙hr^-1^, but was observed at 3.5 km∙hr^-1^. Whilst these findings are largely in accord with those of Hatamoto et al. \[[@pone.0182333.ref017]\], certain methodological differences should be considered, such as the training status of the sample population and turning frequencies utilised, which limit inter-study comparisons. Specifically, there were considerable differences in the number of turns completed, with Hatamoto et al. \[[@pone.0182333.ref017]\] utilising up to 30 turns per minute compared to the 35 turns in 3 minutes at 3.5 km∙hr^-1^ used in the present study. However, not all studies have found a significant influence of turning on energy expenditure, with Zamparo et al. \[[@pone.0182333.ref040]\] reporting no change in $\overset{˙}{V}\text{O}_{2}$ with increasing turn angle from 0 to 180°. This discrepancy may be related to the use of maximal running velocity during this study, thereby minimising the potential for further increases in $\overset{˙}{V}\text{O}_{2}$ to be elicited with increasing turn angle. Nonetheless, we would concur with Hatamoto et al. \[[@pone.0182333.ref016]\] that the energy expenditure associated with turning should be considered when estimating total energy expenditure during a football game in which more than 700 turns are typically completed per match \[[@pone.0182333.ref019]\]. From a health perspective, one important application of the present findings is in the design and interpretation of physical activity interventions. For example, the majority of energy expenditure prediction algorithms based on accelerometry data are derived from treadmill exercise. Such linear modes of locomotion are not cognisant of the additional metabolic costs associated with turning and this may, to some extent, contribute to the poor accuracy associated with the derived models during free-living conditions \[[@pone.0182333.ref041], [@pone.0182333.ref042]\]. Such inaccuracies are likely to be emphasised in certain populations, such as children, who are characterised by highly sporadic movements \[[@pone.0182333.ref043], [@pone.0182333.ref044]\]. Furthermore, accounting for the energy expenditure of turning could also be important in the evaluation of clinical trial effectiveness. Whilst the six-minute walking test is designed to be conducted over a 30m, straight line course with a 180° turn \[[@pone.0182333.ref020]\], reported distances covered range from 20 to 50 m \[[@pone.0182333.ref021], [@pone.0182333.ref022]\] due to space and resource limitations. Such discrepancies, using reference values reported by Chetta et al. \[[@pone.0182333.ref045]\] could result in the number of turns ranging from 12 to 32, which, according to the present data, may be associated with an additional $\overset{˙}{V}\text{O}_{2}$ expenditure ranging from 118 ml∙min^-1^ to 296 ml∙min^-1^. Swank et al. \[[@pone.0182333.ref046]\] demonstrated that a 6% improvement in peak $\overset{˙}{V}\text{O}_{2}$ was associated with a 5% decrease in risk of all-cause mortality in Congestive Heart Failure patients. Given the significantly lower peak aerobic capacity in patients, discrepancies arisen from failing to account for the energy expenditure of turning, which could be as much as 20% of a patients peak $\overset{˙}{V}\text{O}_{2}$, would considerably alter the interpretation of intervention efficacy. Future studies should seek to generate algorithms that account for distance and turns completed during a six-minute walk test, facilitating standardisation between centres. There are certain limitations associated with the current study that should be acknowledged, such as the walking velocities utilised. Previous studies have employed higher running speeds, whereas we employed speeds more typical of habitual physical activity. Whilst this increased the generalisability of our findings to health contexts, caution should be taken when extrapolating these findings to a sporting context. Furthermore, although a strength of the study to optimise interpretation of our results, the controlled nature of the protocol limits ecological validity. Finally, although the walking speeds were associated with a moderate intensity of exercise for most of the participants, some may not have achieved a steady state $\overset{˙}{V}\text{O}_{2}$ within the 3-minute bout, thereby influencing the mean $\overset{˙}{V}\text{O}_{2}$ observed. In conclusion, the present study demonstrated a synergistic interaction between speed and angle in determining the energy expenditure associated with walking. Specifically, 90° and 180° turns are associated with significant additional metabolic costs at 4.5 km∙hr^-1^ and above. These findings therefore highlight the importance of accounting for the quantity and magnitude of turns completed when estimating energy expenditure and have significant implications within both sport and health contexts. Supporting information {#sec012} ====================== ###### Minimum required dataset. (XLSX) ###### Click here for additional data file. [^1]: Competing Interests:The authors have declared that no competing interests exist.	{ "pile_set_name": "PubMed Central" }
Introduction {#sec1} ============ Selective serotonin reuptake inhibitors (SSRIs), including fluoxetine (Flx), are the most commonly prescribed antidepressants given their improved side effect profile compared with other major classes of antidepressants \[[@B1]\]. However, it is important to note that treatment with SSRIs such as Flx is known to exert potent anxiogenic behavioral effects, in particular during the acute phase of treatment, an observation that is also noted in animal models \[[@B2],[@B3]\]. The acute effects of Flx treatment differ from those noted following chronic administration, namely increased anxiety in the acute phase of treatment and a shift to anxiolytic effects observed on sustained treatment \[[@B3]\]. Furthermore, acute Flx is known to evoke enhanced activity of the stress-responsive, hypothalamus--pituitary--adrenocortical (HPA) axis leading to elevated circulating corticosterone levels, whereas sustained Flx administration is reported to normalize HPA axis dysfunction in animal models of depression and in clinical studies \[[@B4],[@B5]\]. The acute phase of Flx treatment through induction of a 'hyperserotonergic' state \[[@B6]\] may overlap to a certain degree with the nature of state induced by acute stress, which is known to activate the HPA axis, enhance serotonin release, and induce anxiety-like behavior. The contribution of specific serotonergic receptors to the acute anxiogenic effects of Flx treatment remains at present unclear, though prior studies have implicated the 5-HT~2~ receptor subtypes in contributing to the anxiogenic effects of acute SSRI treatment \[[@B7]\]. The G~q/11~-coupled serotonin~2A~ (5-HT~2A~) and serotonin~2C~ (5-HT~2C~)receptor are highly expressed in limbic neurocircuitry, and have been strongly implicated in the regulation of anxiety-like behavior \[[@B9]\]. Studies have raised the possibility that the acute anxiogenic effects of SSRIs, may involve a contribution of the 5-HT~2C~ receptor in the bed nucleus of the stria terminalis (BNST) \[[@B6]\] and basolateral nucleus of the amygdala (BLA) \[[@B8]\], with a down-regulation of the 5-HT~2C~ receptor suggested to mediate a shift toward anxiolysis following sustained SSRI administration \[[@B12]\]. Adjunct administration of the 5-HT~2A~ receptor selective antagonist (R)-(+)-α-(2,3-Dimethoxyphenyl)-1-\[2-(4-fluorophenyl)ethyl\]-4-piperinemethanol (MDL100907) along with SSRIs is reported to improve SSRI efficacy on specific behavioral tasks \[[@B13]\]. However, the contribution of the 5-HT~2A~ receptor to the acute anxiogenic effects of SSRIs still remains unclear. Here, we sought to examine the contribution of the 5-HT~2A~ receptor to the acute effects of the SSRI, Flx, on anxiety-like behavior and circulating corticosterone levels, using 5-HT~2A~ receptor knockout mice (5-HT~2A~^−/−^) \[[@B10]\]. Further, we addressed the degree of neural activation evoked by acute Flx in the paraventricular nucleus (PVN) of the hypothalamus and the prefrontal cortex (PFC) in 5-HT~2A~^−/−^ and WT mice, and gene expression of several activity-regulated, immediate early genes (IEGs) in the PFC. 5-HT~2A~^−/−^ mice have been previously reported to exhibit a baseline anxiolytic phenotype \[[@B10],[@B14]\], higher firing rates of dorsal raphe 5-hydroxytryptamine or serotonin (5-HT) neurones \[[@B15]\] and also to display treatment resistance to chronic Flx administration \[[@B16]\]. Our results reveal that the anxiogenic effects of acute Flx treatment, as well as the enhanced serum corticosterone levels are unaltered in 5-HT~2A~^−/−^ mice of both sexes. However, 5-HT~2A~ receptor deficiency does alter the pattern of neural activation within the PVN following acute Flx, and the baseline expression of several IEGs in the PFC in a sexually dimorphic manner. Experimental procedures {#sec2} ======================= Animals and drug treatment {#sec2-1} -------------------------- Serotonin~2A~ receptor (5-HT~2A~) knockout mice (5-HT~2A~^−/−^) \[[@B10]\] and wild-type (WT) littermate controls of both sexes (4--7 months) were maintained on a 129S6/SvEv background, and group housed on a 12-h normal light--dark cycle, with access to food and water ad libitum in the Tata Institute of Fundamental Research (TIFR) animal house facility. Genotypes were confirmed using PCR analysis as described previously \[[@B14]\]. All experimental procedures followed the guidelines of the Committee for Supervision and Care of Experimental Animals (CPCSEA), Government of India, and were approved by the TIFR Institutional Animal Ethics committee in accordance with the National Institutes of Health guide for the care and use of Laboratory animals (NIH Publication No. 8023, revised 1978). The experimental groups were as follows for both sexes: WT + saline (WT Sal), WT + Flx (WT Flx), 5-HT~2A~^−/−^ + saline (5-HT~2A~^−/−^ Sal) and 5-HT~2A~^−/−^ + Flx (5-HT~2A~^−/−^ Flx). Different cohorts were used for each behavioral (open-field test (OFT): males: n=10/group; females: n=9--10/group; elevated plus-maze test (EPM): males: n=11--15/group; females: n=12--14/group), cellular (FBJ murine osteosarcoma viral oncogene homolog peptide (c-Fos) profiling) (males: n=4--6/group; females: n=5--6/group) and molecular (quantitative real-time PCR (qPCR)) (males: n=8--12/group; females: n=7--10/group) characterization. Serum corticosterone analysis was performed on the same cohort used for cellular analysis (males: n=3--6/group; females: n=5--6/group). Mice were administered a single intraperitoneal injection of Flx (15 mg/kg, Sigma, U.S.A.) or vehicle (saline), and were subjected to behavioral testing or killed for gene expression analysis 2 h later, or serum corticosterone quantitation and immunohistochemistry analysis after 30 min. OFT and EPM {#sec2-2} ----------- Animals were assessed for anxiety-like response on the OFT as described previously \[[@B14]\]. Briefly, mice were placed in one corner of the arena and allowed to explore the arena for 10 min under dim light conditions. Behavioral tracking was done using Panlab SMART video tracking software (SMART 3.0) and percent distance in center, percent time in center, entries to center and total distances traversed were assessed. A separate cohort of mice was assessed for anxiety-like behavioral responses on the EPM consisting of adjacent open and closed arms for a duration of 10 min. The mice were placed at the intersection of the four arms facing the open arm, and allowed to explore the arena under dim light conditions. The movement was tracked using an overhead camera and Ethovision 3.1 (Noldus, The Netherlands) video tracking software. The total distance traversed in the EPM arena, percent distance traversed in the open arms, percent time spent in the open arms, and number of entries to the open arms were analyzed. Serum corticosterone assay {#sec2-3} -------------------------- Blood was collected by cardiac puncture 30 min post a single Flx (15 mg/kg) or saline injection from 5-HT~2A~^−/−^ and WT littermate control mice of both sexes. Serum was collected by transferring the supernatant into a fresh tube and stored at −80°C till further analysis. ELISA was performed to measure serum corticosterone levels using a commercially available ELISA kit (Abcam, Catalog number ab108821) as per the manual (male: n=3--6/group; female: n=5--6/group). c-Fos immunohistochemistry {#sec2-4} -------------------------- Animals were perfused with saline, followed by 4% paraformaldehyde (PFA) solution and brains were dissected out and stored in 4% PFA. Brains were sectioned on a vibratome (Leica VT1000 S, U.S.A.) at 40-μm thickness. Free-floating sections spanning the PVN of the hypothalamus (−0.58 to −1.22 mm from Bregma, five to six sections per animal) and PFC (2.46--1.34 mm from Bregma, five to six sections per animal) were subjected to processing for c-Fos immunohistochemistry. Briefly, sections were blocked in 10% horse serum, followed by overnight incubation with rabbit-anti-c-Fos antibody (1:500, Catalog number CST-2250S, Cell Signaling Technology). Sections were washed and incubated with a biotinylated horse-anti-rabbit IgG secondary antibody (1:500, Catalog number BA-1100, Vector Laboratories). Signal amplification and visualization involved exposure to an avidin--biotin complex solution (ABC solution, Vector Laboratories) followed by 3,3′-Diaminobenzidine (DAB) substrate (Sigma, U.S.A.). Sections were mounted and visualized at 200× magnification (Zeiss Axioskop, Germany). c-Fos positive cells were counted per section by an experimenter blinded to the treatment conditions. qPCR analysis {#sec2-5} ------------- Animals were killed 2 h post a single Flx (15 mg/kg) or saline injection by decapitation; the PFC was dissected out and snap-frozen in liquid nitrogen. RNA was extracted using Tri reagent (Sigma) and reverse transcribed using a cDNA synthesis kit (PrimeScript First Strand cDNA Synthesis Kit, Takara Bio). qPCR was performed with primers for the genes of interest (Supplementary Table S1) using a Bio-Rad CFX96 real-time PCR machine. Data were analyzed using the ΔΔC~t~ method, as described previously \[[@B17]\]. Normalization was done using hypoxanthine guanine phosphoribosyl transferase (Hprt), whose level was unaltered across experimental groups. Gene expression of IEGs analyzed included activity regulated cytoskeleton associated protein (Arc), brain-derived neurotrophic factor (Bdnf), early growth response gene 1-4 (Egr 1-4), FBJ osteosarcoma gene (Fos), FBJ murine osteosarcoma viral oncogene homolog B (Fosb), Fos-like antigen 2 (Fosl2), Homer scaffolding protein 1-3 (Homer 1-3), Jun proto-oncogene (Jun), Jun B proto-oncogene (Junb), Jun D proto-oncogene (Jund). qPCR analysis was also performed for the 5-HT~2A~ gene (Htr2a) to further confirm our genotyping results. Statistical analysis {#sec2-6} -------------------- Statistical analysis was carried out using Prism 6 (GraphPad Software Inc, U.S.A.). Data were assessed using two-way ANOVA analysis followed by Bonferroni post-hoc group comparisons. Normality of data was verified using the Kolmogorov and Smirnov method. Significance was determined at P\<0.05. Results {#sec3} ======= 5-HT~2A~ ^−/−^ mice exhibit acute Flx-induced enhanced anxiety-like behavior in the OFT {#sec3-1} --------------------------------------------------------------------------------------- 5-HT~2A~ receptor knockout (5-HT~2A~^−/−^) mice have been reported to exhibit reduced anxiety-like behavior across both sexes (Supplementary Figure S1) \[[@B10],[@B14]\]. Acute Flx treatment is known to enhance anxiety-like behavioral responses \[[@B3]\]. Here, we sought to address whether the acute Flx-mediated increase in anxiety-like behavior is altered in 5-HT~2A~^−/−^ mice. Male and female 5-HT~2A~^−/−^ and WT littermate control mice were subjected to a single administration of Flx or saline, and were assessed for anxiety-like behavior in the OFT arena for 10 min ([Figure 1](#F1){ref-type="fig"}A). Two-way ANOVA analysis of OFT behavior in male 5-HT~2A~^−/−^ and WT mice revealed a significant main effect of acute Flx administration on percent distance traveled in center (F~(1,36)~ = 9.313, P=0.0043) ([Figure 1](#F1){ref-type="fig"}C), percent time in center (F~(1,36)~ = 4.164, P=0.048) ([Figure 1](#F1){ref-type="fig"}D), number of entries to the center (F~(1,36)~ = 7.608, P=0.0091) ([Figure 1](#F1){ref-type="fig"}E), and the total distance traveled in the OFT arena (F~(1,36)~ = 5.17, P=0.029) ([Figure 1](#F1){ref-type="fig"}F). Our results indicate a significant effect of acute Flx treatment on anxiety-like behavior in both WT and 5-HT~2A~^−/−^ mice ([Figure 1](#F1){ref-type="fig"}B--F). Interestingly, the baseline decrease in anxiety-like behavior previously noted in 5-HT~2A~^−/−^ mice \[[@B10],[@B14]\], was not observed following administration of a single saline injection, suggesting that the anxiety-like behaviors in 5-HT~2A~^−/−^ mice may be sensitive to both handling and injection stress. There was no significant main effect of genotype and no interaction effect (genotype × Flx). ![5-HT2A receptor knockout mice exhibit acute Flx-induced enhanced anxiety-like behavior in OFT\ Shown is a schematic of the experimental paradigm (A). 5-HT~2A~^−/−^ and WT littermate control male and female mice were administered acute Flx or saline (Sal), and were subjected to behavioral testing on the OFT 2 h post treatment (A). Shown are representative traces in the OFT arena from saline and Flx treated, WT and 5-HT2A^−/−^ male (B) and female (G) mice. Both 5-HT2A^−/−^ and WT male (B,C--F) and female (G,H--K) mice exhibited significant increases in anxiety-like behavior on the OFT following acute Flx treatment, with a reduction noted in percent distance traveled in center (males (C); females (H)), percent time in center (males (D); females (I)) and number of entries to the center (males (E); females (J)) of the OFT arena, as well as the total distance traveled in the OFT arena (males (F); females (K)). Results are expressed as the mean ± S.E.M. (n=9--10/group). Two-way ANOVA analysis, \*P\<0.05, significant main effect of acute Flx treatment.](ns-03-ns20180205-g1){#F1} Similar to our observations in male mice, female 5-HT~2A~^−/−^ and WT littermate controls both showed enhanced anxiety-like behavior on the OFT following acute Flx treatment ([Figure 1](#F1){ref-type="fig"}G--K). Two-way ANOVA analysis of OFT behavior indicated a significant main effect of acute Flx treatment on percent distance traveled in center (F~(1,35)~ = 13.54, P=0.0008) ([Figure 1](#F1){ref-type="fig"}H), percent time in center (F~(1,35)~ = 7.454, P=0.0098) ([Figure 1](#F1){ref-type="fig"}I), number of entries to the center (F~(1,35)~ = 15.09, P=0.0004) ([Figure 1](#F1){ref-type="fig"}J), and total distance traveled (F~(1,35)~ = 5.592, P=0.0237) in the OFT arena ([Figure 1](#F1){ref-type="fig"}K). We noted no main effect of genotype and no significant interaction (genotype × Flx) effects. This indicates that the baseline anxiolytic responses on OFT reported in both sexes of 5-HT~2A~^−/−^ mice are not observed following handling and injection stress. Taken together, our findings indicate that 5-HT~2A~ receptor deficiency does not influence the increased anxiety-like behavior noted following acute treatment with the SSRI, Flx. 5-HT~2A~ receptor knockout mice exhibit acute Flx-induced enhanced anxiety-like behavior in the EPM {#sec3-2} --------------------------------------------------------------------------------------------------- We next examined the influence of acute Flx on anxiety-like behavior in 5-HT~2A~^−/−^ mice and their WT controls, in the EPM ([Figure 2](#F2){ref-type="fig"}A--K). ![5-HT~2A~^−/−^ mice exhibit acute Flx-induced enhanced anxiety-like behavior on EPM\ Shown is a schematic of the experimental paradigm (A). 5-HT~2A~^−/−^ and WT control male and female mice were administered saline (Sal) or acute Flx, and were subjected to behavioral testing on the EPM 2 h post treatment (A). Shown are representative EPM arena tracks from saline and Flx treated, WT and 5-HT~2A~^−/−^ male (B) and female (G) mice. Both 5-HT~2A~^−/−^ and WT male (B,C--F) and female (G,H--K) mice exhibited significant increases in anxiety-like behavior on the EPM following acute Flx treatment, with a reduction noted in total distance traveled in the EPM arena while percent distance traveled in open arms (males (C); females (H)), percent time in open arms (males (D); females (I)) were unaltered by Flx administration. The number of entries to the open arms of the EPM arena was significantly reduced by Flx in females (J) but not in males (E). Results are expressed as the mean ± S.E.M. (n=11--15/group). Two-way ANOVA analysis, \*P\<0.05, significant main effect of acute Flx treatment.](ns-03-ns20180205-g2){#F2} Two-way ANOVA analysis of EPM behavior in males revealed no significant main effect of acute Flx on percent distance traveled ([Figure 2](#F2){ref-type="fig"}C), percent time spent ([Figure 2](#F2){ref-type="fig"}D), and number of entries ([Figure 2](#F2){ref-type="fig"}E) in the open arms. For the percent time spent in the open arms for males, we did note a trend toward an interaction of genotype × Flx (F~(1,50)~ = 3.39, P=0.0714) ([Figure 2](#F2){ref-type="fig"}D), which did not reach statistical significance. Flx treatment significantly reduced the total distance traveled in the EPM arena (F~(1,50)~ = 6.42, P=0.0145) ([Figure 2](#F2){ref-type="fig"}F), which is considered to be indicative of an enhanced anxiety-like behavioral response. We noted no main effect of genotype and no significant genotype × Flx interaction on any of the EPM parameters in male mice. Comparable with the male EPM behavioral response, female WT and 5-HT~2A~^−/−^ mice also exhibited enhanced acute Flx-mediated anxiety in the EPM ([Figure 2](#F2){ref-type="fig"}G--K). Two-way ANOVA analysis of EPM in females revealed no significant main effect of acute Flx on percent distance traveled ([Figure 2](#F2){ref-type="fig"}H), and percent time spent ([Figure 2](#F2){ref-type="fig"}I) in the open arms. Acute Flx treatment significantly reduced the number of entries in the open arms (F~(1,46)~ = 7.86, P=0.0074) ([Figure 2](#F2){ref-type="fig"}J) and the total distance traveled in the EPM arena (F~(1,46)~ = 25.3, P\<0.0001) ([Figure 2](#F2){ref-type="fig"}K), suggestive of enhanced anxiety-like behavior. We noted no main effect of genotype and no significant genotype × Flx interaction for the behavioral measures assayed on the EPM in females. 5-HT~2A~ receptor knockout mice exhibit elevated serum corticosterone levels and sexually dimorphic PVN neural activation following acute Flx administration {#sec3-3} ------------------------------------------------------------------------------------------------------------------------------------------------------------ Acute Flx treatment is known to increase serum corticosterone levels through activation of the HPA axis \[[@B18],[@B19]\]. Given that the 5-HT~2A~ receptor has been implicated in regulation of HPA axis activity via effects on stimulation of PVN neurones in the hypothalamus \[[@B20]\], we sought to address whether the acute Flx-mediated effects on serum corticosterone and neural activation in the PVN are altered in 5-HT~2A~^−/−^ mice ([Figure 3](#F3){ref-type="fig"}A). ![5-HT~2A~ receptor knockout mice exhibit elevated serum corticosterone levels and sexually dimorphic PVN c-Fos positive cell number following acute Flx administration\ Shown is a schematic of the experimental paradigm (A). 5-HT~2A~^−/−^ and WT control male and female mice were administered acute Flx or saline (Sal), and were killed 30 min later for serum corticosterone analysis or immunohistochemical analysis of c-Fos positive cells in the PVN of the hypothalamus (A). Male (B) and female (E) 5-HT~2A~^−/−^ and WT mice showed an increase in the serum corticosterone levels post Flx treatment. Shown are representative sections of c-Fos positive cells within the PVN of 5-HT~2A~^−/−^ and WT mice treated with saline or Flx (male (C); female (F)). 5-HT~2A~^−/−^ and WT male mice exhibited a significant increase in c-Fos positive cell numbers in the PVN (D). A significant genotype × Flx interaction effect was noted for c-Fos positive cell numbers in the PVN of 5-HT~2A~^−/−^ and WT female mice administered Flx or saline (G). Post-hoc Bonferroni multiple comparisons revealed a significant increase in PVN c-Fos positive cell numbers following acute Flx in WT, but not in 5-HT~2A~^−/−^, female mice (G). Results are expressed as the mean ± S.E.M. (n=3--6/group). Two-way ANOVA analysis, \*P\<0.05, significant main effect of acute Flx treatment, ^§^P\<0.05, significant genotype × Flx interaction, Bonferroni post-hoc test, ^¢^P\<0.05, as compared with saline-treated WT female mice.](ns-03-ns20180205-g3){#F3} Two-way ANOVA analysis for serum corticosterone levels in males revealed a significant main effect of acute Flx treatment (F~(1,15)~ = 15.09, P=0.002) and genotype (F~(1,15)~ = 5.18, P=0.04), but no significance in interaction of genotype and Flx ([Figure 3](#F3){ref-type="fig"}B). We next assessed the number of c-Fos^+^ cells within the PVN of 5-HT~2A~^−/−^ and WT littermate control male mice as a measure of the degree of neural activation following acute Flx treatment. Two-way ANOVA analysis for number of c-Fos positive cells indicated a significant main effect of acute Flx treatment (F~(1,18)~ = 24.89, P\<0.0001) ([Figure 3](#F3){ref-type="fig"}C,D), with no main effect noted for genotype and no genotype × Flx interaction. This indicates that 5-HT~2A~ receptor deficiency in male mice does not alter either the acute Flx-mediated increase in serum corticosterone levels or the increased neural activation, as assessed using c-Fos positive cell numbers, within the PVN. Similar analysis for serum corticosterone levels carried out in female 5-HT~2A~^−/−^ and WT mice indicated a significant main effect of acute Flx treatment (F~(1,17)~ = 13.21, P=0.002) with elevated levels of circulating corticosterone noted following acute Flx administration ([Figure 3](#F3){ref-type="fig"}E). No main effect of genotype, or interaction between genotype and Flx treatment, was noted in the two-way ANOVA analysis. The degree of neural activation in the PVN of female 5-HT~2A~^−/−^ and WT mice following acute Flx treatment was assessed using c-Fos cell counting analysis ([Figure 3](#F3){ref-type="fig"}F,G). Two-way ANOVA analysis indicated a significant genotype × Flx interaction (F~(1,18)~ = 6.58, P=0.02), and significant main effects for acute Flx treatment (F~(1,18)~ = 20.57, P=0.0003) and genotype (F~(1,18)~ = 6.53, P=0.02) ([Figure 3](#F3){ref-type="fig"}G). Post-hoc group comparisons revealed a significant increase in c-Fos positive cell numbers in the PVN following acute Flx administration to WT, but not 5-HT~2A~^−/−^, female mice ([Figure 3](#F3){ref-type="fig"}G). This suggests that 5-HT~2A~ receptor deficiency in female mice results in a blunting of the degree of neural activation in the PVN following acute Flx administration. 5-HT~2A~ receptor deficiency does not alter neural activation patterns in the PFC following acute Flx administration {#sec3-4} -------------------------------------------------------------------------------------------------------------------- We next examined neural activation, using c-Fos cell counting analysis, within the subdivisions of the PFC, which are known to exert a strong top-down control on HPA axis activity \[[@B21]\], exhibit robust 5-HT~2A~ receptor expression and are known to regulate anxiety-like behavior \[[@B10],[@B22]\]. We assessed c-Fos positive cell numbers within the cingulate (Cg), prelimbic (PrL) and infralimbic (IL) subdivisions of the PFC in male and female 5-HT~2A~^−/−^ and WT littermate control mice treated with acute Flx or saline ([Figure 4](#F4){ref-type="fig"}A). ![Acute Flx treatment does not alter c-Fos positive cell number in the PFC of 5-HT~2A~^−/−^ and WT male and female mice\ Shown is a schematic of the experimental paradigm (A). 5-HT~2A~^−/−^ and WT control male and female mice were administered acute Flx or saline (Sal), and were killed 30 min later for immunohistochemical analysis of c-Fos positive cell number in the subdivisions of the PFC, namely the Cg, PrL, and IL cortex (A). Shown are representative sections of c-Fos positive cells within the PFC of 5-HT~2A~^−/−^ and WT mice treated with saline or Flx (male (B); female (G)). Cell counting analyses revealed no significant main effect of Flx treatment, genotype or genotype × Flx interaction on c-Fos positive cell numbers within the Cg (male (C); female (H)), PrL (male (D); female (I)), IL (male (E); female (J)) and total PFC (male (F); female (K)). Results are expressed as the mean ± S.E.M. (n=5--6/group). Two-way ANOVA analysis.](ns-03-ns20180205-g4){#F4} Two-way ANOVA analysis for c-Fos positive cell numbers in the Cg, PrL, IL and cumulatively for the entire PFC, in experiments done in 5-HT~2A~^−/−^ and WT male mice indicated no significant main effects of Flx, genotype or genotype × Flx interaction ([Figure 4](#F4){ref-type="fig"}B--F). Similarly, c-Fos positive cell numbers in the Cg, PrL, IL and for the total PFC were unaltered in 5-HT~2A~^−/−^ and WT female mice following acute Flx or saline treatment ([Figure 4](#F4){ref-type="fig"}G--K). Collectively, these findings indicate that neural activation patterns in the PFC subdivisions are not altered in 5-HT~2A~ receptor deficient mice of both sexes. 5-HT~2A~ receptor knockout mice exhibit sexually dimorphic alterations in PFC IEG expression {#sec3-5} -------------------------------------------------------------------------------------------- IEG expression within the neocortex, including the PFC, has been reported to be regulated by 5-HT~2A~ receptors \[[@B14],[@B23]\], hence we sought to address the effects of acute Flx on expression of several IEGs in the PFC of 5-HT~2A~^−/−^ and WT mice of both sexes ([Figure 5](#F5){ref-type="fig"}A). ![5-HT~2A~ receptor knockout mice exhibit sexually dimorphic changes in PFC IEG expression\ Shown is a schematic of the experimental paradigm (A). 5-HT~2A~^−/−^ and WT control male and female mice were administered acute Flx or saline (Sal), and were killed 2 h post treatment for qPCR analysis of IEG expression in the PFC (A). The heat map denotes the magnitude of regulation represented as percent of WT saline group (WT Sal), with up-regulated genes shown in red and down-regulated genes shown in green (key (A)). qPCR analysis for gene expression of the IEGs examined represented as percent of WT Sal ± S.E.M. (male (B); female (C)) (male: n=8--12/group; female: n=7--10/group). Two-way ANOVA analysis with significance determined at P\<0.05. IEGs analyzed included the Arc, Bdnf, Egr 1-4, Fos, Fosb, Fosl2, Homer 1-3, Jun, Junb, and Jund. We also performed qpCR analysis for the Htr2a to further confirm our genotyping results.](ns-03-ns20180205-g5){#F5} In the male PFC, two-way ANOVA analysis of qPCR results indicated significant main effects of genotype for several IEGs: Bdnf, Egr2, Egr4, Fos, Fosb, Fosl2, Homer1-3, and Jun ([Figure 5](#F5){ref-type="fig"}B). No significant main effects of genotype were observed for Arc, Egr1, Egr3, Junb, and Jund gene expression ([Figure 5](#F5){ref-type="fig"}B). 5-HT~2A~^−/−^ male mice exhibited a decline in PFC gene expression for several IEGs as compared with WT littermate control male mice ([Figure 5](#F5){ref-type="fig"}B). Acute Flx treatment did not alter the mRNA expression levels of the IEGs assessed, except for a strong trend for decline observed in gene expression levels of Bdnf and Homer3 in the male PFC ([Figure 5](#F5){ref-type="fig"}B). We observed no significant interactions of genotype × Flx for gene expression of the IEGs in the PFC in 5-HT~2A~^−/−^ and WT male mice. In striking contrast with the males, PFC IEG expression was largely unperturbed in the 5-HT~2A~^−/−^ female mice indicating a sexual dimorphism in the effects of the 5-HT~2A~ receptor on PFC IEG expression. Two-way ANOVA analysis indicated no significant main effect of genotype on the gene expression of IEGs in the PFC in females, with the exception of a main effect of genotype observed for Egr1 gene expression, and a strong trend toward a main effect of genotype in the expression of Egr2 and Junb ([Figure 5](#F5){ref-type="fig"}C). Two-way ANOVA analysis revealed a significant main effect of Flx for Bdnf, Egr1, Egr4, and Homer3 gene expression in the PFC ([Figure 5](#F5){ref-type="fig"}C), and a strong trend for a main effect of Flx in the prefrontal expression of Fosl2 and Homer1 ([Figure 5](#F5){ref-type="fig"}C). We observed no significant genotype × Flx interaction for gene expression of the IEGs in the PFC in 5-HT~2A~^−/−^ and WT female mice. Collectively, our results indicate that 5-HT~2A~ receptor deficiency exerts sexually dimorphic consequences on the expression of several IEGs in the PFC, with a decline in PFC expression of Bdnf, Egr2, Egr4, Fos, Fosb, Fosl2, Homer1-3, and Jun noted in 5-HT~2A~^−/−^ male, but not female mice. Discussion {#sec4} ========== Here, we report that 5-HT~2A~ receptor loss of function does not alter either the acute Flx-mediated increase in anxiety-like behavior, or the enhanced circulating corticosterone, in both male and female mice. However, 5-HT~2A~ receptor knockout mice do exhibit sexually dimorphic effects in the acute Flx-induced neural activation in the PVN with a blunted response to acute Flx noted in 5-HT~2A~^−/−^ female, but not male, mice. Furthermore, we noted sex-dependent changes in expression of several IEGs (Bdnf, Egr2, Egr4, Fos, Fosb, Fosl2, Homer1-3 and Jun) in the PFC of 5-HT~2A~^−/−^ male, but not female, mice. 5-HT~2A~ receptor deficiency and acute Flx evoked anxiety-like behavior {#sec4-1} ----------------------------------------------------------------------- The SSRI Flx exerts paradoxical anxiogenic effects during the acute phase of treatment \[[@B3],[@B4]\]. These aversive effects of acute Flx on anxiety-like behavior are suggested to involve a role for the 5-HT~2~ receptor subtypes, with the 5-HT~2C~ receptor in the BNST \[[@B6]\] and BLA \[[@B8],[@B12]\], implicated in these anxiogenic effects. The 5-HT~2A~ receptor also plays an important role in the modulation of anxiety-like behavior \[[@B9]\], and co-administration of a 5-HT~2A~ receptor antagonist is reported to accelerate and augment antidepressant effects of SSRIs \[[@B13],[@B24]\]. Here, we examined whether the increased anxiety-like behavior evoked by acute Flx is altered in 5-HT~2A~^−/−^ male and female mice. Our results indicate that similar to WT mice, adult 5-HT~2A~^−/−^ male and female mice both exhibit acute Flx-induced increases in anxiety-like behavioral responses on the OFT and EPM, indicating that 5-HT~2A~ receptor deficiency does not influence the anxiogenic effects of acute Flx. It would be important in future experiments to address whether 5-HT~2A~ receptor deficient mice which continue to exhibit the acute anxiogenic effects of Flx, also respond in a similar manner to acute treatment with other SSRIs. Interestingly, while we did observe the previously reported baseline anxiolytic behavioral response in 5-HT~2A~^−/−^ male and female mice (Supplementary Figure S1) \[[@B10],[@B14]\], the vehicle-treated 5-HT~2A~^−/−^ cohorts used for the acute Flx experiment did not exhibit this baseline anxiolytic phenotype. The cohorts used for the present study were subjected to handling and injection of vehicle/Flx, and since the anxiolytic behavioral phenotype observed in naïve, unhandled 5-HT~2A~^−/−^ mice was lost, this suggests the possibility that 5-HT~2A~ receptor deficient mice may be highly sensitive to effects of handling and injection stress. Prior reports do indicate that handling and injection stress can alter anxiety-like behaviors \[[@B25],[@B26]\]. Further experiments are required to directly test whether 5-HT~2A~ receptor deficient mice exhibit such enhanced sensitivity to handling and injection stress. A previous report indicates that 5-HT~2A~^−/−^ mice exhibit treatment resistance to chronic Flx, failing to exhibit both antidepressant-like behavioral responses and enhanced hippocampal neurogenesis following chronic Flx treatment \[[@B16]\]. Taken together, these findings indicate that while the anxiogenic effects noted during the acute phase of Flx treatment are unaltered by 5-HT~2A~ receptor deficiency, 5-HT~2A~^−/−^ mice exhibit a treatment-resistant phenotype to the effects of chronic Flx \[[@B16]\]. Role of the 5-HT~2A~ receptor in the effects of acute Flx on the HPA axis {#sec4-2} ------------------------------------------------------------------------- Acute Flx treatment is known to influence the HPA axis, increasing circulating levels of corticosterone via enhanced activity of PVN neurones in the hypothalamus \[[@B18],[@B19]\]. 5-HT~2A~ receptors are densely expressed by PVN neurones, and their activation leads to release of adrenocorticotrophic hormone and enhanced circulating corticosterone \[[@B20],[@B27]\]. We find that the acute Flx-mediated elevation of serum corticosterone levels is observed in both WT and 5-HT~2A~^−/−^ male and female mice, indicating that the effects of acute Flx on enhanced corticosterone levels do not appear to involve a role for the 5-HT~2A~ receptor. Acute Flx administration also resulted in a clear increase in neural activation within the PVN, as revealed by enhanced numbers of c-Fos positive cells. This acute Flx-evoked increase in PVN c-Fos positive cells was noted in both WT and 5-HT~2A~^−/−^ male mice, and the magnitude of this effect was unaltered by 5-HT~2A~ receptor deficiency in males. Strikingly, 5-HT~2A~^−/−^ female mice showed a significant blunting of the acute Flx-mediated increase in c-Fos positive cells in the PVN. Previous studies indicate an interaction of sex hormones such as estrogen with SSRIs such as Flx \[[@B28],[@B29]\], and also with the 5-HT~2A~ receptor \[[@B30],[@B31]\]. This in turn provides support for the possibility that the effects of acute Flx on neural activation in the PVN may exhibit sex differences, possibly through interactions between the 5-HT~2A~ receptor, estrogen and acute Flx treatment. While we noted that the acute Flx-mediated increase in c-Fos positive cell number in the PVN was blunted in 5-HT~2A~^−/−^ female mice, we did not observe any significant genotype × Flx interaction for corticosterone levels in 5-HT~2A~^−/−^ female mice. This suggests that despite a substantial blunting of the neural activation in the PVN evoked by acute Flx in 5-HT~2A~^−/−^ female mice, this does not translate to a failure of acute Flx to increase circulating corticosterone in 5-HT~2A~^−/−^ female mice. Collectively, our findings indicate that the effects of acute Flx on enhancing corticosterone levels are unaltered by 5-HT~2A~ receptor deficiency, however 5-HT~2A~ receptor loss of function does alter the degree of neural activation induced by acute Flx in the PVN in a sex-dependent manner. 5-H~T2A~ receptor and the acute Flx-evoked regulation of neural activation, and IEG expression, in the PFC {#sec4-3} ---------------------------------------------------------------------------------------------------------- Acute Flx treatment is known to rapidly elevate serotonin levels within the PFC \[[@B32]\], a limbic brain region that exerts top-down control over the HPA axis and regulates anxiety behaviors \[[@B21]\]. The 5-HT~2A~ receptor is expressed at particularly high levels within the PFC, and the cortical 5-HT~2A~ receptor plays a key role in the modulation of anxiety behavior \[[@B10],[@B22],[@B33]\]. We sought to address whether 5-HT~2A~ receptor deficiency altered neural activation measured via assessing c-Fos positive cell number, or activity-dependent IEG expression, in the PFC, both under baseline conditions and in response to acute Flx administration. We observed no difference in c-Fos positive cell numbers in the PFC across groups 30 min post acute Flx administration. Our results do not preclude the possibility of differences in the nature and degree of neural activation of the PFC in WT and 5-HT~2A~^−/−^ mice at later time-points post Flx treatment. We also examined the expression of several IEGs in WT and 5-HT~2A~^−/−^ male and female mice, both baseline and post Flx treatment. While we did not observe any major effect of Flx treatment on IEG expression in the PFC, we noted robust baseline down-regulation of several IEGs (Bdnf, Egr2, Egr4, Fos, Fosb, Fosl2, Homer1-3, Jun) in the PFC of 5-HT~2A~^−/−^ male, but not female mice. Our results are largely in agreement with prior reports that acute Flx, unlike chronic treatment, does not appear to alter cortical IEG and plasticity-associated gene expression \[[@B34]\]. We have previously reported a sex-dependent, baseline regulation of several IEG and plasticity-associated genes in both the PFC and hippocampus of 5-HT~2A~^−/−^ mice \[[@B14]\]. This decline in basal expression of several IEGs in the PFC of 5-HT~2A~^−/−^ male mice may arise as a consequence of reduced excitability of PFC pyramidal neurones following 5-HT~2A~ receptor deficiency, however the fact that this change is not observed in 5-HT~2A~^−/−^ female mice is intriguing. Prior reports suggest an interaction between estrogen and the 5-HT~2A~ receptor with regard to the transcriptional regulation of Bdnf \[[@B31]\]. In particular, Bdnf expression is reported to be regulated by estrogen receptors via an estrogen response element \[[@B35]\], it is possible that such a regulation may contribute to the differential effects of 5-HT~2A~ receptor deficiency on cortical Bdnf expression in male compared with female mice \[[@B36]\]. Our findings suggest a sexual dimorphism in the 5-HT~2A~ receptor-mediated regulation of activity-dependent IEG expression, and motivate future experiments to address the nature of interaction between sex hormones and the 5-HT~2A~ receptor in this regard. Conclusion {#sec5} ========== The findings of our study indicate that the 5-HT~2A~ receptor does not appear to play a major role in the effects of acute Flx on anxiety-like behavior and corticosterone secretion. However, 5-HT~2A~ receptor deficiency evokes sex-dependent differences in both baseline cortical IEG expression, and in the acute Flx-mediated neural activation of the PVN of the hypothalamus. These findings highlight the sex differences in the effects of the 5-HT~2A~ receptor in the regulation of several genes, implicated in neuronal plasticity, within key brain regions such as the PFC. This suggests the possibility of an important interaction between the 5-HT~2A~ receptor, estrogen and activity-dependent IEG and plasticity-associated gene expression, motivating future experiments to uncover the mechanistic details of such an interaction. Supplementary Material ====================== ###### Supplementary Figures ###### Click here for additional data file. ###### Supplementary Table ###### Click here for additional data file. We thank Dr. Shital Suryavanshi for assistance with breeding and maintenance of the 5-HT~2A~ receptor knockout mouse colony. Competing interests {#sec6} =================== The authors declare that there are no competing interests associated with the manuscript. Funding {#sec7} ======= This work was supported by the intramural funds from the Tata Institute of Fundamental Research \[grant number 12P-0118\]. Author contribution {#sec8} =================== M.J. and V.A.V. conceptualized the experiments, provided critical inputs, and wrote the paper. M.J. performed most experiments, all analyses, and made the figures. T.B. performed elevated plus-maze experiment and helped in manuscript rebuttal. N.W. and J.A.G. made and provided the 5-HT~2A~ knockout mouse line. 5-HT~2A~ : serotonin~2A~ receptor 5-HT~2A~^−/−^ : 5-HT~2A~ developmental knockout 5-HT~2C~ : serotonin~2C~ receptor Arc : activity regulated cytoskeletal associated protein Avg : average Bdnf/BDNF : brain derived neurotrophic factor BLA : basolateral nucleus of the amygdala BNST : bed nucleus of the stria terminalis c-Fos : FBJ murine osteosarcoma viral oncogene homolog peptide Cg : cingulate cortex Flx : fluoxetine Fos : FBJ osteosarcoma oncogene Fosb : FBJ murine osteosarcoma viral oncogene homolog B Fosl2 : Fos-like antigen 2 Homer : Homer scaffolding protein HPA axis : hypothalamus--pituitary--adrenocortical axis IEG : immediate early gene IL : infralimbic cortex Jun : Jun proto-oncogene OFT : open-field test PFA : paraformaldehyde PFC : prefrontal cortex PrL : prelimbic cortex PVN : paraventricular nucleus qPCR : quantitative real-time PCR SSRI : selective serotonin reuptake inhibitor WT : wild-type	{ "pile_set_name": "PubMed Central" }
1. Introduction =============== Multiple sclerosis (MS) is a chronic neurological disease affecting a significant number of young adults \[[@B1-ijms-16-13885],[@B2-ijms-16-13885]\]. Although the exact cause of MS is unknown, it is the most common multifocal inflammatory demyelinating disease of the central nervous system (CNS) \[[@B3-ijms-16-13885],[@B4-ijms-16-13885],[@B5-ijms-16-13885]\]. The neurodegenerative process with damage to axons and oligodendrocytes is thought to be the cause of permanent neurological impairment and disability \[[@B6-ijms-16-13885],[@B7-ijms-16-13885]\]. Due to the progressive neurodegenerative nature of MS, developing treatments that exhibit direct neuroprotective effects are needed. Currently, most available MS therapies are thought to exert their effects via immunomodulatory or immunosuppressive functions \[[@B5-ijms-16-13885],[@B8-ijms-16-13885],[@B9-ijms-16-13885]\]. Although these treatments are effective at inhibiting immune cell-driven inflammation and reducing the relapse rate, they are ineffective at controlling the predominantly neurodegenerative processes that occur later in the disease course \[[@B10-ijms-16-13885]\]. Methods of enhancing remyelination and decreasing axonal damage (i.e., promoting neuroprotection) would be particularly useful for later stage and primary progressive MS forms. Fumaric acid esters (FAE) have been used since 1959 as a treatment for psoriasis \[[@B11-ijms-16-13885]\]. Tecfidera™ (BG-12) is an oral formulation of the FAE, containing the active metabolite dimethyl fumarate (DMF), which showed remarkable efficacy in lowering relapse rates in two phase III clinical trials of relapsing-remitting MS (RR-MS) treatment \[[@B12-ijms-16-13885],[@B13-ijms-16-13885],[@B14-ijms-16-13885],[@B15-ijms-16-13885],[@B16-ijms-16-13885]\]. However, its mechanism of action is not yet well understood. It has been shown that FAE and its primary metabolite DMF and monomethyl fumarate (MMF) were able to activate the transcription factor nuclear factor-erythroid 2-related factor 2 (Nrf2) pathway and subsequently induce the expression of antioxidant proteins \[[@B17-ijms-16-13885],[@B18-ijms-16-13885],[@B19-ijms-16-13885]\]. Oxidative stress plays a major role in multiple sclerosis (MS) and is readily apparent within experimental autoimmune encephalomyelitis (EAE), mouse model of MS, and MS lesions \[[@B15-ijms-16-13885],[@B20-ijms-16-13885]\]. Invading leukocytes contribute to cell damage and demyelination by producing excessive amounts of cytotoxic mediators, including reactive oxygen species (ROS). Under normal conditions, ARE/EpRE (antioxidant/electrophile response element) is not active, but under stress Nrf2 activates ARE/EpRE \[[@B21-ijms-16-13885]\]. Nrf2 is rapidly ubiquitinated and degraded by the help of the adaptor protein, kelch-like ECH-associated protein 1 (Keap1) \[[@B22-ijms-16-13885]\]. Glutathione, a major intracellular ROS scavenger, is decreased within CNS inflammatory foci, and a variety of antioxidant proteins are increased in MS lesions \[[@B20-ijms-16-13885]\]. Additionally, EAE in Nrf2^(−/−)^ mice is more severe and neuroprotective effects of DMF are absent in Nrf2^(−/−)^ mice \[[@B23-ijms-16-13885]\]. Furthermore, it has been shown that administration of MMF could protect motor neurons and astrocytes against hydrogen peroxide (H~2~O~2~)-induced oxidative stress \[[@B23-ijms-16-13885]\]. In a noninflammatory demylination model, DMF/MMF have little to no impact on demyelination or remyelination \[[@B24-ijms-16-13885]\]. Other studies also have shown Nrf2-dependent cytoprotection of neurons and astrocytes, but did not demonstrate decreased infiltrates \[[@B25-ijms-16-13885],[@B26-ijms-16-13885],[@B27-ijms-16-13885]\]. The therapeutic application of DMF in chronic EAE leading to a reduction in demyelination with relative preservation of myelin and axons suggests that the rescue of neurons and glial cells from oxidative stress-induced cell death is most likely through neuroprotection mediated by DMF-induced activation of Nrf2 pathway \[[@B23-ijms-16-13885]\]. The observation of proliferating neural cells in the adult rat brain made the possibility of CNS regeneration more feasible. Neural stem/progenitor cells (NPCs) reside in the adult CNS and support neurogenesis and gliogenesis throughout life \[[@B8-ijms-16-13885],[@B28-ijms-16-13885]\]. They are a heterogeneous population of perpetually self-renewing and multipotent cells that can spontaneously differentiate into neurons, astrocytes, or oligodendrocytes (post-mitotic daughter cells) \[[@B29-ijms-16-13885]\]. The ability of NPCs to reconstitute neurons makes them an important target for therapy in MS, as various drugs may interact with stem cells to promote neural protection and/or remyelination. It has been thought that the benefits seen in the clinical symptoms of EAE mice after stem cell transplantation are mediated by neural stem cells and immune cell interactions that promote neuroprotection. NPCs transplantation holds significant promise as a novel treatment strategy for MS \[[@B30-ijms-16-13885]\]. Neuroprotection has also been seen in excitotoxic neurodegeneration and Huntington's disease. DMF has also been shown to improve lifespan, reduce behavioral deficits, and preserve striatal and motor cortex neurons in two different genetic models of Huntington's disease in mice \[[@B31-ijms-16-13885]\], suggesting broad neuroprotective properties. Damage of cortical neurons and motor neurons are the main cause of dysfunction and disability in MS patients. MS disease occurs early and progresses later in life. It is currently unknown whether DMF exerts its neuroprotective effects directly on differentiated neuronal cells and/or through activation of NPCs. Furthermore, the effect of DMF at the level of NPCs has not yet been reported. However, the mechanism of DMF action in these neurons, in terms of signaling pathways in MS, is not clearly understood. In this study, we sought to investigate the neuroprotective effect of DMF in the context of two different neuronal cellular systems: NPCs and motor neurons. Our data showed that DMF protected NPCs in addition to differentiated motor neurons from oxidative damage (H~2~O~2~) through regulation of Nrf2, as well as several other novel genes involved in superoxide metabolism. Furthermore, the involvement of the signal transduction ERK1/2 MAPK pathway was identified in DMF-mediated neuroprotection. 2. Results ========== 2.1. DMF Increased Frequency of Mouse NPCs in Vitro and Protected Rat NPCs and Motor Neurons from Oxidative Damage ------------------------------------------------------------------------------------------------------------------ To examine the neuroprotective effects of DMF, we first used mouse NPCs to determine whether DMF has any effects on neurosphere formation and self-renewal in vitro. In the neurosphere formation assay, we found that DMF increased the frequency of the multipotent neurospheres in culture with the frequency of 30 ± 2 comparing to 22 ± 2.6 in control per 1000 cells ([Figure 1](#ijms-16-13885-f001){ref-type="fig"}A). Next we tested whether DMF plays a role in regulation of stem/progenitor cells self-renewal. We calculated the number of secondary neurospheres clonally derived from one primary neurosphere plated under low cell density condition and found that DMF treatment increased the number of secondary neurospheres in the self-renewal assay with 302 ± 11 comparing to 204 ± 27 in control ([Figure 1](#ijms-16-13885-f001){ref-type="fig"}B). To test whether DMF has neuroprotective effects with respect to oxidative damage, rat NPCs were under oxidative stress with hydrogen peroxide (H~2~O~2~) with or without DMF treatment, then an apoptosis assay was performed. DMF indeed decreased H~2~O~2~-induced apoptosis from 27.1% ± 3% to 12.6% ± 1.8%, effectively increasing the survival of rNPCs following oxidative stress treatment with H~2~O~2~. DMF treatment alone did not have any effect (11% ± 1%, [Figure 1](#ijms-16-13885-f001){ref-type="fig"}C). In addition, utilizing reactive oxygen species (ROS) assay, we showed that DMF reduced ROS production induced by H~2~O~2~ from 98 ± 0.002 to 0.73 ± 0.07 in rat NPCs ([Figure 1](#ijms-16-13885-f001){ref-type="fig"}D). To observe the efficacy of the neuroprotective features of DMF, dissociated rat motor neurons were cultured on electrospun poly-[l]{.smallcaps}-lactic acid (PLLA) nanofibers and the following features were measured: average number of neurites per image field, total neurite outgrowth, mean neurite length, and cell body area. The fibers serve as a substrate for growing the neurons since they direct neurite growth in a straight line, simplifying identification of neurites and neurite measurement. Neurons were exposed to oxidative stress with H~2~O~2~ after one day of culture and grown for either two or four days; half of these samples were treated with DMF. Both were compared to positive control samples that were unstressed and untreated. After two days of growth, total neurite outgrowth decreased on stress-treated neurons from 531 ± 16 to 438 ± 14 ([Figure 2](#ijms-16-13885-f002){ref-type="fig"}A), indicating poorer health of the neurons. However, addition of DMF abrogated this decrease and raised total neurite outgrowth from 438 ± 14 to 569 ± 16 ([Figure 2](#ijms-16-13885-f002){ref-type="fig"}A). No difference was observed amongst samples for cell body area or number of neurites per image field. This establishes that after one day of culture, neurons had begun to succumb to the stress of H~2~O~2~ treatment. After four days of growth (with three days of stress), no neurites were found on stressed motor neurons ([Figure 2](#ijms-16-13885-f002){ref-type="fig"}D). Addition of DMF protected neurons from stress as measured by average neurite number per image field (4 ± 0.3, n = 81, [Figure 2](#ijms-16-13885-f002){ref-type="fig"}C). This effect was also seen in all of our other measures: total neurite outgrowth, mean neurite length, and cell body area (data not shown). This substantiates the neuroprotective effects of DMF. ![Dimethyl fumarate (DMF) increased self-renewal and reduced oxidative stress-induced apoptosis and ROS levels in neural stem/progenitor cells (NPCs). (A) Frequency of mouse neuronal stem/progenitor cells was calculated by dividing the number of neurospheres in the well by the number of plated cells with (DMF) or without (C) DMF treatment (Mean ± SEM were determined from seven replicate samples, \* p = 0.027); (B) Self-renewal potential of mouse NPCs was calculated as number of secondary (2nd) neurospheres (NS) clonally derived from one primary (1st) neurosphere plated with (DMF) or without (C) DMF treatment (Mean ± SEM were determined from three replicate samples, \* p = 0.03); (C) Rat NPCs were grown in adherent cultures in self-renewal medium. After treatment under indicated conditions with control (C), DMF alone (DMF) for overnight, then H~2~O~2~ alone (H~2~O~2~) and DMF together with H~2~O~2~ (H~2~O~2~ + DMF) for another 6 h, the NPC cells were trypsinized, then apoptosis assay was performed as described under "Material and methods". Mean ± SEM were determined from three replicatesamples, \* p = 0.015, *\\*** p = 0.007); (D) Cells treated same as in "C" and then incubated with 5 µM 2′,7′-dichlorofluorescein diacetate (DCFDA) for 15 min at 37 °C and oxidative stress was measured as ROS binding to DCFDA and quantified by flow cytometry. Mean ± SEM were determined from three replicate samples, \* p = 0.027.](ijms-16-13885-g001){#ijms-16-13885-f001} ![DMF promoted survival of rat motor neurons under oxidative stress. Isolated rat motor neurons were cultured on aligned, electrospun poly-[l]{.smallcaps}-lactic acid (PLLA) nanofiber scaffolds that direct neurite growth in straight lines along the fibers, allowing straightforward measurement of neurite length. Treatment with vehicle (control; labeled C), H~2~O~2~ alone (H~2~O~2~) and DMF together with H~2~O~2~ (H~2~O~2~ + DMF) was added after one day in vitro. Following fixation and staining, neurons were identified and their neurites were then counted and measured after two and four days in vitro (resulting in exposures of one and three days, respectively). Motor neuron survival was assessed by measuring the neurite length and counting the number of survived motor neurons on the slide view field. (A,C) mean ± SEM of three replicate measurements (n = 60--81); *\\*** p \< 0.01, *\\\** p \< 0.001 (B,D) representative images from Days 2 and 4 of culture. Scale bars = 200 µm.](ijms-16-13885-g002){#ijms-16-13885-f002} 2.2. Mechanism of Neuroprotective Effects of DMF May Be through Regulation of the Anti-Oxidative Stress Gene Nrf2 ----------------------------------------------------------------------------------------------------------------- The main mechanism of action that has been attributed to DMF involves direct effects on neuroprotection through activation of antioxidant response elements by the nuclear factor (erythroid-derived 2)-like 2 (Nrf2) pathway \[[@B23-ijms-16-13885]\]. To determine whether Nrf2 gene and protein were regulated by DMF in rNPCs, we used RT-qPCR and Western blot analysis to evaluate Nrf2 expression levels of DMF-treated rNPCs in vitro. Using cDNA or whole cell lysates prepared from rNPCs treated under different conditions with or without oxidative stress, total RNA and protein lysates were prepared for RT-qPCR and Western blot analysis for Nrf2 expression. We found that Nrf2 at both mRNA and protein levels was increased by DMF treatment and this effect was significantly enhanced in the presence of oxidative stress induced by H~2~O~2~ treatment ([Figure 3](#ijms-16-13885-f003){ref-type="fig"}). This suggests that the neuroprotective effects of DMF may be through regulation of the anti-oxidative stress gene Nrf2. ![DMF increased Nrf2 RNA and protein expression in rat NPCs in the presence of oxidative stress. After treatment under indicated conditions with DMSO (C), DMF or H~2~O~2~ alone, or DMF and H~2~O~2~ together, Nrf2 RNA expression was determined using cDNAs prepared from rNPCs culture by RT-qPCR (A), *\\\** p \< 0.001; Nrf2 protein expression was assessed using whole cell lysates prepared from rNPCs cells by Western blot analysis (B) as described under "Material and Methods"; (C) Mean ± SEM were determined from 3--4 individual experiments. \* p = 0.026.](ijms-16-13885-g003){#ijms-16-13885-f003} 2.3. Mechanism of Involvement of ERK1/2 MAPK in DMF-Mediated Neuroprotection ---------------------------------------------------------------------------- The ERK1/2 MAPK pathway plays an important role in cell growth and survival \[[@B32-ijms-16-13885]\]. To determine whether the ERK1/2 MAPK pathway is involved in DMF-mediated neuroprotection, we treated NPCs with H~2~O~2~ to induce oxidative stress. Due to the fact that Nrf2 is degraded through proteasomal cleavage, the proteasomal inhibitor, MG132, was used as a control \[[@B26-ijms-16-13885],[@B33-ijms-16-13885]\]. MG132 dramatically increased ERK1/2 MAPK phosphorylation as expected ([Figure 4](#ijms-16-13885-f004){ref-type="fig"}A). DMF treatment alone increased phosphorylation of ERK1/2 MAPK from 1.0 ± 0.09 to 1.7 ± 0.23 and further induction was seen when treated with a combination of DMF and H~2~O~2~ for 4 h (2.7 ± 0.33). We also compared serum-starved NPCs that were treated for 4 h without B-27 supplement with NPCs treated overnight with B-27 supplement. Data showed that overnight treatment did not have an effect on ERK1/2 MAPK phosphorylation as seen with 4 h treatment except with the MG132 positive control. For the subsequent assay ([Figure 4](#ijms-16-13885-f004){ref-type="fig"}B,C), we performed only 4 h treatment. Since different kinases have been reported to regulate Nrf2 expression \[[@B16-ijms-16-13885],[@B34-ijms-16-13885]\], to confirm the involvement of ERK1/2 MAPK pathway, we used MEK1 pathway (ERK1/2) MAPK inhibitor PD98059 \[[@B34-ijms-16-13885]\], which completely blocked DMF- and H~2~O~2~-induced ERK1/2 phosphorylation including DMSO control ([Figure 4](#ijms-16-13885-f004){ref-type="fig"}B,C). The total ERK was not changed indicating that cell viability was not reduced. To further confirm the ERK1/2 MAPK involved in the Nrf2 pathway, the effect of the ERK1/2 inhibitor PD98059 on blocking Nrf2 protein level was investigated ([Figure 5](#ijms-16-13885-f005){ref-type="fig"}). PD98059 completely blocked DMF-induced Nrf2 protein expression. These data indicated that the mechanism of DMF-mediated neuroprotection may be through the involvement of ERK1/2 MAPK upstream of Nrf2 expression. ![Involvement of ERK1/2 MAPK in DMF-mediated neuroprotection. (A) rNPCs cultures were serum starved in neurobasal medium with (24 h) or without (4 h) B-27 supplement and treated with medium alone (--), DMSO (C), MG132 (MG), DMF, H~2~O~2~, or both H~2~O~2~ and DMF (H + D). ERK1/2 phosphorylation was determined by Western blot analysis; (B) rNPCs cells were serum starved in neurobasal medium and treated with (PD98059) or without (V) the ERK1/2 inhibitor under the same treatment as "A" for 4 h; (C) ERK1/2 phosphorylation was presented as a ratio of normalized arbitrary units (a.u.) of phosphorylated ERK1/2 (P-ERK) over total ERK1/2 (T-ERK) in four individual experiments under same treatment as in "B". \* p = 0.028, *\\*** p = 0.0042.](ijms-16-13885-g004){#ijms-16-13885-f004} ![Effect of ERK1/2 inhibitor on Nrf2 protein level. (A) Whole cell lysates were prepared from the rNPCs cultures treated under various conditions with (PD98059) or without (V) the ERK1/2 inhibitor for overnight. Western blot analysis for Nrf2 protein expression was performed as described under "Material and Methods"; (B) Mean ± SEM were determined from 3--6 experiments. \* p = 0.01, *\\*** p = 0.0068.](ijms-16-13885-g005){#ijms-16-13885-f005} 2.4. Identification of Additional Anti-Oxidative Stress Genes by SuperArray Gene Screen Technology -------------------------------------------------------------------------------------------------- When activated, Nrf2 protein dissociates from Keap1 complex, translocates into nucleus, and then activates ARE-mediated phase II antioxidant gene expression \[[@B35-ijms-16-13885]\]. To further determine whether DMF has an effect on up- or down-regulation of antioxidant gene(s) expression, we utilized Oxidative Stress PCR SuperArray gene screen technology to identify additional anti-oxidative stress genes under our conditions with the treatment of DMF and oxidative stress. Our results showed that under oxidative stress with H~2~O~2~, DMF significantly up-regulated gene expression of genes including Gstp1 (Glutathione S-transferase pi 1-, 5-fold increase), Nqo1 \[NAD(P)H dehydrogenase, quinone 1-, 2-fold increase\], Sod2 (Superoxide dismutase 2, mitochondrial, 2-fold increase), Srxn1 (Sulfiredoxin1 homolog, 4-fold increase), and Fth1 (Ferritin, 2-fold increase) ([Figure 6](#ijms-16-13885-f006){ref-type="fig"}A), which all play roles in reducing oxidative stress by breaking down ROS. Intriguingly, in NPC cultures, DMF also showed a trend of down-regulating genes including Ccl5 \[Chemokine (C--C motif) ligand 5-, 3-fold decrease\], which is a ligand of CCR1, CCR3, and CCR5 with the potential to reduce the recruitment of leukocytes into inflammatory sites ([Figure 6](#ijms-16-13885-f006){ref-type="fig"}A). Using Real-Time qRT-PCR analysis, Fth1 and Gstp1 gene up-regulation were further confirmed ([Figure 6](#ijms-16-13885-f006){ref-type="fig"}B). ![Identification of additional anti-oxidative stress genes by SuperArray gene screening technology. (A) SuperArray gene screen technology for anti-oxidative stress genes was performed as described under "Material and methods". DMF significantly up-regulates gene expression of Gstp1, Nqo1, Sod2, Srxn1, and Fth1, with down-regulation of Ccl5; (B) Confirmation Real-Time qRT-PCR. Up-regulation of Gstp1 and Fth1 genes were further confirmed by Real-Time qRT-PCR. Mean ± SEM were determined from 3 replicate samples. \\ p = 0.0015--0.0031.](ijms-16-13885-g006){#ijms-16-13885-f006} 3. Discussion ============= Our current study has shown that DMF protects different neuronal cell types including neural stem/progenitor cells (NPCs) and differentiated neurons (motor neurons) from oxidative damage through regulation of Nrf2 involving ERK1/2 MAPK pathway. Key Nrf2 target genes and genes involved in superoxide metabolism were demonstrated to be induced by DMF in rNPCs. The possible role of this mechanism in reducing oxidative stress by breaking down ROS suggests that analysis of anti-oxidative stress mechanisms in NPCs may yield further insights into new targets for treatment of MS. Inflammation and oxidative stress are thought to promote CNS tissue damage in multiple sclerosis by affecting demyelination. Recent studies had described a potential neuroprotective mechanism of action for the fumaric acid ester, dimethyl fumarate (DMF), and its primary metabolite, monomethyl fumarate (MMF), on neurons, astrocytes and glial cells demonstrated by in vitro and in vivo data \[[@B23-ijms-16-13885],[@B31-ijms-16-13885]\]. Protection of oligodendrocyte and neurons from irreversible damage in CNS may exert neuroprotective effects both directly and indirectly in the periphery and CNS. It has been shown that DMF and MMF are able to activate Nrf2, the key transcription factor regulating the antioxidant response and induce expression of antioxidant proteins \[[@B19-ijms-16-13885]\]. Furthermore, administration of MMF could protect motor neurons and astrocytes against H~2~O~2~-induced oxidative stress \[[@B23-ijms-16-13885]\]. These results may point to an underlying functional cellular substrate for the neuroprotective response observed in mouse EAE, in which DMF treatment reduced oxidative damage and consequential nerve fiber demyelination, resulting in greater axonal preservation and improved motor function \[[@B23-ijms-16-13885],[@B36-ijms-16-13885]\]. Different CNS cell types were used to demonstrate the DMF- and MMF-dependent stabilization of Nrf2, indicating the relevance of these compounds in treating diseases that result in the degeneration of CNS cells. Reports have shown that DMF exerted anti-inflammatory and prometabolic effects in a variety of cell types in primary astrocytes and C6 glioma cells by modifying glutathione (GSH) levels that can induce expression of the anti-inflammatory protein HO-1 (heme oxygenase-1) \[[@B19-ijms-16-13885]\]. DMF treatment increased nuclear levels of Nrf2 in astrocytes but not C6 cells. DMF suppressed inflammatory activation in astrocytes and C6 glioma cells, but with distinct mechanisms, different dependence on GSH, and different effects on transcription factor activation. The effect of DMF on NPCs was not previously demonstrated. We add to this data that DMF directly enhanced NPCs neurosphere growth, and decreased oxidative stress-induced apoptosis in both NPCs and neurons treated with H~2~O~2~. It has been suggested that NPCs and perhaps neurons are capable of responding to DMF treatment, leading to induction of a prototypical Nrf2-dependent antioxidant response. One report also showed that DMF promoted cytoprotection of central nervous system cells against oxidative stress via Nrf2 \[[@B26-ijms-16-13885]\]. Some other studies have shown that antioxidant response proteins like Nrf2 are elevated in MS lesions \[[@B20-ijms-16-13885],[@B37-ijms-16-13885]\]. However, the degree of intrinsic antioxidant induction is apparently insufficient to ultimately prevent disease progression. Neural stem/progenitor cells (NPCs) proliferate and produce new neurons in neurogenic areas throughout the lifetime. While these cells represent potential therapeutic treatment of neurodegenerative diseases, regulation of neurogenesis is not completely understood \[[@B38-ijms-16-13885]\]. Our observations have confirmed that DMF can activate the Nrf2 pathway in rat NPCs and stabilize levels of Nrf2 protein, resulting in the accumulation of active transcription factor, thereby conferring neuroprotective responses against oxidative stress. All cells have an intrinsic mechanism for combating reactive oxygen species (ROS) that is dynamically controlled through the actions of Nrf2, a transcription factor that is the principal regulator of the phase II cellular antioxidant response \[[@B39-ijms-16-13885]\]. The studies described here demonstrated that DMF treatment can decrease apoptosis and increase cell survival in motor neurons under H~2~O~2~ oxidative stress which further supported the DMF neuroprotective effects on primary cultures of motor neurons and suggests a potential role for cell type-specific neuroprotective responses. In addition to characterization of Nrf2 pathway activation in NPC cells, the potential upstream and downstream functional mechanism by which DMF may promote resistance to oxidative stress has not been identified. Though the involvement of mitogen-activated protein kinases (MAPKs) in the Nrf2 pathway has been reported, different subgroups of MAPKs are involved, and their exact functions differ depending on cell types \[[@B35-ijms-16-13885],[@B40-ijms-16-13885]\]. MAPKs play crucial roles in cellular processes, including proliferation, stress responses, apoptosis, and immune defense \[[@B41-ijms-16-13885]\]. Some reports have shown that Nrf2 protein accumulation has been induced by activation of MAPKs. ERK1/2 is one of three major subgroups of MAPKs. An ERK pathway inhibitor prevented Nrf2 phosphorylation at Ser40 to retard Nrf2 nuclear translocation, thus decreasing antioxidant gene transcription \[[@B35-ijms-16-13885],[@B40-ijms-16-13885],[@B42-ijms-16-13885]\]. There is no previous report on DMF neuroprotection via ERK1/2 MAPK-induced Nrf2 accumulation in neurons and NPCs. Here we showed that DMF increased Nrf2 accumulation and oxidative stress enhanced this effect through ERK1/2 phosphorylation which was completely abolished upon treatment with the ERK1/2 inhibitor, PD98059 \[[@B43-ijms-16-13885],[@B44-ijms-16-13885]\]. The neuroprotective action of DMF on NPCs is therefore likely through increasing Nrf2 accumulation by activation of the ERK1/2 signaling pathway. This does not exclude the possibility that other signaling pathways may also be involved, such as the NFκB pathway, which also decreases apoptosis of neurons, astrocytes, microglia and oligodendrocytes \[[@B19-ijms-16-13885],[@B45-ijms-16-13885],[@B46-ijms-16-13885]\]. Currently, we cannot rule out the possibility that DMF may act as a Keap1 inhibitor that leads to the accumulation of Nrf2 protein by decreasing proteasomal degradation \[[@B47-ijms-16-13885],[@B48-ijms-16-13885],[@B49-ijms-16-13885],[@B50-ijms-16-13885]\]. This neuroprotective role of DMF may explain the decreased rate of brain atrophy and the slowed rate of disease progression in patients with MS treated with Tecfidera™ \[[@B51-ijms-16-13885],[@B52-ijms-16-13885]\]. Elucidation of this pathway suggests a potential role for cell type-specific neuroprotective responses. Multiple genes have been reported to be candidates for direct targets of Nrf2 activation in different cell types and tissues which are enhanced only in the presence of the Nrf2 gene \[[@B53-ijms-16-13885],[@B54-ijms-16-13885]\]. The roles of these Nrf2 target genes in protection from oxidative stress were unknown for NPCs. To our knowledge, this is the first report to screen Nrf2 target genes and genes involved in superoxide metabolism in rat NPCs. Hence this is the first report in which five Nrf2 targeted genes (Fth1, Gstp1, Sod2, Nqo1 and Srxn1) have been identified as strong candidates for DMF-mediated Nrf2 activation in rat NPCs. Under oxidative stress, DMF significantly up-regulated the expression of these genes, which all play roles in reducing oxidative stress by breaking down ROS. Sod2 is a critical component of the mitochondrial pathway for detoxification of O~2~^−^ and targeted disruption of this locus leads to embryonic or neonatal lethality in mice \[[@B55-ijms-16-13885]\]. Gstp1 is a phase II detoxification enzyme that conjugates both endogenous and exogenous compounds to glutathione to reduce cellular oxidative stress. Decreased expression of Gstp1 has recently been implicated in dopamine neuron degeneration in Parkinson's disease progression \[[@B56-ijms-16-13885]\]. Nqo1 has been shown to be upregulated in Sulforaphane protection of cortical neurons against 5-S-cysteinyl-dopamine-induced toxicity through the activation of the ERK1/2-Nrf2 pathway \[[@B57-ijms-16-13885]\]. Srxn1 is an endogenous antioxidant protein that has neuroprotective effects. While the mechanisms of Srxn1 in oxidative stress have not been well studied, it has been shown that Srxn1 protects PC12 cells against oxidative stress induced by hydrogen peroxide \[[@B58-ijms-16-13885]\]. Srxn1 expression has also been shown to mediate protective effects of brain-derived neurotrophic factor (BDNF) against neurotoxicity in primary rat cortical cultures induced by 3-nitropropionic acid (3-NP), a mitochondrial complex II inhibitor \[[@B59-ijms-16-13885]\]. Fth1 gene has been implicated in protection against neuronal cell death under various neurodegenerative stresses, such as hypoxia-ischemia and oxidative stress and its increased expression may have protective effects against neural damage that could otherwise result in cell death \[[@B60-ijms-16-13885]\]. DMF also showed a trend of down-regulation of Ccl5, a ligand of CCR1, CCR3, and CCR5, which could potentially reduce the recruitment of leukocytes into CNS inflammatory sites, thereby decreasing cell trafficking and inflammation \[[@B61-ijms-16-13885]\]. In particular, abnormal Ccl5 expression was detected in the cerebrospinal fluid (CSF) of patients suffering from MS and in the CNS of EAE mice \[[@B62-ijms-16-13885],[@B63-ijms-16-13885],[@B64-ijms-16-13885],[@B65-ijms-16-13885],[@B66-ijms-16-13885]\] implicating this chemokine in the onset of the demyelinating disease(s). 4. Materials and Methods ======================== All experiments were done in accordance with the NIH Guide for Care and Use of Laboratory Animals, as approved by the R&D Service of the VA Ann Arbor Healthcare Center on 21 November 2012 (ACORP 2012-090604) and the University Committee on Use and Care of Animals (UCUCA) of the University of Michigan on 6 June 2013 (PRO 00004815). 4.1. Preparation of Rat NPCs and Mouse Neurospheres --------------------------------------------------- NPCs were isolated based on neurosphere formation and confirmed through staining with the NPC marker nestin as previously described \[[@B67-ijms-16-13885],[@B68-ijms-16-13885],[@B69-ijms-16-13885]\]. Rat embryonic E15-18 pups were decapitated, and the heads were placed into a petri dish with ice-cold HBSS (Ca^2+^, Mg^2+^ free). The whole brain without cerebellum was dissected and washed with fresh HBSS solution. It was then digested with 25 µg/mL DNase I in ice-cold HBSS and filtered through a cell strainer (70 µm, Falcon, BD Biosciences, San Jose, CA, USA). Cells were counted and plated onto 100 mm dishes pre-equilibrated with self-renewal media (SRM) containing DMEM, Neurobasal-A media, 2-mercaptoethanol, Chick Embryo Extract (CEE), 1× N-2 supplement, 2× B-27 supplement, FGF (20 ng/mL), EGF (20 ng/mL), and Penicillin-Streptomycin. Mouse neurosphere preparation and growth were carried out as described previously \[[@B69-ijms-16-13885]\]. Briefly, neonatal mouse brain was dissected and washed with fresh cold HBSS solution. Two coronal slices (as thin as possible, \~1--2 mm) were cut starting at the olfactory bulb. The lateral wall of the lateral ventricle was removed and mechanically triturated in a 15 mL conical tube with 1:3 dilution of trypsin. These cells were incubated for 4 min at 37 °C, quenched in two volumes of medium containing 25 µg/mL DNase I and spun at 210× g at 4 °C for 5 min. The pellets were re-suspended in ice-cold HBSS, counted, and plated at a density of 1000 cells/well onto 6-well low binding tissue culture plates pre-equilibrated with SRM. After 8--10 days in culture the neurospheres were counted, measured, and selected for the self-renewal assay. The frequency of neurospheres was calculated with the numbers of neurospheres formation from originally plated 1000 cells. To assay self-renewal potential, individual primary neurospheres (\>50 μm in diameter) were dissociated and re-plated at clonal density in non-adherent secondary cultures. 4.2. Primary Rat Motor Neurons Preparation and Culture ------------------------------------------------------ Primary motor neurons were cultured as previously described \[[@B70-ijms-16-13885]\]. Briefly, perineural membranes were removed from spinal cords of E15 Sprague-Dawley pregnant rats and the tissue was chopped into 2-mm pieces. Cells were dissociated by incubating in 0.25% trypsin/EDTA for 15 min at 37 °C followed by gentle trituration for one minute with a serum-coated, fire-polished glass pasteur pipette. Motor neurons were isolated over 9% Optiprep in L-15 media by centrifugation at 1000× g for 15 min and then collected by taking 2 mL of supernatant. Cells were washed in L-15 media, then resuspended and plated in culture medium. Defined culture media consisted of Neurobasal (Invitrogen, Carlsbad, CA, USA) supplemented with 2× B27 and the following additives: 2.5 mg/mL albumin, 2.5 µg/mL catalase, 2.5 µg/mL superoxide dismutase, 0.01 mg/mL transferrin, 15 µg/mL galactose, 6.3 ng/mL progesterone, 16 µg/mL putrescine, 4 ng/mL selenium, 3 ng/mL b-estradiol, 4 ng/mL hydrocortisone, and 1× penicillin/streptomycin/neomycin. [l]{.smallcaps}-Glutamine (2 µM) was added to the medium immediately before plating. Cells were counted and plated at a density of 25 cells/mm^2^, so that neurons would not contact one another. 4.3. Immunocytochemistry and Morphological Analysis --------------------------------------------------- Immunohistochemistry techniques were performed as previously described \[[@B70-ijms-16-13885]\]. Briefly, cells were fixed in 4% paraformaldehyde at room temperature for 30 min. To block nonspecific antibody binding, samples were incubated in 2% goat serum/2.5% BSA/0.05% Triton-X-100 in 1× PBS for 30 min. Primary antibody (anti-β-Tubulin III diluted 1:1000, Sigma, St. Louis, MO, USA) was prepared in 10% goat serum/2.5% BSA/0.05% Triton-X-100/0.1% sodium azide in 1× PBS and incubated with cells overnight. Cells were then washed in 1× PBS and incubated in the secondary antibody Goat anti-mouse IgG (FITC) diluted 1:200 (Sigma) for 4 h. Prolong Gold (Molecular Probes/Invitrogen, Carlsbad, CA, USA), an antifade agent with 4′,6-diamidino-2-phenylindole (DAPI), was used to stain nuclei. The neurons were imaged using an EVOS fluorescent microscope and analyzed with Metamorph imaging software (Molecular Devices, LLC. Sunnyvale, CA, USA). Neurite number, total neurite growth, mean neurite growth, and cell body area were all recorded. 4.4. Treatments --------------- Dimethyl fumarate (DMF) and Hydrogen peroxide solution (H~2~O~2~, 30% w/w in H~2~O, contains stabilizer) was purchased from Sigma--Aldrich (St. Louis, MO, USA). DMF was dissolved in DMSO and aliquots were kept in dark in −20 °C freezer. H~2~O~2~ was kept at 4 °C refrigerator. Newly opened bottle of H~2~O~2~ was used and freshly prepared and diluted H~2~O~2~ stock solution was immediately added into cell culture medium to achieve the desired working concentration. The concentration responses for both H~2~O~2~ (ranging from 0, 10, 20, 40, 80, and 100 μM) and DMF (ranging from 0, 10, 20, and 40 μM) and time course for both H~2~O~2~ and DMF (ranging from 4, 6, 18, 24, 48, and 72 h) were studied in the culture medium and then the optimal conditions (40 µm H~2~O~2~ and 10 µm DMF) for studying DMF neuroprotective effect under H~2~O~2~-induced oxidative stress were determined for subsequent experiments. In most experiments unless otherwise noted, the cultures were divided into four different groups which were treated in the following way: (1) Control with same concentration of DMSO used to dissolve DMF; (2) DMF alone; (3) H~2~O~2~ alone with same concentration of DMSO; (4) DMF + H~2~O~2~. At least three separate cultures were set up for each experiment. 4.5. Survival Assay ------------------- Motor neurons were isolated from E15 pregnant Sprague Dawley rats and plated onto poly (lactic-co-glycolic acid) (PLGA) solvent-casted, poly-[l]{.smallcaps}-lactic acid (PLLA) electrospun nanofibers coated with poly-[l]{.smallcaps}-lysine (PLL) \[[@B70-ijms-16-13885]\]. Three separate groups of cells were cultured: a control group of motor neurons on fibers treated with 0.1% DMSO, a group of cells cultured with 40 µM H~2~O~2~ (the stress group), and a group of cells cultured with both 40 µM H~2~O~2~ and 20 µM DMF (the treatment for this experiment). Cells were cultured for one day before any addition of treatment (DMF) or stress H~2~O~2~. In a separate experiment, cells were subjected to 20 µM DMF alone and did not show any difference in growth when compared to controls. Concentrations of H~2~O~2~ and DMF were determined in pilot experiments on neurons for optimal destructive/neuroprotective results. 4.6. Electrospinning -------------------- The electrospinning methods were similar to those previously described \[[@B70-ijms-16-13885]\]. Briefly, 0.4 g of poly-[l]{.smallcaps}-lactic acid (PLLA) was dissolved in a total volume of 10 mL in parts 9:1 chloroform: dimethylformamide, to yield a 4.4% w/v solution. The mixture was loaded into a 3 mL syringe with a 22 gauge blunt-tip needle and placed on a syringe pump set to 0.22 mL/h. The tip of the needle was protruded through the center of a 10 × 10 cm folded aluminum sheet. The rotating disc collector was placed 30 cm away, such that the axis of rotation was perpendicular to the syringe. A high voltage power supply applied a 20 kV charge to the aluminum sheet via an alligator clip. A second power supply applied a counter charge of −2 kV to the rotating disc via a wire brush. Fibers were collected for 5 min or until a desired density was obtained. 4.7. Apoptosis Assay -------------------- Rat NPCs were grown in adherent cultures in self-renewal medium. Cultures were pre-treated with indicated concentration of DMF (or DMSO control) overnight. After H~2~O~2~ treatment, the NPC cells were trypsinized, washed with 1× Annexin binding buffer and stained with APC-Annexin V kit (BD Biosciences, San Jose, CA, USA), according to the manufacturer's instructions. 7-AAD viability staining solution (BioLegend, San Diego, CA, USA) was used to determine the necrotic cells. Staining was analyzed by flow cytometry and the apoptotic cell population was calculated by the percent of Annexin V positive cells. 4.8. ROS Assay -------------- Rat NPCs were grown in adherent cultures overnight following pretreatment with DMF or DMSO (control) in minimal medium containing Neurobasal-A medium plus 1:1:2 ratio of penicillin/streptomycin:[l]{.smallcaps}-Glutamine:glucose at 1:100 dilution. After induction of oxidative stress by addition of H~2~O~2~, cells were incubated with 5 µM 2′-7′-dichlorofluorescein diacetate (CM-H2DCFDA, Life Technology, Grand Island, NY, USA) for 15 min at 37 °C, washed, and the ROS levels were measured by flow cytometry \[[@B71-ijms-16-13885]\]. 4.9. Western Blot Analysis for Nrf2 Protein ------------------------------------------- After treatment, whole cell lysates were prepared in RIPA lysis buffer containing protease inhibitors, subjected to SDS-PAGE on an 8% mini-gel and transferred to Immobilon™-P transfer membrane as previously described \[[@B33-ijms-16-13885]\]. The membrane was blocked and cut at 75 kDa marker. The upper part of membrane was incubated with 1 µg/mL anti-Nrf2 antibody (R&D Systems Tools for Cell Biology Research) \[[@B72-ijms-16-13885]\] and the bottom part of membrane was incubated with anti-α tubulin (Sigma) as a loading control at a 1:10,000 dilution for overnight in a 4 °C cold room. After three consecutive washes with TBS plus 0.1% Tween 20 (10 min for each), the membranes were incubated with 1:10,000 secondary antibody (anti-mouse IgG-HRP, Sigma) for 1 h at room temperature. Prestained SDS-PAGE protein standards (Bio-Rad, Hercules, CA, USA) were used to determine the size of detected proteins. Proteins were visualized by chemiluminescence with SuperSignal West Pico (Pierce, Grand Island, NY, USA) and exposed to X-ray film. 4.10. ERK1/2 Phosphorylation Assay ---------------------------------- Rat NPCs were plated in a 24-well plate and changed to serum-free medium for 4 h before the assay. The ERK1/2 phosphorylation assay was performed as previously described \[[@B43-ijms-16-13885]\]. Briefly, cells were washed and removed from the wells, boiled, and then subjected to electrophoresis using a 12% SDS-PAGE mini-gel, followed by transfer to an Immobilon™-FL membrane (Fisher, Loughborough, UK) for Western blotting. After blocking, the blot was probed with mouse anti-phospho-p44/42MAPK (Thr202/Tyr204) antibody and rabbit anti-p44/42 MAPK antibody (Cell Signaling Technology, Danvers, MA, USA) overnight on a rocking shaker in a 4 °C cold room. Then, the blot was incubated with the secondary antibodies, anti-mouse IRdye 680 RD and anti-rabbit IRdye 800 CW, for 1 h at room temperature. Finally, images were acquired using an Odyssey FC imaging system (Li-COR Biosciences, Lincoln, NE, USA) and quantified using the manufacturer's analysis program. The ERK1/2 phosphorylation was calculated as the ratio of normalized arbitrary units (a.u.) of phosphorylated ERK1/2 over total ERK1/2. 4.11. RNA Purification and Gene Expression Analysis --------------------------------------------------- Rat NPCs were grown in adherent cultures and treated under various conditions described above. Then the total RNAs were prepared using RNeasy RNA purification kit (Qiagen, Valencia, CA, USA) according to the manufacturer's instruction. RNA integrity was assessed with the Agilent 2100 Bioanalyze at the University of Michigan microarray facility. RNA concentration was quantified using Nanodrop (Thermo Scientific, Grand Island, NY, USA) and the samples were tested for absence of genomic DNA contamination. Gene expression of oxidative stress related genes were assessed using Oxidative Stress PCR SuperArray \[[@B73-ijms-16-13885]\]. RNA samples (80 ng each) were reverse-transcribed to cDNA using the SuperArray RT^2^ First Strand Kit (Qiagen, Valencia, CA, USA). cDNA samples were next subjected to Oxidative Stress PCR SuperArray. PCR amplifications (Applied Biosystems 7900HT, Thermo Scientific, Grand Island, NY, USA) were completed with PCR plates (Superarray Cat No. PAHS-031A) assaying for 84 neurotrophic factors and receptors and five houkeeping genes. Results were quantified using (2^−ΔΔCt^) method and expressed as a fold difference up/down with respect to the control condition. 4.12. Real-Time RT-qPCR ----------------------- The candidate genes that were either up- or down-regulated by DMF treatment were confirmed by real time-quantitative polymerase chain reaction (RT-qPCR) using Applied Biosystems 7500 (Thermo Scientific, Grand Island, NY, USA). Amplification consisted of 40 cycles (95 °C for 15 s and 60 °C for 1 min) with approximately 15 ng/μL cDNA, TaqMan Master Mix and specific primer pairs were purchased from Applied Biosystems (assay on demands primers). 4.13. Statistical Analysis -------------------------- Experimental data was analyzed using GraphPad Prism version 6.0 (GraphPad Software Inc., San Diego, CA, USA). The unpaired t-tests were performed for all experiments and values were expressed as mean ± SEM from at least three individual experiments. p-Values were considered significant at \* p \< 0.05 and highly significant at *\\*** p \< 0.01 or *\\\** p \< 0.001. ![Schematic model of DMF-induced neuroprotection in rat NPCs and differentiated neurons. Under oxidative stress (H~2~O~2~), DMF enhances Nrf2-mediated antioxidant response gene transcription through activation of ERK1/2 MAPK signaling pathway leading to neuroprotection in rNPCs and differentiated neurons.](ijms-16-13885-g007){#ijms-16-13885-f007} 5. Conclusions ============== DMF protected neural stem/progenitor cells and differentiated neurons from oxidative damage through regulation of Nrf2 involving the ERK1/2 MAPK pathway. We propose that under oxidative stress (H~2~O~2~), DMF enhances Nrf2-mediated antioxidant response gene transcription through activation of the ERK1/2 signaling pathway leading to neuroprotection in rat NPCs and differentiated neurons ([Figure 7](#ijms-16-13885-f007){ref-type="fig"}). The key Nrf2 target genes and genes involved in superoxide metabolism are involved in reducing oxidative stress by breaking down ROS, decreasing apoptosis and increasing survival. Investigating the anti-oxidative stress mechanisms may yield further insights into new targets for treatment of multiple sclerosis (MS). The authors thank Catherine Dowling and Kelly VanDerberg for their critical reading the manuscript and also thank our collaborator Benjamin Segal for the support. The work of Y.M.D., Qin W., Qi W., S.C., and S.L. were supported by NIH (NINDS) 1R01NS080821-01A1 and NIH (NIAID) UM1 AI110557; A.R., S.J.T., and J.M.C. were supported by VA Rehab R&D Project B7025-R; S.T. was supported by NIH (NIAID) training grant T32 AI007413. Qin Wang, Sophina Taitano, Sergei Chuikov, Arjun Rastogi, Joseph M. Corey, and Yang Mao-Draayer. participated in research design; Qin Wang, Sophina Taitano, Sergei Chuikov, Arjun Rastogi, Qi Wu, and Samuel J. Tuck conducted experiments and data analysis; Qin Wang, Arjun Rastogi, Joseph M. Corey, Steven K. Lundy, and Yang Mao-Draayer wrote or contributed to the writing of the manuscript. The authors declare no conflict of interest. MS : multiple sclerosis MAPK : mitogen-activated protein kinase ERK1/2 : extracellular-signal-regulated kinases 1/2 FGF : recombinant human basic fibroblast growth factor 146 aa EGF : epidermal growth factor RIPA : radio-immunoprecipitation assay rNPCs : rat neural stem/progenitor cells ROS : reactive oxygen pecies CNS : central nervous system FAE : Fumaric acid esters MMF : monomethyl fumarate DMF : dimethyl fumarate H~2~O~2~ : hydrogen peroxide Nrf2 : transcription factor nuclear factor-erythroid 2-related factor 2 EAE : experimental autoimmune encephalomyelitis MOG : myelin oligodendrocyte glycoprotein ARE/EpRE : antioxidant/electrophile response element IL-10 : interlukin-10 Keap1 : kelch-like ECH-associated protein 1 CEE : chick embryo extract PLGA : poly lactic-co-glycolic acid PLLA : poly-[l]{.smallcaps}-lactic acid PLL : poly-[l]{.smallcaps}-lysine SRM : self-renewal medium EGFP : enhanced green fluorescent protein CM-H2DCFDA : chloromethyl derivative of 2′,7′-dichlorofluorescein diacetate RT-qPCR : real time quantitative reverse transcription polymerase chain reaction GSH : glutathione HO-1 : heme oxygenase-1 DMSO : dimethyl sulfoxide FITC : fluorescein isothiocyanate DAPI : 4′,6-diamidino-2-phenylindole Gstp1 : glutathione S-transferase pi 1 Nqo1 : NAD(P)H dehydrogenase, quinone 1 Sod2 : superoxide dismutase 2 Srxn1 : sulfiredoxin 1 homolog Fth1 : ferritin 1 Ccl5 : chemokine (C--C motif) ligand 5	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1} =============== Odontogenic myxoma (OM) is a benign, locally invasive, aggressive, nonmetastasizing neoplasm of the jaw bones. It is the third most common odontogenic lesion with an annual incidence of \~0.07 new patients per million people \[[@B1], [@B2]\]. Although OM is a benign tumor, radical treatment is needed because the rate of local recurrence ranges from 10% - 33% \[[@B3]-[@B5]\]. Surgical resection with a minimum bone margin of 1 cm has been advocated \[[@B6]-[@B9]\], but emerging evidence suggests that a more conservative surgical approach will result in acceptable recurrence rates with less patient morbidity if follow-up can be over the long term \[[@B7], [@B10]-[@B12]\]. However, little has been reported about the actual long-term follow-up of patients with OM after conservative surgery. Here, we describe a mandibular odontogenic myxoma that was treated by conservative surgery and followed up for 100 months thereafter. 2. Case Report {#sec2} ============== 2.1. Patient Information {#sec2.1} ------------------------ A 56-year-old Japanese man was referred from a dental clinic for further examination of a radiolucent finding on the left side of his mandible in January 2010. The patient reported having no symptoms in his mouth including the left mandible area. His medical and dental history was noncontributory. An extraoral examination upon initial presentation revealed unremarkable findings and no complaints of paresthesia. An intraoral examination also confirmed the absence of redness and tender swelling of the left mandibular mucosa ([Figure 1](#fig1){ref-type="fig"}). Panoramic radiography revealed an extensive multilocular radiolucent area with imprecise borders and a "soap bubble appearance" ([Figure 2](#fig2){ref-type="fig"}). Computed tomography showed an approximately 39 × 19 × 11 mm tumor that extended to the roots of four teeth (\#33 - 36; Figures [3(a)](#fig3){ref-type="fig"} and [3(b)](#fig3){ref-type="fig"}). We considered that the odontogenic tumor was benign and an incisional biopsy was performed under local anesthesia. The histopathological findings revealed loosely arranged spindle-shaped cells in a myxoid fibrous stroma, indicating a clinical diagnosis of an odontogenic tumor. Segmental resection of the mandible was planned. The patient was given repeated and careful explanations about the high likelihood of recurrence, but he insisted upon a more conservative approach as he desired functional and cosmetic preservation. Conservative surgery then proceeded under general anesthesia after endodontic treatment of \#33 -- 36 was completed. The surgery consisted of extracting the second premolar from the left mandible, followed by total enucleation and vigorous curettage of the bone ([Figure 4(a)](#fig4){ref-type="fig"}). The surgical specimen ([Figure 4(b)](#fig4){ref-type="fig"}) revealed apparently benign, spindled-shaped cells in a loose and abundant mucoid stroma (Figures [5(a)](#fig5){ref-type="fig"} and [5(b)](#fig5){ref-type="fig"}). These findings confirmed the diagnosis of odontogenic myxoma. The immediate postoperative period and wound healing were uneventful. The patient underwent monthly clinical examinations for the first year thereafter, then every two months during the second year. Panoramic X-rays were obtained every three months for the first two years. Annual follow-up for eight years included panoramic X-rays and CT imaging (Figures [6(a)](#fig6){ref-type="fig"} and [6(b)](#fig6){ref-type="fig"}, respectively), which showed no clinical or radiological signs of recurrence. 3. Discussion {#sec3} ============= Odontogenic myxomas are very rare (\<10% all odontogenic tumors) benign tumors of the ectomesenchymal and/or mesenchymal origin \[[@B13], [@B14]\]. They are locally invasive and slow-growing, and their pathological characteristics in the tooth-bearing areas of the mandible and maxilla are well defined \[[@B13]-[@B17]\]. The radiographic features are described as those of a multilocular lesion with a "soap bubble" or "spider web" appearance \[[@B14]-[@B17]\]. Such lesions are often discovered incidentally during routine dental checkups, and about 60% of patients are in their second or third decade of life \[[@B4], [@B18]\]. Although the present patient was asymptomatic, panoramic radiography of the left mandible revealed an extensive radiolucent and multilocular area with imprecise borders that extended from the root of tooth \#33 to that of tooth \#36. The treatment strategy for OM remains controversial. Because OM is locally aggressive and it can potentially cause extensive bony destruction, the treatment of choice seems to be radical surgery such as segmental resection. Indeed, complete surgical removal with curettage and peripheral ostectomy alone seems insufficient since OM is not encapsulated and myxomatous tissue infiltrates adjacent bone \[[@B3], [@B6]-[@B9]\]. Thus, the only initial treatment option for extended OM in principle is surgical resection because of the high risk of recurrence which reportedly ranges from 10% to 33% \[[@B4], [@B5]\]. In addition, recurrence rates can reach 25% after simple enucleation and curettage alone \[[@B19], [@B20]\]. On the other hand, some reports described that the choice of current recommended therapy depends on the size of the lesion and its nature and behavior between curettage and radical excision \[[@B3], [@B7]\]. In the present case, CT had shown that the tumor included the \#33 - \#36 apex and extended to the inferior border of the mandible. Therefore, radical resection such as block resection for mandible was strongly recommended. However, the patient rejected this strategy because he desired functional and cosmetic preservation, and in fact, conservative treatment with enucleation, curettage, and marginal resection would confer several advantages over the radical approach. It is substantially less invasive, it can be surgically implemented via an intraoral approach, and it offers the possibility of preserved nerve function and aesthetics and a shorter stay in the hospital. Recent evidence suggests that a more conservative approach will result in acceptable recurrence rates with less morbidity to patients if long-term follow-up is provided \[[@B7], [@B10]-[@B12]\]. [Table 1](#tab1){ref-type="table"} describes 20 reports of 37 patients with that was treated by conservative surgery \[[@B4], [@B6], [@B7], [@B10], [@B11], [@B17], [@B18], [@B21]-[@B33]\]. The mean age of the patients was 27.0 ± 15.2, and 20 (54%) were female. The tumors in almost all of them were managed by enucleation and curettage and recurred in 7 (18.9%). One and six of the procedures with recurrence had undergone marginal resection and enucleation with curettage, respectively. The patients were followed up for a mean of 49.2 ± 42.9 (range, 2 -- 180) months. Including the present patient, only five have been followed up for over 100 months and the tumor recurred in one of them. The recurrence rate among all patients who were treated by conservative surgery was 19.0%. This rate is relatively lower than that previously described \[[@B19], [@B20]\]. Furthermore, our investigation of the literature indicated that the rate of recurrence decreases from 24.0% to 8.3% when the follow-up period exceeds 60 months. The main reason for recurrence is thought to be incomplete removal rather than the intrinsic biological behavior of the OM \[[@B34]\]. Although the tendency is towards a more conservative surgical approach for children and a more radical approach for adults, Kansy et al. does not support this management strategy because the recurrence rates between enucleation and segmental resection are similar \[[@B17]\]. Tumor size has recently been considered to be a factor in the choice of a radical or more conservative surgical approach \[[@B4], [@B10], [@B11]\]. Conservative surgical procedures such as enucleation and curettage are recommended when the diameter of OM is \<3 cm, whereas a radical approach such as segmental resection with immediate reconstruction is preferred when patients have larger tumors \[[@B11]\]. The conservative surgical recommendation is to enucleate a lesion with wide curettage of normal tissue or a generous amount of apparently intact tissue or even marginal resection of the mandible \[[@B12]\]. This approach has the advantage of preserving vital structures and maintaining oral function, and it could be applied to treat OM that recur after simple surgery \[[@B12]\]. However, one patient who developed recurrent OM 15 years after the initial procedure including tumor resection with the preservation of mandibular continuity has been described \[[@B17]\]. More radical surgery is inevitable for a large number of patients due to a tendency towards more extensive OM with significant destruction of key structures \[[@B17], [@B27]\]. Moreover, the rate of OM recurrence remains vague because few reports described long-term follow-up after conservative surgery. Thus, more evidence about long-term outcomes after conservative surgical treatment of OM is needed. 4. Conclusions {#sec4} ============== A conservative surgical approach comprising enucleation and curettage can be effective for OM management. Recurrence rates decreased from 24.0% to 8.3% among patients who were treated with conservative surgery and followed up for over 60 months. The risk of recurrence is likely to be considerable, and long-term follow-up is indispensable for the conservative management of OM. Conflicts of Interest ===================== The authors have no conflicts of interest to declare. ![Intraoral photography before surgery. The image shows no symptoms such as redness or swelling of mucosa in the mandible.](CRID2019-1634842.001){#fig1} ![Panoramic dental radiography before surgery. The image shows extensive multilocular radiolucent area with imprecise borders and "soap bubble appearance".](CRID2019-1634842.002){#fig2} ![Computed tomography (CT) image before surgery. The axial (a) and coronal (b) CT images show tumor infiltration of cortical bone extending to the inferior mandibular border.](CRID2019-1634842.003){#fig3} ![Surgical procedures and resected specimen. The total enucleation and wide curettage of the surrounding bone (a) and resected specimen (b).](CRID2019-1634842.004){#fig4} ![Pathophysiological findings of stained specimen. Hematoxylin and eosin stain ×100 (a) and ×400 (b) magnification.](CRID2019-1634842.005){#fig5} ![Images of the jaw at 96 months after surgery. The panoramic dental radiograph (a) and computed tomography image (b) show no signs of recurrence.](CRID2019-1634842.006){#fig6} ###### Clinical reports of mandibular odontogenic myxoma published after 1990. No. Author Year Age Sex Follow-up period (months) Treatment Recurrence Size (mm) ----- --------------------------------- ------ ----- ----- --------------------------- ----------- ------------ -------------- 1 Oliveira et al. \[[@B21]\] 2018 9 F 6 E/C None NA 2 Albanese et al. \[[@B22]\] 2012 25 F 6 E None 21.2 × 47.6 3 Mauro et al. \[[@B23]\] 2012 6 M 6 E/C None NA 4 Subramaniam et al. \[[@B7]\] 2016 16 --- 7 M None NA 5 Shivashankara et al. \[[@B24]\] 2017 13 M 12 E/N None 40 × 20 6 Subramaniam et al. \[[@B7]\] 2016 18 --- 12 E None NA 7 Miranda Rius et al. \[[@B25]\] 2013 55 M 12 E/C None 33 × 28 8 Hammad et al. \[[@B26]\] 2016 45 F 13 M None 50 × 30 9 Francisco et al. \[[@B27]\] 2017 9 F 14 E/C Recurrence NA 10 Francisco et al. \[[@B27]\] 2017 12 F 16 E/C Recurrence NA 11 Sumi et al. \[[@B28]\] 2000 48 M 22 E/C None 70 × 25 × 15 12 Mittal et al. \[[@B29]\] 2016 48 F 24 E/C Recurrence 25 × 20 13 Lin and Basile \[[@B30]\] 2010 25 F 24 E None NA 14 Lo Muzio et al. \[[@B4]\] 1996 21 M 24 E/C None NA 15 Lo Muzio et al. \[[@B4]\] 1996 28 M 24 E/C Recurrence NA 16 Bucci et al. \[[@B31]\] 1993 28 M 24 E/C None 43 × 40 17 Francisco et al. \[[@B27]\] 2017 7 F 26 E/C None NA 18 Francisco et al. \[[@B27]\] 2017 15 F 27 E/C Recurrence NA 19 Lo Muzio et al. \[[@B4]\] 1996 16 F 31 E/C None NA 20 Francisco et al. \[[@B27]\] 2017 30 F 34 E/C None NA 21 Boffano et al. \[[@B11]\] 2011 38 M 38 E/C None 25 22 Boffano et al. \[[@B11]\] 2011 42 F 40 E/C None 30 23 Boffano et al. \[[@B11]\] 2011 20 M 42 E/C None 20 24 Rajasekhar et al. \[[@B32]\] 2008 17 F 48 M None 70 × 30 25 Lo Muzio et al. \[[@B4]\] 1996 22 F 48 E/C Recurrence NA 26 Takahashi et al. \[[@B6]\] 2018 37 F 73 E/C None 40 × 19 × 12 27 Chaudhary et al. \[[@B33]\] 2015 25 F 84 E/C/M None NA 28 Li et al. \[[@B18]\] 2006 7 M 84 E/C None NA 29 Li et al. \[[@B18]\] 2006 32 M 84 E/C None NA 30 Lo Muzio et al. \[[@B4]\] 1996 65 F 84 E/C None NA 31 Francisco et al. \[[@B27]\] 2017 17 M 85 E/C None NA 32 Francisco et al. \[[@B27]\] 2017 11 F 98 E/C None NA 33 Present case 56 M 100 E/C None 39 × 19 × 11 34 Francisco et al. \[[@B27]\] 2017 27 F 117 E/C None NA 35 Kawase-Koga et al. \[[@B10]\] 2014 40 M 120 E/C None 40 × 30 × 15 36 Li et al. \[[@B18]\] 2006 37 M 132 C None NA 37 Kansy et al. \[[@B17]\] 2012 33 F 180 M Recurrence NA Abbreviations: M: male, F: female, E: enucleation, C: curettage, M: marginal resection, and NA: not available. [^1]: Academic Editor: Eugenio Maiorano	{ "pile_set_name": "PubMed Central" }
Introduction ============ Chronic insomnia is one of the most common sleep disorders, affecting one third of the population, and should receive greater recognition.[@B01],[@B02] Symptoms are considered clinically relevant if they occur at least three times/week and persist for 3 or more months despite adequate sleep opportunities, leading to sleep dissatisfaction.[@B03] Some insomnia patients underestimate total sleep time (TST). It is hypothesized that these individuals have higher cortical arousability and self-reported rates of insomnia, as well as other psychiatric symptoms. Other patients, however, might complain less but present objective sleep disturbance, being at higher risk of metabolic or cardiovascular outcomes.[@B04] From a mental health point of view, it is unlikely that insomnia, either chronic or recurrent, is the sole disorder, but is rather part of a complex multimorbidity. Untreated individuals are in a state of chronic sleep deprivation and have an increased risk of accidents,[@B05] work or social problems, and compromised health and quality of life.[@B01],[@B02] According to regulatory agencies, non-benzodiazepine GABA-A receptor agonists are a standard insomnia treatment, and prescription rates are on the rise.[@B01],[@B02],[@B06],[@B07] Zolpidem is the most popular non-benzodiazepine GABA-A receptor agonist. In therapeutic dosages, it reduces sleep latency and the frequency of nocturnal awakenings and increases TST and sleep efficiency. Pharmacokinetic parameters could differ for women and older adults, and have been associated with increased residual impairment and adverse events.[@B08] These findings have led the U.S. Food and Drug Administration to recommend lower initial dosages for these patients.[@B09] Nevertheless, mid- and long-term trials focusing on objective and subjective measures of effectiveness are scarce,[@B10] as are descriptions of complementary therapeutic schemes or dosing regimens.[@B11] The purpose of our study was to evaluate the mid-term safety and efficacy of a 5 mg sublingual dose of zolpidem for reducing sleep latency when administered at bedtime, and "as needed" following middle-of-the-night (MOTN) awakenings, compared to the traditional oral 10 mg dose. To achieve these goals, we designed a three-month randomized, controlled, and double-dummy trial, in which we tested the patients' clinical improvement, reported adverse events, and changes in subjective and objective sleep measures. Methods ======= Sampling and design ------------------- The prospective participants were adults (20-64 years) who reported nocturnal awakenings predominantly before 3:00 a.m. and who had not used psychoactive drugs in the 30 preceding days. Referrals and online subscriptions to the Sleep Institute's website were channels for recruitment. Exclusion criteria were: 1) a history of major neurological, psychiatric, or other medical disorders; 2) a history of substance abuse/dependence; 3) a history of daily alcohol use; 4) pregnancy, lactation, or intending to conceive; 5) an apnea and hypopnea index (AHI) \> 10 events/hour; 6) a periodic limb movement index \> 15 events/hour; and 7) blood alterations suggestive of an undiagnosed medical condition. The protocol included five onsite visits. The first was for screening, when participants gave written informed consent and were examined. The second was for randomization, scheduled the morning after polysomnography (PSG). The third, fourth, and fifth were follow-up visits. Performance was measured with the Psychomotor Vigilance Test (PVT) which was self-applied twice during the day: 30 minutes and 2 hours after awakening. After lab and tests results, the patients were reassessed for continued inclusion. Eligible participants were assigned (1:1) to receive active doses of zolpidem either as sublingual 5 mg or oral 10 mg tablets. Because sublingual and oral tablets cannot be made to look and taste alike, a double-dummy technique was used with matching placebos of both formulations, which assured participant and investigator blinding. The oral group received kits consisting of oral zolpidem 10 mg and sublingual placebo, which was used at bedtime and "as-needed." The sublingual group received sublingual zolpidem 5 mg and an oral placebo at bedtime, and sublingual zolpidem 5 mg "as-needed." Placebos were indistinguishable in size, shape, color, and taste. Visits 3, 4, and 5 were follow-ups scheduled at weeks 2, 6, and 13, when the kits were dispensed, and unused medication was collected and counted. The bottles provided to participants contained enough tablets for the period between visits. The sponsor provided boxed kits numbered according to a randomization list with sealed envelopes identifying treatment. Procedures and instruments -------------------------- All visits included a medical evaluation, application of questionnaires, and the distribution and collection of sleep diaries. PSG, the PVT, electrocardiogram, and blood testing were performed twice, once before randomization and again during the final visit. Patients were instructed to take a maximum of one tablet from each bottle per night (bedtime and as needed) and record it in their sleep diaries. Two sleep specialists performed the evaluations. At screening, the participants' history of insomnia, medication use, comorbidities, and related symptomatology were investigated, and adverse events and clinical improvement were assessed during follow-up. Heart rate, blood pressure, and body mass index were assessed in physical examinations. A 12-lead digital electrocardiogram system (Wincardio System, Micromed, Brasília, Brazil) was used. Blood tests included a complete cell count and assessed liver, renal, and thyroid functioning. ### Sleep diaries Sleep diaries were used to collect data on medication use and sleep schedules. Based on their data, the following were estimated: 1) medication time; 2) bedtime; 3) sleep duration; 4) time of rescue dose ("as-needed" tablet); 5) wake time; 6) rise time; 7) number of nights in which the regular dose was taken ; 8) number of nights in which a rescue dose was taken; 9) number of nights with sleep-onset latency \> 30 minutes; 10) number of nights that MOTN awakenings occurred; 11) perceived TST; 12) calculated TST (sleep time minus wake time); 13) sleep efficiency (TST/time-in-bed); 14) alcohol consumption; and 15) mood and sleep quality ratings.[@B12] ### Questionnaires The Pittsburgh Sleep Quality Index evaluates sleep schedule, frequency of insomnia/nighttime symptoms, and level of sleep-related problems in the last month. The global score varies from 0-21, and scores \> 5 indicate poor sleep quality.[@B12],[@B13] The Insomnia Severity Index includes seven Likert scales (scored 0-4) that evaluate symptoms and dissatisfaction/distress in the past two weeks. Scores \> 7 (up to 28) indicate clinically meaningful insomnia. A reduction of ≥ 8 points suggests at least moderate improvement after treatment.[@B12],[@B14] The Karolinska Sleepiness Scale measures alertness in the present moment on a scale from 1-9 ("extremely alert" to "very sleepy/great effort staying awake/fighting sleep"). Electrophysiological signs of sleepiness, such as slow eye movement and alpha-power density, correlate with ratings ≥ 7.[@B15] ### Objective sleep, alertness, and performance measures PSG was performed with a digital system (EMBLA^®^S7000, Embla-Systems, Broomfield, CO, USA). Sleep stages, respiratory events, and leg movements were scored according to American Academy of Sleep Medicine criteria.[@B16] The selected variables were TST, sleep-onset latency, wake after sleep onset, sleep efficiency, arousal index, percentage of each sleep stage, rapid eye movement (REM) sleep latency, AHI, minimum oxyhemoglobin saturation (SaO~2~), and periodic limb movement index. The PVT measures the response to visual stimuli (PVT-192, CWE, Inc., Ardmore, PA, USA). During the 10-minute test, patients are instructed to press a button when a light randomly appears on the screen every few seconds. Data are collected on sleepiness, number of errors, hits, lapses, and reaction time.[@B17] Outcome measures ---------------- Treatment compliance was defined as the use of bedtime medication at least 75% of the nights, which was verified using sleep diary data and tablet counts during follow-up visits. Clinicians rated adverse events by their level of association with the study medication (unrelated, possibly related, or probably/certainly related), and decided whether to continue or discontinue treatment. Means and frequencies from the diary data were computed weekly and for the entire treatment period. For delta values, means from the third and final phase of the treatment period (between weeks 6 and 13) were subtracted from baseline means, or from the first and second phase means (i.e., day 1 until week 6). Ordinal responses to the questionnaires were dichotomized to estimate changes in sleep satisfaction. Clinical improvement referred to positive changes in sleep and functioning, as well as symptom remission as rated by physicians during follow-up visits. Partial or no improvement included persistent symptomatology and/or sleep-related problems. Statistical analysis -------------------- Sample size estimation (30/group) aimed at finding differences in the proportion of patients with decreased MOTN awakenings and sleep onset latency (20% in oral 10 mg and 70% in sublingual 5 mg) with 80% power and a 5% error rate. Attrition was expected to be 10%. Descriptive statistics are presented as absolute frequency and percentage for categorical variables and mean ± standard deviation for continuous variables. Associations between the categorical variables were measured in cross-tabulations with chi-square or Fisher's exact tests. Analysis of variance/covariance was used to compare group means and were performed through general linear models. Confidence intervals were adjusted with Bonferroni correction, and effect size (Eta^2^) and observed power were used to evaluate type I and II errors. Mauchly's test of sphericity, with the Greenhouse-Geisse correction for degrees of freedom, was used for repeated measures without assuming equal variance. The frequencies of medication use and nights in which a rescue dose was taken were used as covariates in general linear models for PSG and PVT parameters, controlling for possible residual effects. Intention-to-treat analysis was performed and every randomized participant was included in the analyses. Missing data were replaced by using the last observation carried forward method, in which the last available measurement for each individual prior to discontinuation from the study was retained in the analysis.[@B18] Sleep diary data analysis included use of general estimating equations, which do not assume the normality of the dependent measure and allow the selection of different types of distribution probabilities. Since they also allow all recorded daily values to be examined, they were used to evaluate "daily" variations in perceived TST across all treatment phases and to control for possible interaction effects with bedtime, rescue medication, baseline insomnia severity, and age. Ethics statement ---------------- The study protocol was approved by the Universidade Federal de São Paulo research ethics committee (protocol 184.648), and the study was registered in ClinicalTrials.gov (NCT01896336). Results ======= The mean ± standard deviation number of days between screening and randomization was 6±4. Follow-ups at weeks 2, 6, and 13 occurred 15±2, 31±4, and 46±4 days after the previous visit, respectively, totaling 92±5 days of treatment and 98±5 days in the protocol (between March and November 2013). Study population ---------------- Of the 85 screened individuals, 18 (21%) were not randomized due to reasons presented in the study flow diagram ([Figure 1](#f01){ref-type="fig"}). The 67 participants were middle-aged (48±10 years), and most were non-obese women (n=53; 79%) whose average insomnia duration was 13±9 years and who took at least one dose of study medication. All participants attended the week 2 follow-up and were evaluated by clinicians, but not all had completed their sleep diaries and questionnaires or had PSG or PVT measures collected. Participants who completed the protocol (n=46; 69%) and those who did not had similar baseline characteristics. However, those who did not complete the protocol had more adverse events and fewer signs of improvement. The treatment groups also presented similar baseline characteristics ([Table 1](#t01){ref-type="table"}). Adverse events -------------- Seventeen of the 21 participants who did not complete the protocol reported adverse events (25% of the total group), seven did not wish to continue (10%), and 10 were discontinued on medical advice (15%). Of those who did not wish to continue, three were from the sublingual group and four were from the oral group, mostly women with mild-to-moderate headaches, dizziness and/or gastrointestinal symptoms who had lower compliance rates and concurrent life events. Of the 10 who discontinued based on medical advice, four presented moderate to severe events associated with emotional instability or gastrointestinal symptoms, but recovered within a few days of discontinuation: three of these were in the oral group and one was in the sublingual group. The other six presented persistent mild-to-moderate adverse events or were in need of medication that could interfere with the study's outcomes. The difference in discontinuation frequency between treatment groups was not significant ([Figure 1](#f01){ref-type="fig"}). The investigators registered 152 adverse events, 58 (38%) unrelated to the study medication, 69 (45%) possibly related, and 25 (16%) probably/certainly related. Headache, sleepiness, and dizziness were the most likely events to be treatment-related. The proportion of patients with adverse events did not differ between groups, but the number of treatment-related adverse events leading to discontinuation was higher in the oral group than the sublingual group (29%, n=24 vs. 13%, n=9; two-sided p = 0.02). See [Table 2](#t02){ref-type="table"} for further information. Efficacy analyses ----------------- ### Inter- and intragroup sleep diary variance before and after treatment Both treatments significantly decreased the number of nights per week with MOTN awakenings, on average by -3.1±2.3 days. Before treatment, the sublingual group reported MOTN awakenings an average of 75% (95%CI 62-88) of the nights each week, which was reduced to 36% (20-53) in the third phase of the treatment period. The oral group reported MOTN awakenings an average of 87% (71-100) of the nights before treatment, which was reduced to 43% (24-62) in the third phase of the treatment period. Accordingly, a mean increase of 1.5 hours perceived TST was reported by both groups, with a proportional increase in sleep efficiency ([Table 3](#t03){ref-type="table"}). Sleep-onset latency was the only variable from the sleep diary with an interaction effect and an intergroup difference. There was a greater decrease in sleep-onset latency in the sublingual group than the oral group, even though its baseline values were higher (78 vs. 51; p = 0.03; effect size = 0.10; observed power = 0.61) ([Table 3](#t03){ref-type="table"}). When comparing group means from the first and second phases of treatment vs. those from the third phase, no intergroup differences were detected in the percentage of nights in which a rescue dose was taken or the number of "as needed" tablets used per week of treatment. Complementary and detailed analyses of sleep diary data are presented in [Table 4](#t04){ref-type="table"}. ### Changes in insomnia severity and sleep quality indexes No significant differences were found in the general linear model between mean questionnaire scores at baseline and the third phase. The mean reduction in Pittsburgh Sleep Quality Index scores was modest but significant, both in the sublingual (mean difference: -3.4±4.1) and the oral group (-2.6±3.4). However, the changes in Insomnia Severity Index scores were more marked, -11.3±5.5 and -10.4±6.1, respectively ([Table 3](#t03){ref-type="table"}). In the per-protocol analysis, including only individuals who fully completed the trial, the frequency of sleep dissatisfaction according to the Insomnia Severity Index at week 13 was significantly lower in the sublingual group (16 vs. 43%; chi-square: p = 0.04). ### Objective sleep and signs of residual effects Findings for the PSG variables were similar to and corroborated those from the sleep diaries. For PSG TST, both groups had a mean increase of 30 minutes from baseline ([Table 3](#t03){ref-type="table"}). Differences in wake after sleep onset were not significant. As seen by the large standard deviations, individual changes varied consistently, particularly in the sublingual group (mean difference: -57±226) and in contrast with the oral group (-26±65). Sleep-onset latency was the only statistically significant factor: the sublingual group decreased by -14±42 minutes, while the oral group increased by 10±29 minutes (p = 0.03). The findings were similar after controlling for the percentage of rescue dose nights. Additionally, use of a rescue dose was found to have an interaction effect with changes in TST, sleep efficiency, REM latency, and the respiratory-related variables ([Table 3](#t03){ref-type="table"}). PVT scores showed minor signs of residual effects. Both groups had a slight increase in sleepiness 30 minutes after awakening, which reversed after 2 hours. A small group difference was observed in the number of lapses after 30 minutes, which also reversed after 2 hours. Sleepiness ratings in the Karolinska Sleepiness Scale corroborated the PVT results ([Table 3](#t03){ref-type="table"}). We further compared PVT variables between patients who had taken (n=8) as needed medication or not (n=38) on the night of PSG and found no differences in sleepiness (p = 0.47) or lapses (p = 0.37) 30 minutes after awakening. ### Intragroup sleep diary variability Patients handed in a total of 5,053 sleep diaries: 2,761 (55%) from the sublingual group, 2,292 (45%) from the oral group, and 578 (11%) from participants of both groups who discontinued the treatment protocol. All available data are described below. Of all the reported nights, 269 (5%) were from the baseline period (6±4 days), 1,027 (20%) from baseline until the week 2 visit (after 15±2 days), 1,657 (33%) from week 2 to week 6 (after 31±4 days), and 2,100 (42%) from week 6 until the final visit (after 46±4 days). Bedtime (11.1 p.m. ± 1.2h) and "as needed" (1.9 a.m. ± 2.8h) medication were reported for 96% and 12% of the nights, respectively. The total number of nights with reported "as needed" rescue doses were higher in the sublingual group than the oral group (n=366 nights \[14%\] vs. n=193 nights \[9%\]; p \< 0.05). [Table 4](#t04){ref-type="table"} presents the number of sleep diaries collected at each visit and the valid percentages of complete data reported for each variable, such as the percentage of nights with sleep-onset latency \> 30 minutes (39%), MOTN awakenings (45%), TST \> 6 hours (59%), and ratings of good mood and having had a good night's sleep after awakening (58%). [Figure 2](#f02){ref-type="fig"} compares histograms of sleep diary data throughout the protocol period for both treatment groups, including nightly variations in the number of patients reporting MOTN awakenings or not and whether they took a rescue dose or not. The pattern of rescue dose usage tended to decrease between the first and third phases of treatment in both groups, suggesting no dose escalation or abuse. According to repeated analysis of the sleep diary data with general estimating equations, controlling for bedtime, the use of rescue doses, age, and baseline insomnia severity, there was no difference between groups in baseline perceived TST. The TST increase in both groups differed somewhat over the treatment phases, there was an interaction effect (Wald χ^2^: 94.1; df: 7; p \< 0.001). In the oral group, there was a TST gain of 1.3 hours in the first phase of treatment, which was sustained until week 13. In the sublingual group, there was a TST gain of 1.2 hours in the first phase, followed by a mean, progressive, and cumulative increase of 0.2 hour (12 minutes) between the first and second phases, as well as between the second and third phases. MOTN awakenings that led to a rescue dose of medication were associated with a small decrease in perceived TST in the oral group (B: -0.11; Wald χ^2^: 11.3; degrees of freedom: 1; p = 0.001) but not in the sublingual group (B: -0.01; Wald χ^2^: 0.1; degrees of freedom: 1; p = 0.77). Bedtime (B: -0.03; Wald χ^2^: 23.6; degrees of freedom: 1; p \< 0.001), but not age or baseline insomnia severity, also exerted a small independent effect on perceived TST. There was clinical improvement in 37 (55%) participants after completing the protocol: 23 (68%) in the sublingual group and 14 (42%) in the oral group (p = 0.05). Most of these participants (n=23) improved after week 6. Additionally, 11 (16%) participants presented early improvement followed by symptom recurrence, and there was no clinical improvement in 19 (28%) participants. There was no symptom remission in 30 of the randomized participants by the end of the study. Discussion ========== Our main objective was to describe the safety and efficacy of different dosages and treatment schemes of zolpidem in a 3-month trial. Patients from both groups presented no abuse-related behavior and no adverse events severe enough to discourage either drug formulation, which agrees with previous findings.[@B08],[@B09],[@B19]-[@B21] Both treatments increased TST and sleep efficiency and decreased wake after sleep onset. Regarding sleep-onset latency, the sublingual version had faster absorption and distribution rates, reaching higher concentrations and inducing sleep more rapidly.[@B22]-[@B26] However, an important factor that we failed to control for was the impact of food/ingestion on pharmacokinetics.[@B27] Greenblatt et al. showed that taking zolpidem on a full stomach decreases the maximum concentration and increases the chance of residual sedation. The influence of gender is also significant.[@B08],[@B28]-[@B30] Our findings suggested a higher dropout rate among men, although we did not have a representative number of men, and thus could not detect whether this was an interaction effect. In women, especially older women, perhaps 10 mg was a high dose, as has been stated by the U.S. Food and Drug Administration. Moreover, a 5 mg dose may not have been effective in men. Taking rescue doses might have led to small changes in sleep architecture, such as changing REM latency or slow-wave sleep, or to changes in sleep quantity and quality, increasing TST and sleep efficiency. These results are supported by the findings of Roth et al.,[@B31] who found a dose-dependent effect on the same sleep parameters with even lower doses. These authors reported that improved sleep architecture led to better subjective outcomes, and our findings point in the same direction. It should be mentioned that our use of the last observation carried forward method to deal with missing values or intention-to-treat analyses was a conservative measure. It could have underestimated associations, not only with PSG results, but with PVT scores as well. Nevertheless, the participants' performance and sleepiness results did not suggest unmanageable residual effects. Studies using different methods, mostly in healthy individuals, have also found no signs of relevant residual sedation, though patients should always be advised about the risks of activities that require alertness early in the morning.[@B08],[@B21],[@B24],[@B26],[@B31],[@B32] Regarding clinical improvement, even though we performed no inter-rater agreement tests between clinicians, they were trained and experienced, as well as blinded. About half of the patients ended treatment with clinically significant symptom remission. Of those whose symptoms recurred, most were in the oral 10 mg group, which was also reflected in sleep dissatisfaction ratings in the questionnaires. The majority of those who had no meaningful improvement had discontinued treatment due to adverse events. A randomized trial by Morin et al.[@B20] evaluated cognitive behavioral therapy (CBT) for insomnia, applied either alone or associated with 10 mg oral zolpidem. After 6 weeks, the CBT-only group remained in the same treatment protocol. Patients in the combined therapy group either continued in the same treatment protocol or switched to CBT alone. After 12 weeks, symptom remission was higher among patients from the combined therapy group that switched to CBT alone than in those who remained in combined therapy (68 vs. 42%). We observed the same success rates in our treatment groups: 42% in the oral group (n=14) and 68% in the sublingual group (n=23) after the same treatment time, only without CBT. It is possible that the availability of rescue medication gave some patients a sense of control that contributed to lower sleep anxiety and MOTN awakenings. Perlis et al.[@B11] suggested something similar in a trial testing intermittent zolpidem doses with and without placebo. The present study has certain limitations. The heterogeneity and the night-to-night variability in sleep patterns of insomnia patients is hard to address and has been heavily investigated.[@B02],[@B04],[@B33],[@B34] Our sample included patients with different durations of insomnia and different histories of psychoactive drug use, and we did not control for possible exposure to previous behavioral interventions, although none of the participants reported previous CBT. Moreover, there was a higher proportion of women, which could increase the probability of slower pharmacokinetics, and we did not control for menopausal status. It is difficult to recruit patients with untreated insomnia, which contributed to our small sample size. We sought to address this limitation by applying general estimating equation models. Our study is one of the few to have used more robust statistical procedures that can adequately handle a large number of repeated measures, such as the data from the sleep diaries. With general estimating equations, we were able to identify "daily" predictors of changes in perceived TST throughout treatment and the "size" of those changes. In summary, and in agreement with recommended guidelines, our findings reaffirm that initial lower doses of zolpidem can achieve better overall treatment outcomes while lowering the occurrence of adverse events. For some patients, higher dosing regimens may be needed, for others, having rescue medication on hand might be helpful. In a meta-analysis on the effectiveness of non-benzodiazepine GABA-A receptor agonists, Huedo-Medina et al.[@B07] found that placebo response accounts for, on average, about half of the drug response. Similarly, according to findings by Perlis et al.,[@B11] alternating nights of drug-therapy and placebo, or stipulating the number of nights per week that a rescue dose can be taken, or other innovative strategies, are potential avenues for clinical and behavioral research and interventions among insomnia sufferers. In conclusion, both sublingual zolpidem 5 mg (with "as needed" medication) and oral zolpidem 10 mg showed comparable safety and efficacy. Both treatments reduced the number of nights with MOTN awakenings and increased perceived TST. Clinical improvement occurred in 55% of the participants, of whom 68% received the sublingual formulation. Sublingual zolpidem was more rapidly absorbed, induced sleep faster, and resulted in a higher frequency of sleep satisfaction. Disclosure ========== This study was conducted at Centro de Desenvolvimento de Estudos Clínicos Brasil and the Sleep Research Center of the Associação Fundo de Incentivo a Pesquisa (AFIP) and the Universidade Federal de São Paulo, and it was sponsored by EMS Pharmaceutical. EMS had no role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; preparation or approval of the manuscript; or the decision to submit the manuscript for publication. All authors have filled out a disclosure form: LSC has received honorariums and consults for EMS; CFS was a clinician involved in evaluating participants; ST presides over AFIP, which has received a grant from EMS for the present study; DP received a grant from EMS for this study and has received honorariums as a speaker from Sanofi-Aventis and Libbs Pharmaceuticals, for developing and delivering educational presentations and other expenses beyond this study. The authors report no other conflicts of interest. The authors would like to acknowledge the work of Magda Bignotto, study administrator and leader of the study's monitors and coordinators: Graciela Santos (research analyst), Alexandre Ito (pharmacist), and Merielen CPP Silva (research assistant). The authors would also like to thank Dr. Luciano Rotella and Dr. Erika Tremptow, clinicians involved in patient evaluation and screening, as well as Dr. Altay Alves Lino de Sousa, professor of statistics (psychobiology), for his advice. How to cite this article: Castro LS, Otuyama LJ, Fumo-dos-Santos C, Tufik S, Poyares D. Sublingual and oral zolpidem for insomnia disorder: a 3-month randomized trial. Braz J Psychiatry. 2020;42:175-184. <http://dx.doi.org/10.1590/1516-4446-2018-0389> ![Study flow diagram outlining subject selection and study design. The safety population included all randomized participants who took at least one dose of the study medication. The efficacy population included those who provided any information on treatment effect. Fisher's exact test: no group differences in discontinuation frequency (24 vs. 39%; chi-square = 2.08; two-sided p = 0.18). ECG = electrocardiogram; PSG = polysomnography.](bjp-42-02-175-g001){#f01} ![Histogram of sleep diary data: number of daily reports during the treatment period stacked according to middle-of-the-night awakenings with or without the use of "as needed" tablets. MOTN: middle-of-the-night awakenings.](bjp-42-02-175-g002){#f02} ###### Patient sociodemographic and clinical characteristics at baseline Sublingual zolpidem 5 mg (n=34) Oral zolpidem 10 mg (n=33) Total (n=67) Discontinued (n=21) ---------------------------------------------------------- --------------------------------- -------------------------------------------- -------------- --------------------- Age, mean (SD), years 46 (11) 49 (10) 48 (10) 48 (9) BMI, mean (SD), kg/m^2^ 26 (3) 26 (4) 26 (4) 26 (4) Women 27 (79) 26 (79) 53 (79) 16 (76) Marital status Married 15 (44) 19 (58) 34 (51) 10 (48) Single 11 (32) 6 (18) 17 (25) 5 (24) Divorced/widowed 8 (24) 8 (24) 16 (24) 6 (28) Education level Primary and secondary 22 (65) 21 (64) 43 (64) 12 (57) Higher 12 (35) 12 (36) 24 (36) 9 (43) Smoking status Never smoked 31 (91) 30 (91) 61 (91) 17 (81) Current or former smoker 3 (9) 3 (9) 6 (9) 4 (19) Clinical history Treated for insomnia 13 (38) 13 (41) 26 (39) 8 (38) Psychoactive drug use 3 (9) 9 (27)[\](#TFN04t01){ref-type="table-fn"} 12 (18) 4 (19) Headache 15 (44) 16 (49) 31 (46) 7 (33) Muscle pain 10 (29) 5 (15) 15 (22) 5 (24) Gastrointestinal symptoms 10 (29) 9 (27) 19 (28) 4 (19) Type of symptoms DIS, DMS or EMA 14 (41) 13 (39) 27 (40) 10 (48) Two+ symptoms 20 (59) 20 (61) 40 (60) 11 (52) AHI, mean (SD) events/hour 2 (2) 4 (6) 3 (4) 3 (3) SaO~2~min, mean (SD), % 91 (3) 90 (4) 90 (3) 90 (3) Duration of insomnia, mean (SD), years 12 (9) 15 (10) 13 (9) 14 (8) ISI score, mean (SD)[†](#TFN05t01){ref-type="table-fn"} 17 (4) 18 (5) 17 (5) 18 (5) PSQI score, mean (SD)[‡](#TFN06t01){ref-type="table-fn"} 11 (3) 12 (3) 12 (3) 12 (3) Data presented as n (%), unless otherwise specified. AHI = apnea and hypopnea index; BMI = body mass index; DIS = difficulty initiating sleep; DMS = difficulty maintaining sleep; EMA = early-morning awakening; ISI = Insomnia Severity Index; PSQI = Pittsburgh Sleep Quality Index; SaO~2~min = minimum oxyhemoglobin saturation; SD = standard deviation. Statistical analyses were performed with two independent binary variables (sublingual zolpidem 5 mg vs. oral zolpidem 10 mg, and discontinued protocol vs. completed protocol). Statistically significant at p \< 0.05. ISI scores range from 0-28 (higher scores indicate more severe insomnia; clinical relevance \> 7). PSQI scores range from 0-21 (higher scores indicate poorer sleep quality; clinical relevance \> 5). ###### Number of reported adverse events in association with treatment discontinuation Adverse event types Sublingual zolpidem 5 mg 70 (31) Oral zolpidem 10 mg 82 (30) Total 152 (61) ---------------------------- ---------------------------------- ----------------------------- ---------------- -------- ---------- --------- Central nervous system 23 (16) 7 (3) 19 (16) 16 (7) 42 (32) 23 (10) Respiratory 15 (11) 0 (0) 15 (11) 0 (0) 30 (22) 0 (0) Gastrointestinal 7 (7) 2 (1) 14 (9) 3 (3) 21 (16) 5 (4) Neuromuscular and skeletal 10 (10) 0 (0) 4 (4) 2 (2) 14 (14) 2 (2) Cardiovascular 2 (2) 0 (0) 1 (1) 1 (1) 3 (3) 1 (1) Other 4 (4) 0 (0) 5 (3) 2 (2) 9 (7) 2 (2) Total 61 (27) 9 (4) 58 (22) 24 (8) 119 (49) 33 (12) Data presented as number of events (number of patients). Continued = continued in the protocol; Discontinued = discontinued participation in the study under medical advice. In the Total column, the numbers in parentheses may be higher than the total number of participants, since participants could report more than one adverse event during the study. ###### Inter- and intragroup variance in repeated subjective and objective measures Sublingual zolpidem 5 mg (n=34) Oral zolpidem 10 mg (n=33) Within subjects (p-values) --------------------------------------------------------------- --------------------------------- ---------------------------- ---------------------------- --------------- -------------- ----------- Sleep diaries Sleep onset latency, minutes 78 (57-99) 25 (17-34) 51 (27-75) 31 (20-41) \< 0.001* 0.031 Nights with MOTN awakenings, % 75 (62-88) 36 (20-53) 87 (71-100) 43 (24-62) \< 0.001 0.707 Rescue dose nights[\](#TFN05t03){ref-type="table-fn"}, % 22 (11-33) 15 (5-26) 20 (7-33) 10 (-3-22) 0.223 0.799 Total sleep time, hours 4.8 (4.3-5.3) 6.4 (5.9-6.9) 4.5 (3.9-5.0) 5.7 (5.2-6.3) \< 0.001* 0.390 Sleep efficiency, % 62 (55-70) 81 (75-86) 60 (52-69) 76 (70-83) \< 0.001 0.096 Questionnaire scores ISI 17 (15-18) 6 (4-8) 18 (16-19) 8 (6-10) \< 0.001 0.470 PSQI 11 (10-12) 8 (7-10) 12 (11-13) 10 (9-11) \< 0.001 0.608 KSS, 30 minutes[†](#TFN06t03){ref-type="table-fn"} 5 (4-5) 5 (4-6) 5 (4-6) 5 (5-6) 0.088 0.605 KSS, 2 hours[†](#TFN06t03){ref-type="table-fn"} 4 (3-4) 3 (3-4) 4 (3-4) 4 (3-4) 0.178 0.233 PSG parameters Sleep-onset latency 27 (17-38) 17 (9-24) 18 (8-28) 24 (16-31) 0.502 0.031 Wake after sleep onset 89 (39-139) 46 (34-58) 67 (17-118) 52 (39-64) 0.102 0.440 Total sleep time 326 (306-347) 352 (334-369) 333 (312-353) 350 (332-368) 0.001 0.518 Sleep efficiency 80 (75-84) 83 (79-87) 80 (75-84) 82 (78-87) 0.048 0.824 Arousal index 12 (10-15) 13 (11-16) 16 (13-18) 14 (12-17) 0.828 0.165 REM sleep 20 (18-22) 19 (17-20) 19 (17-21) 19 (17-20) 0.440 0.504 Slow wave sleep 22 (19-25) 24 (21-27) 22 (19-25) 23 (21-26) 0.057 0.708 AHI 2.1 (0.6-3.6) 3.1 (1.7-4.5) 3.6 (2.1-5.1) 3.9 (2.5-5.3) 0.183 0.448 SaO~2~min 91 (89-92) 89 (88-90) 89 (88-91) 88 (87-90) 0.002 0.591 PVT parameters[†](#TFN06t03){ref-type="table-fn"} Sleepiness, 30 minutes 5 (4-6) 6 (5-7) 5 (4-6) 6 (5-7) 0.043 0.869 Sleepiness, 2 hours 3 (2-4) 3 (2-4) 4 (4-5) 4 (3-5) 0.171 0.328 Total hits, 30 minutes 92 (91-94) 90 (85-95) 93 (92-95) 91 (85-96) 0.169 0.885 Total hits, 2 hours 90 (86-94) 94 (90-98) 95 (91-99) 92 (88-96) 0.850 0.097 Lapses, 30 minutes 16 (10-22) 9 (5-12) 5 (0-11) 7 (3-11) 0.151 0.054 Lapses, 2 hours 7 (3-12) 4 (2-7) 4 (0-9) 4 (2-7) 0.279 0.363 Reaction time, 30 minutes 412 (359-464) 368 (320-416) 340 (287-394) 346 (297-395) 0.285 0.162 Reaction time, 2 hours 336 (301-372) 309 (273-346) 300 (263-336) 313 (276-350) 0.625 0.151 Data presented as mean (95% confidence interval), unless otherwise specified. 95%CI = 95% confidence interval; AHI = apnea-and-hypopnea index; ISI = Insomnia Severity Index; KSS = Karolinska Sleepiness Scale; MOTN awakenings = middle-of-the-night awakenings; PSG = polysomnography; PSQI = Pittsburgh Sleep Quality Index; PVT = Psychomotor Vigilance Test; REM = rapid-eye-movement; SaO~2~min = minimum oxyhemoglobin saturation. Last observation carried forward (LOCF) = Missing values for discontinued participants were replaced by baseline or the last observed value. Within-subject statistics are presented for treatment effect alone (Factor) and interaction effect with the treatment group (x Group). Bold font indicates statistical significance. At baseline, rescue dose nights refers to the percentage of nights a rescue dose was taken until the visit at week 6, i.e. half of the treatment period. It was used as a covariate in the PSG and PVT models to control for residual effects. Subjective sleepiness and psychomotor performance were measured twice, 30 minutes and 2 hours after awakening from PSG. ###### Means and frequencies of data reported in all sleep diaries Sublingual zolpidem 5 mg Oral zolpidem 10 mg ------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- Total diaries, n 139 507 912 1,203 130 520 745 897 Weekends, n (%) 48 (34) 220 (43) 395 (43) 510 (42) 42 (32) 231 (44) 324 (43) 377 (42) Nights medication taken \- 97 93 96 \- 84 95 99 Rescue dose nights \- 20 13 13 \- 14 7 9 Perceived TST ≥ 6 hours 33 65[\](#TFN04t04){ref-type="table-fn"} 69[\](#TFN04t04){ref-type="table-fn"} 68[\](#TFN04t04){ref-type="table-fn"} 23 53[\](#TFN04t04){ref-type="table-fn"} 49[\](#TFN04t04){ref-type="table-fn"} 54[\](#TFN04t04){ref-type="table-fn"} SOL ≥ 30 minutes 72 39 33[\](#TFN04t04){ref-type="table-fn"} 35[\](#TFN04t04){ref-type="table-fn"} 65 41 42[\](#TFN04t04){ref-type="table-fn"} 42[\](#TFN04t04){ref-type="table-fn"} MOTN awakenings 67[\](#TFN04t04){ref-type="table-fn"} 52[†](#TFN05t04){ref-type="table-fn"} 41[\](#TFN04t04){ref-type="table-fn"} 33[\](#TFN04t04){ref-type="table-fn"} 89[\](#TFN04t04){ref-type="table-fn"} 58[†](#TFN05t04){ref-type="table-fn"} 48[\](#TFN04t04){ref-type="table-fn"} 41[\](#TFN04t04){ref-type="table-fn"} Sleep efficiency ≥ 85% 37 64[†](#TFN05t04){ref-type="table-fn"} 71[\](#TFN04t04){ref-type="table-fn"} 69[\](#TFN04t04){ref-type="table-fn"} 44 57[†](#TFN05t04){ref-type="table-fn"} 59[\*](#TFN04t04){ref-type="table-fn"} 57%[†](#TFN05t04){ref-type="table-fn"} Data presented as %, unless otherwise specified. The chi-square and z-test of column proportions were adjusted for all pairwise comparisons within a row using the Bonferroni correction. Proportions were compared between treatment groups at each phase of the protocol period. MOTN awakenings = middle of the night awakening; SOL = sleep-onset latency; TST = total sleep time. Statistically significant at p \< 0.01. Statistically significant at p \< 0.05.	{ "pile_set_name": "PubMed Central" }
Introduction {#sec1-1} ============ The algorithm used for dose calculation plays a very important role in delivery of dose to patients undergoing radiation treatment.\[[@ref1]\] The beam configuration of analytical anisotropic algorithm (AAA) involves precalculated Monte Carlo data to determine all parameters to match the measured beam data.\[[@ref2]\] The AAA calculation uses accurate Monte Carlo-based three-dimensional pencil beam convolution superimposition for inhomogeneity correction. Algorithms can be divided into three classes based on their accuracy. Type "A" was proposed by Knöös et al.\[[@ref3]\] which were based on measurements and accounts for correction for patient contours and heterogeneities. Type "B" was proposed by Ojala et al.\[[@ref4]\] which were based on superposition and convolution techniques. The algorithm, Acuros XB (AXB), based on linear Boltzmann transport equation which was first implemented in the Eclipse treatment planning system (TPS) by Varian Medical Systems, Inc., (Palo Alto, CA, USA).\[[@ref5]\] comes under more accurate type "C"\[[@ref6][@ref7]\] algorithms. The origin of AXB algorithm designed for radiotherapy dose calculations is Transpire, Inc., (Gig Harbor, WA, USA).\[[@ref8][@ref9][@ref10][@ref11][@ref12]\] Lloyd and Ansbacher\[[@ref13]\] showed that the performance of AXB was better than existing clinical algorithm, AAA in high-density materials. Kan et al.\[[@ref14]\] showed the importance of options for dose to medium and dose to water in AXB in bone in the treatment of nasopharyngeal carcinoma by intensity modulated radiotherapy (IMRT) and RapidArc. Tomiyama et al.\[[@ref15]\] showed that Voxel Monte Carlo algorithm and AXB were better than AAA in stereotactic body radiotherapy (SBRT) of the lung. Similarly, Liu et al.\[[@ref16]\] showed that advantage of Acuros over AAA for treatment of SBRT lung. There were contradictory results on studies of lung cancers for planning target volume (PTV) coverage as well as sparing of the normal lung. Jiang et al.\[[@ref17]\] reported that volumetric modulated arc therapy (VMAT) was better than IMRT with better PTV coverage and sparing of normal lung by evaluating V20, V30, and mean dose (D~mean~) for the lung. Rao et al.\[[@ref18]\] showed comparable PTV coverage using VMAT and IMRT. Ong et al.\[[@ref19]\] reported that IMRT was better than VMAT by evaluating V5 for the lung. The current study evaluated the differences in PTV coverage and in sparing of OARs when using the two different dose calculation algorithms AAA and AXB in IMRT. Materials and Methods {#sec1-2} ===================== A 6 MV linear accelerator, Clinac 600C (Varian, Palo Alto, CA, USA) having forty-pairs multi-leaf collimator (MLC) with each leaf projecting 1 cm width at isocenter was used for the delivery of radiation treatments. A cohort of 15 patients diagnosed with non-small cell carcinoma of the left lung was taken for the study. All patients received a prescription dose of 50 Gy in 25 fractions. Thermoplastic sheet (Orfit) was used for immobilizing the patients. A Philips (big bore) computed tomography (CT) scanner was utilized for imaging patients, and CT images of 3-mm slice thickness were acquired with the patients lying in supine position. The CT images were transferred to the Eclipse TPS, version 13.6 (Varian Medical Systems, Palo Alto, CA, USA). The gross target volume (GTV), clinical target volume, the PTV, and the organs at risks (OARs) were contoured on the CT images by a qualified radiation oncologist following the guidelines of the International Commission on Radiation Units and Measurements report 83.\[[@ref20]\] Since the tumor was at the left side, the beams were selected from the left side only. Initially, the plans were created in AAA using IMRT technique with a grid size of 2.5 mm, and 7 mm margin was given around the GTV to account for lung motion during the treatment. For this, a template of dose constraints was created, and plan optimization was done using the created template. In addition, these constraints were changed to obtain minimum possible dose to critical organs without compromising the PTV coverage of at least 95% dose to 95% of PTV volume. One more set of plans was created using AXB algorithm, only by performing recalculation without doing any reoptimization. Thereby, in both the plans (AAA and AXB), all the plan parameters, namely, beam shape, beam angle, beam weight, number of beams, prescribed dose, normalization point, region of interest constraints, number of monitor units (MUs), and plan optimization were kept constant, and only the calculation algorithm was changed. The dose distributions in the PTV and the OARs in both the sets of plans were compared. [Table 1](#T1){ref-type="table"} shows dose constraints\[[@ref21]\] for PTV and OARs used in treatment planning. ###### Planning target volume and organs at risks-dose constraints for treatment planning ![](JMP-41-219-g001) Plan analysis for dose-volume histogram parameters {#sec2-1} -------------------------------------------------- The plans were evaluated and compared on the basis of following dosimetric parameters, namely, target coverage, dose to OARs, homogeneity index (HI), and conformity index (CI). In the current study, HI was evaluated using the formula:\[[@ref22][@ref23][@ref24][@ref25]\] HI = D5/D95 where D5 is dose to 5% of volume of PTV and D95 is dose to 95% of volume of PTV. A value of HI closer to 1 points to a more homogeneous dose inside the PTV. CI was evaluated using the formula\[[@ref26][@ref27][@ref28][@ref29]\] CI = (TV95/V95) × (TV95/TV). where TV95 is the volume of PTV receiving 95% of prescribed dose, V95 is the volume of tissue receiving 95% of prescribed dose, and TV is the volume of PTV. A value of CI closer to 1 represents more a conformal dose around PTV with less spillage of dose. Statistical analysis of the data sets for the two algorithms was done using paired, two-tailed t-test and computing the P value. When P \< 0.05, the difference between the two algorithms was considered as significant. Results {#sec1-3} ======= [Table 2](#T2){ref-type="table"} shows the location and dimensions of PTV for 15 cases. Mean of PTV was found to be 203.69 cc with standard deviation (SD: 84 cc). Mean values of OARs were left lung-GTV, 1163 cc (SD: 363 cc), contralateral lung, 1612 cc (SD: 564 cc), heart, 482 cc (SD: 202 cc), and spinal cord, 35.5 cc (SD: 18 cc), respectively. ###### Location and volume of planning target volume for 15 cases ![](JMP-41-219-g002) Tables [3](#T3){ref-type="table"} and [4](#T4){ref-type="table"} show the differences between the doses calculated by AAA and by AXB. [Figure 1](#F1){ref-type="fig"} shows the difference between dose coverage to PTV for a representative patient. [Figure 2](#F2){ref-type="fig"} shows the deviation in mean dose (D~mean~) for both the algorithms. ###### Comparison of planning target volume coverage for analytical anisotropic algorithm versus Acuros XB algorithm using intensity modulated radiotherapy technique ![](JMP-41-219-g003) ###### Comparison of organs at risk doses for analytical anisotropic algorithm versus Acuros XB algorithm using intensity modulated radiotherapy technique ![](JMP-41-219-g004) ![The difference between dose coverage to planning target volume for a representative patient (left side: Analytical anisotropic algorithm and right side: Acuros)](JMP-41-219-g005){#F1} ![Percent deviation in mean dose for Analytical Anisotropic Algorithm versus Acuros for 15 cases. The X-axis represents case number](JMP-41-219-g006){#F2} Of the various assessed dose-volume parameters for PTV, D95, maximum dose D~max~, TV95%, and HI showed statistical significance between AAA and AXB algorithms (P = 0.01, 0.00, 0.01 and 0.04). Similarly, among the several assessed dose-volume parameters for various OARs, significant differences between the two algorithms were observed only for V5, V20 and V30 in left lung-GTV, D~mean~ in right lung, D33 in heart, and D~max~ in spine (P = 0.00, 0.00, 0.00, 0.00 and 0.00 respectively) as shown in [Table 4](#T4){ref-type="table"}. Discussions {#sec1-4} =========== Dose at 95% of PTV, volume of PTV receiving 95% of dose, and HI were overestimated in AAA compared to AXB as reported previously.\[[@ref15][@ref16]\] Our results are comparable to that of Rana et al.\[[@ref30]\] who showed that more number of MUs are required for AXB to achieve similar target coverage in comparison to that of AAA. This typical dose coverage to PTV showing the difference in both algorithms is shown in [Figure 1](#F1){ref-type="fig"} for a representative patient. As the plans were recalculated with the same number of MUs in both the algorithms (AXB and AAA), PTV coverage at 95% level with AAA was good whereas it was not so with AXB. This can be also seen by observing the variation in percent deviation in D~mean~ for AAA versus AXB from [Figure 2](#F2){ref-type="fig"}. Even though there is a statistical significance between AAA versus AXB, the mean difference is very small for PTV coverage as shown in [Table 3](#T3){ref-type="table"}. Similarly, from [Table 4](#T4){ref-type="table"}, it can be observed that the OAR doses were slightly underestimated in case of normal lung (represents air) and slightly overestimated in case of the heart (represents muscle) and spine (represents bone) in AAA compared to AXB. Even though statistical significance does exist for V5, V10, V30 of (Lt.Lung-GTV), Rt.Lung (D~mean~), Heart (D33), Heart (D~mean~) and Spine (D~max~), the difference between the two algorithms is less than 3%. Accuracy of AXB over AAA is well documented in literature.\[[@ref13][@ref14][@ref15][@ref16]\] A limited study done by us, varying the size of PTV and OAR did not show any significant difference between the two algorithms. Conclusions {#sec1-5} =========== This study concludes that overall minor overestimation of PTV coverage in AAA compared to AXB algorithm. In case of OAR doses, mixed results were observed. Doses to the normal lung were slightly underestimated, and doses to the heart and spine were slightly overestimated in AAA compared to AXB. Even though statistically significant differences were observed between the two algorithms, the magnitude of the dose difference is too small to cause any clinical significance. Financial support and sponsorship {#sec2-2} --------------------------------- Nil. Conflicts of interest {#sec2-3} --------------------- There are no conflicts of interest. The authors wish to thank Prof. Komanduri Ayyangar, Emeritus Professor, University of Nebraska, Omaha, NE. USA, for helpful suggestions.	{ "pile_set_name": "PubMed Central" }
Introduction {#Sec1} ============ Hand, foot and mouth disease (HFMD) is a common childhood disease, that is characterized by rapidly ulcerating vesicles in the mouth and vesicular lesions on the hands, feet, and buttocks. In China, HFMD has exhibited the highest morbidity and mortality among the group C notifiable infectious diseases in the last decade (<http://www.nhfpc.gov.cn/jkj/s3578/new_list.shtml>). Enterovirus A71 (EV-A71) is a member of family Picornavieidae that is one of major etiological agent of HFMD and causes a variety of clinical manifestations including severe neurological complications including aseptic meningitis, brainstem encephalitis, neurogenic pulmonary edema, acute flaccid paralysis, and death^[@CR1]--[@CR3]^. EV-A71 was first isolated from a child with encephalitis in California, USA, in 1969^[@CR4]^. In the Asian-Pacific region, EV-A71-related outbreaks have been reported frequently and are associated with increased neurovirulence and fatalities^[@CR5],[@CR6]^. The dominant EV-A71 strains circulating in different countries and regions vary genetically. In mainland China, more than 600,000 HFMD cases and 126 deaths were reported in Fuyang City of Anhui province from March 2008 to June 2009. Subsequently this outbreak spread quickly to other regions and caused nationwide HFMD epidemics. An emerging recombinant EV-A71 C4 genotype was reported to be responsible for this outbreak^[@CR7]^. We have reported on the prevalence of EV-A71 during the HFMD epidemic in Beijing from 2007 to 2009, including the isolation of the CMU4232 strain, one of the circulating strains identified in 2008^[@CR8]^. Xing et al. reported that more than 90% of deaths were associated with EV-A71 in China from 2008 to 2012^[@CR9]^. However, the pathogenesis of EV-A71 infections is not fully understood, especially the cause of the severe neurological manifestation. Numerous animal models have been developed to study the pathogenesis of EV-A71 infection, and appropriate animal models are needed to better understand EV-A71-associated neuropathogenesis and to facilitate the development of effective vaccines and drugs. EV-A71-infected cynomolgus and rhesus monkeys develop similar clinical symptoms as those observed in humans, including neurological complications^[@CR10]--[@CR12]^. However, the application of EV-A71 nonhuman primate models is limited because of ethical and economic reasons. Mice models with the strategies of virus adaption and immunodeficient hosts are also developed, although they do not closely mimic human disease^[@CR13]--[@CR15]^. Moreover, mice that are more than 2-week-old are generally not susceptible to EV-A71. Human scavenger receptor class B member 2 (hSCARB2) is widely expressed in many human tissues and cell types, including neurons in the central nervous system (CNS)^[@CR16]^. Studies have showed that hSCARB2 serves as a functional receptor for EV-A71 in vitro and in vivo^[@CR17]--[@CR21]^. Although the mouse SCARB2 shares 85.8% homology to the human protein, it does not function as a receptor for EV-A71^[@CR18]^. To date, four reported transgenic (Tg) mice carrying hSCARB2^[@CR19],[@CR20],[@CR22]--[@CR24]^ have been established. Tg mice developed by Lin et al., appeared to be susceptible to EV-A71 up to 2 weeks old and exhibited pathological features similar to a suckling mouse model^[@CR20]^. In their model, B genotypes of clinically isolated EV-A71 led to HFMD-like diseases while C genotypes led to neuropathogenesis, such as limb paralysis (LP). Another hybrid (hSCARB2+/+/stat-1−/−) mouse strain established from crossbreeding SCARB2 transgenic and stat-1 knockout (KO) mice was also susceptible to EV-A71 infection up to 2 weeks of age^[@CR23]^. In the murine model developed by Fujii et al., young hSCARB2 Tg mice (3-week-old) were susceptible to infection by the EV-A71 Isehara strain (C genotype) and displayed features of neuropathology^[@CR19]^. A recent hSCARB2 knock-in (KI) mouse model established by Zhou et al. showed that the KI mice are not only susceptible to luciferase (Luc) recombinant EV-A71 pseudovirus infection (up to 12 weeks of age) but also exhibited partial HFMD clinical symptoms present in humans, combining with bioluminescent imaging (BLI) technique^[@CR24]^. In this study, we characterized of this hSCARB2 KI mice challenged by two clinical strains, including EV-A71-CMU4232, an endemic strain isolated in mainland China in 2008 and a clone-derived virus (CDV)-Isehara, whose original strain (Isehara/Japan/99) was among the circulating strains identified in Japan in 1999^[@CR25]^ and was used in another hSCARB2 Tg mouse model to study EV-A71 neuropathogenesis^[@CR19]^. We assessed the utility of these mice as appropriate small animal models for evaluating the pathogenicity and virulence of different EV-A71 strains. Results {#Sec2} ======= Characterization of CMU4232 and CDV-Isehara {#Sec3} ------------------------------------------- At 24-h post infection, the rhabdomyosarcoma (RD) cells infected with CMU4232 and CDV-Isehara displayed typical CPE, such as cell rounding, aggregation, detachment, and floatation (Fig. [1a](#Fig1){ref-type="fig"}). The production of specific VP0 and VP2 proteins (Fig. [1b](#Fig1){ref-type="fig"}) was detected in the lysates of the cells infected by CMU4232 or CDV-Isehara.Fig. 1Characterization of CMU4232 and CDV-Isehara.a Cytopathic effects at 1dpi displayed on RD cells infected with CMU4232 or CDV-Isehara. b Detection of viral structural proteins VP0 and VP2 of CMU4232 and CDV-Isehara by western blotting with antibody specific for EV-A71. β-actin was used as a internal control. c Growth curves of CMU4232 and CDV-Isehara in RD cells. RD cells were inoculated with CMU4232 or CDV-Isehara as indicated at either MOI of 1 or 0.1. Samples were collected at the times indicated and titrated by PFU assay. All assays were performed in triplicate. At each time point, titer values are means of three samples; error bars represent SEM. dpi days post infection, PFU plaque forming unit, SEM standard error of mean We next determined the growth kinetics of the two viruses in RD cells. The CMU4232 strain showed a similar proliferation rate to that of CDV-Isehara during the 24 h post-inoculation period when an multiplicity of infection (MOI) of 1 was used to inoculate the RD cells (Fig. [1c](#Fig1){ref-type="fig"}). The CMU4232 strain reached a peak viral titer (3.8 × 10^7^ plaque forming unit (pfu)) after 24 h and then decreased, whereas CDV-Isehara reached a peak viral titer (4.3 × 10^7^ pfu) at 48 h post-inoculation. We also determined the growth curve for both viruses at an MOI of 0.1. At a low MOI, CMU4232 and CDV-Isehara reached a peak viral titer at 48 h postinoculation. The maximum viral titer observed for CMU4232 and CDV-Isehara was 2.6 × 10^7^ pfu and 3.55 × 10^7^ pfu, respectively (Fig. [1c](#Fig1){ref-type="fig"}). These data indicated that the CDV-Isehara strain replicated relatively faster than that of the stain CMU4232 in RD cells. Sequences and phylogenetic analyses of the strains CMU4232 and CDV-Isehara {#Sec4} -------------------------------------------------------------------------- The complete genomes of CMU4232 and CDV-Isehara exhibited a nucleotide similarity of 83%. The nucleotide and amino acid similarities of the four structural genes (VP1 to VP4) between CMU4232 and CDV-Isehara were observed to be 89--90.8% and 98.3--100%, respectively ([S1 Table](#MOESM1){ref-type="media"}). Of the 297 amino acids of VP1 protein, five substitutions were identified, including H22Q, A145E, N237T, V249I, and A289T (Fig. [2](#Fig2){ref-type="fig"}), whereas only one substitution (I31M) was observed in the 254 amino acids of the VP2 protein. No substitutions were observed in the VP3 and VP4 proteins between the two strains ([S1 Table](#MOESM1){ref-type="media"}).Fig. 2Variation of the VP1 amino acid sequences between CMU4232 and CDV-Isehara strains.In total, there were 297 amino acids in the VP1 protein of EV-A71, and five residue variations (22 H/Q, 145 A/E, 237 N/T, 249 V/I, and 289 A/T) were found in the VP1 amino acid sequence between CMU4232 and CDV-Isehara strains In the phylogenetic analysis, the sequences of 30 prevalent EV-A71 strains reported from different regions of the world were compared, including the strains CMU4232 and CDV-Isehara used in this study. Consistent with the results of previous studies, all the representative strains isolated in China during 1998--2017 were clustered into the C4 genotype. The C4 cluster showed a stepwise evolutionary trajectory over time. There are two branches included in the genotype C4, the early C4b (1998--2004) and the later C4a (2005--2017). The C4a branch could be further divided into two separate groups, C4a-1 and C4a-2, which include strains isolated from 2005--2007 and 2008--2017, respectively. The results clearly show that CMU4232 belongs to the C4a-2 of genotype C4 and CDV-Isehara belongs to genotype C2 (Fig. [3](#Fig3){ref-type="fig"}).Fig. 3Phylogenetic relationship among 30 worldwide EV-A71 strains based on the complete VP1 gene.CMU4232, CDV-Isehara, 18 representative CHN strains, and 10 international representative strains of different subgenotypes are included in this dendrogram. The 18 representative CHN strains were selected from different isolation places in mainland China from 1998 to 2017 according to available data on internet. Details of all the EV-A71 strains included in the dendrogram are provided in Table [S1](#MOESM1){ref-type="media"}. CHN Chinese hSCARB2 KI mice are susceptible to CDV-Isehara intracerebrally (I.C.) infection {#Sec5} ------------------------------------------------------------------------------- When hSCARB2 KI mice of different ages were challenged, 76/105 showed ataxia, paralysis, or death at days 2--14 after infection (Figs. [4a, b](#Fig4){ref-type="fig"} and Supporting videos). A significant difference in the disease incidence and death rate was observed between KI mice younger and older than 3 weeks (P \< 0.0001) (Figs. [4c, d](#Fig4){ref-type="fig"}). In addition, the death rate of 1-week-old-infected KI mice was 100% (20/20) from 3 to 5 days post inoculation (dpi), while that of 2-week-old-infected Tg mice was 77.3% (17/22) from 4 to 11 dpi (P \< 0.0001) (Fig. [4d](#Fig4){ref-type="fig"}). Paralysis occurred in both the forelimb and hindlimb post infection, particularly in the hindlimb. Some of the KI mice died within 1--3 days after LP symptoms appeared, while some were still alive when the symptoms gradually decreased. The susceptibility of the hSCARB2 KI mice to CDV-Isehara infection decreased as the age of mice increased, although no significant difference was observed among groups of KI mice older than 3 weeks (P \> 0.05) (Figs. [4b--d](#Fig4){ref-type="fig"}). The development of disease symptoms in 4-week-old-infected KI mice was very similar to that of 3-week-old (Fig. [4b](#Fig4){ref-type="fig"}). In 4-week-old-infected KI mice, 33.3% (8/24) and 20.8% (5/24) experienced mild and severe symptoms, respectively, while 12.5% (3/24) died, whereas 17.2% (5/29) and 27.6% (8/29) of 3-week-old-infected KI mice exhibited mild and severe symptoms, respectively, while 13.8% (4/29) died.Fig. 4Susceptibility of hSCARB2 KI mice to CDV-Isehara infection.The 1, 2, 3, 4, and 6-week-old hSCARB2 Tg mice were infected intracerebral (I.C.) with CDV-Isehara strain at a dose of 4.8 × 10^6^ pfu and monitored daily after infection. a The CNS-like hindlimb paralysis, ruffled fur, and shrinking. b The clinical symptoms (asymptomatic, mild, severe, and dead), followed by the criteria described in the materials and methods section (c) asymptomatic (%) and (d) survival (%). Significant difference of asymptomatic (%) and survival (%) with different weeks old Tg mice infected with CDV-Isehara was shown as \\\\p,0.0001. pfu plaque forming unit, CNS central nervous system, I.C. intracerebral Difference in clinical features of hSCARB2 KI mice challenged by CMU4232 and CDV-Isehara {#Sec6} ---------------------------------------------------------------------------------------- Differences in the clinical features of hSCARB2 KI mice caused by the CMU4232 and CDV-Isehara strains when using the same route of infection (I.C. or I.V.) were assayed for using 3-week-old hSCARB2 KI mice. Some of the KI mice infected with both EV-A71 strains led to diverse clinical signs and various degrees of manifestations, especially in the development of neurological symptoms. The peak period of disease onset was 3--7 dpi, and death was typically delayed by 1--2 days after this period. No clinical signs were observed in the mock-infected KI mice (Fig [5](#Fig5){ref-type="fig"}a[1](#Fig1){ref-type="fig"}, a[2](#Fig2){ref-type="fig"}, b[1,](#Fig1){ref-type="fig"} and b[2](#Fig2){ref-type="fig"}).Fig. 5The clinical symptoms, asymptomatic (%), survival (%), and body weight change (%) of different EV-A71 Tg mouse models.Three-week-old hSCARB2 KI mice were inoculated I.C. with DMEM/ EV-A71 (a) or inoculated intravenous (I.V.) with DMEM/EV-A71 (b) respectively. Three-week-old hSCARB2 KI mice were inoculated with CMU4232 (c) /CDV-Isehara (d) strain via I.C. or I.V. route respectively. (1) The clinical symptoms, (2) asymptomatic (%), (3) survival (%), and (4) body weight change (%) of the Tg mice were observed and recorded daily for 2 weeks after viral infection. \\\p,0.000. I.C. intracerebral, I.V. intravenous The death rate of the CDV-Isehara/I.C. group was 17.6% (6/34), and 41.2% (14/34) of mice displayed mild or severe clinical signs, including ruffled fur, hunchbacked appearance, easily frightened, trembling, shrinking, limb weakness, mental bluntness, dispirited, slow in action, emaciation, and LP, whereas only mild symptoms (19%) occurred in the CMU4232/I.C. group (N* = 21), with trembling, shrinking, and limb weakness observed, and these mild symptoms completely disappeared soon thereafter (Fig [5](#Fig5){ref-type="fig"}a[1](#Fig1){ref-type="fig"}). The clinical signs of the CDV-Isehara/I.C. group were observed in 20 of 34 mice from 3 to 11 dpi, while clinical signs were observed in 4 of 21 mice from 3 to 7 dpi for the CMU4232/I.C. group (Fig [5](#Fig5){ref-type="fig"}a[2](#Fig2){ref-type="fig"}). A significant difference in disease incidence was observed between these two groups (P \< 0.05) (Fig [5](#Fig5){ref-type="fig"}a[2](#Fig2){ref-type="fig"}). However, no significant differences were found in death rate between them (P \> 0.05) (Fig [5](#Fig5){ref-type="fig"}a[3](#Fig3){ref-type="fig"}). Among the groups infected via the I.C. route, a significant difference in body weight change was also observed (P = 0.023) (Fig [5](#Fig5){ref-type="fig"}a[4](#Fig4){ref-type="fig"}). Furthermore, significant differences were observed between the CDV-Isehara/I.C. and CMU4232/I.C. or DMEM/I.C groups (P = 0.042 and P = 0.047, respectively) (Fig [5](#Fig5){ref-type="fig"}a[4](#Fig4){ref-type="fig"}). The body weights of mice in the CDV-Isehara/I.C. group increased much slower than those of in the CMU4232/I.C. and DMEM/I.C. groups, demonstrating that CDV-Isehara infection affected the growth of infected KI mice more than CMU4232 when administered via the same I.C. route. Taken together, the disease severity induced by the strain CMU4232 in 3-week-old hSCARB2 KI mice was less than that caused by the strain CDV-Isehara. These results suggested that these KI mice were susceptible to EV-A71 and that there was a difference in sensitivity of the hSCARB2 KI mice to the two strains. When the onset of typical symptoms was compared, hSCARB2 KI mice infected with the strain CMU4232 via the I.V. route exhibited typical symptoms from 2 to 11 dpi, including ruffled fur, hunchbacked appearance, easily frightened, trembling, shrinking, limb weakness, mental bluntness, dispirited, slow in action, emaciation, and death, but no severe symptoms or paralysis occurred when the mice were followed up to 14 dpi. However, the symptoms and consequences of CDV-Isehara/I.V. infection were more severe from 2 to 14 dpi, with paralysis in the limbs and death observed in this group (Fig [5](#Fig5){ref-type="fig"}b[1](#Fig1){ref-type="fig"}, b[2](#Fig2){ref-type="fig"}). In the hSCARB2 KI mice I.V. infected with CMU4232, 86.4% (19/22) developed mild symptoms and 4.5% (1/22) died whereas 50% (9/18) and 27.8% (5/18) of mice in CDV-Isehara/I.V. group exhibited mild and severe symptoms, respectively, while 11.1% (2/18) died (Fig [5](#Fig5){ref-type="fig"}b[1](#Fig1){ref-type="fig"}). A significant difference in the disease incidence was observed between the CMU4232/I.V. or CDV-Isehara/I.V. groups and the DMEM/I.V. group (P \< 0.0005 and P \< 0.005, respectively) (Fig [5](#Fig5){ref-type="fig"}b[2](#Fig2){ref-type="fig"}). However, no significant differences in disease incidence and death rate were observed between CMU4232/I.V. and CDV-Isehara/I.V. groups. In the CDV-Isehara/I.V. group, two of 18 infected Tg mice died from 3 to 10 dpi, while one death was observed in the CMU4232/I.C. group at 4 dpi (Fig [5](#Fig5){ref-type="fig"}b[3](#Fig3){ref-type="fig"}). Furthermore, no difference in body weight change was observed between these two groups (P = 0.468), although a significant difference was observed between the CDV-Isehara/I.V. and DMEM/I.V. groups (P = 0.012) (Fig [5](#Fig5){ref-type="fig"}b[4](#Fig4){ref-type="fig"}). We next compared the disease development of hSCARB2 KI mice infected with the same strain via different routes (I.C. or I.V.). For KI mice infected with CMU4232, distinct symptoms were observed and were strongly dependent on the infection route. The severity of the disease caused by the I.V. route was more severe than that of I.C. route based on the observed clinical symptoms, disease incidence (P \< 0.005) and the body weight change (P \< 0.005) (Fig [5](#Fig5){ref-type="fig"}c[1](#Fig1){ref-type="fig"}--c[4](#Fig4){ref-type="fig"}). However, the severity of the disease induced by CDV-Isehara was similar for both infection routes, irrespective to the observed clinical manifestations, disease incidence, death rate, and the body weight change (P \> 0.05) (Fig [5](#Fig5){ref-type="fig"}d[1](#Fig1){ref-type="fig"}--d[4](#Fig4){ref-type="fig"}). Viral proliferation and pathological features in hSCARB2 KI mice {#Sec7} ---------------------------------------------------------------- To assess viral proliferation in vivo, the viral RNA in tissues of infected 3-week-old KI mice was quantified by real-time PCR (Fig. [6](#Fig6){ref-type="fig"}). Specific viral RNAs were detected in the brain of CMU4232/ I.C. KI mice (Fig. [6a](#Fig6){ref-type="fig"}), while in the CDV-Isehara/I.C. group, viral RNAs were detected in the brain, intestine, muscle, and heart tissues of mice, but no viral RNAs were identified in the lungs (Fig. [6b](#Fig6){ref-type="fig"}). The distribution of both EV-A71 strains in tissues of KI mice infected via the I.V. route was highly consistent. However, higher viral loads were detected in CMU4232-infected KI mice than was observed in CDV-Isehara-infected mice. Furthermore, the viral load in the muscle was higher than that observed in the brain, heart, lung, and intestine, demonstrating the susceptibility of different tissues of hSCARB2 KI mice to EV-A71 when infected via the I.V. route (Figs. [6c, d](#Fig6){ref-type="fig"}). No viruses were detected in any tissue of the control group, which was inoculated with RD cell lysate supernatant.Fig. 6Virus replication in some tissues of hSCARB2 KI mice by EV-A71 infection.Three-week-old hSCARB2 KI mice were inoculated with CMU4232/ CDV-Isehara via I.C. or I.V. route. EV-A71 genome copy numbers in the brain, heart, lung, intestine, and muscle of EV-A71-infected Tg mice were determined at several time points by quantitative RT-PCR. Mouse β-actin gene expression in each tissue was used as the internal control. Results represent the mean of three to five samples; error bars represent STDEV. a CMU4232/I.C. (N = 3); b CDV-Isehara/I.C. (N = 4); c CMU4232 /I.V. (N = 5); d CDV-Isehara/I.V. (N = 4) Next, we observed the pathological changes in the two major replication sites, the brain and skeletal muscle (Fig. [6](#Fig6){ref-type="fig"}). Different degrees of pathological changes were evident in the KI mice with clinical signs. These changes primarily included cellular damage (such as necrosis and neuronophagia) and inflammatory changes (such as gliosis and perivascular cuffing) (Fig [7.1--7.3](#Fig7){ref-type="fig"}) in the brain. In addition, degeneration and necrosis in muscle bundles and severe necrotizing myositis with a mass of inflammatory cell infiltration was observed (Fig [7.5--7.7](#Fig7){ref-type="fig"}), especially in KI mice infected with CDV-Isehara via the I.V. route. Viral antigens were also detected in the above affected tissues. Immunohistochemistry (IHC) staining revealed the presence of viral antigen in the tissues with pathological changes, indicating that the pathological findings were caused by EV-A71 infection. All mock mice tested negative for viral antigens, regardless of the infection routes (Fig. [8](#Fig8){ref-type="fig"}).Fig. 7Pathological analysis of brain and the skeletal muscle of EV-A71-infected mouse models.Sporadic neuronal necrosis and neuronophagia (arrowhead) in the brain (1) and Inflammatory cell infiltration (arrowhead) in the skeletal muscle (5) of mouse model with CDV-Isehara/I.C.; glial nodule (arrow) and neuronophagia (arrowhead) in the brain (2) and severe necrotizing myositis with a mass of inflammatory cell infiltration (arrow) in the skeletal muscle (6) of mouse model with CDV-Isehara/I.V.; perivascular cuffing (arrowhead), partial neuronal necrosis and neuronophagia in the brain (3) and degeneration and necrosis in muscle bundles with inflammatory cell infiltration (arrowhead) (7) in mouse model with CMU4232/I.V.; the brain (4) and the skeletal muscle (8) of MOCK mouse with DMEM/I.V. (magnification: 200 × ). Representative samples are shown for each groupFig. 8VP2 expression in the brain and the skeletal muscle of EV-A71-infected mouse models.The brain (1) and the skeletal muscle (5) of mouse model with Isehara/I.C.; the brain (2) and the skeletal muscle (6) of mouse model with CDV-Isehara/I.V.; the brain (3) and the skeletal muscle (7) of mouse model with CMU0804/I.V.; the brain (4) and the skeletal muscle (8) of MOCK mouse with DMEM/I.V. Arrowheads indicate EV-A71-positive cells (IHC staining against EV-A71 VP2, magnification: 200 × ). Representative samples are shown for each group Discussion {#Sec8} ========== For robust results, an adequate number of hSCARB2 KI mice (N = 181) were used to establish a mouse model that is sensitive to EV-A71 infection in this study. In this mouse model, we observed that 3 weeks is a crucial age, as infected mice younger than 3-week-old became extremely ill. The susceptibility of the KI mice to EV-A71 infection could be observed until 8-week in mice with viral challenged by EV-A71 at 6-week-old age. Fujii et al. reported that their Tg mice older than 6 weeks were also susceptible to EV-A71 infection, and the development of disease was similar to that of 3-week-old mice^[@CR19]^. The model based on 3-week-old weaning hSCARB2 KI mice in the present study displayed diverse clinical symptoms and was characterized as pathological tropism in the CNS and peripheral sites. Based on the reliable EV-A71 disease model, we evaluated the pathogenicity of an endemic strain (CMU4232) in mainland China, and compared it with an internationally published strain (Isehara)^[@CR19]^. We observed that disease manifestations induced by CDV-Isehara were more severe than those induced by CMU4232. Both the biological characteristics of specific viral strains and infection routes were observed to be important for the pathogenicity of EV-A71 in this hSCARB2 KI mouse model. Previous studies reported that different EV-A71 strains could display varying pathogenic effects^[@CR26]--[@CR28]^. When infected via the I.C. route, the disease induced by CDV-Isehara was more severe than that induced by CMU4232 at the same dose (Fig [5](#Fig5){ref-type="fig"}a[1](#Fig1){ref-type="fig"}) in our KI mice model. A significant difference in disease incidence between these two groups was observed (P \< 0.05) (Fig [5](#Fig5){ref-type="fig"}a[2](#Fig2){ref-type="fig"}). For mice in the CMU4232/I.C. group, only mild clinical symptoms (4/21) developed from 3 to 7 dpi, and these symptoms subsequently completely disappeared. In contrast, mice in the CDV-Isehara/I.C. group exhibited a longer onset (from 3 to 11 dpi) of various mild manifestations, paralysis, and death (6/34). In addition to the difference in disease incidence caused by the two strains, we also observed a difference in the presence of the viruses infected mouse tissues. Using qRT-PCR, viral RNAs were detected in the intestine, muscle, heart, and brain tissues of mice in the CDV-Isehara/I.C. group. We speculated that CDV-Isehara replicated well in the brain after I.C. inoculation and then entered into multiple organs via viremia, leading to viral infections that caused clinical manifestations (Fig. [6b](#Fig6){ref-type="fig"}). In contrast, viral replication of the strain CMU4232 was only detected in the brains of mice after infected via the I.C. route, and viral titers in other non-CNS organs were below detectable levels via qRT-PCR. This difference suggested that the strain CMU4232 may be less sensitive to or replicated less efficiently in the brain of this mouse model, resulting in the eventual elimination of the virus. Thus, low viremia failed to induce an effective infection in other non-CNS tissues. With I.C. administration, the brain is the most direct injury site. Both CMU4232 and CDV-Isehara strains can replicate at this site, indicating that both strains possess the characteristics of neurotropism. However, the neurovirulence of the strain CMU4232 appeared to be weaker than that of CDV-Isehara in this model. Considering that the CDV-Isehara strain originated from an infectious clone, it is unclear whether there could be some change in virulence compared with the original strain. However, the neurovirulence of the original Isehara/Japan/99 isolate in another hSCARB2 Tg mouse model was significantly superior to those of other EV-A71 strains from encephalitis and death cases^[@CR19]^. Based on a phylogenetic analysis, the two strains used in this study were determined to have the genotypes C4 and C2, respectively (Fig. [3](#Fig3){ref-type="fig"}). The different genetic backgrounds of the two strains may explain the difference in virulence observed between the two EV-A71 strains in our mouse model. We observed that the nucleotide similarity of the CMU4232 and CDV-Isehara complete genomes was only 83%. VP1 of EV-A71 is known to be an important immunodominant protein that mediates the binding of this virus to host receptor and determines its tissue tropism. Elizabeth et al. reported that a single mutation (K244E) in the VP1 is responsible for increased virulence and neurotropism in adult interferon-deficient mice^[@CR29]^. When the deduced amino acid sequences were compared in this study, difference in five residues (H22Q, A145E, N237T, V249I, and A289T) were observed between the two strains CMU4232 and CDV-Isehara (Fig. [3](#Fig3){ref-type="fig"}). The substitution of VP1-145G or -145Q for -145E has been shown to be responsible for murine adaptation and/or virulence, either alone or in combination with changes in other amino acids in different mouse models^[@CR30]--[@CR33]^. In our hSCARB2 KI mice, the strain CDV-Isehara harbors VP1-145E and induce more severe neurological manifestations and myositis than those induced by the strain CMU4232 which harbors VP1-145A (Figs. [7](#Fig7){ref-type="fig"} and [8](#Fig8){ref-type="fig"}). These results suggest that EV-A71 with VP1-145E can confer better in vivo fitness of the virus, and viruse carrying this mutation are more virulent than strains with other amino acids at this position (VP1-145A/G/Q) in mouse models. Furthermore, Kataoka et al. identified VP1-145 as a critical molecular determinant for the binding of EV-A71 to another specific cellular receptor, human P-selectin glycoprotein ligand-1 (PSGL-1)^[@CR34]^. The strong in vivo selection of VP1-145E variants and CNS spread have been suggested to occur in a PSGL-1-independent manner. The VP1-145E variant was primarily responsible for the development of viremia and neuropathogenesis in a cynomolgus monkey model^[@CR35]^. Previous studies showed that the amino acid VP1-145 is a variable residue among the clinical EV-A71 isolates, and it serves as a major site of positive selection in the evolution of EV-A71^[@CR3],[@CR36],[@CR37]^. Molecular epidemiological studies have shown that strains with VP1-145G/Q/R are more frequently observed in severe neurological cases in humans than VP1-145E isolates^[@CR8],[@CR38]--[@CR40]^. Our finding also suggested that VP1-145E may be a crucial residue that is responsible for increased virulence in mice through unclear mechanisms, which may involve the other receptors and proteins required for viral entry and host immune responses to EV-A71. Two latest studies conducted in Japan reported that the amino acid VP1-145E contributed to virulence determination by controlling attachment receptor and antibody sensitivity^[@CR41],[@CR42]^. The virus Isehara with VP1-145E replicated efficiently in the CNS and acted as a virulent phenotype in their hSCARB2 Tg mice. However, VP1-145G viruses were adsorbed by attachment receptors such as heparan sulfate during circulation in vivo, leading to abortive infection and less virulent phenotype in the same Tg mice model. The maintenance of viremia titer at a high level is necessary for virus entrance into CNS via the intravenous route^[@CR41]^. When cynomolgus monkey model was tested, VP1-145E, but not VP1-145G viruses induced neurological symptoms. VP1-145E viruses were frequently detected in the tissues of infected monkeys, but VP1-145G viruses were detected less frequently and disappeared quickly. Clearly, VP1-145E viruses were more resistant to neutralizing antibodies than VP1-145G viruses^[@CR42]^. Our results are in line with the two studies. However, the exact mechanism of the CMU4232 with VP1-145A in the infected hSCARB2 KI mice remains to be elucidated. The difference between the pathogenicity of EV-A71 in animals and humans caused by the VP1-145 variation may partly explain why the Isehara/Japan/99 strain from a HFMD patient was observed to be more pathogenic than other clinical strains from encephalitis and death cases in another hSCARB2 Tg mouse model^[@CR19]^. In addition to the VP1-145, two other variable sites observed in this study (VP1-22 and VP1-237) have been reported to be under positive selection pressure^[@CR3],[@CR36],[@CR39]^. We also observed a difference at VP2-I31M between these two strains. Huang et al. reported that amino acid changes in both A145E of VP1 and K149M of VP2 contributed to viral infectivity in vitro and mouse lethality in vivo^[@CR32]^. However, we did not observe an amino acid change at 149 K of VP2. It is known that both alanine (A) and glycine (G) are nonpolar neutral amino acids. However, it is not known whether CMU4232 VP1-145A could be adsorbed by heparan sulfate in tissues with abundant blood supply under the super higher viremia titer (10^8^ pfu), as VP1-145G viruses were tested by Kobayashi et al.^[@CR41]^ When assessing differences arising from the routes of viral infection (I.C. vs. I.V.), the CDV-Isehara models displayed no significant difference in disease incidence, survival rate or body weight change (Fig [5](#Fig5){ref-type="fig"}d[2](#Fig2){ref-type="fig"}--[5](#Fig5){ref-type="fig"}d[4](#Fig4){ref-type="fig"}), although differences in the distribution of the virus in the tissues of infected KI mice were observed (Figs. [6b](#Fig6){ref-type="fig"}, [6d](#Fig6){ref-type="fig"}). The replication of CDV-Isehara was detected in all tested tissues of mice via the I.V. route, which differed a report by Fujii et al. in which the Isehara/Japan/99 strain was only detected in CNS in another Tg mouse model when using the same I.V. infection route. However, the dose used by Fujii et al. was 1 × 10^6^ TCID~50~^[@CR19]^, whereas the dose used in our study was about 100 times higher. Therefore, the results obtained by Fujii et al. cannot be directly compared with those obtained in this study. It should be kept in mind that pathological tropisms observed in the CNS and peripheral sites cannot only be distinguished by clinical symptoms. The identification of virus distribution and viral load in different tissues still needs a combination of different assays used in this study. In contrast to the CDV-Isehara models, significant differences between the CMU4232/I.C. and CMU4232/I.V. group were observed with respect to disease incidence, body weight change (Fig [5](#Fig5){ref-type="fig"}c[2,](#Fig2){ref-type="fig"} c[4](#Fig4){ref-type="fig"}) and the distribution of the virus in tissues (Figs. [6a](#Fig6){ref-type="fig"}, [c](#Fig6){ref-type="fig"}). CMU4232 could only replicate in the brains of mice when 3.2--6.7 × 10^6^ pfu was administered via the I.C. route. However, via an I.V. administration of 3.2--7.5 × 10^8^ pfu, CMU4232 could infect multiple organs. Following viremia, CMU4232 disseminated from the peripheral to central, leading to the observed CNS manifestations finally and demonstrating that the viral dose/hyperviremia is crucial to the in pathogenicity of EV-A71. Our results are in agreement with a recent study in which epidemic Chinese EV-A71 strains isolated from 2008--2010 were compared in neonatal ICR mice and neonatal rhesus monkeys^[@CR27]^. It appears that the CMU4232/IV mouse model mimics not only the neurological manifestation and infection of multiple organs injury following hyperviremia but also the morbidity of EV-A71 infection in human beings to some extent. A previous study reported that the risk of neurological complications was 1.1% and the severe-case fatality risk was 3.0%, with \>90% of deaths associated with EV-A71 in the research of HFMD in China from 2008--2012^[@CR9]^. In this study, we also compared the replication capacity of the strains CMU4232 and CDV-Isehara in RD cells through a plaque-forming assay. The growth curves showed that CDV-Isehara replicated faster and generated more virions than CMU4232 (Fig. [1c](#Fig1){ref-type="fig"}). As reported by Sun et al., a difference in the replication capacity of EV-A71 strains in RD cells could indicate their pathogenicity in humans^[@CR43]^. There appeared to be some differences in viral loads of the two strains detected in vitro and in vivo. Under the I.C. inoculation, the virus CDV-Isehara was detected in different tissues including the brain, intestine, muscle, and heart (Fig. [6b](#Fig6){ref-type="fig"}) whereas the virus CMU4232 was only detected in the brain (Fig. [6a](#Fig6){ref-type="fig"}). Furthermore, the viral load in tissues infected by CDV-Isehara was higher than that infected by CMU4232. The finding was consistent to the growth curves in RD cells (Fig. [1c](#Fig1){ref-type="fig"}). Under the I.V. inoculation, both viruses were detected in the brain, heart, lung, intestine, and muscle (Figs. [6c,d](#Fig6){ref-type="fig"}). The highest viral loads of both viruses were observed in muscles, dissimilar to those observed under the I.C. route infection. In addition, the viral load in tissues infected by CMU4232 was higher than that by CDV-Isehara. However, we did not observe differences in rates of morbidity and mortality and in change of the body weight of mice between these two viruses after I.V. inoculation. In contrast, CDV-Isehara-infected mice appeared to have a high proportion of severe symptoms compared with those of CMU4232-infected mice. The result indicated that the virus CDV-Isehara seems to be more virulent than the CMU4232 in the hSCARB2 KI mouse model. On the other hand, although having had high viral load in different tissues, the virus CMU4232 seemed to cause mild symptoms, suggesting that the virus CMU4232 is less virulent than that of CDV-Isehara. In short, one possible explanation for the difference observed may be related to the infectious routes used. Another explanation may be due to characteristics of the same strain in vitro and in vivo. In addition, we could not exclude the effect of timing of disease onset when tissue samples were taken from the infected mice. EV-A71 strains have been shown to be continuously evolving. The strain CMU4232 used in this study is a member of the recent C4a-2 cluster, suggesting that CMU4232 could be a representative endemic strain after the HFMD outbreak in mainland China in 2008 (Fig. [3](#Fig3){ref-type="fig"}). EV-A71 strains with the genotype C4 have been endemic and epidemic in mainland China for 20 years. The prevalence of neutralizing antibodies against the EV-A71 strains has been high in China and the selection pressure for the virus from population immunity is strong^[@CR44],[@CR45]^. Indeed, as reported by Zhang et al. the evolution rate of C4a EV-A71 appeared to be faster than the average for all other EV-A71 genotypes^[@CR46]^. Therefore, persistent attention should be paid to the changes in virulence resulting from evolution of EV-A71 viruses and their related pathogenic characteristics. In summary, we showed that hSCARB2 KI mice are a sensitive and useful model for investigating the clinical outcomes caused by different EV-A71 strains. Diverse clinical symptoms characterized as bi-pathological tropism in the CNS and peripheral sites can be observed even when the "representative" Chinese strain CMU4232 was tested. The intracranial infection model appears to be suitable for evaluating of neurovirulence caused by EV-A71, whereas the venous infection model is useful for studying the pathogenicity of EV-A71. Moreover, impact of the change in key amino acids site (e.g., VP1-145) on pathogenesis in different hosts should be taken into account in future studies. Materials and methods {#Sec9} ===================== hSCARB2 KI mice and ethics statements {#Sec10} ------------------------------------- The human SCARB2 KI mice^[@CR24]^ used in this study were provided by the Institute for Laboratory Animal Resources, National Institute for Food and Drug Control, Beijing, China. The animal experiments were performed according to the recommendations in the national guidelines for the care and use of animals in scientific research "Regulations for the Administration of Affairs Concerning Experimental Animals". The study protocol was approved by the Capital Medical University Animal Experiments and Experimental Animals Management Committee (AEEI-2016-121). Termination of mice was performed under anesthesia, and all efforts were made to minimize suffering. All samples used in this study were anonymized. EV-A71 viruses and cells {#Sec11} ------------------------ CMU4232 (EV71/CMU4232-1/BJ/CHN/2008) was originally isolated from a HFMD patient by our group during the 2008 outbreak in Beijing^[@CR8]^, and it was passaged in RD cells (National Infrastructure of Cell Line Resource, Beijing, China) and stocked in our laboratory. An infectious cDNA clone of EV-A71 (pSVA14-Isehara ver4) was kindly provided by Professor Satoshi Koike, Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan. The RD cell line was cultured in DMEM containing 4.5 g/L glucose, L-glutamine, and sodium pyruvate (Corning, 10-013-CVR, USA) supplemented with 10% fetal bovine serum (FBS) (Corning, 35-076-CV, USA), 100 IU of penicillin, and 100 μg of streptomycin per ml. The cells were incubated at 37 ℃ and with 5% CO~2~. CMU4232 and the rescued virus, CDV-Isehara were harvested from RD cultures by freezing and thawing three times and were stored at −80 ℃. The titers of the virus stocks were tested using a modified plaque-forming assay and determining the CCID50. Growth curves of CMU4232 and CDV-Isehara in RD cells {#Sec12} ---------------------------------------------------- To assess the proliferation dynamics of both EV-A71 strains, the growth curves were determined as follows: RD cells cultured in 24-well plates were inoculated at an MOI of 1 or 0.1 in triplicate. After being allowed to attach for 2 h, the unbound viral particles were washed off with PBS. Next, the infected cells were cultured at 37 ℃ for 0, 8, 24, 48, 72, and 96 h and were titrated to determine pfu values after three consecutive freeze--thaw cycles. All assays were performed in triplicate. Western blot analysis {#Sec13} --------------------- Infected RD cells were collected and lysed using RIPA buffer. Viral and cellular proteins were separated by SDS-PAGE on a 12% gel and were electrotransferred onto a PVDF membrane. After blocking, the membrane was probed with a mouse anti-EV-A71 antibody (Millipore, MAB979, 1:1000 dilution) and an HRP-conjugated secondary antibody (Transgen Biotech, HS201-1, 1:5000 dilution). Reverse transcription (RT)-PCR and nucleotide sequencing of CMU4232 {#Sec14} ------------------------------------------------------------------- Viral RNA of the CMU4232 strain was extracted from infectious RD cell supernatant using a QIAmp^®^ Viral RNA Kit (Qiagen, Hilden, Germany). RNA was reverse transcribed into cDNA using a Superscript II Kit (Invitrogen, USA) according to the manufacturer's instructions. All sequencing was performed by Guangzhou Darui Biotechnology Co., Ltd. The accession number of the complete genome sequence of CMU4232 is MH373639. Nucleotide and deduced amino acid sequence identity analysis between CMU4232 and CDV-Isehara {#Sec15} -------------------------------------------------------------------------------------------- The complete genome sequence alignment of CMU4232 and CDV-Isehara was performed using the Bioedit and Clustal W programs in MEGA 4.0, and the deduced amino acid sequences of structural and nonstructural proteins were aligned and shown. Phylogenetic analysis {#Sec16} --------------------- Representative strains of each EV-A71 subgenotypes were chosen based on a previously published study, especially representative strains available at the Genbank reported in mainland China that were first screened from those isolated from different regions and time periods from 1998 to 2017. Phylogenetic analysis was based on viral VP1 protein-coding nucleotide sequences, and the phylogenetic tree was constructed using the neighbor-joining method using MEGA 4.0 software with 1000 bootstrap replications. Details of all the EV-A71 strains included in the phylogenetic tree are provided in Supplementary Table [S2](#MOESM1){ref-type="media"}. Infection of EV-A71 in hSCARB2 KI mice {#Sec17} -------------------------------------- To evaluate the susceptibility of the hSCARB2 KI mice to EV-A71 infection, we challenged 1-, 2-, 3-, 4-, and 6-week-old KI mice I.C. with the CDV-Isehara strain at a dose of 4.8 × 10^6^ pfu and observed the development of clinical symptoms daily for 2 weeks. In addition, we chose 3-week-old hSCARB2 KI mice to be inoculated with EV-A71 (CMU4232/ CDV-Isehara) at the indicated doses and routes (I.C. infection at a dose of 3.2--6.7 × 10^6^ pfu or I.V. infection at a dose of 3.2--7.5 × 10^8^ pfu per animal). Control mice were mock injected with an equivalent volume of RD cell lysate supernatant. All mice were monitored daily for clinical signs, survival, and body weight for 2 weeks. The activity level, mental status, degree of limb paralysis, and other symptoms were observed and recorded daily for each mouse. Furthermore, the clinical signs were scored as follows: asymptomatic (0), ruffled fur and/or hunchbacked appearance (1--2), easily frightened/ trembling/shrinking/mental bluntness/dispirited/slow in action (3--5), limb-shake weakness (6), wasting \> 3 consecutive days (7), limb paralysis (8), and moribund and death (9). Based on the scores, we classified the clinical manifestations at four levels, asymptomatic (0--2); mild symptoms (3--5); severe symptoms (6--8); and death (9). Experiments for group of 3-week-old mice were performed in triplicate. Quantitative detection of EV-A71 RNA by real-time RT-PCR in different tissues {#Sec18} ----------------------------------------------------------------------------- The infected mice were killed at the peak onset of disease development and were immediately processed for sample collection. Tissues were collected, including brains, hearts, lungs, skeletal muscle, and intestines. Total RNA was extracted from different tissues using TRIzol reagent (Invitrogen, 15596026, USA) following the manufacturer's instructions, and then was converted into cDNA with a RT-PCR kit (Transgen Biotech, AE301-02) followed by qPCR using a PCR system (Transgen Biotech, AQ101-02). PCR amplifications were performed at 94 ◦C for 30 s followed by 40 cycles of 94 ◦C for 5 s, 60 ◦C for 15 s, and 72 ◦C for 10 s. SYBR Green PCR Master Mix (Transgen Biotech, AQ131-02) was used to detect of viral load in different tissues using the primers of EV71-S /A (S3 Table). Mouse β-actin cDNA was used as an internal control. Each assay was carried out in triplicate. Histopathology examination (HE) and immunohistochemistry (IHC) analysis {#Sec19} ----------------------------------------------------------------------- Tissue samples from control and EV-A71-infected hSCARB2 KI mice were fixed with formalin buffered saline by perfusion and then dehydrated and embedded in paraffin and sectioned according to conventional procedures. Some sections were either stained with hematoxylin--eosin or were used for IHC analysis. Tissue sections were dewaxed and rehydrated in graded ethanol, and IHC was performed using a standard avidin--biotin immunoperoxidase technique. An anti-EV-A71 monoclonal antibody (MAB979; Millipore, USA), was used as a primary antibody. Electronic supplementary material ================================= {#Sec20} Supplementary Tables Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Electronic supplementary material ================================= Supplementary Information accompanies this paper at (10.1038/s41426-018-0201-3). The cDNA clone of EV-A71, pSVA14-Isehara ver4 was kindly provided by Prof. Satoshi Koike, Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan. This work was partly supported by National Key Technologies R&D Program for the 13th Five-year Plan (2017ZX10202101-004), Beijing Natural Science Foundation (7172016) and the National Science and Technology Major Projects of Infectious Disease (grants 2017ZX103304402). J.Z., C.F., and Q.H. conceived and designed the study. J.Z., N.C., K.Z., L.S., Y.Z., L.C., X.Z., Q.X., and Z.C. performed the experiments. S.Z. and C.W. participated in experiments. C.F. participated in the writing of the manuscript. J.Z. and Q.H. analyzed data and wrote the paper. All authors reviewed and approved the final paper. The authors declare that they have no conflict of interest.	{ "pile_set_name": "PubMed Central" }
Background ========== Alzheimer's disease (AD) afflicts millions of elderly people worldwide and is the most common cause of dementia among neurodegenerative diseases. AD symptoms include impaired memory, aphasia, and visuospatial deficits while the histopathological changes of AD are characterized by senile plaques, neurofibrillary tangles, and lipid granule accumulation. The major molecular component of senile plaques is aggregated amyloid-β (Aβ) peptide, which is generated from the amyloid precursor protein (APP) via sequential cleavage by β-secretase and the γ-secretase complex \[[@B1]\]. AD onset is assumed to be related to Aβ peptide polymerization. On the other hand, neurofibrillary tangles are composed of aberrantly phosphorylated tau protein \[[@B2]\]. In addition to the above pathological features, numerous studies reported a link between lipids and AD. In general, membrane lipids provide a milieu for transmembrane proteins and can modulate their function. Indeed, γ-secretase activity is affected by the lipid composition of the membrane with sphingolipids and cholesterol increasing and phosphatidylinositols decreasing its activity \[[@B3],[@B4]\]. Lipid carbon chain length and double bond position also affect γ-secretase activity. Hypercholesterolemia reportedly accelerates intraneuronal accumulation of Aβ oligomers and subsequent synapse loss, resulting in memory impairment in AD model mice \[[@B5]\]. Moreover, ω-3 polyunsaturated fatty acids were shown to have possible roles in AD prevention \[[@B6]\]. The docosahexaenoic acid (DHA) content of phospholipids (PLs) was lower in brain tissue and plasma of AD patients compared to those without cognitive impairment \[[@B7]\]. In addition, neuroprotectin D1, which is derived from DHA, was related to suppression of Aβ~42~-induced neurotoxicity \[[@B8]\]. Several lipidomic analyses for AD were recently reported. For instance, an analysis of postmortem brain tissues obtained from patients with late-onset AD showed enrichment in lysobisphosphatidic acids, sphingomyelins (SMs), ganglioside GM3, and cholesterol esters (ChEs) as well as novel region-specific lipid anomalies that were potentially linked to AD pathogenesis \[[@B9]\]. However, our current knowledge of the overall changes in lipids related to AD, especially for the time-course of their changes, remains low. Animal models are extensively used to study AD pathogenesis since brain samples can be obtained from pre-symptomatic to late stages of AD, which allows analysis of pathophysiological changes at different disease stages. A recent study showed that some lipid dysregulation occurred in postmortem AD brain tissues, including increased levels of ganglioside GM3 and ChEs, which was consistent with changes seen in AD mouse models \[[@B9]\], and supports the validity of using animal models to study AD disease mechanisms. In this study, we performed a global lipid metabolomic analysis using brain tissues and plasma obtained from transgenic mice expressing both mutated human APP and tau proteins \[[@B10]\] at 4 months (pre-symptomatic), 10 months (early disease stage), and 15 months (late stage). We focused not only on lipid changes observed in the brain but also on those that were consistently changed in both brain and plasma because these molecules could be candidate biomarkers for AD diagnosis. Materials and methods ===================== Animals ------- We used female transgenic mice hemizygous for transgenes encoding both human mutated APP and tau (APP/tau mice) as AD model mice. The mice were generated by crossing Tg2576 male transgenic mice hemizygously expressing human APP containing Swedish type mutations (Lys670Asn, Met671Leu) with JNPL3 female transgenic mice homozygously expressing mutant P301L four-repeat tau proteins. Wild-type tau females crossed with wild-type APP male mice were used as a control (no transgenes). Two-month-old mice were purchased from Taconics Farms, Inc. (Hudson, NY, USA) and raised to 4 (pre-symptomatic), 10 (early stage of symptoms), and 15 months (late stage) of age. At the given time point, mice were anesthetized with 2.0-2.5% isoflurane in the atmosphere and blood from the inferior vena cava was collected in tubes containing EDTA-2Na for plasma preparation. After blood collection, mouse brains were excised, and the cerebral hemispheres without the cerebellum were weighed and frozen in liquid nitrogen before storage at −80°C. Each group consisted of 5 mice except for the 15-month-old APP/tau mice (N=3). All animal experiments were performed in accordance with the Guidelines for Animal Experimentation and with approval from the Ethics Committee of Animal Care and Experimentation of the National Center for Geriatrics and Gerontology, Japan. Assay of human Aβ~40~ and tau proteins -------------------------------------- The right cerebral hemisphere from each animal (ca. 150 mg) was homogenized as described below in the section describing lipid extraction. After removal of methanol with a centrifugal vacuum concentrator (TOMY), Aβ~40~ in brain tissues (2 mg) was extracted by homogenization in 0.2 ml 70% formic acid with a small Dounce homogenizer and centrifugation at 100,000 × g for 1 h at 4°C in an ultracentrifuge (Optima MAX-E, Beckman Coulter, Tokyo, Japan). The supernatants were recovered and stored at −80°C until analysis. Frozen formic acid extracts were thawed on ice and neutralized with 20 volumes of 1 M Tris base. Within 1 h of neutralization, Aβ~40~ levels in the extracts were determined by an Enzyme-Linked Immunosorbent Assay (ELISA) kit for human Aβ~40~ (Wako Pure Chemical Industries, Ltd., Osaka, Japan) according to the manufacturer's instructions. Total protein in the extracts was assayed by a Pierce® BCA™ Protein Assay Kit (Thermo Fisher Scientific Ltd., Yokohama, Japan). Human tau protein in brain tissue from each mouse was analyzed according to the method described by Lewis et al. \[[@B10]\]. Two mg of EtOH-free homogenized brain tissue from each mouse was homogenized with a small Dounce homogenizer in 0.2 ml of tris-buffered saline (TBS) and centrifuged at 100,000 × g for 1 h at 4°C. The supernatant was collected as the TBS-soluble fraction. The pellet was homogenized similarly in 0.2 ml of 0.8 M NaCl and 10% sucrose in TBS, and after centrifugation at 150,000 × g for 15 min at 4°C, the supernatant was brought to 1% sarkosyl and incubated at 37°C for 1 h. The mixture was then centrifuged at 150,000 × g for 30 min at 4°C whereupon the supernatant and precipitate were collected as the sarkosyl-soluble and -insoluble fraction, respectively. TBS-soluble and sarkosyl-soluble fractions were combined and defined as the soluble fraction. Human tau levels in the soluble and sarkosyl-insoluble fractions were determined using an ELISA kit for human tau (Lifetechnology Japan Ltd., Tokyo, Japan). The sarkosyl-insoluble fraction was dissolved in 0.1 ml of 70% formic acid and neutralized with 20 volumes of 1M Tris base immediately before ELISA determination. Lipid extraction ---------------- Total lipids from right cerebral hemisphere tissues or plasma were extracted using Bligh & Dyer's (BD) method with minor modifications \[[@B11]\]. The right cerebral hemisphere from each animal (ca. 150 mg) was homogenized with zirconia beads in 1 mL of methanol at 4°C using a cell disruptor (TOMY), and the equivalent of 10 mg tissue was transferred into glass tubes. Plasma (100 μl) diluted with methanol was also transferred into glass tubes. A mixture of internal standards (ISs) was added to tissue homogenates or plasma: 1, 2-dipalmitoyl D6-3-sn glycerophosphatidylcholine (16:0/16:0PC-d6, 40 nmol/10 mg tissue or 20 nmol/100 μl plasma, Larodan, Malmo, Sweden), 1,2-caprylin-3-linolein (250 pmol and 20 nmol, Larodan), ^13^C-labelled tripalmitin (tripalmitin-1,1,1-^13^C~3~, 250 pmol and 2 nmol, Larodan), deuterated prostaglandin E~2~ (PGE~2~-d4, 5 ng in both tissue and plasma, Cayman Chemical, Ann Arbor, MI, USA) and deuterated leukotriene B~4~ (leukotriene B~4~-d4, 5 ng in both tissue and plasma, Cayman Chemical). Then, chloroform, methanol and 20 mM potassium phosphate (Kpi) buffer were added to achieve a volume ratio of buffer/methanol/chloroform = 0.8/2/1, and mixed vigorously for 5 min. Phase separation was achieved by adding 1 ml each of chloroform and 20 mM Kpi buffer and vortexing. The mixture was centrifuged at 1,000 × g for 10 min. For brain tissue, both the upper aqueous layer and bottom organic layer were collected. For plasma, 100 mM KCl/methanol/chloroform (48/47/3) was further added to the remaining organic samples after the upper aqueous layer was collected. After mixing vigorously for 5 min and centrifugation at 1,000 x g for 10 min, only the bottom organic layer was collected. Samples of the organic layer were dried under a gentle stream of nitrogen, and then dissolved in chloroform-methanol (1:1) at a concentration of 10 mg (tissue weight)/ml or 100 μL (plasma volume)/ml, and were stored at −90°C until use (BD sample). To distinguish alkenylacyl and alkyl PL species with the same exact mass as 34:1 ethanolamine plasmalogen (pPE) and 34:2 alkylacyl PE, a small aliquot of each BD sample was hydrolyzed using 0.5 N HCl as described previously (BD acid hydrolysis sample) \[[@B12]\]. Samples of the aqueous layer (3.2 ml) were subjected to solid extraction to obtain oxidative fatty acids. First, samples were diluted 10-fold using water adjusted to pH 3.0 with 1N HCl, and then applied to Oasis SPE cartridges (60 mg, Waters, Millford, MA, USA) preconditioned with 3 ml each of methanol and Milli Q water. After columns were washed with 3 ml of Milli-Q water followed by 3 ml hexane, the oxidative fatty acids were eluted with 3 ml methyl formate (MF). The MF fraction was dried under nitrogen, dissolved in chloroform-methanol (1:1) at a concentration of 10 mg (tissue weight)/ml or 100 μL (plasma volume)/ml, and stored at −90°C until use. Lipid metabolite analysis by reverse-phase liquid chromatography (RPLC)-time of flight mass spectrometry (TOFMS) ---------------------------------------------------------------------------------------------------------------- Total lipid samples (BD samples and BD acid hydrolysis samples) were analyzed by electrospray ionization (ESI)-TOFMS (LCT Premier XE; Waters Micro-mass, Waters) interfaced with an Acquity UPLC System (Waters). The MS was operated in a W-optics mode with 10,000 and 8,000 resolution in the positive and negative modes, respectively, using dynamic range extension. The scan range of the instrument was set at a mass-to-charge ratio (m/z) of 150--1200, and three functions (positive and negative ion modes, and another negative ion mode for in-source fragmentation) were recorded simultaneously for lipid species identification and quantification. The capillary voltage in the positive ion mode was set at 3.0 kV and the cone voltage 40 V, whereas in the negative ion mode the capillary and cone voltage was 2.5 kV and -40V, respectively. In in-source fragmentation of the negative ion mode, the aperture 1 voltage was set at 60 V. The desolvation gas was set to 600 L h^−1^ at a temperature of 350°C; the cone gas was set to 50 L/h and the source temperature was 120°C. The data acquisition rate was 0.5 s with a 0.01 s interscan delay. All analyses were acquired using the lock spray to ensure accuracy and reproducibility. A lock-mass of leucine enkephalin at 0.6 ng/mL in 50:50 acetonitrile:water containing 0.1% formic acid was used for the positive and negative ion modes (\[M+H\]^+^ = 556.2771 and \[M+H\]^-^ = 554.2615) with a flow rate of 5.0 μL/min via a lock spray ionization source. The data were collected in the centroid mode using MassLynx (Waters). Before the RPLC-ESI-TOFMS analysis, the dried BD and BD acid hydrolysis samples were redissolved in chloroform/methanol/isopropanol (1:2:4) supplemented with 0.2% formic acid and 0.028% ammonia at a final concentration of 10 mg tissue/ml solution or 100 μl plasma/ml solution, and filtered through a 0.2 μm pore size polyvinylidene difluoride membrane filter (Millipore, Bedford, MA, USA). The injection volume was fixed at 2 μL, and an ACQUITY UPLC BEH C18 column (1.7 mm i.d. × 150 mm, 1.7 μm) was used for separation. The column temperature was maintained at 45°C. The flow rate of the mobile phase was 70 μL/min. Mobile phase A consisted of acetonitrile/methanol/water = 18/18/4 (0.1% formic acid and 0.028% ammonia), while mobile phase B consisted of isopropanol (0.1% formic acid and 0.028% ammonia). The linear gradient increased from 0 to 15% B over 5 min, from 15% to 30% B for the next 20 min, from 30% to 45% B during the next 15 min, from 45% to 60% B for the next 10 min, from 60% to 80% B for the next 5 min, and immediately ramped to 85% B for the final 12 min. Blank runs were carried out randomly between samples to confirm that there was no chromatographic carryover. Structural analysis of lipid metabolites by RPLC-ESI-linear ion trap-MS ----------------------------------------------------------------------- For structural analysis of PL, SM and triacylglycerol (TAGs) species, BD samples were also analyzed using a Finnigan LTQ linear ion trap mass spectrometer (Thermo Fisher Scientific Inc. San Jose, CA, USA) interfaced with a Shimadzu Prominence HPLC system (Shimadzu, Kyoto, Japan). To analyze the fatty acid composition of PLs and SMs, data-dependent MS/MS (MS2) or MS3 analysis was performed in negative ion mode with an ion spray voltage of 4 kV and a scan range of m/z = 400--1000 as described previously \[[@B12]\]. For structural analysis of TAGs, data-dependent MS2 analysis was performed in the positive ion mode with an ion spray voltage of 4.5 kV and a scan range of m/z = 300--2000. The trap fill-time was 50 ms (full mass analysis) and 100 ms (MS^n^ analysis) in both ion modes. Nitrogen was used as a sheath gas (25 arbitrary units), and helium as a collision gas at a collision energy setting of 35%. BD samples were treated in the same way as RPLC-ESI-TOFMS analysis and applied onto an ACQUITY UPLC BEH C18 column (1.7 mm i.d. × 100 mm, 1.7 μm, Waters) kept at 45°C. LC conditions were the same as those in RPLC-ESI-TOFMS analysis. The classes of individual lipid species and their fatty acid compositions were determined with these data, together with RPLC-ESI-TOFMS data with exact mass and in-source fragmentation information. Targeted analysis of oxidative fatty acids by RPLC-ESI-triple quadrupole MS/MS ------------------------------------------------------------------------------ Oxidative fatty acids were measured as reported previously \[[@B13]\]. Briefly, MF fractions in solid extraction were analyzed by a 5500QTRAP quadrupole-linear ion trap hybrid mass spectrometer (AB Sciex, Framingham, MA, USA) interfaced with an ACQUITY UPLC System (Waters) equipped with an ACQUITY BEH C18 column (1.7 mm i.d. × 150 mm, 1.7 μm, waters), kept at 40°C. Samples were eluted with a mobile phase of water/acetate (100:0.1, v/v, solvent A') and acetonitrile/methanol (4:1, v/v, solvent B'). The gradient increased from 27 to 50% B' in five min, from 50 to 80% B' in the next 30 min, from 80 to 100% B' in the next five min, and kept at 100% B' for ten min with flow rates of 50 μL/min (0 -- 35 min), 50--100 μL/min (35 -- 40 min), and 100 μL/min (40 -- 50min). Analysis was performed with multiple reaction monitoring in the negative ion mode, with ion spray voltage, -4500 V; curtain (nitrogen), 10 arbitrary units; collision gas (nitrogen), 6 arbitrary units and a gas temperature of 350°C. We simultaneously measured 31 arachidonic acid metabolites, 14 eicosapentaenoic acid (EPA) metabolites, 13 DHA metabolites, 2 docosapentaenoic acid metabolites, and 2 linolenic acid metabolites. Declustering potential, collision energy and collision cell exit potential were optimized for each individual metabolite using authentic standard compounds (Cayman Chemical). In plasma samples from 15 month old control mice, one sample was omitted from the measurement due to instrumental problems (N=4). Data processing --------------- RPLC-ESI-TOFMS data were subsequently processed using the 2DICAL software package (2 Dimensional Image Converted Analysis of LCMS, Mitsui Knowledge Industry Co. Ltd., Tokyo, Japan) \[[@B14]\]. This software allowed detection and alignment of the ion-peak intensity derived from biomolecules identified with m/z and column retention times (RTs) in an amount of collected data. The main parameters of 2DICAL were set as follows: RT range 0.1--60.0 min and 0--37.5 min in the positive and negative ion modes, respectively, a mass range 400 to 1200 m/z, mass tolerance 0.3 m/z, mass window 0.3 m/z, an RT window of 0.60 min, and noise elimination level 2. The resulting three-dimensional matrix contains an arbitrarily assigned peak index (RT-m/z pairs) and ion intensity information (variables). The extracted ion peaks of each metabolite were normalized to that of the ISs. 16:0/16:0PC-d6 (Larodan Fine Chemicals) was used as an IS to normalize metabolites of PLs, lysoPLs, SMs, ceramides (Cers) and diacylglycerols (DAGs), which eluted at 0.1 to 37.5 min in RPLC. To normalize the TAG and ChE metabolites that eluted at 37.5 to 60 min, Tripalmitin-1,1,1-^13^C~3~ (Larodan Fine Chemicals) was used. Data for oxidative fatty acids were subsequently processed with Analyst™ (ver. 1.5.1, AB Sciex) software or MultiQuant™ Software (Version 2.1, AB Sciex). The integrated peak area of each metabolite was normalized to that of the IS (leukotriene B~4~-d4). Statistical and multiple classification analysis ------------------------------------------------ The IS-normalized peak list from RPLC-TOFMS analysis was first exported to the SIMCA-P+ for multivariate data analysis (Infocom Corp., Tokyo, Japan). Orthogonal partial least-squares-discriminant analysis (OPLS-DA) was used to select variables that strongly contributed to the difference between the two mice groups of the same ages. The Welch's t-test was then applied to assess statistical measures between the APP/tau and wild-type mice. For oxidative fatty acid analysis, Welch's t-tests were directly applied to the results with normalized peak areas. Results ======= Expression levels of Aβ~40~ and tau proteins in APP/tau mice ------------------------------------------------------------ We measured the protein expression levels of Aβ~40~ and tau in APP/tau mice by ELISA. In APP/tau mice, Aβ~40~ accumulated in the brain from 10 months of age (Additional file [1](#S1){ref-type="supplementary-material"}: Figure S1). On the other hand, soluble human tau protein was expressed at similar levels from 4 through 15 months of age. The levels of sarkosyl-insoluble tau appeared to be slightly higher at 10 and 15 months than at 4 months, but these differences did not reach statistical significance among all ages tested. In wild-type mice, both Aβ~40~ and human tau were under the detection limit at all tested ages. Analysis of brain tissues ------------------------- We first performed lipidomic analysis of BD samples extracted from brain tissues of APP/tau and wild-type mice at 4, 10, and 15 months of age using RPLC-ESI-TOFMS. When creating a two-dimensional map of RT versus m/z values of the individual precursor ion peaks in the positive ion mode, the intensities of a few spots that were apparently identified as ChEs were higher in APP/tau mice compared to wild-type mice at 4 months (Additional file [2](#S2){ref-type="supplementary-material"}: Figure S2). Likewise, at 10 and 15 months, these ChE spots retained the higher intensities in APP/tau mice compared to age-matched wild-type mice (data not shown). In the positive ion mode, a total of 803 peaks were detected and divided into two groups based on their retention time (RT). For RTs of 0.1--37.5 min, the 695 peaks contained mainly glycerophospholipids such as PLs, lysoPLs and sphingolipids such as SMs and Cers. For RTs of 37.5--60 min, the 108 peaks represented neutral lipids such as TAGs and ChEs. Peaks in each RT category were then separately processed by OPLS-DA (Additional file [3](#S3){ref-type="supplementary-material"}: Figure S3) because glycerophospholipids and neutral lipids are the two major classes of lipid metabolites with different physical properties. The OPLS-DA score plots for brain tissues indicated that APP/tau and wild-type mice could be reliably discriminated using established models at 4 (Additional file [3](#S3){ref-type="supplementary-material"}: Figure S3), 10 and 15 months (data not shown). To detect with high confidence the metabolites contributing to the discrimination between the two groups (APP/tau vs. wild-type) that were located at extreme ends of "S" of the OPLS-DA loading S-plot, we screened variables from this plot with threshold values of w\[[@B1]\] \> \\|0.05\\|, p(corr)\[[@B1]\] \> \\|0.6\\|, which represent the magnitude of contribution (weight) and reliability (correlation), respectively. For brain tissue from 4-month-old animals (pre-symptomatic phase), 9 and 12 metabolites in total (RT, 0.1-60 min) that exceeded this threshold value showed higher and lower levels in APP/tau compared to wild-type mice, respectively (Additional file [3](#S3){ref-type="supplementary-material"}: Figure S3). In addition, 8 and 8 metabolites at 10 months (early disease stage) and 5 and 58 metabolites at 15 months (late stage) had higher and lower levels in APP/tau compared to wild-type mice, respectively, and met the criteria (data not shown). The relative quantification data of the identified metabolites that meet the OPLS-DA loading S-plot criteria (unidentified peaks were omitted) were summarized in Additional file [4](#S4){ref-type="supplementary-material"}: Table S1. Of these metabolites, the levels of 2, 3 and 2 species were significantly increased and 4, 3 and 24 were significantly decreased in APP/tau mice compared to wild-type mice at 4, 10 and 15 months, respectively, by Welch's t-test (Table [1](#T1){ref-type="table"}). Particularly, docosahexaenoyl-ChE (22:6 ChE) and 38:2 pPE (p18:1/20:1) were consistently increased and decreased, respectively, across the tested ages in APP/tau mice compared to wild-type mice (Figure [1](#F1){ref-type="fig"}). Other ChE species such as 18:1ChE were under the detection limit (data not shown). Notably, many pPE species (e.g., 36:2pPE (p18:1/18:1), 36:4pPE (p16:0/20:4), 38:4pPE (p16:0/22:4), 38:4pPE (p18:0/20:4), 40:4pPE (p18:0/22:4), 40:6pPE (p18:0/22:6)) and SM species (e.g., 36:2SM (d18:2/N18:0), 38:1SM (d18:1/N20:0), 34:1SM (d18:1/N16:0)) were decreased at 15 months (Table [1](#T1){ref-type="table"}). Decreases in several phosphatidylcholines (PCs) and PEs mainly with polyunsatulated fatty acids (PUFAs) were also detected in the late symptomatic phase (Table [1](#T1){ref-type="table"}). As for triacylglycerol (TAG), the level of 60:12TAG (16:0/22:6/22:6) was increased (1.33-fold) in 10 month old animals, whereas most TAG levels were decreased in APP/tau mice compared to wild type mice at 15 months, and showed statistical significance for 52:3TAG (16:0/18:1/18:2, 16:1/18:1/18:1), 53:6TAG and 54:4TAG (16:0/18:0/20:4) (0.40, 0.54 and 0.31-fold, respectively) (Table [1](#T1){ref-type="table"} and Additional file [4](#S4){ref-type="supplementary-material"}: Table S1). ###### Fold changes in significantly changed lipids in brain tissues Molecular species 4 months 10 months 15 months ------------------------------------------- -------------- --------------- --------------- ----------- ------ --------------- *Phosphatidylcholine (PC)* 30:0PC (14:0/16:0) 1.35 0.026\* 1.14 0.200 1.03 0.198 32:0PC (16:0/16:0) 1.02 0.815 0.87 0.093 0.83 0.011\* 34:2PC (16:0/18:2) 1.32 0.190 1.63 0.014\* 1.03 0.282 36:3PC (16:0/20:3) 1.06 0.550 1.07 0.441 1.28 0.029\* 38:1PC (18:0/20:1) 0.84 0.092 0.87 0.155 0.77 0.010\\ 38:4PC (16:0/22:4) 1.07 0.462 0.92 0.279 0.86 0.011\* 38:5PC (18:1/20:4) 1.10 0.104 0.93 0.426 0.88 0.021\* 40:4PC (18:0/22:4) 1.05 0.596 0.91 0.188 0.84 0.003\\ 40:7PC (18:1/22:6) 1.00 0.984 0.90 0.615 0.87 0.021\* 34:1ePC 1.04 0.542 0.92 0.338 0.86 0.029\* *Phosphatidylethanolamine (PE)* 38:1PE (20:0/18:1) 0.74 0.021\* 0.89 0.413 0.66 0.024\* 38:5PE (18:1/20:4) 0.99 0.942 0.94 0.352 0.85 0.023\* 40:4PE (18:0/22:4) 1.10 0.265 0.91 0.269 0.89 0.035\* *Ethanolamine plasmalogen (pPE)* 36:2pPE (p18:1/18:1) 0.76 0.299 0.88 0.183 0.86 0.042\* 36:4pPE (p16:0/20:4) 1.02 0.810 0.91 0.359 0.89 0.027\* 38:2pPE (p18:1/20:1) 0.78 0.037\* 0.83 0.037\* 0.76 0.034\* 38:4pPE (p16:0/22:4) 1.02 0.682 0.92 0.310 0.82 0.009\\ 38:4pPE (p18:0/20:4) 0.94 0.819 0.93 0.796 0.87 0.029\* 40:4pPE (p18:0/22:4) 0.96 0.611 0.94 0.399 0.88 0.013\* 40:6pPE (p18:0/22:6) 1.02 0.822 0.89 0.182 0.87 0.043\* *Sphingomyelin (SM)* 34:1SM (d18:1/N16:0) 0.98 0.863 0.97 0.675 0.76 0.007\\ 36:2SM (d18:2/N18:0) 0.94 0.335 0.81 0.010\* 0.86 0.027\* 38:1SM (d18:1/N20:0) 0.90 0.278 0.85 0.067 0.79 0.006\\ 42:2SM (d18:1/N24:1) 0.78 0.048\* 0.86 0.388 0.60 0.066 *Cerebroside* 42:2cerebroside 0.81 0.045\* 0.92 0.334 0.84 0.137 *Diacylglycerol (DAG)* 38:4DAG 0.94 0.420 0.74 0.016\* 0.78 0.043\* *Triacylglycerol (TAG)* 52:3TAG (16:0/18:1/18:2, 16:1/18:1/18:1) 0.87 0.353 0.68 0.143 0.40 0.002\\ 53:6TAG 1.05 0.765 0.91 0.648 0.54 0.003\\ 54:4TAG (16:0/18:0/20:4) 1.09 0.463 1.04 0.801 0.31 \<0.001\\\* 60:12TAG (16:0/22:6/22:6) 0.99 0.935 1.33 0.011\* 0.87 0.544 *Cholesterol ester (ChE)* 22:6ChE 5.22 0.002\\ 6.96 0.004\\ 5.43 0.032\* ^1^ Fold change (ratio of APP/tau to wild-type) is the mean of N=3-5. ^2^ Welch's t-test (\*p\<0.05, \\p\<0.01,\\\*p\<0.001). ![Changes in 22:6ChE and 38:2pPE (p18:1/20:1) levels in brain tissue from APP/tau and wild-type mice. The relative levels of metabolites are represented by the IS normalized peak heights using 16:0/16:0PC-d6 for 38:2pPE and tripalmitin-1,1,1-13C3 for 22:6ChE as ISs. These two metabolites were significantly changed in the brain across the tested ages. \*p\< 0.05, \\p\< 0.01 by Welch's t-test. Values are mean ± SD of N = 3--5.](1476-511X-12-68-1){#F1} Free PUFAs and their oxidized metabolites extracted from brain tissues were then measured using a multiple reaction monitoring system, and 16 species were detected (Additional file [5](#S5){ref-type="supplementary-material"}: Table S2). Of these, 9 species were significantly changed between APP/tau and wild-type mice as shown in Table [2](#T2){ref-type="table"}. The numbers of variable metabolites were 2, 7 and 0 species at 4, 10 and 15 months, respectively. No free PUFAs or their oxidized metabolites were consistently changed between APP/tau and wild-type mice across the ages tested (Additional file [5](#S5){ref-type="supplementary-material"}: Table S2). Pathway analysis of arachidonic acid metabolites is shown in Additional file [6](#S6){ref-type="supplementary-material"}: Figure S4. A marked decrease in prostaglandin D~2~ and 15-hydroxyeicosatetraenoic acid (HETE) was observed at 10 months (Figure [2](#F2){ref-type="fig"}). ###### Fold changes of the significantly changed oxidized fatty acids in brain tissues Molecular species 4 months 10 months 15 months ------------------------------------------------------ -------------- ---------------- --------------- ---------------- ------ ------- *Metabolites derived from arachidonic acid* Prostaglandin D~2~ 0.99 0.941 0.46 \< 0.001\\\* 0.63 0.059 Thromboxane B~2~ 1.24 0.113 0.66 0.013\* 1.07 0.845 12-HHT 1.19 0.149 0.67 0.005\\ 0.98 0.952 12-HETE 0.90 0.777 1.77 0.049\* 1.03 0.931 15-HETE 1.03 0.747 0.71 \<0.001\\\* 0.87 0.604 11,12-EpETrE 0.84 0.225 0.83 0.007\\ 1.04 0.886 14,15-EpETrE 0.85 0.053 0.79 0.029\* 0.97 0.810 *Metabolites derived from eicosapentaenoic acid* 17,18-diHETE 2.12 \< 0.001\\\* 1.10 0.602 1.42 0.300 *Metabolites derived from docosahexaenoic acid* 19,20-diHDoPE 1.72 0.002\\ 1.20 0.251 1.38 0.261 ^1^ Fold changes (ratio of APP/tau to wild-type) are means of N=3-5. ^2^ Welch's t-test (\*p\<0.05, \\p\<0.01,\\\*p\<0.001). HHT, hydroxy-heptadecatrienoic acid; HETE, hydroxyeicosatetraenoic acid; EpETrE, epoxyeicosatrienoic acid; diHETE, dihydroxyeicosatetraenoic acid; diHDoPE, dihydroxydocosapentaenoic acid. ![Changes of prostaglandin D~2~and 15-HETE levels in brain tissue from APP/tau and wild-type mice. The relative levels of metabolites are represented by the IS normalized peak areas using leukotriene B~4~-d4 as an IS. \\\*p\< 0.001 by Welch's t-test. Values are mean ± SD of N = 3--5.](1476-511X-12-68-2){#F2} Analysis of plasma ------------------ In the positive ion mode of BD samples from plasma, 385 and 308 peaks were detected on the basis of their RT, 0.1-37.5 min and 37.5-60 min, respectively. Next, we analyzed changes in lipid species levels in plasma by OPLS-DA loading S-plot (data not shown) in the same manner as for the brain tissues. At 4 months of age, 35 and 3 metabolites (RT, 0.1-60 min) with higher and lower levels in APP/tau mice than in wild-type mice, respectively, were detected. Forty-five and 9 metabolites displayed higher and lower levels in APP/tau mice compared to wild-type mice, respectively, at 10 months, and 32 and 22 were detected at 15 months. Relative quantification results of these metabolites are shown in Additional file [7](#S7){ref-type="supplementary-material"}: Table S3 (unidentified peaks were omitted) and the fold changes of metabolites with significantly different levels between APP/tau and wild-type mice are listed in Table [3](#T3){ref-type="table"}. The levels of 23 metabolites (22 were increased and 1 was decreased in APP/tau mice) at 4 months, 32 metabolites (26 and 6) at 10 months and 12 metabolites (7 and 5) at 15 months were significantly different between APP/tau and wild-type mice. As for PUFAs and their oxidized metabolites, the 34 species were detected (Additional file [8](#S8){ref-type="supplementary-material"}: Table S4) and, 7, 3 and 9 species levels were also significantly changed between the two mice groups at 4, 10, and 15 months, respectively (Table [4](#T4){ref-type="table"}). ###### Fold changes of the significantly changed lipids in plasma Molecular species 4 months 10 months 15 months -------------------------------------------- -------------- --------------- --------------- --------------- ------ ----------- *Phosphatidylcholine (PC)* 16:1LPC 0.99 0.960 0.59 0.026\* 1.07 0.724 18:0LPC 1.50 0.019\* 1.27 0.281 1.01 0.921 20:5LPC ^3^ 1.45 0.011\* 0.82 0.440 1.26 0.190 22:6LPC 1.45 0.011\* 1.03 0.889 1.10 0.619 34:1PC (16:0/18:1) 1.24 0.037\* 1.00 0.983 1.27 0.071 34:2PC (16:0/18:2) 1.27 0.053 1.17 0.007\\ 1.21 0.015\* 34:3PC (16:0/18:3) 1.06 0.698 0.65 0.034\* 0.98 0.936 35:2PC 1.68 0.030\* 1.55 0.013\* 1.27 0.035\* 36:1PC (18:0/18:1) 1.55 0.010\\ 1.16 0.256 1.19 0.109 36:2PC (18:0/18:2) 1.48 0.024\* 1.35 0.017\* 1.11 0.139 36:5PC (16:0/20:5) 1.51 0.011\* 0.81 0.509 1.40 0.374 37:2PC (19:0/18:2) 1.33 0.144 1.47 0.008\\ 1.26 0.071 38:3PC (18:0/20:3) 1.52 0.045\* 1.38 0.209 1.17 0.497 38:4PC (18:1/20:3) 0.99 0.957 0.63 0.031\* 0.82 0.137 38:5PC (18:0/20:5) 1.74 0.008\\ 1.08 0.783 1.24 0.266 38:5PC (18:1/20:4) 1.13 0.331 0.79 0.018\* 0.84 0.149 38:7PC 1.08 0.644 0.48 0.019\* 0.94 0.807 40:4PC (20:0/20:4) 1.50 0.018\* 1.47 0.077 1.12 0.313 40:5PC (18:0/22:5) 1.71 0.007\\ 1.25 0.182 1.30 0.349 40:6PC (18:0/22:6) 1.64 0.033\* 1.39 0.099 1.06 0.629 40:7PC (18:1/22:6) 1.06 0.642 0.72 0.022\* 0.73 0.072 42:6PC (20:0/22:6) 1.41 0.002\\ 1.52 0.012\* 1.38 0.110 34:2ePC (16:0e/18:2) 1.52 \<0.001\\\* 1.81 \<0.001\\\* 1.47 0.011\* *phosphatidylinositol (PI)* 38:5PI (18:0/20:5) 1.53 0.024\* 0.92 0.748 2.00 0.002\\ *Sphingomyelin (SM)* 34:1SM (d18:1/N16:0) 1.25 0.057 1.35 0.027\* 1.30 0.032\* 42:2SM 1.37 0.051 1.39 0.024\* 1.38 0.069 42:3SM (d18:2/N24:1) 1.48 0.030\* 1.22 0.061 1.30 0.141 *Triacylglycerol (TAG)* 52:3TAG (16:0/18:1/18:2) 1.21 0.557 1.21 0.558 0.58 0.045\* 52:4TAG (16:0/18:2/18:2) 1.57 0.250 1.41 0.421 0.40 0.032\* 52:5TAG (16:0/18:2/18:3) 1.04 0.903 0.74 0.314 0.49 0.011\* 54:3TAG (18:0/18:1/18:2) 0.64 0.553 0.62 0.411 0.48 0.043\* 54:4TAG (18:1/18:1/18:2) 0.93 0.919 0.97 0.950 0.37 0.038\* 56:6TAG (18:0/18:2/20:4) 2.00 0.209 2.75 0.016\* 0.82 0.500 58:10TAG (16:0/20:4/22:6) 1.40 0.311 2.07 0.020\* 0.88 0.509 58:11TAG 2.44 0.038\* 1.92 0.027\* 1.70 0.205 58:11TAG 1.87 0.030\* 1.69 0.176 1.64 0.383 58:12TAG (18:3/20:4/20:5) 2.06 0.032\* 1.91 0.168 2.65 0.005\\ 60:10TAG 1.68 0.131 2.24 0.021\* 0.90 0.762 60:11TAG (16:2/22:4/22:5) 1.43 0.178 2.04 \<0.001\\\* 1.51 0.168 60:11TAG (16:0/22:5/22:6) 1.66 0.335 1.98 0.007\\ 1.39 0.401 60:11TAG (18:2/20:3/22:6) 1.80 0.018\* 2.13 0.005\\ 1.42 0.304 60:12TAG (18:2/20:5/22:5) 1.67 0.106 1.75 0.177 2.80 0.011\* 60:12TAG (18:3/20:4/22:5) 1.38 0.375 2.85 0.007\\ 1.37 0.356 60:12TAG (18:3/20:4/22:5) 2.25 0.039\* 2.08 0.068 2.81 0.087 60:12TAG (18:3/20:4/22:5) 2.29 0.154 2.04 0.021\* 1.64 0.308 60:13TAG (18:2/20:5/22:6) 1.66 0.086 2.00 0.015\* 2.58 0.067 62:11TAG (18:2/22:4/22:5, 16:1/22:5/24:5) 1.64 0.063 1.77 0.031\* 1.58 0.232 62:12TAG (18:3/22:4/22:5) 1.56 0.147 2.27 0.003\\ 1.35 0.370 62:13TAG (16:1/22:6/24:6, 20:2/20:5/22:6) 2.28 0.121 1.77 0.003\\ 1.82 0.363 62:13TAG (18:1/22:6/22:6) 2.09 0.106 2.36 0.029\* 2.53 0.272 62:14TAG (18:2/22:6/22:6) 2.25 0.119 2.45 \<0.001\\\* 2.20 0.274 64:14TAG 1.57 0.215 2.06 0.008\\ 1.95 0.251 65:14TAG 1.83 0.052 2.03 \<0.001\\\* 1.52 0.299 *Cholesterol ester (ChE)* 18:1ChE 0.74 0.016\* 0.78 0.089 1.12 0.423 18:2ChE 1.16 0.170 1.17 0.029\* 1.16 0.307 20:5ChE 1.38 0.041\* 0.93 0.762 1.33 0.138 ^1^ Fold changes (ratio of APP/tau to wild-type) are means of N=3-5. ^2^ Welch's t-test (\*p\<0.05, \\p\<0.01,\\\*p\<0.001). ^3^ This specie was picked up not from OPLS-DA loading S-plot but manually. ###### Fold changes of the significantly changed oxidized fatty acids in plasma Molecular species 4 months 10 months 15 months ------------------------------------------------------ -------------- --------------- --------------- --------- ------ ----------- *Metabolites derived from arachidonic acid* 17-HETE 1.59 0.018\* 1.41 0.112 1.36 0.111 18-HETE 1.46 0.021\* 1.48 0.105 1.67 0.011\* 5,15-diHETE 2.02 0.002\\ 1.42 0.159 1.97 0.002\\ *Metabolites derived from eicosapentaenoic acid* 8-HEPE 1.55 0.125 1.65 0.095 2.00 0.006\\ 15-HEPE 1.20 0.314 1.47 0.197 1.51 0.020\* 17,18-EpETE 1.66 0.165 1.46 0.157 1.85 0.004\\ 14,15-diHETE 2.16 0.032\* 1.80 0.051 2.21 0.122 17,18-diHETE 2.11 0.003\\ 1.51 0.116 2.22 0.004\\ *Metabolites derived from docosahexaenoic acid* 7-HDoHE 1.28 0.336 1.46 0.102 1.46 0.044\* 10-HDoHE 1.62 0.047\* 1.11 0.550 1.52 0.252 19,20-diHDoPE 1.67 0.010\* 1.42 0.027\* 1.80 0.033\* *Metabolites derived from linolenic acid* 9-HOTrE(alpha) 1.64 0.202 2.53 0.023\* 2.08 0.108 13-HOTrE(alpha) 1.50 0.147 1.87 0.041\* 1.73 0.008\\ ^1^ Fold changes (ratio of APP/tau to wild-type) are means of N=3-5. ^2^ Welch's t-test (\*p\<0.05, \\p\<0.01,\\\*p\<0.001). HETE, hydroxyeicosatetraenoic acid; diHETE, dihydroxyeicosatetraenoic acid; HEPE, hydroxyeicosapentaenoic acid; EpETE, epoxyeicosatetraenoic acid; HDoHE, hydroxydocosahexaenoic acid; diHDoPE, dihydroxydocosapentaenoic acid; HOTrE, hydroxyoctadecatrienoic acid . Concomitant changes in brain tissues and plasma ----------------------------------------------- Finally, we assessed whether the significant changes in lipid species were consistent for brain tissues and plasma. We found that 34:2PC (16:0/18:2) was concomitantly increased in brains (1.63-fold, p = 0.014) and plasma (1.17-fold, p = 0.007) of APP/tau mice compared to wild-type mice at 10 months (Tables [1](#T1){ref-type="table"} and [3](#T3){ref-type="table"}). Furthermore, 52:3TAG was decreased both in brain (0.4-fold, p=0.002) and plasma (0.58-fold, p=0.045) at 15 months (Tables [1](#T1){ref-type="table"} and [3](#T3){ref-type="table"}). The levels of 19,20-dihydroxy-docosapentaenoic acid (19,20-diHDoPE) and 17,18-dihydroxy-eicosatetraenoic acid (17,18-diHETE) in APP/tau mice at 4 months were significantly higher both in brain (1.72-fold, p = 0.002 for 19,20-diHDoPE and 2.12-fold, p \< 0.001 for 17,18-diHETE) and plasma (1.67-fold, p = 0.010 for 19,20-diHDoPE and 2.11-fold, p = 0.003 for 17,18-diHETE) than those from wild-type mice (Figure [3](#F3){ref-type="fig"}). ![Consistent changes in the levels of two ω-3 fatty acid metabolites in brain tissue and plasma. The relative levels of 19,20-diHDoPE and 17,18-diHETE in brain tissue (A) and plasma (B) are represented by the IS normalized peak areas using leukotriene B~4~-d4 as an IS. \*p \< 0.05, \\p\< 0.01, \\\*p \< 0.001 by Welch's t-test. Values are mean ± SD of N = 3--5.](1476-511X-12-68-3){#F3} Discussion ========== In the present study we performed a lipidomic analysis of brain tissue and plasma from APP/tau mice and corresponding control (wild-type) mice at both presymptomatic and post-symptomatic phases using RPLC-ESI-TOFMS, RPLC-ESI-linear ion trap-MS, and RPLC-ESI-triple quadrupole MS/MS. AD brains display several characteristic pathological features, including accumulation of amyloid plaques composed of Aβ and neurofibrillary tangles composed of hyperphosphorylated tau protein. APP/tau mice are suitable animal models to study lipid dysregulation over very early to late stages of AD because these animals are destined to develop the disease based on currently proposed pathogenesis mechanisms. In brain tissue from our model mice, 22:6ChE levels were significantly increased in APP/tau mice compared to controls at all stages examined. In agreement with our results, 22:6ChE was reported to be increased in forebrains of mutated APP/presenilin-1 mice (11-fold, p \< 0.01) and mutated APP mice (3-fold, borderline significance), but not of presenilin-1 mice (0.9-fold) at 9--11.5 months of ages, compared with wild-type mice \[[@B9]\]. In postmortem entorhinal cortex of AD patients, increased levels of ChE (1.7-fold, p = 0.027) were also observed, although the predominantly increased species was different between AD humans (16:1ChE, 16:0ChE and 18:1ChE) and mice with mutated APP (22:6ChE) \[[@B9]\]. The elevation in ChE levels in our mouse model had already occurred at 4 months of age (Figure [1](#F1){ref-type="fig"}) when Aβ accumulation was not yet observed (Additional file [1](#S1){ref-type="supplementary-material"}: Figure S1), which supports the idea that intracellular ChE levels regulate Aβ generation \[[@B15]\]. Other notable lipid dysregulation observed in brains of APP/tau mice was the reduced levels of several pPE species, particularly at the late stage of AD (15 months). In brains from AD patients, plasmalogen levels were reported to be decreased \[[@B16],[@B17]\], suggesting that the mechanisms for this reduction could be common between humans and APP/tau mice. Marked increases in Aβ at 15 months of age might promote the formation of reactive oxygen species, which oxidize pPE, resulting in reduced pPE levels. It was reported that pPE decreases γ-secretase activities \[[@B18]\], and that Aβ destabilized the rate-limiting enzyme of plasmalogen synthesis, alkyl-dihydroxyacetonephosphate synthase, through peroxisomal dysfunction \[[@B19]\]. The levels of 38:2pPE (p18:1/20:1) were consistently decreased across the tested ages in APP/tau mice, although the pathophysiological significance of this lipid species in AD progression awaits evaluation. The elevated levels of 12-HETE observed in APP/tau mice at early stages of AD (10 months, Table [2](#T2){ref-type="table"}) were also consistent with a previous report showing that brains of AD patients exhibited higher levels of 12-HETE than control subjects \[[@B20]\]. As suggested in human cases, 12-HETE may be linked to oxidative stress and neurodegeneration in AD as well as being involved in learning and memory processes \[[@B21]\]. The decreased levels of SM species with medium-chain fatty acids at late stages of AD were also in agreement with a previous study using human postmortem AD brains in which medium-length SMs such as d18:0/N20:0 and d18:1/N20:0 were shown to be slightly down-regulated (p \< 0.05) \[[@B9]\]. These decreases might reflect demyelination as in the case of human AD brains described by Svennerholm and Gottfries \[[@B22]\]. Aβ42 was reported to accelerate SM degradation by activating neutral sphingomyelinase \[[@B23]\]. Increased levels of 60:12TAG bearing two DHAs (16:0/22:6/22:6, 1.33-fold) in 10 month old animals were in agreement with the previous report in which the levels of 56:7TAG containing docosahexaenoic acid (22:6FA) were increased in entorhinal cortices from AD patients. Since DHA-conjugated ChE was also increased in brains from APP/Tau mice at all ages tested (Figure [1](#F1){ref-type="fig"}), the AD brain appears to accumulate DHA in both TAG and ChE forms during the early symptomatic phase. These lipids might be involved in the production of DHA-derived messenger molecules such as neuroprotectin D1, which are reported to play neuroprotective roles in AD brain \[[@B8]\]. On the other hand, most APP/tau mice TAG levels were decreased compared with wild type mice at 15 months, particularly 52:3TAG (16:0/18:1/18:2, 16:1/18:1/18:1), 53:6TAG and 54:4TAG (16:0/18:0/20:4). A decrease in AA containing TAG (58:7TAG) was also observed in the prefrontal cortices of AD patients and in forebrains from 9--11.5 month old presenilin 1-APP mice compared to control animals \[[@B9]\]. Hydrolysis of these TAGs might thus be enhanced for their conversion into various structural and signaling lipid molecules, including eicosanoids, DAGs, monoacylglycerols (MAG) and phospholipids. In addition to the changes described above, significant decreases in PCs and PEs mainly with PUFAs were also detected during the late symptomatic phase. These phenomena might reflect fatty acid chain oxidation caused by the increased oxidative stress derived from increased Aβ peptide levels \[[@B24]\]. Several arachidonic acid metabolites in the brain also showed different levels when comparing APP/tau mice with their controls. Levels of PGD~2~, which mediates neural damage by Aβ~42~\[[@B25]\], were lowered at 10 months (Figure [2](#F2){ref-type="fig"}), reflecting a possible protective reaction in the brain. Inflammation is characteristic of AD brain tissues as aggregated Aβ and phosphorylated tau proteins are associated with increased levels of inflammatory cytokines such as interleukin-6 and tumor necrosis factor-α \[[@B26]\]. While levels of some cyclooxygenase (COX)-mediated metabolites such as TXB~2~ and 12-HHT were also decreased, others (PGE~2~, 6-keto PGF~1α~ and PGF~2α~) were not (Additional file [6](#S6){ref-type="supplementary-material"}: Figure S4). Overall, these results suggest that not all COX pathways but instead several downstream pathways (such as PGD~2~ synthesis) might be impaired in AD brains during the early symptomatic phase. 15-HETE is known to be a ligand for peroxisome proliferator-activated receptor γ (PPARγ) \[[@B27]\] and might have anti-inflammatory activity in the AD brain. In this context, decreases in 15-HETE levels in the early phase of AD might promote lowered resistance to AD progression. We also observed decreases in both 11,12-epoxy-5,8,11-eicosatrienoic acid (11,12-EpETrE) and 14,15-EpETrE during the early symptomatic periods (Table [2](#T2){ref-type="table"}). The roles of these decreases in AD progression should be evaluated in the future. Because metabolites that are consistently changed in both brain tissues and plasma from APP/tau mice compared to controls could be candidate diagnostic biomarkers for AD, we examined changes in lipid species levels in plasma. In both brain and plasma, 34:2PC (16:0/18:2) and 52:3TAG were consistently increased and decreased, respectively, in APP/tau mice compared with wild-type mice only after disease onset (10 or 15 months). Regarding oxidative fatty acid metabolites, 19,20-diHDoPE and 17,18-diHETE were consistently increased in both brain tissues and plasma from APP/tau mice before symptoms appeared at 4 months (Tables [2](#T2){ref-type="table"} and [4](#T4){ref-type="table"}, Figure [3](#F3){ref-type="fig"}), and these increases were essentially maintained at 10 and 15 months of age in plasma but not the brain. 19,20-diHDoPE and 17,18-diHETE are stable metabolites of DHA and EPA, respectively, that are mediated by cytochrome P450 (CYP) and soluble epoxide hydrolase (EH). In mouse brain, CYP2C29, CYP2C37, CYP2C38, and CYP2C40 were identified as arachidonic acid epoxygenases \[[@B28]\] and appear to be responsible for EPA and DHA epoxidation. Furthermore, soluble EH was reported to be expressed in various regions of mouse brain, including the cerebral cortex and hippocampus \[[@B29]\]. Levels of the intermediate metabolites, 19,20-epoxy-docosapentaenoic acid (19,20-EpDPE) and 17,18-epoxy-eicosatetraenoic acid (17,18-EpETE), which are generated by P450 epoxygenase in neuronal and glial cells, were below the quantification limit in brain (data not shown) and thus appeared to be quickly metabolized into their stable metabolites, 19,20-diHDoPE and 17,18-diHETE, respectively. Although the physiological functions of these metabolites in the central nervous system are unknown, 19,20-EpDPE and 17,18-EpETE were shown to act as vasodilators in cerebral arteries \[[@B30]\]. Since increased blood pressure is a risk factor for AD \[[@B31]\], these metabolites might counteract the vasoconstrictive effects of Aβ. Furthermore, 17,18-EpETE was reported to have anti-inflammatory properties mediated through PPARγ in lung tissues \[[@B32]\]. Therefore, 19,20-EpDPE and 17,18-EpETE in the brain might play a protective role against the onset of AD, but they are rapidly metabolized into their less active metabolites, 19,20-DiHDoPE and 17,18-diHETE. In this context, the expression of microsomal EH was reported to be significantly elevated in the hippocampus and associated cortex in AD patients \[[@B33]\]. However, the exact mechanisms leading to the consistent increases in these two compounds in brain and plasma remain unknown. During the pre-symptomatic phase, docosahexaenoyl lysoPC (22:6LPC) and eicosapentaenoy-lysoPC (20:5LPC) in plasma were significantly higher in APP/tau mice than in control mice (Table [3](#T3){ref-type="table"}). LysoPC may represent a preferred physiological carrier of fatty acids to allow passage across the blood--brain barrier \[[@B34]\], and thus brain DHA and EPA might be at least partly supplemented by plasma 22:6LPC and 20:5LPC. The contribution of the lysoPC species to supplementation of DHA and EPA in the brain needs to be investigated. A previous report showed that serum pPE levels were higher in Alzheimer's disease patients than in age-matched controls \[[@B35]\] and increased Cer and decreased SM levels in plasma were observed in patients with early-stage AD \[[@B36]\]. However, these changes were not observed in our APP/tau mice (Table [3](#T3){ref-type="table"}). Conclusions =========== We performed a lipidomic analysis using brain tissue and plasma obtained from APP/tau mice at three AD stages. Two lipid species, 38:2pPE (p18:1/20:1) and 22:6ChE, were consistently decreased and increased, respectively, across the tested ages in APP/tau mice compared to wild-type mice. As reported for human AD brains, brain tissue from APP/tau mice showed increased 12-HETE levels at the early symptomatic phase and decreased levels of both pPEs and SMs with medium chain fatty acids at the late symptomatic phase compared with control mice. In addition to these previously reported changes, 19,20-diHDoPE and 17,18-diHETE, the stable metabolites of 19,20-EpDPE and 17,18-EpETE that have a potential counteracting role for AD, were increased in APP/tau brain tissue at the pre-symptomatic phase, and concomitant increases in plasma were observed. Furthermore, 15-HETE, which has a potential protective role in AD, was decreased during the early symptomatic phase in APP/tau mice. Our results provide fundamental information on the further precise analysis of lipids in AD pathogenesis. These changes detected in APP/tau mice models of AD should be examined in patients in the near future. Abbreviations ============= AD: Alzheimer's disease; Aβ: Amyloid β; APP: Amyloid precursor protein; BD: Bligh & Dyer; Cer: Ceramide; ChE: Cholesterol ester; CYP: Cytochrome P450; DAG: Diacylglycerol; DHA: Docosahexaenoic acid; 19,20-diHDoPE: 19,20-dihydroxy-4Z,7Z,10Z,13Z,16Z-docosapentaenoic acid; 17,18-diHETE: 17,18-dihydroxy-5Z,8Z,11Z,14Z-eicosatetraenoic acid; EH: Epoxide hydrolase; ELISA: Enzyme-linked immunosorbent assay; 19,20-EpDPE: 19,20-epoxy-4Z,7Z,10Z,13Z,16Z-docosapentaenoic acid; 17,18-EpETE: 17,18-epoxy-5Z,8Z,11Z,14Z-eicosatetraenoic acid; EPA: Eicosapentaenoic acid; ESI: Electrospray ionization; GM3: Ceramide-lactose-N-acetylneuraminic acid; HETE: Hydroxyeicosatetraenoic acid; MAG: Monoacylglycerols; PC: Phosphatidylcholine; PE: Phosphatidylethanolamine; pPE: Ethanolamine plasmalogen; PUFA: Polyunsatulated fatty acid; MF: Methyl formate; m/z: Mass-to-charge ratio; OPLS-DA: Orthogonal partial least-squares-discriminant analysis; PL: Phospholipid; PPAR: Proliferator-activated receptor; RPLC: Reversed-phase liquid chromatography; RT: Retention time; SM: Sphingomyelin; TAG: Triacylglycerol; TOFMS: Time of flight mass spectrometry. Competing interests =================== The authors declare that they have no competing interests. Authors' contributions ====================== Conception and design: MK, TR, NS, TO, SY. Execution of the experiments: TY, IM, MK, MM, SY, OA, MR. Data analysis and interpretation: MK, NMT, NH, IK, AM, TR, NS, TO, SY. Manuscript writing: TY, MK, NS, TO, SY. All authors have read and approved the manuscript and agree with submission to your journal. Supplementary Material ====================== ###### Additional file 1: Figure S1 Expression levels of Aβ~40~ and tau proteins in APP/tau mice. The levels of human Aβ~40~, and soluble (TBS-soluble plus sarkosyl-soluble) and sarkosyl-insoluble APP/tau mice at 4, 10, and 15 months of age were determined by ELISA as described in Materials and Methods. Values are mean ± SD of N = 3--5. ###### Click here for file ###### Additional file 2: Figure S2 RPLC-ESI-TOFMS analysis of brain lipids from APP/tau and wild-type mice at 4 months of age. Total ion counts chromatograph (TIC) and two-dimensional map with retention time (RT) versus mass-charge (m/z) values of brain lipids measured by RPLC-ESI-TOFMS in the positive ion mode from wild-type (A) and APP/tau mice (B) at 4 months. The intensity of peaks is represented by color density spots. Lipid metabolites were eluted in the following order: lysophospholipids (Lyso PLs, LPLs) \> phospholipids (PLs) = sphingomyelins (SMs) = ceramides (Cers) \> triacylglycerols (TAGs) = cholesterol esters (ChEs), as indicated in the areas surrounded by the ellipses. Spots for ChEs were visible in APP/tau (B) but not wild-type mice (A). ###### Click here for file ###### Additional file 3: Figure S3 Multivariate statistical analysis of brain lipids between APP/tau and wild-type mice at 4 months of age. OPLS-DA score plots and loading S-plots for brain tissues from APP/tau vs. wild-type mice at 4 months, derived from the RPLC-ESI-TOFMS data set (A: 0.1-37.5min in RT; B: 37.5-60min in RT). In score plots (wild-type is shown in open circle vs. APP/tau in closed circle), the goodness-of-fit parameter R2 and the predictive ability parameter Q2 were 1.000 and 0.840, respectively, for (A) and 0.897 and 0.619, respectively, for (B). Loading S-plot showed covariance w against correlation p (corr) of variables for discriminating components of OPLS-DA model. Cut-off values for w\[1\] \> \\|0.05\\| and p (corr) \> \\|0.6\\| were used to select metabolites that strongly contributed to the discrimination between APP/tau and wild-type mice, which are surrounded by the red dotted line. ###### Click here for file ###### Additional file 4: Table S1 Relative quantification of lipid species meeting the OPLS-DA loading S-plot criteria in brain tissues. ###### Click here for file ###### Additional file 5: Table S2 Relative quantification of the detected oxidized fatty acids in brain tissues). ###### Click here for file ###### Additional file 6: Figure S4 Pathway analysis of arachidonic acid metabolites. HHT, hydroxy-heptadecatrienoic acid; HETE, hydroxyeicosatetraenoic acid; EpETrE, epoxyeicosatrienoic acid; TX, Thromboxane; COX, cyclooxygenase; LOX, lipoxygenase; CYP, cytochrome P450; PGFS, prostamide/prostaglandin F synthase; PTGIS, Prostacyclin synthase; PTGES, prostaglandin E synthase; PTGDS, prostaglandin D synthase; TBXAS, thromboxane-A synthase. ###### Click here for file ###### Additional file 7: Table S3 Relative quantification of lipid species meeting the OPLS-DA loading S-plot criteria in plasma. ###### Click here for file ###### Additional file 8: Table S4 Relative quantification of detected oxidized fatty acids in plasma. ###### Click here for file Acknowledgment ============== This work was supported by the Advanced Research for Products Mining Programme \[grant number 10--45\] from the National Institute of Biomedical Innovation of Japan.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1-molecules-23-02894} =============== Osimertinib (AZD9291, Merelitinib, Tagriiso^©^), N-(2-{2dimethylaminoethyl-methylamino}-4-methoxy-5-{\[4-(1-methylindol-3-yl)pyrimidin-2yl\]amino}phenyl)prop-2-enamide ([Figure 1](#molecules-23-02894-f001){ref-type="fig"}), a third-generation, highly selective, irreversible covalent inhibitor has been created by AstraZeneca for the clinical therapy of advanced non-small cell lung cancer (NSCLC) \[[@B1-molecules-23-02894],[@B2-molecules-23-02894],[@B3-molecules-23-02894],[@B4-molecules-23-02894],[@B5-molecules-23-02894],[@B6-molecules-23-02894]\]. NSCLC patients, who have epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor (TKI) resistance, are mostly subject to a mutation of EGFR. Therefore, this has been the active target for the osimertinib \[[@B7-molecules-23-02894],[@B8-molecules-23-02894],[@B9-molecules-23-02894]\]. In addition, the tablet formulation of osimertinib has been approved by the FDA (Food and Drug Administration of the USA) for NSCLC patients, who have progressed to or completed EGFR TKI therapy in 2015. To the best of our knowledge, several papers have established the methods for the determination of osimertinib in biological samples, and the utilized apparatuses are all ultra-performance liquid chromatography coupled with tandem mass spectrometry (UPLC-MS/MS) \[[@B10-molecules-23-02894],[@B11-molecules-23-02894],[@B12-molecules-23-02894],[@B13-molecules-23-02894]\]. There are some advantages of UPLC-MS/MS, including its high sensitivity, high stability, and short analytic time. Unlike UPLC-MS/MS, UPLC-TOF-MS has specific advantages such as its high working efficiency, wide measurable mass range, and high ratio of resolution \[[@B14-molecules-23-02894],[@B15-molecules-23-02894],[@B16-molecules-23-02894],[@B17-molecules-23-02894]\]. In addition, the capability of simultaneous quantitative analysis and qualitative analysis greatly benefits the analysts, and it is very useful for the further study of agents, such as their metabolism, enzymology, transportation and so on \[[@B18-molecules-23-02894]\]. However, to date, methods using UPLC-TOF-MS for the determination of osimertinib have not been reported. In the present study, we are the first to quantify osimertinib in rat plasma using UPLC-TOF-MS. The objective of this study was to investigate a specific, sensitive, rapid and reliable UPLC-TOF-MS method for osimertinib quantification in rat plasma samples. Meanwhile, we have successfully investigated the pharmacokinetic study of osimertinib in rats using this UPLC-TOF-MS method. 2. Results and Discussion {#sec2-molecules-23-02894} ========================= 2.1. UPLC-TOF-MS Method Development {#sec2dot1-molecules-23-02894} ----------------------------------- By using the product scan mode, we could find the method of pyrolysis of osimertinib and ion of sorafinib (IS) in rat plasma under the UPLC-TOF-MS condition that we have optimized (see [Section 2.3](#sec2dot3-molecules-23-02894){ref-type="sec"}) ([Figure 2](#molecules-23-02894-f002){ref-type="fig"}). The parent ion m/z of osimertinib was 500.2768 and the characteristic product ion was 72.0810. In addition, the m/z of the parent ion and product ion of sorafenib (IS) were 465.0953 and 270.0882, respectively. All these results are consistent with previous studies \[[@B4-molecules-23-02894],[@B5-molecules-23-02894],[@B6-molecules-23-02894],[@B7-molecules-23-02894],[@B19-molecules-23-02894]\]. However, the optimization of the UPLC condition became the key point in the process of investigating our UPLC-TOF-MS method. We used various proportions and gradients of water (containing 0.1% formic acid)--acetonitrile and water (containing 0.1% formic-ammonia formate)--acetonitrile to optimize the chromatographic conditions. However, the chromatographic peaks all displayed the trailing phenomenon ([Figure S1](#app1-molecules-23-02894){ref-type="app"}). Since the pKa of osimertinib was 13.64, it needs a basic environment to keep its molecular state. Therefore, we decided to use 0.1% ammonia water--acetonitrile as the mobile phase, which enabled us to finally obtain the symmetrical chromatographic peak ([Figure 3](#molecules-23-02894-f003){ref-type="fig"}). 2.2. Method Validation {#sec2dot2-molecules-23-02894} ---------------------- ### 2.2.1. Specificity and Selectivity {#sec2dot2dot1-molecules-23-02894} The typical chromatograms of blank plasma, the plasma sample spiked with osimertinib and IS and rat plasma after treatment are shown in [Figure 3](#molecules-23-02894-f003){ref-type="fig"}. [Figure 3](#molecules-23-02894-f003){ref-type="fig"}A--C displays the characteristic peaks of standard osimertinib in rat plasma, the pharmacokinetic plasma sample and blank plasma, and the chromatograms showed a good specificity. In addition, [Figure 3](#molecules-23-02894-f003){ref-type="fig"}D--F exhibits the good specificity of standard sorafenib (IS). Moreover, these results also showed that there was no significant chromatographic interference with the osimertinib and IS in rat plasma. ### 2.2.2. Calibration and Lower Limit of Quantification (LLOQ) {#sec2dot2dot2-molecules-23-02894} To investigate the linearity of osimertinib and IS, nine calibration concentrations were analyzed in each validation batch, and the results showed the good linearity (r^2^ \> 0.99) of the calibration curve of osimertinib and IS over the range of 1 to 500 ng/mL. The lower limits of quantification (LLOQ) of osimertinib and IS were both 1 ng/mL, and the ratios of signal-to-noise were considerably higher than 5. In addition, the LLOQ of this UPLC-TOF-MS method was sufficient for the determination of the osimertinib pharmacokinetic study. ### 2.2.3. Precision and Accuracy {#sec2dot2dot3-molecules-23-02894} The intra- and inter-day precision and accuracy are shown in [Table 1](#molecules-23-02894-t001){ref-type="table"}. In addition, the intra- and inter-day results of the HQC, MQC and LQC are investigated with the RSD ranging from 5.38--9.76% (intra-day) and 6.02--9.46% (inter-day), respectively. The results ranged within the standard acceptance limit of 15% and demonstrated the good accuracy and precision of osimertinib. ### 2.2.4. Extraction Recovery and Matrix Effect {#sec2dot2dot4-molecules-23-02894} [Table 2](#molecules-23-02894-t002){ref-type="table"} shows the extraction recovery of osimertinib and IS, and these results are sufficient for quantification. The matrix effects of osimertinib and IS range from 0.810 to 0.926 and from 0.798 to 0.934, respectively ([Table 3](#molecules-23-02894-t003){ref-type="table"}). The calibration curves of the final concentrations of IS were the same as those of osimertinib. The results showed the high extraction recovery and lack of significant matrix effect of this method for osimertinib and IS in the rat plasma. ### 2.2.5. Stability {#sec2dot2dot5-molecules-23-02894} [Table 4](#molecules-23-02894-t004){ref-type="table"} shows the stability of HQC, MQC and LQC of osimertinib under four different storage conditions. According to the results, osimertinib was found to have good stability at room temperature (25 °C) and the autosampler temperature (4 °C) for 24 h and remained stable following three freeze (−80 °C) and thaw (0 °C) cycles. Moreover, the plasma samples of osimertinib were also stable at the storage temperature (−80 °C) for at least 30 days. 2.3. Pharmacokinetic Application {#sec2dot3-molecules-23-02894} -------------------------------- After successfully establishing the analysis method of osimertinib by UPLC-TOF-MS, we applied this method into the pharmacokinetic study in SD rats. The dosage of the oral administration was 4.5 mg/kg and the mean plasma time-concentration of osimertinib in seven rats is shown in [Figure 4](#molecules-23-02894-f004){ref-type="fig"}. This showed that the osimertinib had the highest plasma concentration at 4.5 h after oral administration and the C~max~ was 28.49 ng/mL. In addition, [Table 5](#molecules-23-02894-t005){ref-type="table"} shows the following non-compartmental parameters of osimertinib in SD rats: a terminal half-life of (14.96 ± 3.44) h, a distribution volume of (233.82 ± 66.68) L/kg, and a clearance of about (10.84 ± 1.94) L/h/kg ([Table 5](#molecules-23-02894-t005){ref-type="table"}). These pharmacokinetic data of osimertinib can provide more information regarding its application in clinical treatment. 3. Materials and Methods {#sec3-molecules-23-02894} ======================== 3.1. Drugs and Materials {#sec3dot1-molecules-23-02894} ------------------------ Osimertinib (purity \>99%) and sorafenib (internal standard, IS) were purchased from Stanford Analytical Chemicals Inc. (Eugene, OR, USA). Dimethyl sulfoxide (DMSO) was purchased from Beijing Solarbio Science and Technology Co., Ltd. (Beijing, China). Ammonia and acetonitrile (HPLC grade) were provided by Tianjin Kemiou Chemical Reagent Co., Ltd. (Tianjin, China) and the Fisher Scientific Co., Ltd. (Pittsburgh, PA, USA), respectively. Ultrapure water was obtained from a milli-Q reagent water purification system (Millipore, Bedford, MA, USA). 3.2. Apparatus {#sec3dot2-molecules-23-02894} -------------- The UPLC-TOF-MS/MS method was performed on a system that includes a Qtrap 5600-TOF mass spectrometer (AB Sciex, MA, USA) and an UPLC chromatographic analysis system (Shimadzu, Kyoto, Japan). An Xterra MS C~18~ column (100 × 2.1 mm, 3.5 μm) (Waters Corp., Milford, MA, USA) was used for the analytical separation at the temperature of 40 °C. A TGL-16M high speed centrifuge was purchased from Cence Co., Ltd. (Changsha, China). 3.3. Solution Preparation {#sec3dot3-molecules-23-02894} ------------------------- DMSO was used to dissolve the accurately weighed standard of osimertinib and sorafenib (IS) to obtain the stock solutions of 1.0 mg/mL. Then, the working solutions of osimertinib were diluted serially with 50% acetonitrile in water to achieve 10, 20, 50, 100, 200, 500, 1000, 2000 and 5000 ng/mL. Next, 10 μL diluted solutions were diluted in 100 μL blank plasma to obtain the final calibration standard samples, and the range of the final concentrations of the calibration standards was from 1 to 500 ng/mL. The working solution concentration of IS was 500 ng/mL, which was dissolved using 50% acetonitrile water (v/v). The quality controls (QCs) were diluted to achieve the lower limit of the quantification (10 ng/mL, LLOQ), low (20 ng/mL, LQC), medium (200 ng/mL, MQC), and high (4000 ng/mL, HQC) concentrations of osimertinib. After that step, 10 μL of the QC solutions were dissolved by 100 μL of blank plasma to get the final concentrations of LLOQ (1 ng/mL), LQC (2 ng/mL), MQC (20 ng/mL), HQC (400 ng/mL) and 50 ng/mL of IS. All samples and working solutions were kept at −20 °C before use. 3.4. UPLC-TOF-MS Condition {#sec3dot4-molecules-23-02894} -------------------------- The chromatographic separation was performed using a C~18~ column, and its temperature was kept at 40 °C. The chromatographic separation consisted of a 0.1% ammonia (A) and acetonitrile (B) mixture and a 0.4 mL/min flow rate was maintained. The gradient ran linearly from 10% to 95% between 0 and 1.5 min, and then the mobile phase was kept at 95% for 5.0 min. 3.0 μL of samples was injected into the analysis system and the total analytical time of one sample was 5.1 min. The temperature of the autosampler was maintained at 4 °C. The TOF-MS spectrometer was set up in the positive ion full scan electrospray and high sensitivity mode with an m/z range from 100 to 1000 Da, and the accumulation time was set as 0.25s. The parameters of TOF-MS were as follows: nebulizer gas (gas 1), 55 psi; heater gas (gas 2), 55 psi; curtain gas, 35 psi; ion spray voltage, 5500 V; turbo spray temperature, 550 °C; declustering potential (DP), 100 V; and collision energy (CE), 35 eV. The conditions of the information-dependent data acquisition (IDA) criteria were as follows. The eight most intense fragment ions of each analyte in 100 cps were chosen as the product ions, and the m/z of the product ions ranged from 600 to 1300 over a 0.08 s accumulation time. In addition, the CE and collision energy scope (CES) of the product ions scan were set at 35 eV and 15 eV, respectively. 3.5. Pharmacokinetic Application {#sec3dot5-molecules-23-02894} -------------------------------- The pharmacokinetic study of osimertinib was applied to seven male SD rats using oral administration by gavage and the dosage was 4.5 mg/kg. Osimertinib was diluted and suspended by 0.5% sodium carboxymethylcellulose. No less than 0.3 mL rat plasma was sampled at 0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 24, 36, and 48 h after oral administration through the oculi chorioideae vein under the condition of light ether anesthesia. After 10 min centrifuging of all analytes at 5000× g, the supernatant was collected and frozen at −40 °C until analysis. The use of animals in the presented study was permitted by the Ethics Committee of the Hebei Medical University, and all animal studies were carried out according to the Guidance for the Care and Use of Laboratory Animals of the US National Institute of Health. 3.6. Sample Preparation {#sec3dot6-molecules-23-02894} ----------------------- Ten microliters of IS solution (500 ng/mL) was prepared in 100 μL rat plasma and vortex-mixed for 20 s. After that step, 500 μL of acetonitrile was added to precipitate the protein. Then, the mixture was vortexed for 1 min and centrifuged at 12,000 rpm for 10 min. The supernatant was transferred into a new Eppendorf tube and evaporated to dryness through nitrogen gas at 45 °C. After 100 μL of acetonitrile was added to reconstitute the residue and vortexed for 1 min, all samples were centrifuged at 12,000× g for 10 min. Finally, 80 μL of the supernatant was collected and injected into UPLC sample vials before use. 3.7. Method Validation {#sec3dot7-molecules-23-02894} ---------------------- According to the US Food and Drug Administration (FDA) guidelines regarding the bioanalytical method's validation \[[@B20-molecules-23-02894]\], the method validation was investigated and established, including the specificity and selectivity, the linearity and sensitivity, the recovery, the stability, and the precision, accuracy and matrix effect. ### 3.7.1. Selectivity and Specificity {#sec3dot7dot1-molecules-23-02894} The selectivity and specificity of the developed method was assessed by comparing the chromatography of the blank plasma and blank plasma spiked with the targets. ### 3.7.2. Linearity and Sensitivity {#sec3dot7dot2-molecules-23-02894} A series of calibration analytes from 1 to 500 ng/mL consisted of the calibration curve of osimertinib. In addition, the method to determine the linearity of calibration curve was the peak area ratios of osimertinib and IS, and these ratios were used to get a least-squares weighted regression (the weighting factor was 1/y, and y = peak area ratio of osimertinib/IS). The correlation coefficient (r^2^) of all calibration curves, which were desirable for this method, were better than or equal to 0.99. LLOQ was used to investigate the method sensitivity and follows these two criteria: (1) the comparison of the LLOQ and blank response should occur at least 5 times; and (2) the analyte peak of the LLOQ should be discrete, reproducible and identifiable, and its accuracy and precision should be at least 20%. ### 3.7.3. Precision, Accuracy and Matrix Effect {#sec3dot7dot3-molecules-23-02894} The intra-day accuracy and precision were investigated through the determination of six QC analytes of the high (HQC = 400 ng/mL), medium (MQC = 20 ng/mL), low (LQC = 2 ng/mL), and LLOQ (LLOQ = 1 ng/mL) concentrations. The inter-day accuracy and precision were conducted through the determination of the six replicates of the four levels of QCs using the same preparation on 3 separate days. The assay accuracy of the QC samples was compared to the corresponding standard calibration concentration. The precision of the replicates was evaluated by the RSD (relative standard deviation). It can be accepted that the mean values of accuracy should not exceed 15% at the HQC and MQC, and LQC concentrations and LLOQ should not exceed 20%. Similarly, the relative standard deviation of the precision for HQC, MQC, and LQC concentration levels should not exceed 15% and the limitation of LLOQ was 20%. The matrix effect was determined by dividing slopes of calibration curves of osimertinib in the rat blood matrix and mobile phase. ### 3.7.4. Recovery {#sec3dot7dot4-molecules-23-02894} The extraction recovery of osimertinib was evaluated by comparing the peak area ratios of standard solution samples and the same concentrations rat plasma samples through five replicates at HQC, MQC and LQC concentrations. ### 3.7.5. Stability {#sec3dot7dot5-molecules-23-02894} The stabilities of this method, which include the freeze--thaw stability, autosampler stability, short-term stability and long-term stability, were evaluated through three QC samples after the sample preparation method. The freeze (−80 °C)--thaw (room temperature) stability was conducted under the conditions of three free--thaw cycles. The autosampler stability of the plasma samples was investigated by the extracted QC samples that were kept in an autosampler (4 °C) for 12 h. In addition, the long-term stability was evaluated through the determination of three QC samples that were kept at −80 °C for 30 days. All samples were considered stable with RSDs \< ±15%. 3.8. Data Analysis {#sec3dot8-molecules-23-02894} ------------------ Dynamic background subtraction is a novel technique performed using the Analyst software (AB Sciex, Foster City, CA, USA). In addition, the data of the pharmacokinetic study of osimertinib were collected and calculated by the DAS 2.1.1 software in the non-compartmental mode (Mathematical Pharmacology Professional Committee of China, Shanghai, China). 4. Conclusions {#sec4-molecules-23-02894} ============== A sensitive UPLC-TOF-MS method for the determination of osimertinib has been established in this research. The method exhibited excellent precision, recovery and sensitivity. The results indicated that UPLC-TOF-MS could serve as a highly interesting analytical alternative for bioanalysis. This work was supported by the Liaoning Planning Program of Philosophy and Social Science (NO. L17BGL034), the Research Program on the Reform of Undergraduate Teaching in General Higher Schools in Liaoning (NO. 2016-346), the Key Program of the Natural Science Foundation of Liaoning Province of China (NO. 20170541027) and the Medical Education Scientific Research on the "12th Five-Year Plan" (Fifth Batch of Subjects) funded by China Medical University (NO. YDJK2015005). Sample Availability: Samples of the compounds are available from the authors. The following are available online. ###### Click here for additional data file. Conceived and designed the experiments: Y.Z., Z.-J.D. and S.-T.D. Performed the experiments: S.-T.D., Y.L., H.-T.Y., Y.W., Y.-J.L., C.-Y.D., L.M. and Z.-J.D. Analyzed the data: S.-T.D. Wrote the paper: Y.Z. and S.-T.D. All authors read and approved the final manuscript. The authors declared no conflict of interest. ![Structures of osimertinib (molecular weight = 499.619 Da) and sorafenib (molecular weight = 464.825 Da).](molecules-23-02894-g001){#molecules-23-02894-f001} ![Product ion scan of osimertinib 500.2768 → 72.0810 (A), and IS 465.0953 → 270.0882 (B).](molecules-23-02894-g002){#molecules-23-02894-f002} ###### Typical chromatograms of (A) standard osimertinib (20 ng/mL) in rat plasma, (B) pharmacokinetic plasma sample, (C) blank plasma, (D) standard ion of sorafenib (IS) (500 ng/mL) in rat plasma, (E) pharmacokinetic plasma sample, and (F) blank plasma. ![](molecules-23-02894-g003a) ![](molecules-23-02894-g003b) ![Plasma concentration-time profile after single oral administration of osimertinib (4.5 mg/kg) to rats. Data are expressed as the mean ± SD (n = 7).](molecules-23-02894-g004){#molecules-23-02894-f004} molecules-23-02894-t001_Table 1 ###### Intra- and inter-day precision and accuracy of osimertinib in rat plasma. RSD: relative standard deviation. Concentration (ng/mL) Intra-Day (n = 7) Inter-Day (n = 7) ----------------------- ------------------- ------------------- -------- ---------------- ------ -------- 400 395.12 ± 21.27 5.38 98.78 401.13 ± 24.15 6.02 100.28 20 21.08 ± 1.42 6.74 105.40 19.34 ± 1.83 9.46 96.70 2 2.05 ± 0.20 9.76 102.50 1.98 ± 0.17 8.59 99.00 molecules-23-02894-t002_Table 2 ###### Extraction recovery and matrix effect of osimertinib and IS in rat plasma. Analyte Concentration (ng/mL) Extraction Recovery (%) ------------- ----------------------- ------------------------- ------ Osimertinib 400 95.24 ± 3.01 3.16 20 96.14 ± 1.83 1.90 2 84.31 ± 3.18 3.77 Sorafinib 500 87.22 ± 4.23 4.85 molecules-23-02894-t003_Table 3 ###### The slope ratio of the solvent linear equation and the matrix linear equation. Analyte Calibration Curve R^2^ R~slope~ ------------- --------------------- -------- ---------- ------- ------- Osimertinib Y = 0.085X + 0.1102 0.9997 0.810 0.926 0.116 Sorafinib Y= 0.0781X + 0.2314 0.9996 0.798 0.934 0.136 R~slope~ = Slope of matrix standard calibration curve/slope of mobile phase standard calibration curve. molecules-23-02894-t004_Table 4 ###### Stability of osimertinib in rat plasma under various storage conditions. Storage Condition Concentration (ng/L) Mean ± SD RSD% ----------------------------------------------------- ---------------------- ------------- ------- Autosampler (4 °C) temperature for 24 h 2 2.12 ± 0.23 10.85 20 21.45 ± 1.81 8.44 400 406.81 ± 5.64 1.39 Room temperature (25 °C) for 24 h 2 2.21 ± 0.26 3.66 20 22.45 ± 2.18 9.71 400 407.28 ± 5.12 1.26 Storage temperature (−80 °C) for 30 days 2 2.29 ± 0.25 10.92 20 21.33 ± 1.74 8.16 400 406.34 ± 7.51 1.85 Three freeze--thaw cycles (each at −80 °C for 24 h) 2 2.27 ± 0.19 8.37 20 22.20 ± 1.92 8.65 400 406.17 ± 6.19 1.52 molecules-23-02894-t005_Table 5 ###### Pharmacokinetic parameters of osimertinib after oral administration of 4.5 mg/kg to rats. Pharmacokinetic Parameter Osimertinib --------------------------- ---------------- AUC~(0--t)~, ng/mL·h 382.00 ± 69.00 AUC~(0--∞)~, ng/mL·h 426.01 ± 81.73 MRT, h 14.51 ± 1.91 t~1/2z~, h 14.96 ± 3.44 t~max~, h 4.80 ± 1.10 C~max~, ng/mL 28.49 ± 3.97 V~z~/F, L/kg 233.82 ± 66.68 CL~z~/F, L/h/kg 10.84 ± 1.94 Data are expressed as the mean ± SD (n = 7).	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1} =============== For any nation, understanding the epidemiology and clinical status of young people with diabetes is essential for training health professionals and also for conducting advocacy with the respective government to plan and improve the appropriate clinical services. Very little published material is available on diabetes in youth in Bolivia. The International Diabetes Federation (IDF) Atlas estimates of children \< 15 years with type 1 diabetes in Bolivia \[[@B1]\] are extrapolated from the data from Peru that is over two decades old \[[@B2], [@B3]\], a time during which the incidence of type 1 diabetes has been increasing at 2-3% or more per annum in many countries \[[@B1]\]. In 2005, the IDF Life for a Child (LFAC) Program \[[@B4]\], with the assistance of Rotary International, commenced support for the Centro Vivir con Diabetes (CVcD), a multidisciplinary clinic in Cochabamba. Satellite networks were established in five other cities. LFAC Program commenced professional mentoring and provision of insulin (initially with support of Insulin for Life Australia \[[@B5]\]). Later, blood glucose meters and test strips (three tests/day), HbA1c testing, and educational materials were provided. This study presents epidemiological data on all young people with diabetes known to CVcD from January 2005 to February 2017 and the clinical features of the cohort receiving care at the end of 2015 and documents the initial impact of the LFAC Program. 2. Materials and Methods {#sec2} ======================== The study was approved by the Board of Centro Vivir con Diabetes. 2.1. Epidemiology Data {#sec2.1} ---------------------- Data (age, gender, and diabetes type) were compiled on all people diagnosed with diabetes at age \< 25 years (y) and seen by CVcD and its five satellite clinics from January 2005 to February 2017. Diabetes was diagnosed according to the standard World Health Organization (WHO) criteria \[[@B6]\], and the diabetes type was determined by the local investigators based on clinical features. It is not possible to determine the exact national incidence as other diabetes care providers exist in Bolivia and no secondary ascertainment sources are available. However, in Cochabamba Province---comprising Cochabamba city and its immediate environs---where 63 (43.8%) of cases were living at diagnosis, CVcD is well-known and the ascertainment rate is estimated at 80%. A minimum incidence for Cercado Province was calculated according to census data \[[@B7]\], with population interpolated or extrapolated according to the annual growth rate of 1.6% \[[@B8]\]. The crude mortality rate was calculated as the total number of deaths in the cohort of 144 subjects divided by the total number of years from diagnosis until the last follow-up (February 2017) or, if they had died, date of death. Results are expressed as mortality per 1000 patient years. 2.2. Clinical Status {#sec2.2} -------------------- Clinical information was collated on all 141 subjects being followed from September to December 2015, who aged \<25 y at diabetes diagnosis and ≤35 y at the time of this assessment. These 141 subjects included some cases diagnosed before January 2005 and excluded some cases included in the incidence study who were now being seen by other centres. The data were collected via LFAC Annual Clinical Data Sheets. Data included sex, date of birth, diabetes diagnosis date, details of diabetes treatment regimen, and physical and biochemistry measures. Also, social parameters as to whether diabetes was limiting school attendance, if subjects were in the age-appropriate grade, and how well overall the young person was, in their doctor\'s opinion, psychologically coping with their diabetes (rated as good, with some problems, or poor) were also recorded. Body weight and height were measured by electronic scales and a stadiometer, respectively, with subjects wearing light-weight clothing and without shoes. Body mass index (BMI) was then calculated. BMI SD scores were calculated using the WHO standards for \<5 years \[[@B9]\], \>5 years, and \<19 years \[[@B10]\]. Blood pressure SD score was calculated from published data from the United States \[[@B11]\]. The presence of cataracts, retinopathy, and peripheral neuropathy was recorded. However, the methodology was not prescribed and therefore could have varied across centres and physicians. 2.3. Biochemistry {#sec2.3} ----------------- Glycosylated haemoglobin (HbA1c), serum creatinine, total cholesterol, triglycerides, and HDL cholesterol were measured in local biochemistry laboratories, with the exception of HbA1c at Cochabamba where it was measured with a Siemens DCA Vantage machine (Erlangen, Germany). Lipid tests were done on fasting samples. LDL cholesterol was calculated using the Friedewald equation and hence is only available on the subset of patients who had HDL cholesterol levels \[[@B12]\]. Estimated glomerular filtration rate (eGFR) was calculated for subjects ≥ 14 y by the CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) formula \[[@B13]\], adjusted for individual surface area. Initial impact of Life for a Child Program was measured by assessing all HbA1c levels of all subjects having data at baseline (program commencement in 2004) and then at either six- or 12-month follow-up. 2.4. Statistics {#sec2.4} --------------- Data and descriptive statistics were managed in Excel and analysed in R 3.3.1 (R Core Team, Vienna, Austria) with RStudio integrated development environment (RStudio Team, Boston, USA). 3. Results {#sec3} ========== 3.1. Epidemiological Data {#sec3.1} ------------------------- ### 3.1.1. Sex Distribution and Age of Onset {#sec3.1.1} Of the 144 subjects diagnosed at age \< 25 y between January 2005 and February 2017, 62 (43.1%) were males. Age of onset ranged from 0.3--22.2 y, with a peak at 11 years (see [Figure 1](#fig1){ref-type="fig"} histogram). Sixteen subjects (11.1%) were diagnosed at age 0--4 y, 42 (29.2%) were diagnosed at age 5--9 y, 62 (43.1%) were diagnosed at age 10--14 y, 19 (13.2%) were diagnosed at age 15--19 y, and 5 (3.5%) were diagnosed at age 20--24 y. Genetic studies are being arranged for the child diagnosed at 3 months to check for a monogenic form of diabetes. This child has no other medical problems. ### 3.1.2. Minimum Incidence {#sec3.1.2} Sixty-three cases of diabetes (diagnosed at \<25 years of age) occurred in the Cercado Province of Cochabamba Department in the 12.2 y study period from January 2005 to February 2017. Forty-nine were \<15 y at diagnosis and 14 were 15--24 y. This is a measured incidence of 2.2 per 100,000 children \< 15 y/y. Ascertainment is estimated as 80%, resulting in an estimated total incidence of 2.7 per 100,000 children \< 15 y/y. ### 3.1.3. Mortality Rate {#sec3.1.3} Of the 144 subjects followed up for 0.1--12.2 y (mean ± SD 5.9 ± 3.4 y), two had died, a male aged 22.8 y from gastric ulcer haemorrhage and a female aged 13.4 y from ketoacidosis (who had been abandoned by her family). The crude mortality rate was 2.3/1000 patient years. 3.2. Clinical Status {#sec3.2} -------------------- ### 3.2.1. Demographics {#sec3.2.1} One hundred and forty-one T1D subjects diagnosed at \<25 years of age and ≤35 y were being followed up by CvCD at the end of 2015. Sixty-four (45%) patients were male. Sixty (43%) were from Cochabamba Department; 31 (22%), Santa Cruz; 27 (19%), La Paz; 12 (8.5%), Sucre; 6 (4.3%), Tarija; and 5 (3.5%), Potosi. One child also had cerebral palsy, one had Down syndrome, and one child also had epilepsy. Age at diagnosis and diabetes duration are shown in [Table 1](#tab1){ref-type="table"}. ### 3.2.2. Diabetes Care {#sec3.2.2} Ninety-four subjects (67%) were adjusting their insulin dose (data available on 141 subjects), with the remainder giving a fixed insulin dose set in agreement with their treating clinicians. Regarding insulin types (n = 138), 99 (72%) were on short- and long-acting human insulin only. Twenty-five (18%) were using analogue insulin, of which nine (7%) were using analogue only. Thirteen (9%) were on premixed insulin, and one patient was receiving long-acting human insulin only. For the number of insulin injections per day (n = 122), 18 subjects (158%) were taking two injections per day; 70 (57%), three injections; 24 (20%), four injections; and the remaining 10 (8%), five injections. Insulin usage ranged 0.35--1.76 units/kg/day (mean ± SD 0.99 ± 0.29) (n = 136). The frequency of self-blood glucose tests per week (n = 139) was 14--28 (mean ± SD 21.3 ± 1.4). The frequency of clinic visits in the last year (n = 131) was 1--8 (mean ± SD 3.5 ± 1.3). ### 3.2.3. Social Parameters {#sec3.2.3} School/college attendance was regular in 93 (96%) of the 97 patients for whom this information was available, and the year of schooling was age-appropriate in 94%. Out of 105 patients for whom a response was recorded on the question of how well they thought their patients were coping with their diabetes, 62 (59%) of the responses were "good," 38 (37%) were with "some problems," and four (4%) were "poor." ### 3.2.4. Physical Measurements {#sec3.2.4} Height and weight information was available for 137 patients, and 81 of these were under 19 years, permitting BMI SD score calculation (shown in [Table 1](#tab1){ref-type="table"}). Blood pressure was measured on 21 subjects \< 20 years, with systolic and diastolic SD scores shown in [Table 1](#tab1){ref-type="table"}. Both systolic and diastolic mean SD scores were above zero (p \< 0.001 for both, by one-sided t-test). None of the patients for whom SD scores were available were noted to have hypertension (as defined by local clinical standards) or be on antihypertensive medications. ### 3.2.5. HbA1c {#sec3.2.5} HbA1c was measured in the 135 subjects with duration of diabetes \> 6 months. Mean and median HbA1c levels were 8.5% (69 mmol/mol) and 8.2% (66 mmol/mol), respectively (see also [Table 1](#tab1){ref-type="table"}). Forty-four subjects (33%) had HbA1c levels in the recommended target range of \<7.5% (58 mmol/mol); 35 (25%), 7.5%--8.49% (58--69 mmol/mol); 48 (36%), 8.5%--12% (69--108 mmol/mol); and the remaining eight subjects (5.8%), \>12% (108 mmol/mol). Twenty-six percent of those aged \<15 y and 30% of those aged 15--24 y had target HbA1c \< 7.5% (58 mmol/mol). A series of one-way ANOVAs on log-transformed HbA1c levels were performed, finding trends towards relationships between HbA1c levels, the number of insulin injections per day (p = 0.07), and the use of analogue insulin (p = 0.06), but not for gender (p = 0.38). A pattern of increased HbA1c levels and variance during adolescence was observed from a plot of HbA1c by age ([Figure 2](#fig2){ref-type="fig"}). For those aged 0--13 y, the HbA1c mean ± SD was 7.9% (63 mmol/mol) ± 1.3% (14 mmol/mol); for those aged 13--20 y, the HbA1c mean ± SD was 9.4% (78 mmol/mol) ± 2.0% and for those aged \>20 y, the HbA1c mean ± SD was 7.9% (63 mmol/mol) ± 1.7% (18 mmol/mol). This indicates a significant difference in HbA1c levels between age groups (p = 0.0002 by one-way ANOVA on log-transformed values). However, Levene\'s test for heterogeneity of variance was not statistically significant (p = 0.24). Mean/median HbA1c levels by a diabetes clinic site were 9.0%/8.4% (75/68 mmol/mol) in Cochabamba, 7.9%/7.6% (63/60 mmol/mol) in La Paz, 8.0%/7.4% (64/57 mmol/mol) in Potosi, 8.5%/8.4% (69/68 mmol/mol) in Santa Cruz, 8.5%/8.3% (69/67 mmol/mol) in Sucre, and 7.1%/6.8% (54/51 mmol/mol) in Tarija, with no statistical difference between clinics on ANOVA. ### 3.2.6. Lipids {#sec3.2.6} Lipid levels are shown in [Table 1](#tab1){ref-type="table"}. No subjects were on lipid-lowering agents. For total cholesterol, 97 patients (80%) had \<200 mg/dL (5.2 mmol/L), 16 patients (13%) had 200--239 mg/dL (5.2--6.2 mmol/L), and nine (7.3%) had ≥240 mg/dL (≥6.2 mmol/L). Total cholesterol levels were not correlated with concurrent HbA1c levels (p = 0.14 by Pearson\'s correlation). Triglyceride levels (n = 121) were \<1.7 mmol/L (150 mg/dL) in 99 patients (82%), 1.7--2.2 mmol/L (150--199 mg/dL) in 11 (9%) patients, and 2.3--5.6 mmol/L (200--499 mg/dL) in 11 (9%) patients (all subjects had \<4.5 mmol/L (450 mg/dL). Log-transformed triglyceride levels were correlated with concurrent HbA1c levels (p = 0.03 by Pearson\'s correlation). HDL cholesterol was measured in 27 patients from the Department of La Paz (see [Table 1](#tab1){ref-type="table"}). None of these patients were on renal replacement therapy. Fifteen patients (56%) had HDL \< 1.0 mmol/L (40 mg/dL) for men or 1.3 mmol/L (50 mg/dL) for women. There were no significant correlations between HDL cholesterol and either HbA1c or log-transformed triglyceride levels. Seven patients (26%) had calculated LDL cholesterol levels \< 100 mg/dL (\<2.6 mmol/L), 10 patients (37%) had 100--129 mg/dL (2.6--3.3 mmol/L), and 10 patients (37%) had ≥130 mg/dL (3.3 mmol/L). [Table 2](#tab2){ref-type="table"} shows the percentage of subjects meeting target levels for cardiovascular risk factors (BMI, triglycerides, total cholesterol, and HbA1c), for the 115 subjects that had values for all four measurements. ### 3.2.7. Renal Function {#sec3.2.7} Serum creatinine values are shown in [Table 1](#tab1){ref-type="table"}. Five patients (4%) had serum creatinine levels above the recommended level of 1.3 mg/dL (115 μmol/L) for males or 1.1 mg/dL (97 μmol/L) for females. eGFR for the 115 subjects ≥ 14 y ranged from 10.4--178.9 mL/min/1.73m^2^ (mean ± SD 104.9 ± 31.0 mL/min/1.73m^2^). Three patients on renal replacement therapy (see below) were excluded from this analysis. Two subjects were receiving haemodialysis (age 28.5 and 28.8 y) (duration of diabetes to start of haemodialysis 16.5 and 1.25 y), one had received a renal transplant after 0.3 y on haemodialysis (age 30.4 y, duration of diabetes to time of haemodialysis 17.3 y) and the other was noted as having chronic renal failure (24.4 y, duration 20.5 y, eGFR 12.8 mL/min/1.73m^2^). Three other subjects had eGFR \< 50 mL/min/1.73m^2^, all with diabetes duration of at least 16 years. Two of these three were receiving treatment for hypertension, and one was also treated with erythropoietin injections for associated anaemia. Seventeen subjects ≥ 14 y had an eGFR of ≥140 mL/min/1.73m^2^, suggestive of hyperfiltration. ### 3.2.8. Other Complications {#sec3.2.8} One subject had hypothyroidism which was being treated. One subject had experienced a severe hypoglycemia episode in the past resulting in severe cognitive and motor impairment. Three patients were recorded as having diabetic retinopathy, and five subjects had cataracts (two of whom also had retinopathy). Neuropathy (loss of vibration and light touch) was recorded in one subject. 3.3. Initial Impact of Life for a Child Program {#sec3.3} ----------------------------------------------- HbA1c results at the start of the LFAC Program in 2004-2005 were available for 57 subjects. Values ranged from 4.9--15% (30--140 mmol/mol), with mean ± SD 9.6 ± 2.2% (81 ± 24 mmol/mol). For six-month paired data, mean ± SD HbA1c at baseline was 9.4 ± 2.3% (79 ± 25 mmol/mol), and mean ± SD HbA1c at six months was 8.1 ± 1.8% (65 ± 20 mmol/mol) (n = 29, p = 0.006 by one-sided paired t-test). For 12-month paired data, mean ± SD HbA1c at baseline was 9.7 ± 2.1% (83 ± 23 mmol/mol), and mean ± SD HbA1c at twelve months was 8.7 ± 2.5% (72 ± 27 mmol/mol) (n = 36, p = 0.007 by one-sided paired t-test) (see [Figure 3](#fig3){ref-type="fig"}). 4. Discussion {#sec4} ============= Globally, the incidence rates of type 1 diabetes in children vary widely due to differences in genetic susceptibility, combined with inadequately understood but powerful and evolving environmental factors \[[@B1]--[@B3]\]. There have been no previous studies on type 1 diabetes incidence in Bolivia. Rates vary substantially in Central/South American and other Hispanic populations. Estimates for Bolivia in the IDF Atlas \[[@B1]\] are extrapolated from a 1990--1991 study in Lima, Peru, which found an incidence of 0.5 per 100,000 children \< 15 y/y \[[@B2]\]. Other studies in South America in the early 1990s (all \<15 y) showed rates per 100,000 ranging from 0.1 in Caracas, Venezuela, to 8.3 in Montevideo \[[@B2], [@B3]\]. Countries with both recent and older data are seeing sharp rises in type 1 diabetes incidence---in Mexico, from 3.4 (per 100,000) in 2000 to 6.2 in 2012 for young people \< 20 y in the Instituto Mexicano del Seguro Social (IMSS) health insurance service \[[@B14]\]; in Chile, from 5.7 in 2006 to 12.1 in 2012 for those aged \<20 y \[[@B15]\], with the sharpest increase in the \<5 y age group; and in Brazil, a trend for increased incidence in those aged \<15 y from 1986 to 2006 \[[@B16]\]. Further afield for Hispanic populations, rates (\<15 y) are 20.6 in Spain (\<15 y, 1995--2001) \[[@B1]\], 13.2 in Portugal (\<15 y, 1994--1998) \[[@B1]\], and 14.1 in Hispanic Whites in the USA (\<15 y, 2002--2003) \[[@B17]\], with a recent study showing an increase of 4.2% per year in 2003--2012 on numbers \< 20 y \[[@B18]\]. Some of the variance in South American populations is thought to be due to lower rates of type 1 diabetes in Central/South American indigenous peoples as compared to those of Hispanic White/mixed descent \[[@B2], [@B19]\]. Larenas et al. \[[@B20]\] found 3.8 times higher rates in Caucasian Chileans than in Mapuche (native Chileans). Bolivia has, by a substantial margin, the highest proportion of indigenous peoples in Central/South America \[[@B21]\]. This lower rate in indigenous populations is likely largely genetic in origin due to differences in HLA haplotypes \[[@B19], [@B22], [@B23]\]. However, differing exposure to environmental factors may also be potent influencers: rates of type 1 diabetes have increased in many countries as standards of living have increased. Incidence rates were found to be higher in high socioeconomic communes in Santiago de Chile \[[@B24]\]. The observed type 1 diabetes rate in Cercado in this study---2.2 per 100,000 children \< 15 y/y with an estimated ascertainment of 80%, giving an estimated incidence of 2.7 per 100,000---is five times higher than the figure previously used for Bolivia (based on the Peruvian data from 1990--1991 \[[@B2]\]. It is likely that type 1 incidence is increasing in Bolivia; however, further studies will be needed to confirm this. This is still a low incidence rate compared to global figures. The female preponderance seen in Bolivia is common in low incidence studies \[[@B25]\] and is believed to be due to fathers with type 1 susceptibility genes being more likely to survive and pass on diabetogenic genes than mothers, with fathers\' genes more likely to cause type 1 diabetes in female offspring and mothers\' genes to cause diabetes in male offspring. In addition, it is thought that fathers\' genes may be more diabetogenic than similar genes in mothers \[[@B26], [@B27]\]. Monogenic diabetes has been reported from various countries, and it is likely that the child diagnosed at 3 months of age has this form of diabetes---genetic testing is being arranged. If there is a gene defect, alternate noninsulin therapy may be possible depending on the specific gene \[[@B28]\]. Like various other less-resourced countries \[[@B29]\], the Bolivian government health service does not cover the cost of care for people with diabetes at any age. Most families cannot afford the cost of care, which can be prohibitive in such countries for families with lower incomes \[[@B30]\], and this can translate into the premature death of the child or young adult with diabetes, as occurred once in this Bolivian cohort. Furthermore, in Bolivia, there are very few paediatric endocrinologists and, due to its relative rarity, limited knowledge of type 1 among general paediatricians. Therefore, the support from a centre such as CvCD ± LFAC is crucial for families with limited resources. CvCD provides a multidisciplinary service of not just health professionals in diabetology but also health professionals in ophthalmology, nutrition, psychology, social work, foot care and podiatry, physiotherapy and rehabilitation, laboratory services, and pharmacy. There is also a young group or people with type 1 diabetes which encourage and support each other through meetings and the use of the social media platform "WhatsApp." The IDF Life for a Child Program provides insulin and, since 2009, also provides blood glucose test strips, educational materials, and point-of-care HbA1c testing in Cochabamba. The initial impact seen in the LFAC follow-up data from 2004--2005 has been sustained. Mortality is low at 2.3 per 1000 patient years, and the mean/median HbA1c of 8.5/8.2% (69/62 mmol/mol) is surprisingly good, considering the substantial challenges faced in caring for these young people. Twenty-seven percent of those aged \<15 y and 28% of those aged 15--24 y were achieving the recommended HbA1c target of \<7.5% (58 mmol/mol). This is not substantially different from the mean of 29/30% (male/female) for subjects aged \<15 y and from the mean of 24/20% (male/female) for those aged 15--24 y in a large international study in which 15 of 16 countries were high-income nations \[[@B31]\]. The higher HbA1c values seen in adolescents/emerging adults are similar to those of other reports \[[@B32]\] and likely reflect lifestyle and behavioural factors and growth spurt-related insulin resistance. As in other countries, thoughtful medical attention and support are warranted during this period, along with provision of insulin supply security. It is of great concern that three young people with diabetes required renal replacement therapy and four others have substantial renal impairment (eGFR \<50 mL/1.73m^2^). Furthermore, 17 had hyperfiltration, which may be a risk factor for diabetic nephropathy, although recent evidence suggests this may not be the case \[[@B33], [@B34]\]. The presence of ESRD at a young age suggests that blood glucose control must have been poor for some years, and we note that for most of the Bolivian subjects evaluated herein, regular self-blood glucose monitoring has only been possible since 2009. Anderzén et al. \[[@B35]\] has shown the importance of a good early start in terms of blood glucose control, if later complications are to be reduced. Further assessment should be done for early signs of nephropathy such as microalbuminuria and hypertension, with management (with ACE inhibitors or LDL-lowering "statin" drugs) being informed by the forthcoming results of the Adolescent Type 1 Diabetes Cardio-Renal Intervention Trial (ADDIT) \[[@B36]\]. Given the number with renal complications, it is likely that a thorough systematic assessment would lead to more subjects being found with other microvascular complications such as neuropathy and retinopathy, for whom early intervention with risk factor control and increased screening rates is appropriate. The common cause of death of people with type 1 diabetes is cardiovascular disease, for which dyslipidaemia is a major risk factor and is treatable, with significant LDL-C, cardiovascular event, and mortality reduction with statins \[[@B37]\]. The substantial numbers with one or more lipid risk factors such as high triglycerides and total cholesterol and low HDL cholesterol levels, combined with the suggestion of blood pressure elevation, raise the possibility that some patients have features of the metabolic syndrome, which is recognised in type 1 diabetes and associated with higher rates of diabetes vascular complications \[[@B38], [@B39]\]. This "double diabetes" may benefit from the addition of metformin to insulin and risk factor control \[[@B40]\], though clinical trials with vascular end-points are still awaited. The study has three other limitations. It is possible that with further study (autoantibodies and C-peptide, unfortunately not available in this clinical context), some of the subjects diagnosed with type 1 diabetes may be recategorised as having type 2 diabetes. Secondly, HbA1c measurements were not standardised across clinics, and so it is possible that results could be affected by the different methods used. Finally, the 80% ascertainment is an estimated figure, and Cercado Province may not be fully representative of Bolivia---incidence may vary in other parts of the country due to ethnic or socioeconomic factors. 5. Conclusion {#sec5} ============= This study has estimated the incidence of type 1 diabetes in Bolivia, confirming that incidence is relatively low compared to that in many other countries. The results demonstrate that reasonable blood glucose control and low mortality can be achieved in many subjects in a low-resourced setting with international support. However, there is a significant risk of early long-term complications in some subjects, as evidenced by renal impairment and substantial levels of cardiovascular risk factors. The authors acknowledge the dedicated efforts of all at Centro Vivir con Diabetes and the determination of all these young people with diabetes and their families. They also thank the supporters of the IDF Life for a Child and Insulin for Life Programs and Rotary International (Rotary Club of Cochabamba, Quillacollo, District 6940, and the Rotary Foundation). Finally, they thank Sarah Garnett for helping with the SD scores. This work was supported by the Leona M. and Harry B. Helmsley Charitable Trust. Gabriel Andrew Gregory was partly funded by the University of Sydney Summer Research Scholarship. Alicia Josephine Jenkins is funded by the NHMRC (Australia) Practitioner Fellowship and the Sydney Medical School Foundation Fellowship. Ethical Approval ================ The study was reviewed and approved by the Board of Centro Vivir con Diabetes. Conflicts of Interest ===================== None of the authors have any conflicts of interest in regard to this study. Authors\' Contributions ======================= Elizabeth Duarte Gómez collected the data and contributed to the manuscript, Gabriel Andrew Gregory analysed the data and wrote the manuscript, Miriam Castrati Nostas collected the data and contributed to the manuscript, Angela Christine Middlehurst assisted in the data analysis and contributed to the manuscript, Alicia Josephine Jenkins helped with interpretation of the data and contributed to the manuscript, and Graham David Ogle designed the study and cowrote the manuscript. ![Age at diagnosis of type 1 diabetes in Bolivia, 2005--2017.](JDR2017-8454757.001){#fig1} ![Relationship of age to HbA1c.](JDR2017-8454757.002){#fig2} ![Initial impact at commencement of Life for a Child support.](JDR2017-8454757.003){#fig3} ###### Clinical and biochemical characteristics. Parameter n Mean (SD) Range ----------------------------------------- ----- ------------------ ------------------ Age at diagnosis (years) 141 9.9 (4.4) 0.3--22.2 Diabetes duration (years) 141 8.1 (5.7) 0.1--26.9 BMI SD score 137 0.9 (1.1) −2.2--3.7 Systolic SD score 21 0.7 (0.8) −1.4--2.0 Diastolic SD score 21 1.1 (0.7) −0.1--2.6 HbA1c (%)^∗^ 135 8.5 (1.9) % 5.4--14.0% HbA1c (mmol/mol)^∗^ 135 69 (21) mmol/mol 36--130 mmol/mol Serum creatinine (mg/dL) 130 0.92 (0.69) 0.5--6.2 Total cholesterol (mg/dL) 122 167 (43.2) 81--334 Triglycerides (mg/dL) 121 112 (63) 42--497 HDL cholesterol (mg/dL) 27 43 (14.2) 7.9--59 LDL cholesterol (mg/dL, calculated) 27 124 (38) 43--191 Non-HDL cholesterol (mg/dL, calculated) 27 149 (44) 58--216 ^∗^HbA1c not included on six subjects assessed within six months of diagnosis. ###### Risk factors for cardiovascular disease. Risk factor Definition n \% ------------------------------------ -------------------------------------------------------- ----- ------ High BMI For \<19 years, SD score ≥ 2.0; if ≥19 years, BMI ≥ 25 22 19.1 Elevated fasting triglycerides ≥150 mg/dL (≥1.7 mmol/L) 21 18.3 Elevated fasting total cholesterol ≥175 mg/dL (≥4.5 mmol/L) 45 39.1 HbA1c above target range ≥7.5% (58 mmol/mol) 74 64.3 One risk factor --- 44 38.3 Two risk factors --- 34 29.6 Three risk factors --- 14 12.2 Four risk factors --- 2 1.7 (n = 115). [^1]: Academic Editor: Konstantinos Papatheodorou	{ "pile_set_name": "PubMed Central" }
Introduction {#s1} ============ Cells secrete different types of small membrane vesicles. Exosomes are one of the vesicles with 30--100 nm diameter and physicochemically distinct from other secreted vesicles [@pone.0013247-Thry1]. Inside exosomes, cellular gene products including proteins, mRNAs, and microRNAs (miRNAs) are packaged and these molecules can be transferred to recipient cells to deliver their molecular signaling such as oncogenesis [@pone.0013247-Valadi1]--[@pone.0013247-AlNedawi1] and immune response [@pone.0013247-Thry1], [@pone.0013247-Davis1]. Exosomes are released by many types of cells [@pone.0013247-Thry2] which include tumor cells, epithelial cells, and hematopoietic cells such as reticulocytes, cytotoxic T lymphocytes, Epstein-Barr virus-transformed B cells, mastocytes, dendritic cells, and platelets. Secreted exosomes have been isolated and characterized in vitro from cultured cell lines [@pone.0013247-Simpson1] along with in vivo in body fluids [@pone.0013247-Simpson1] including blood [@pone.0013247-Rabinowits1], urine [@pone.0013247-Pisitkun1], saliva [@pone.0013247-Michael1], amniotic fluid [@pone.0013247-Keller1], and malignant pleural effusions [@pone.0013247-Andre1]. Since observed in many proliferating cell types, it is conceivable to exacerbate tumor cells, as evidenced by their increased presence in plasma and pleural effusions of patients with cancer [@pone.0013247-Rabinowits1], [@pone.0013247-Andre1]. This increased presence in non-invasive body fluids of cancer patients has accelerated to profile molecular components in the exosomes for discovering clinically useful tumor markers and biomarkers [@pone.0013247-Skog1], [@pone.0013247-Simpson1], [@pone.0013247-Taylor1]. miRNAs are a class of noncoding small RNAs that are involved in post-translational regulation of gene expression by inhibiting both stability and translation of mRNAs [@pone.0013247-Ghildiyal1]. Recent evidence has shown that miRNA mutations or misexpression correlate with various human cancers and indicate that some miRNAs can function as oncogenes or tumor suppressors [@pone.0013247-Ahmed1], [@pone.0013247-Croce1]. To analyze RNAs, it is always to consider their stability from degradation by RNase. Recent findings indicate that endogenous plasma miRNAs in blood samples are stably detectable in a form that is resistant to RNase activity [@pone.0013247-Mitchell1], evidenced by identification of miRNAs in body fluids such as blood [@pone.0013247-Mitchell1]--[@pone.0013247-Tanaka1], urine [@pone.0013247-Hanke1], and saliva [@pone.0013247-Michael1], [@pone.0013247-Park1]. Cultured cancer cells have been used to search for tumor markers. In particular, identifying proteins and peptides secreted into the culture media has developed by proteomics-based approach [@pone.0013247-Sasaki1], [@pone.0013247-Ogura1]. As for molecular signature in the exosomes, proteomics as well as transcriptomics analyses have been performed to reveal tumorigenesis and identify tumor marker candidates [@pone.0013247-Valadi1]--[@pone.0013247-AlNedawi1], [@pone.0013247-Simpson1], [@pone.0013247-Ji1]. Here, to identify miRNA related to tumorigenesis and metastasis, we performed extensive miRNA analysis in three cellular fractions including cells, exosomes, and medium from cultured cells. Ranking data of these intracellular and extracellular miRNAs obtained by microarray analysis, we found that let-7 miRNA family is rich in all the fractions from AZ-P7a cells, a metastatic gastric cancer cell line, which produces homogeneous and condensed morphology, and high recovery rate of exosomal miRNAs. These findings were distinct from other cell lines including lung cancer cell lines (SBC-3, NCI-H69, and DMS53), colorectal cancer cell lines (SW480 and SW620), and AZ-521, the parental cell line of AZ-P7a. Considering that let-7 miRNA family functions mainly as tumor suppressor genes [@pone.0013247-Roush1] to target oncogenes such as RAS and high mobility group A2 (HMGA2) [@pone.0013247-Boyerinas1], we propose that AZ-P7a cells selectively secrete let-7 miRNAs into the extracellular environment via exosomes to maintain their tumorigenic and metastatic propensities. Results {#s2} ======= Isolation of exosomes from various cancer cell lines {#s2a} ---------------------------------------------------- Exosomes are produced from inner cells to the extracellular environment via an exocytosis-like pathway [@pone.0013247-Thry1]. In addition to body fluids such as serum and plasma from peripheral blood [@pone.0013247-Simpson1], [@pone.0013247-Rabinowits1], exosomes are found in the medium of cultured cells [@pone.0013247-Simpson1], facilitating identification of exosomal molecules including proteins, mRNAs, and miRNAs for the aim for their clinical use. Profiling such exosomal molecules produced in cultured cancer cells have led to discover a novel candidate marker and antigen for cancer diagnostics and immunotherapy [@pone.0013247-Simpson1]. In this study, for the purpose of profiling exosomal miRNAs, we first isolated exosomes from culture media of 46 cancer cell lines with various tissue types, which include 8 for stomach, 16 for lung, 5 for colon, 9 for pancreas, 3 for prostate, and 5 for breast. After cells were grown for 48 h, exosomes were collected with a combination of successive centrifugation and molecular weight cut-off membranes as described in [Materials and Methods](#s4){ref-type="sec"}. Their purity of exosomal fractions was assessed by analyses of transmission electron microscopy and western blotting. Transmission electron microscopy showed round-shaped vesicular membrane structures approximately within the size of 100 nm in diameter. Among three cell lines, SBC-3, AZ-521, and AZ-P7a, it is of interest that the morphology from AZ-P7a cells was more homogeneous and condensed than other cells ([Figure 1A](#pone-0013247-g001){ref-type="fig"}). Immunoelectron microscopy showed that the extracellular particles isolated from the culture medium of AZ-P7a cells had immunoreactivity with an antibody against CD63, one of the exosomal membrane proteins, on the capsular membranes ([Figure 1B](#pone-0013247-g001){ref-type="fig"}). The presence of known exosomal proteins including CD29/β1-integrin, Aip1/Alix and tumor susceptibility gene 101 (Tsg101) was confirmed by western blot analysis, while a protein localized to endoplasmic reticulum, Bip/Grp78, was undetectable ([Figure 1C](#pone-0013247-g001){ref-type="fig"}). This result indicates that the contamination of material derived from other cellular compartments in the exosomal fractions was minimal. The yield of exosomal proteins from AZ-P7a cells was 10 times higher than that from AZ-521 cells; ∼2.5 mg proteins were obtained from 1×10^8^ AZ-P7a cells. ![Characterization of exosomes.\ A. Morphological characterization of exosomes derived from AZ-P7a, AZ-521, and SBC-3 cells by transmission electron microscopy. B. Immunoelectron micrographs of AZ-P7a exosomes labeled with immunogold anti-CD63. Exosomes with black colloidal gold particles on the capsular membranes were observed as positive CD 63 staining under the transmission electron microscope (arrowheads). C. Molecular characterization of exosomes derived from AZ-P7a, AZ-521, and SBC-3 cells by western blot analysis. Protein extracts (10 µg) prepared from cells (C) or exosomes (Ex) were assessed using antibodies against exosomal protein markers (CD29/β1-integrin, Aip1/Alix, and Tsg101) and an endoplasmic reticulum marker (Bip/Grp78).](pone.0013247.g001){#pone-0013247-g001} Enrichment of exosomal RNAs derived from AZ-P7a cells {#s2b} ----------------------------------------------------- After isolation, exosomal total RNAs were isolated from 46 cultured cell lines. Interestingly, the RNA yields from AZ-P7a cells were much greater than those from other cells ([Figure 2A](#pone-0013247-g002){ref-type="fig"}). The distribution in length showed the presence of large amounts of small RNAs in exosomes (center panel, [Figure 2B](#pone-0013247-g002){ref-type="fig"}). It should be noted that there is a significant difference in total exosomes and the exosomal miRNA levels between patients with lung adenocarcinoma and controls [@pone.0013247-Rabinowits1]. AZ-P7a cell line was established by repeated inoculation of the parental AZ-521 cells in mice, resulted in high peritoneal-metastatic potential [@pone.0013247-Nishimori1], [@pone.0013247-Fukui1]. Thus, the high yields of both RNA and protein along with the morphological characteristics in AZ-P7a cells may be attributed to their high tumorigenic and metastatic propensities. Since miRNAs have been detected in cell-free body fluids [@pone.0013247-Mitchell1]--[@pone.0013247-Gilad1], we also performed direct isolation of total RNAs from culture medium without procedure for isolating exosomes. The amounts of total RNAs from culture media were generally greater than those from exosomes ([Figure 2A](#pone-0013247-g002){ref-type="fig"}). RNA distribution in length showed that miRNA fractions were detected in the range from 10 to 40 nucleotide length in RNAs prepared from culture medium (right panel) as well as exosomes (center panel) from AZ-P7a cells ([Figure 2B](#pone-0013247-g002){ref-type="fig"}). According to the manufacture\'s protocol for Bioanalyzer, peaks with longer than 40 nucleotide length contain other small RNAs including tRNAs at 70∼90, 5S ribosomal RNA at 100, and 5.8S ribosomal RNA at 150, as observed in the intracellular fraction (left panel). These results suggest that miRNAs may exist in the culture medium outside the exosomal fraction although RNAs are believed to be protected from RNases as being packaged in exosomes. Particularly, compared to adherent cells, the increment of RNA yields was remarkably greater in floating cells (NCI-H69 and Lu-135) or cells with mix populations of floating and adherent cells (Colo205), which may be reflected by contamination of RNAs shed from lysed cells that have been continuously left in culture media. ![Total RNAs in exosomes and culture media from various cancer cell lines.\ A. Amounts of total RNAs recovered from exosomes (lower panel) and culture media (upper panel). The amount of total RNAs per cell was shown. B. Distribution in length of RNA. Isolated total RNAs from the exosomes and culture medium of AZ-P7a cells were detected using Bioanalyzer. The result of cellular total RNAs was shown for comparison.](pone.0013247.g002){#pone-0013247-g002} miRNA profiling by microarray analysis {#s2c} -------------------------------------- miRNA profiles in the intracellular and extracellular fractions were determined by microarray analysis for seven cancer cell lines including AZ-521, AZ-P7a, SBC-3, NCI-H69, DMS53, SW480, and SW620. Samples from culture media as well as exosomes provided hybridization signals ([Figure 3](#pone-0013247-g003){ref-type="fig"}), indicating the existence of miRNAs in these samples. Up to the present, analyzing microarray data for miRNA among samples, no clear methods have been made for normalization of signal intensity since it is difficult to find an appropriate internal standard material [@pone.0013247-Pradervand1]. Thus, miRNA microarray data have been in general analyzed without normalization, assuming that the total amounts of input RNAs are constant among samples [@pone.0013247-Pradervand1], [@pone.0013247-Sato1]. In this study, after averaging miRNA microarray data between two replicates, we ranked miRNAs according to their signal intensities. We found that let-7 miRNA family such as let-7a, let-7b, let-7c, let-7d, let-7e, let-7f, let-7g, and let-7i were detected with relatively high signals throughout most of the intracellular fractions from the seven cell lines ([Table 1](#pone-0013247-t001){ref-type="table"}, [Figure 3](#pone-0013247-g003){ref-type="fig"}). However, no or little signal intensities for the let-7 miRNAs were detected in the extracellular fractions except both extracellular fractions from AZ-P7a cells and culture medium fraction from NCI-H69 cells ([Table 1](#pone-0013247-t001){ref-type="table"}, [Figure 3](#pone-0013247-g003){ref-type="fig"}). Particularly, in both the extracellular fractions from AZ-P7a cells, all the eight let-7 miRNAs were detected although the rank of let-7g was lower than other seven let-7 miRNAs. Based on the distinct patterns of let-7 miRNA levels among the three cellular fractions, we divided the seven cell lines into three groups ([Figure 3](#pone-0013247-g003){ref-type="fig"}) as follows; (1) AZ-P7a, cells that let-7 miRNAs were found in all the three fractions, (2) AZ-521 along with SBC-3, DMS53, SW480, and SW620, cells that let-7 miRNAs were generally found in only intracellular fractions, (3) NCI-H69, cells that let-7 miRNAs were found in the intracellular fractions as well as in the culture medium to some extent. ![miRNA profiling in the intra- and extra-cellular fractions from AZ-P7a, AZ-521, NCI-H69, SBC-3, DMS53, SW480 and SW620 cells.\ miRNA profiles in cells (C), exosomes (Ex), and culture media (CM) were obtained by miRNA microarray analysis. Y axis represents log~2~ of hybridization signals, shown by bars and filled boxes. The arrows on the left side from the bunch of signals represent signals corresponding to let-7 miRNA family including let-7a (red), let-7b (yellow), let-7c (light green), let-7d (dark green), let-7e (sky blue), let-7f (blue), let-7g (navy blue), and let-7i (purple).](pone.0013247.g003){#pone-0013247-g003} 10.1371/journal.pone.0013247.t001 ###### Ranking of let-7 miRNA family in the intra- and extra-cellular fractions based on signal intensities from microarray data. ![](pone.0013247.t001){#pone-0013247-t001-1} Cell line Sample Rank by signal intensity ----------- -------- -------------------------- ---- ----- ---- ---- ---- ----- ----- ----- AZ-P7a Cells 2 9 8 14 20 6 29 22 202 Ex 10 4 9 34 28 15 90 51 212 CM 30 13 22 71 83 46 147 63 209 AZ-521 Cells 3 4 5 13 17 6 22 18 171 Ex -- 53 130 -- -- -- -- -- 153 CM -- 99 -- -- -- -- -- 158 225 NCI-H69 Cells 1 2 6 25 27 4 44 23 261 Ex -- 41 -- -- -- -- -- -- 102 CM 36 16 24 86 91 55 -- 69 131 SBC-3 Cells 1 31 56 83 17 7 115 -- 132 Ex -- 38 -- -- -- -- -- -- 56 CM -- -- -- -- -- -- -- -- 186 DMS53 Cells 2 8 20 16 -- 9 29 10 85 Ex -- 25 -- -- -- -- -- -- 42 CM -- -- -- -- -- -- -- -- 86 SW480 Cells 1 2 15 13 36 6 23 25 124 Ex -- 33 -- -- -- -- -- -- 47 CM -- 72 -- -- -- -- -- -- 106 SW620 Cells 2 1 13 14 23 7 21 16 107 Ex -- -- -- -- -- -- -- -- 32 CM -- -- -- -- -- -- -- -- 88 Ex = exosomes. CM = culture medium. miRNAs = the total number of miRNAs detected. A dash (-) indicates no detection. RT-PCR analyses of intracellular and extracellular let-7 miRNA family {#s2d} --------------------------------------------------------------------- We then performed quantitative RT-PCR for the eight let-7 miRNA family to validate the microarray data. let-7 miRNAs were readily detected in the intracellular and extracellular fractions from AZ-P7a cells ([Figure 4A](#pone-0013247-g004){ref-type="fig"}) and AZ-521 cells ([Figure 4B](#pone-0013247-g004){ref-type="fig"}). The levels of let-7 miRNAs per input RNAs were generally higher in the extracellular fractions from AZ-P7a cells than from AZ-521 cells. For normalizing the levels of target miRNAs obtained by RT-PCR, U6 snRNA has been generally used as an internal control. We observed the presence of U6 snRNA in all the three fractions from AZ-P7a cells and AZ-521 cells ([Figure 4C](#pone-0013247-g004){ref-type="fig"}) and then compared the levels of let-7 miRNAs between the corresponding fractions from these two cell lines. After normalization, the levels of let-7 miRNAs including let-7a, let-7b, let-7c, let-7d, let-7e, and let-7i in both the fractions of exosomes and culture media from AZ-521 cells were lowered as compared with those from AZ-P7a cells. On the contrary, there was no large difference in the levels of the intracellular let-7 miRNAs. We next compared the level of exosomal let-7a from AZ-P7a cells with those from other cancer cell lines including SBC-3, NCI-H69, DMS53, SW480, and SW620, and found that the let-7a level only from SW620 cells was relatively close (0.7 times) to the level from AZ-P7a cells ([Figure 4E](#pone-0013247-g004){ref-type="fig"}). It should be noticed that the intracellular level of let-7a from SW620 cells was approximately 3.5 times more abundant than that from AZ-P7a cells. Based on these results, we assumed localization of let-7a in the fractions of cells and exosomes ([Figure 5](#pone-0013247-g005){ref-type="fig"}), and conducted further analysis. Normalized by the U6 levels, we calculated relative levels of let-7a packaged in exosomes per the levels of cellular let-7a ([Figure 4F](#pone-0013247-g004){ref-type="fig"}). As a result, the amount of exosomal let-7a was much greater in AZ-P7a cells than other six cells, suggesting that AZ-P7a cells were selectively and actively secreted let-7 miRNA family into the extracellular environment via exosomes. ![RT-PCR analyses of intra- and extra-cellular let-7 miRNA family.\ A, B. Levels of mature let-7 miRNA family including let-7a, let-7b, let-7c, let-7d, let-7e, let-7f, let-7g, and let-7i in cells (shaded bars), exosomes (filled bars), and culture medium (open bars) from AZ-P7a cells (A) and AZ-521 cells (B). Each miRNA level was analyzed using quantitative RT-PCR. The cycle threshold (Ct) value is presented as the mean ± SD (n = 4). C. Levels of U6 snRNA in cells (shaded bars), exosomes (filled bars), and culture medium (open bars) from AZ-P7a and AZ-521 cells detected by quantitative RT-PCR analysis. The cycle threshold (Ct) value is presented as the mean ± SD (n = 4). D. Relative amounts of let-7 miRNA family in cells (shaded bars), exosomes (filled bars), and culture medium (open bars) from AZ-521 cells versus AZ-P7a cells. The dotted line and arrow represents the levels of let-7 miRNA family in AZ-P7a cells, shown as 100%. The levels of U6 snRNA in each sample were used as an internal standard for normalization amounts of let-7 miRNAs. Except let-7f and let-7g, the levels of let-7 miRNAs were reduced in the extracellular fractions. E. Relative amounts of let-7a in cells (shaded bars) and exosomes (filled bars) from 7 cancer cell lines including AZ-P7a, AZ-521, NCI-H69, SBC-3, DMS53, SW480 and SW620 cells. The dotted line represents the levels in AZ-P7a cells, shown as 1. The levels of U6 snRNA in each sample were used for normalized amounts of let-7a. The value of fold change is presented as the mean ± SD (n = 3) from samples independently prepared from cell culture. F. Relative amounts of exosomal let-7a in 7 cancer cell lines. The dotted line represents the level in AZ-P7a cells, shown as 100%.](pone.0013247.g004){#pone-0013247-g004} ![Models on difference in localization of let-7 miRNA family.\ Based on the results obtained by microarray and RT-PCR analyses, these models were drawn. In comparison between AZ-P7a cells (left) and AZ-521cells (center), normalized by the amount of U6 snRNA (open triangles), the amount of exosomal let-7a miRNA (filled triangles) in AZ-521 cells was approximately 3% of that in AZ-P7a cells while the intracellular amount in AZ-521 cells was rather 1.4 times greater than that in AZ-P7a cells ([Figure 4E](#pone-0013247-g004){ref-type="fig"}). These models are applied to six of eight let-7 miRNA family tested, including let-7a, let-7b, let-7c, let-7d, let-7e, and let-7i. In SW620 cells (right), the amount of let-7a was 3.5 times greater than that in AZ-P7a cells while the amount of exosomal let-7a was 0.7 times less than in AZ-P7a cells ([Figure 4E](#pone-0013247-g004){ref-type="fig"}).](pone.0013247.g005){#pone-0013247-g005} Discussion {#s3} ========== In the present study, we revealed that let-7 miRNA family was secreted into the extracellular environment via exosomal transport, which was specific to a metastatic gastric cancer cell line, AZ-P7a. let-7 miRNA family in human consists of 10 sequences from 13 precursors [@pone.0013247-Roush1]. In general, let-7 miRNAs act as a tumor suppressor by targeting oncogenes such as RAS and HMGA2 and let-7 miRNAs are downregulated in many cancers from solid organs [@pone.0013247-Boyerinas1]. AZ-P7a cells possess a metastatic ability with peritoneal dissemination in nude mice [@pone.0013247-Nishimori1], [@pone.0013247-Fukui1]. Thus, we propose that the exosomal release of let-7 miRNAs into the extracellular environment results in decrease of anti-tumorigenic effect within the cells, which lead to maintain their oncogenesis and invasiveness. On the other hand, let-7 miRNAs less frequently play an oncogenic function, due to increase of expression by hypomethylation at the let-7 locus [@pone.0013247-Brueckner1] or targeting caspase-3 mRNA [@pone.0013247-Tsang1]. In this case, the exosomal release of let-7 miRNAs may cause transformation in target cells by transferring their oncogenic properties. In patients with lung cancer, the total levels of exosomes and their miRNAs increase compared to controls [@pone.0013247-Rabinowits1]. It has been demonstrated that tumor exosomes induce immune escape mechanism in cancers by spontaneous T cell apoptosis [@pone.0013247-Poutsiaka1]--[@pone.0013247-Ichim1]. In this study, we have shown homogeneous morphology and enrichment of RNAs in exosomes from metastatic AZ-P7a cells. This may be reflected by their tumorigenicity. miRNAs are very stable in blood plasma and serum since protected from RNases [@pone.0013247-Mitchell1]. Thus, it makes miRNA levels tested for clinical diagnosis as tumor markers and biomarkers. How does it happen? Chin et al. have suggested two hypotheses for the source of miRNAs on circulating blood samples; it may be due to a result of tumor cell death and lyses or release by tumor cells into the extracellular microenvironment of blood vessels [@pone.0013247-Chim1]. On the contrary to adherent cells, we observed that the amounts of total RNAs in culture media were relatively higher than in exosomes for cells with floating property such as NCI-H69, Lu-135, and Colo205. This increment probably resulted in the contamination of RNAs from dead cells in culture media. For the discovery of novel blood tumor markers and biomarkers, omics approaches including proteomics and transcriptomics are conducted either by direct analysis of blood samples or application of profiling in tissue samples. There are contradictory reports if whether the profiles of miRNA and mRNAs in bloodstream are parallel with the tumor\'s profile. The signature of exosomal miRNAs reflects that of the originating tumor cells in patients with lung adenocarcinoma [@pone.0013247-Rabinowits1] and breast cancer [@pone.0013247-Heneghan1]. On the other hand, Skog et al. have shown that microarray analysis for mRNA derived from glioblastoma cells and the corresponding exosomes revealed mRNAs exclusively detected in microvesicles, speculating that mRNAs are localized to a specific region of cytoplasm [@pone.0013247-Skog1]. In addition, Tanaka et al. have demonstrated that the level of miR-92a decreased in the blood plasma of acute leukemia patients while its strong expression was found in the tissue samples [@pone.0013247-Tanaka1]. Our results on selective enrichment of let-7 miRNA family in AZ-P7a cells-derived exosomes fit with latter case. Unless target molecules are miRNAs derived from circulating tumor cells, it suggests careful investigation are needed to use miRNA profiles in tissue samples for detection in serum or plasma samples. Since exosomes are produced in hematopoietic cells as well as epithelial cells, circulating blood contains exosomes derived from both type of cells. In general, to investigate exosomal profiles of proteins, mRNAs, and miRNAs in serum and plasma, exosomes derived from tumors with epithelial characteristics are separated from those from hematopoietic cells [@pone.0013247-Rabinowits1], [@pone.0013247-Taylor1]. This is performed using affinity purification with antibodies against molecules such as EpCAM that express on the surface of epithelial cells. Our findings that there were similar levels of let-7 miRNA family between exosomes and culture media from AZ-P7a cells suggest that these miRNA levels may be measured without isolating exosomes from clinical samples such as serum and plasma. In conclusion, this study demonstrated that let-7 miRNA family is rich in exosomes from a metastatic gastric cancer cell line, AZ-P7a cells. Since let-7 miRNAs are involved in proliferation process [@pone.0013247-Boyerinas1], their exosomal secretion may play an important role in tumorigenesis and metastasis. Materials and Methods {#s4} ===================== Cell culture {#s4a} ------------ Human stomach cancer cell lines (AZ-521, MKN45, KATOIII, and NUGC-3), human lung cancer cell lines (SBC-3, Lu-65, Lu-135, and KNS-62), and human pancreatic cancer cells (Suit-2) were purchased from Japanese Collection of Research Bioresources. Human stomach cancer cells (SNU-1), human lung cancer cell lines (DMS-53, DMS114, NCI-H69, A549, NCI-H1299, NCI-H1155, NCI-H520, NCI-H1560, SK-MES-1, and HCC827), human mesothelioma cells (MSTO-H211), human colorectal cancer cell lines (SW480, SW620, Colo205, LoVo, and HCT116, human pancreatic cancer cell lines (BxPC-3, AsPC-1, Panc-1, MIAPaca-2, HPAF-II, PL45, and Capan-2), human prostate cancer cell lines (PC-3, SU145, and LNCaP), and human breast cancer cell lines (T-47D, McF7, SK-BR-3, BT-474, and MDA-MB-231) were purchased from American Type Culture Collection. MKN28 (human stomach cancer cells) and PSN-1 (human pancreatic cancer cells) were purchased from Immuno-Biological Laboratories and European Collection of Cell Culture, respectively. Human stomach cancer cell lines, AZ-P7a [@pone.0013247-Nishimori1], [@pone.0013247-Fukui1] and MKN45P [@pone.0013247-Iinuma1] were gift from Drs. Yutaka Yonemura and Yoshio Endo. In AZ-521 and AZ-P7a cells, no RT-PCR products were observed for 14 retroviruses including HPV16, EBV, CMV, SV40, HBV, HCV, BKV, JCV, HHV7, HTLV1, HTLV2, HIV1, HIV2, and ADV type 1 (data not shown), excluding infection by these retroviruses in both the cell lines. Cells were maintained in RPMI-1640 medium (Sigma-Aldrich) supplemented with 10% heat-inactivated fetal bovine serum (Invitrogen), 100 U/ml penicillin, and 0.1 mg/ml of streptomycin. Production and isolation of exosomes {#s4b} ------------------------------------ Adherent cells were seeded to 150 mm-dishes at appropriate number of cells ranging from 3×10^6^ to 1×10^7^ cells per dish, while floating cells were inoculated into 75 cm^2^-flask at 2∼2.5×10^7^ cells per flask. Cells were cultured in complete RPMI-1640 medium of 25 ml and 50 ml for dishes and flasks, respectively, for 24 h at 37°C and 5% of CO~2~, washed twice with phosphate-buffered saline (PBS), and incubated 48 h in phenol red-free RPMI-1640 medium (Sigma-Aldrich) containing 10% heat-inactivated fetal bovine serum which was previously depleted contaminating microvesicles by overnight centrifugation at 100,000×g. Exosomes were isolated from the total cells ranging from 6×10^7^ to 3×10^8^ cells as previously described [@pone.0013247-Ji1]. Briefly, culture medium was collected and centrifuged at 800×g for 5 min and additional 2,000×g for 10 min to remove lifted cells. The supernatant was subjected to filtration on a 0.1 µm pore polyethersulfone membrane filter (Corning) to remove cell debris and large vesicles, followed by concentration by a 100,000 Mw cut-off membrane (CentriPlus-70, Millipore). The volume of supernatant was reduced from approximately 250--500 ml to approximately 30 ml. The supernatant was then ultracentrifuged at 100,000×g for 1 h at 4°C using 70Ti rotor (Beckman Coulter). The resulting pellets were resuspended in 6 ml PBS and ultracentrifuged at 100,000×g for 1 h at 4°C using 100Ti rotor (Beckman Coulter). Transmission electron microscopy {#s4c} -------------------------------- The pelleted exosomes were mixed with equal quantities of freshly prepared 2% glutaraldehyde in PBS, incubated overnight at 4°C, postfixed with 1% osmium tetroxide in PBS at 4°C for 2 h, and dehydrated in a graded series of ethanol. Following dehydration, the samples were transferred to propylene oxide and embedded in epoxy resin Quetol 812 (Nisshin EM). Ultrathin sections were cut with Ultracut UCT (LEICA), stained with uranyl acetate and lead citrate, and observed with the JEM-1230 Electron Microscope (JEOL). Immunoelectron microscopic analysis was performed according to the method described by Xiao et al. [@pone.0013247-Xiao1] with minor modifications. In brief, exosomes were mixed and incubated with mouse anti-human CD63 monoclonal antibody (catalog no. sc-51662, Santa Cruz) for 1 h at room temperature. The sample was dropped onto the membrane surface of a copper mesh (collodion/carbon coated 400 mesh, Nisshin EM) and incubated for 1 h at room temperature. After wash with PBS, immunogold conjugated goat anti-mouse IgG (catalog no. EM.GMHL15, BBInternational) diluted at 1∶20 was dropped onto the membrane. The copper mesh was floated in the droplets with its membrane surface faced down at room temperature for 30 min. After wash with PBS, uranyl acetate drops was put onto the copper mesh surface and stained at room temperature for 30 s. Exosomes with black colloidal gold particles on the capsular membranes were marked as positive under the transmission electron microscope. Western blot analysis {#s4d} --------------------- Cells and exosomal fractions were washed, resuspended in a lysis buffer \[7.5 M urea (Sigma-Aldrich), 2.5 M thiourea (Sigma-Aldrich), 12.5% glycerol (Wako), 50 mM Tris, 2.5% n-octyl-beta-D-glucoside (Sigma-Aldrich), 6.25 mM Tris(2-carboxyethyl)phosphine hydrochloride (Sigma-Aldrich), 1.25 mM protease inhibitor (catalog no. P2714, Sigma-Aldrich)\], and incubated for 1 h at 4°C using the Rotator RT-50 (TAITEC). After centrifugation at 14,000×g for 60 min at 4°C, the supernatant was collected and protein concentration was determined by the Bradford Protein Assay Kit (Bio-Rad). A portion of proteins (10 µg) were denatured by boiling in Laemmli sample buffer, run on 10% SDS-polyacrylamide gels, and transferred to Immobilon-P PVDF membranes (0.45 µm pore size, Millipore). The blots were blocked for 1 h with 5% non-fat dry milk in Tris-buffered saline containing Tween 20 (10 mM Tris--HCl, pH 7.5, 150 mM NaCl and 0.01% Tween 20). After blocking the membranes were incubated with each primary antibody or antiserum including rabbit anti-Tsg101 serum (catalog no. T5951, Sigma-Aldrich), goat anti-Aip1/Alix serum (catalog no. sc-49268, Santa Cruz), mouse monoclonal anti-CD29/β1-integrin (catalog no. sc-610468, BD Biosciences), and mouse monoclonal anti-Bip/Grp78 (catalog no. 610979, BD Biosciences) at a dilution of 1∶200 for 1 h at room temperature. The blots were then incubated with the corresponding anti-IgG secondary antibody conjugated with horseradish peroxidase (Jackson Laboratories) at a dilution of 1∶2,000 for 1 h at room temperature. Specific proteins were visualized using the ECL system (GE Healthcare) and the FUJIFILM Luminescent Image Analyzer LAS3000 (Fuji Film). The protein molecular weight was deduced using the Precision Plus Protein Standards (Bio-Rad Laboratories). RNA Isolation {#s4e} ------------- For RNA isolation from culture medium, cells were seeded as described above. 48 h after grown, culture medium was collected and centrifuged at 800×g for 5 min and additional 2,000×g for 10 min. The supernatant was subjected to filtration on a 0.22 µm pore polyethersulfone membrane filter (Corning), followed by concentration with a 5,000 Mw cut-off membrane (Centricon Plus-70, Millipore). The final volume of supernatant was approximately 10--20 ml from the original volume of 250--500 ml. The supernatant was mixed with equal volume of QIAzol Lysis Reagent (Qiagen) and followed the procedure for RNA isolation described below. Total RNA was isolated from cells, exosomes, or culture media using the miRNeasy Mini Kit (Qiagen) according to the manufacture\'s instructions. To remove genomic DNA, 40 µg of cellular total RNA were incubated with 40 units of RQ1 RNase-Free DNase (Promega) at 37°C for 30 min in the presence of 40 units of RNasin Plus RNase Inhibitor (Promega). For total RNA from exosomes and culture media, all amounts of crude products were treated in the same procedure except the use of 5 U DNase. RNA was purified using the RNeasy MinElute Cleanup Kit (Qiagen) according to the manufacturer\'s instructions. The RNA samples were quantified with a NanoDrop spectrophotometer (Thermo Fisher Scientific) and assessed using the Agilent Small RNA Kit and the Agilent 2100 Bioanalyzer (Agilent Technologies). Microarray analysis {#s4f} ------------------- For miRNA profiling, 100 ng and 20 ng of cellular and extracellular total RNAs, respectively, were fluorescence-labeled using the miRNA Complete Labeling Reagent and Hyb Kit (Agilent Technologies) according to manufacture\'s protocol (Agilent miRNA microarrays version 2.2). The comprehensive miRNA analysis was performed using the Human miRNA Microarray Kit (8×15 K) Ver.3.0 (Agilent). Hybridization signals were detected with the DNA Microarray Scanner (Agilent Technologies). The scanned images were analyzed using the Agilent Feature Extraction software and data analysis was performed using the GeneSpring GX software (Agilent Technologies). The data discussed in this manuscript have been deposited in NCBI\'s Gene Expression Omnibus (GEO) and are accessible through GEO Series accession number GSE21350: <http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE21350>). Reverse transcription (RT) and quantitative RT-PCR for miRNA analysis {#s4g} --------------------------------------------------------------------- Quantitative miRNA levels were determined using real-time RT-PCR with the Applied Biosystems 7900 HT Sequence Detection System (Applied Biosystems), TaqMan® Gene Expression Assay (Applied Biosystems) for human let-7a (assay ID 000377), let-7b (assay ID 002619), let-7c (assay ID 000379), let-7d (assay ID 002283), let-7e (assay ID 002406), let-7f (assay ID 000382), let-7g (assay ID 002282), let-7i (assay ID 002221), and U6 snRNA (assay ID 001973) as an endogenous control. Ten ng of total RNA were subjected to reverse transcription with TaqMan® Universal PCR Master Mix No AmpErase (Applied Biosystems) and the respective TaqMan® reagents for target genes. RT-PCR was carried out in a total volume of 20 µl reaction mixture according to the manufacture\'s protocol. Amplification was carried out as follows: 95°C for 10 min, 40 cycles of 95°C for 15 s and 60°C for 60 s. Samples were analyzed in triplicate as biological replicate or quadruplicate as technical replicate. The miRNA levels were defined from the cycle threshold (Ct), the comparative Ct method, and normalization by the level of U6 snRNA in each sample. Fold increases or decreases in each miRNA level of cell lines tested were determined by reference to the level of AZ-P7a cell line. We thank the members of the Shizuoka Cancer Center Research Institute for their many helpful suggestions. We also thank Drs. Yutaka Yonemura and Yoshio Endo for gift of AZ-P7a and MKN45P cell lines. Competing Interests:The authors have declared that no competing interests exist. Funding:The work was supported by a grant in cooperation with Innovative Technology and Advanced Research in Evolutional Area (CITY AREA) from the Ministry of Education, Culture, Sports, Science and Technology of Japan. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: KO. Performed the experiments: KI AF KK YW KM. Analyzed the data: KO KH YF SiO KY TM. Contributed reagents/materials/analysis tools: KY TM. Wrote the paper: KO KY TM. [^2]: Current address: Integrated Research Institute, Tokyo Institute of Technology, Yokohama, Japan	{ "pile_set_name": "PubMed Central" }
Achim Quaiser, Mart Krupovic, Alexis Dufresne, André-Jean Francez, and Simon Roux. Diversity and comparative genomics of chimeric viruses in Sphagnum-dominated peatlands. Virus Evolution, 2016, 2(2): vew025 The publisher wishes to apologize for mistakenly having abbreviated "Chimeric viruses" to "CHIV" throughout the article where this was not always appropriate. The article has been updated accordingly.	{ "pile_set_name": "PubMed Central" }
In an isolated model of myocardial stunning, cardiac myocytes demonstrate significant swelling and reduced contractility during exposure to 3 different stresses: hypothermic hyperkalemic cardioplegia (CPG), hypo-osmotic stress, or metabolic inhibition.[@b1]--[@b3] The ATP-sensitive potassium channel (K~ATP~) opener, diazoxide (DZX), ameliorates these stress-induced derangements in mouse, rabbit, and human tissue.[@b1]--[@b5] A role of K~ATP~ channels in ischemic preconditioning, by coupling membrane conductance of K^+^ ions to the metabolic state of the cell, has also been well defined in multiple animal models.[@b6]--[@b10] However, no precise mechanism of cardioprotection has been identified. K~ATP~ channels are formed as hetero-octamers of subunits that include 4 potassium inward rectifying (Kir) subunits and 4 regulatory subunits from the sulfonylurea receptor family (SUR1/2).[@b11]--[@b13] Different combinations of Kir and SUR isoforms exist in different tissues, each with unique properties and function. For example, the predominant combination of subunits found in murine ventricle is Kir6.2 and SUR2A, whereas murine atria mainly consist of Kir6.2 and SUR1.[@b14] Identification of these subunits has allowed for genetic manipulation to more specifically evaluate the mechanism of DZX. Using K~ATP~ subunit-knockout animal models, we have shown that the cardioprotection afforded by DZX requires the regulatory subunit, SUR1, but the pore-forming subunit involved remains elusive.[@b15],[@b16] We have also demonstrated that the sarcolemmal K~ATP~ channel (sK~ATP~) is implicated in myocyte swelling secondary to stress, and that knockout of the pore-forming subunit of this channel (Kir6.2) provides resistance to myocyte swelling.[@b2],[@b17] However, there is evidence that DZX cardioprotection does not require the sK~ATP~ channel, leading us to propose mechanisms that may include non-K~ATP~ channel or mitochondrial K~ATP~ channel (mK~ATP~) locations of action.[@b15],[@b16],[@b18]--[@b21] Regulation of mitochondrial volume through a purported mK~ATP~ channel is one of the proposed mechanisms of action of DZX.[@b22]--[@b24] Because no molecular composition of mK~ATP~ has been determined, identification of mK~ATP~ channel subunits has to date involved indirect methods, such as pharmacological inhibition or observed changes in mitochondrial volume (proposed assay of mK~ATP~ activity).[@b24],[@b25] Using such strategies, the inward rectifier potassium channel subunit, Kir1.1 (renal outer medullary kidney channel; ROMK) has been implicated as a pore-forming subunit of the mK~ATP~ channel utilizing the nonspecific inhibitor Tertiapin Q (TPN-Q).[@b24] The present study was conducted to assess potential roles of Kir1.1, Kir3.1, and Kir3.4 in the cardioprotection of DZX at the cellular and mitochondrial levels using the Kir subunit inhibitor, TPN-Q. Methods ======= All animal procedures were approved by the Animal Studies Committee at Washington University School of Medicine, and all animals received humane care in compliance with the National Institute of Health's (NIH) Guide to Care and Use of Laboratory Animals.[@b26] Mitochondrial Isolation ----------------------- Mice (both sexes, 6 weeks to 5 months and 15 to 30 g in weight) were anesthetized with 3% Avertin (0.3 g of 2,2,2-tribromoethanol, 0.186 mL of 2-methyl2-butanol, and 9.814 mL of sterile water) intraperitoneally, and rapid cardiectomy was performed. To isolate mitochondria, atria were excised and discarded whereas ventricular tissue was rapidly minced in cold buffer (in mmol/L: 10 HEPES (N-\[2-hydroxyethyl\][@b1],[@b3]--[@b5] piperazine-N-\[4-butanesulfonic acid\]), 1 EDTA-potassium, 250 sucrose, adjusted to a pH of 7.1 with 20% KOH and transferred to a 10-mL glass tube with a Teflon pestle (Glas-Col Homogenizer, Terre Haute, IN), and volume was adjusted to 7 mL with buffer. Tissue was mechanically homogenized with a Teflon pestle driven by a low-speed motor drive shaft set at 120 rpm. The homogenate was transferred to 6 microcentrifuge tubes and centrifuged at 900g for 10 minutes at 4°C. The supernatant was combined into a clean test tube and mixed to get a homogeneous solution that was divided equally between 6 clean microcentrifuge tubes and centrifuged at 5000g for 15 minutes. One pellet was resuspended in 100 μL of buffer, and 10 μL was taken in duplicate for the Bradford protein assay (Thermo Scientific; Rockford, IL) to determine total protein. Each pellet was maintained on ice and was resuspended in test media volume to equal 0.3 μg/μL in order to standardize protein content. Mitochondrial Volume Measurement -------------------------------- The volume of isolated mitochondria was measured after suspension in test solution: (1) isolation buffer (No ATP; n=13); (2) 200 μmol/L of ATP (n=12); (3) 200 μmol/L of ATP and 100 μmol/L of DZX (Sigma-Aldrich, St. Louis, MO) (n=12); (4) 200 μmol/L of ATP, 100 μmol/L of DZX, and 500 nmol/L of TPN-Q (n=7); or (5) 200 μmol/L of ATP, 100 μmol/L of DZX, and 100 nmol/L of TPN-Q (n=6). Isolation buffer (10 mmol/L of HEPES, 200 mmol/L of mannitol, 50 mmol/L of sucrose, 1 mmol/L of EGTA; pH 7.2) was used as a control solution. ATP has been shown to close mitochondrial K~ATP~ channels, so 200 μmol/L of ATP was used to slow the initial rate of mitochondrial swelling (0% Mito K~ATP~ activity).[@b22] Conversely, DZX activates "Mito K~ATP~" channels, and 100 μmol/L of DZX was added to achieve maximal activation of mitochondrial K~ATP~ channels (100% Mito K~ATP~ activity). We were unable to reproduce the TPN-Q dose-response relationships on mitochondrial volume demonstrated by other investigators using 4 TPN-Q concentrations (0.5, 10, 90, or 1000 pmol/L) despite attempts utilizing 3 different vehicles for TPN-Q: 20% acetonitrile, water, and HEPES.[@b24] In the present study, water was used as a vehicle for TPN-Q because of its stability in the medium and because 20% acetonitrile alone resulted in myocyte swelling owing to its cyanide moiety. Mitochondrial matrix volume measurements were obtained using a light-scattering technique,[@b27] where the absorbance, at 520 nm, of a solution of isolated mitochondria was obtained every 14 seconds for the period of 3 minutes using UV Probe 2.33 (Shimadzu Scientific Instruments, Columbia, MD) and a spectrophotometer (UV-1700 Spectrophotometer; Shimadzu Scientific Instruments, Columbia, MD). Myocyte Isolation ----------------- Ventricular myocytes were isolated from C57BL/6J mice of either sex (age 6 weeks to 5 months and 15 to 30 g in weight), as previously described.[@b15] Mice were anesthetized with 2.5% Avertin intraperitoneally. Heparin (0.1 mL) was administered intraperitoneally. Rapid cardiectomy was performed and solution A (described below) was perfused through the aorta for 5 minutes. The heart was then perfused at 37°C for 12 to 20 minutes with solution B (described below). Ventricles were removed, minced, and placed into solution C (described below) and gently dispersed by glass pipette. Cells were allowed to centrifuge by gravity, and serial washings were performed every 10 minutes for 15 to 20 minutes. Cells were used within 5 hours and randomized to treatment group. A typical yield of viable myocytes was 65% to 75%. Solution A consisted of (in mmol/L, except as noted) 116 NaCl; 5.36 KCl; 0.97 Na~2~HPO~4~; 1.47 KH~2~PO~4~; 21.10 HEPES (N-\[2-hydroxyethyl\] piperazine-N′-\[4-butanesulfonic acid\]); 11.65 glucose; 26.50 μmol/L of phenol red (Sigma-Aldrich); 3.72 MgCl~2~; 4.40 NaHCO~3~; essential vitamins (100×, 10 mL; GIBCO, Grand Island, NY); and amino acids (50×, 20 mL; GIBCO). Solution B consisted of solution A plus 10 μmol/L of CaCl~2~; 1.2 mg/mL of collagenase (Type 2; Worthington Biochemical Corporation, Freehold, NJ). Solution C consisted of solution A plus 5 mg/mL of BSA (Sigma-Aldrich); 1.25 mg/mL of taurine; and 150 μmol/L of CaCl~2~. Myocytes were exposed to 37°C control TYR for 5 minutes to obtain baseline volume. Any changes in cell volume secondary to the isolation would be evident during this period. Myocytes were then exposed to test solution (10 minutes) followed by re-exposure to TYR solution (5 minutes). Test solutions included the following groups (n=12 for each): (1) TYR; (2) CPG; (3) CPG+100 μmol/L of DZX; (4) CPG+100 μmol/L of DZX+TPN-Q 200 nm/L; (5) TYR+TPN-Q; and (6) CPG+TPN-Q. CPG consisted of (in mmol/L): NaCl 110, NaHCO~3~ 10, KCl 16, MgCl~2~ 16, and CaCl~2~ 1.2 and was equilibrated with 95% O~2~ to 5% CO~2~ and titrated to the pH of 7.3 with 10% NaHCO~3~ solution. Diazoxide (7-chloro-3-methyl-1,2,4-benzothiadiazine-1,1-dioxide[@b4]; Sigma-Aldrich) dose of 100 μmol/L was utilized because it was effective in ameliorating cell swelling secondary to stress in previous studies.[@b1]--[@b4] A stock solution of DZX was made by dissolving DZX in 0.1% dimethyl sulfoxide (DMSO), at which concentration DMSO has no effect on cell volume.[@b28] Myocyte Volume Measurement -------------------------- Myocytes were used on the day of isolation and were not cultured. Myocytes were visualized on an inverted microscope stage (IX-51; Olympus, Tokyo, Japan), as previously described.[@b15] After 5 minutes, the chamber was perfused at a rate of 3 mL/min with TYR (in mmol/L): NaCl 130, KCl 5, CaCl~2~ 2.5, MgSO~4~ 1.2, NaHCO~3~ 24, Na~2~HPO~4~ 1.75, and glucose 10 (buffered to a pH of 7.4 using 95% O~2~ to 5% CO~2~). After viability was confirmed, myocyte images were captured using video-based edge detection software (IonOptix, Milton, MA) and volume measured every 5 minutes, as previously described.[@b5] Myocyte Contractility --------------------- Myocyte contractility was measured using a video-based edge detection system (IonOptix). Cells were paced using a field stimulator (MyoPacer; IonOptix) at a voltage of 105 above threshold at a frequency of 1 Hz with a 5-ms duration to avoid occurrence of fusion beats. After 5 minutes of stimulation, data were obtained from 12 to 30 consecutive beats and averaged. Parameters of contractility included percentage of cell shortening, maximal velocity of shortening, and percentage of cell relengthening, as previously described.[@b5] Contractility was measured at baseline and after 5 minutes of re-exposure to TYR. Cells that showed less than 7% cell shortening at baseline were excluded. Statistical Analysis -------------------- Data were analyzed using SYSTAT 13 (Systat Software, Inc, Point Richmond, CA). All data are presented as mean±SEM relative to baseline. A repeated-measures ANOVA was used for sequential time-based measurements for each test solution against its own baseline and control values. Using Fisher's least significant different test, post-hoc multiple comparisons were done between different test groups at different time points during test solution and re-exposure periods. Probability values \<0.05 were considered significant. A Shapiro-Wilk test was used to test for normality. If the data failed the normality test, a nonparametric (Friedman's nonparametric repeated-measures comparison) was used. Statistical analysis was performed using SyStat 10.2 (Systat Software). Results ======= Mitochondrial Matrix Volume --------------------------- Isolated mitochondrial volume is represented in [Figure 1](#fig01){ref-type="fig"}. There were no statistically significant differences in mitochondrial volume between any of the groups. ![Mitochondrial matrix volume. Isolated mouse mitochondria were exposed to: (1) isolation buffer (no ATP); (2) 200 μmol/L of ATP; (3) 200 μmol/L of ATP and 100 μmol/L of DZX; (4) 200 μmol/L of ATP, 100 μmol/L of DZX, and 500 nmol/L of TPN-Q; or (5) 200 μmol/L of ATP, 100 μmol/L of DZX, and 100 nmol/L of TPN-Q and volume measured using light scattering (at 520 nm) for 2 minutes. Mitochondrial matrix volume is represented as 1/percent change in absorbance (mitochondrial matrix swelling is inversely related to absorbance measured at 520 nm) over time. Data are represented as mean±SEM. DZX indicates diazoxide; TPN-Q, tertiapin Q.](jah30004-e002016-f1){#fig01} Myocyte Volume -------------- Myocyte volume over time is represented in [Figure 2](#fig02){ref-type="fig"}. Myocytes demonstrated no significant volume change from baseline during exposure to TYR with or without TPN-Q. Myocytes did demonstrate significant swelling when exposed to CPG (P\<0.05 vs. TYR). This swelling was prevented by addition of DZX (P\<0.05 vs. CPG). The further addition of TPN-Q prevented this benefit (volume homeostasis) of DZX (CPG+DZX+TPN-Q P\<0.05 vs. CPG+DZX). As expected, the interaction between experimental groups and time was significant (P\<0.001). ![TPN-Q prevents beneficial maintenance of myocyte volume during stress by DZX. Isolated myocytes were exposed to control Tyrode's physiologic solution (TYR) for 5 minutes (time 0 to 5), stress (CPG, CPG+DZX, CPG+DZX+TPN-Q, CPG+TPN-Q, and TYR+TPN-Q) for 10 minutes (time 5 to 15), followed by TYR for 5 minutes (time 15 to 20). Myocyte volume was compared. Data are represented as mean±SEM. \*P\<0.05 vs. TYR; ^Ψ^P\<0.05 vs. CPG+DZX. CPG indicates hypothermic hyperkalemic cardioplegia; DZX, diazoxide; TPN-Q, tertiapin Q.](jah30004-e002016-f2){#fig02} Myocyte Contractility --------------------- Myocyte contractility is presented in [Figure 3](#fig03){ref-type="fig"}. Contractility remained unaltered in the TYR group. Myocytes demonstrated a significant decline in all 3 parameters of contractility after exposure to CPG (P\<0.05 vs. TYR) that was worsened by the addition of TPN-Q. The addition of DZX to CPG prevented the significant reduction in contractility observed in the CPG group. TPN-Q inhibited the protection provided by DZX. ![TPN-Q prevents beneficial maintenance of myocyte contractility after stress by DZX. Isolated mouse myocytes were exposed to control Tyrode's physiologic solution (TYR) for 5 minutes (time 0 to 5), stress (CPG, CPG+DZX, CPG+DZX+TPN-Q, CPG+TPN-Q, and TYR+TPN-Q) for 10 minutes (time 5 to 15), followed by TYR for 5 minutes (time 15 to 20). Myocyte contractility was compared. Data are represented as mean±SEM. \*P\<0.05 vs. TYR; ^Ψ^P\<0.05 vs. CPG+DZX. CPG indicates hypothermic hyperkalemic cardioplegia; DZX, diazoxide; TPN-Q, tertiapin Q](jah30004-e002016-f3){#fig03} Discussion ========== In previous studies, we have shown that myocyte swelling and reduced contractility in response to stress (hyperkalemic CPG, hypoosmotic stress, and metabolic inhibition) are ameliorated by addition of the K~ATP~ channel opener, DZX.[@b1]--[@b5] Our efforts to localize the precise mechanism of cardioprotection afforded by DZX have implicated an involvement of the K~ATP~ channel, SUR1,[@b16],[@b19] but not the Kir6.2 subunit.[@b2] Other investigators have provided evidence in support of the idea that Kir1.1 (the ROMK) is a pore-forming subunit of the mK~ATP~ channel.[@b24] In the current study, changes in mitochondrial and myocyte volume were observed in the presence of ROMK blocker TPN-Q in the presence of DZX. A dose effect on mitochondrial volume was not detectable using TPN-Q in the presence of DZX and multiple doses of TPN-Q. The stability of TPN-Q in the vehicle medium was thus evaluated in the media, in 20% acetonitrile, HEPES, and in water. However, no significant change in mitochondrial volume was observed after the addition of TPN-Q, even at high concentrations, in contrast to other investigators.[@b24] Thus, we conclude that TPN-Q had no effect on mitochondrial volume in the presence of DZX, suggesting a location of action of DZX that was distinct from the channels targeted by TPN-Q. Myocytes demonstrated significant swelling in response to CPG that was prevented by the K~ATP~ channel opener, DZX, which is consistent with previous results.[@b2],[@b3] Similarly, a correlation between myocyte volume changes and contractility changes was consistent with previous work, supporting this myocyte model of stunning.[@b4],[@b5] TPN-Q prevented this beneficial effect of DZX, thus implicating TPN-Q targets---Kir1.1, Kir3.1, or Kir3.4---in the cardioprotection of DZX at the cellular level. Other non--Kir TPN-Q targets or undefined channel subunits sensitive to TPN-Q may play a role (GIRK1/4, K~ACH~, and voltage-dependent Ca^2+^-activated K^+^ channels) in cardioprotection and will be the subject of future investigations.[@b29]--[@b31] The present study further characterizes the mechanism of cardioprotection provided by DZX. TPN-Q did not alter DZX-induced mitochondrial swelling, but it did inhibit myocyte cardioprotection provided by DZX during stress. Given that TPN-Q inhibits Kir1.1-, Kir3.1-, and Kir3.4-dependent K-channel activities, these data support that any of these subunits (as well as undefined subunits sensitive to TPN-Q) could be involved in the cardioprotection afforded by DZX. However, these data also suggest that mitochondrial swelling by DZX does not involve Kir1.1, Kir3.1, or Kir3.4 and likely results from a yet to be identified mechanism. Future work to identify the site and mechanism of action of K~ATP~ channel openers will involve indirect methods until genetic identification is accomplished. Sources of Funding ================== This study was supported by NIH RO1 HL098182-01A1 (Lawton), NIH 5T32HL007776 (Henn, Janjua), and the Barnes Jewish Hospital Foundation (Lawton). Disclosures =========== None. [^1]: Presented at the American Heart Association Scientific Sessions, November 15--19, 2014 in Chicago, IL.	{ "pile_set_name": "PubMed Central" }
Background {#Sec1} ========== Suicide is a particular concern in mental health settings because of its strong association with mental illness \[[@CR1]\]. Although suicides rarely occur during in-patient care, these events are clinically important and are among the most concerning patient safety incidents in the mental health sector \[[@CR2]--[@CR4]\]. Suicide prevention is one of the primary tasks of health care professionals practicing in psychiatric wards \[[@CR4]\]. In-patient suicide prevention is a high-priority in many countries \[[@CR5]--[@CR7]\]; however, its practice remains poorly understood. The ethical and pragmatic problems posed by including suicidal patients in research have contributed to the currently limited research regarding the treatment of high-risk and hospitalized suicidal patients \[[@CR8]\]. To understand safety in health care services, information must be obtained from multiple sources, including the patient's perspective. As such, patients can provide insight regarding care and can contribute important information when other sources of evidence are limited \[[@CR9]\]. Patients can also provide unique information on adverse events in hospitals \[[@CR10], [@CR11]\] as well as useful descriptive feedback regarding safety, in particular sensitive safety-related topics \[[@CR12]\]. Patient experiences are considered one of the three pillars of health care quality, along with clinical safety and effectiveness of outcomes \[[@CR13]\]. Qualitative studies of patient experiences with psychiatric in-patient care have been reviewed within certain areas, such as involuntary hospitalizations \[[@CR14]\], physical restraint \[[@CR15]\], acute wards \[[@CR16]\], seclusion practices \[[@CR17]\], locked doors \[[@CR18]\] and service user expectations \[[@CR19]\]. However, no reviews to date have examined studies regarding suicidal in-patients. Therefore, this review aimed to summarize empirical qualitative studies by exploring suicidal patients' psychiatric in-patient care experiences to better understand their perspectives toward safety. Review question {#Sec2} --------------- A literature review was conducted to answer the following review question: How can we describe suicidal patients' experiences regarding safety during psychiatric in-patient care? Methods {#Sec3} ======= The selected studies were systematically reviewed using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines \[[@CR20]\]; the articles were then synthesized using thematic analysis \[[@CR21]\] and assessed further via quality appraisal \[[@CR22]\]. The objectives, inclusion criteria, analysis methods and search strategy were specified and documented in a protocol reviewed by the three authors prior to the database search. The authors are researchers with backgrounds in psychology (SHB), mental health nursing (KR) and safety science (SHB and KAA). Inclusion and exclusion criteria {#Sec4} -------------------------------- The eligibility criteria for inclusion in the review pertained to the following three characteristics: Type of study: Qualitative peer-reviewed studies in English with empirical data on patients' experiences regarding safety were eligible. Participants: Studies examining a sample of suicidal in-patients who were interviewed during their hospitalizations or after discharge were eligible. "Suicidal in-patients" included patients hospitalized after a recent suicide attempt, described as suicidal during hospitalization or with serious suicidal thoughts or ideations; self-harming behaviour was excluded. The final criteria related to Setting: Experiences regarding care in psychiatric hospital wards, including psychiatric emergency wards and psychiatric long-term in-patient care, were eligible. Studies in multiple hospital settings were included if information regarding psychiatric in-patient care experiences could be extracted. Patient experiences pertaining to outpatient clinics, community mental health care, home care, forensic psychiatric services, emergency care and medical care were excluded. Studies describing patient experiences with adverse side effects from pharmacological treatment were excluded. Studies with mixed patient samples and studies involving health care professionals' experiences were included if information regarding patient experiences could be extracted. Search strategy and study selection {#Sec5} ----------------------------------- To increase sensitivity, limitations on publication date were not imposed during the database search. The selection of databases, search terms and search methodology were determined in collaboration with a university librarian. The databases included in the systematic search were MEDLINE and the Academic Search Premier, CINAHL, SocINDEX with Full-Text and PsycINFO Ovid databases. Systematic database searches were conducted between June and December 2014 and in July 2016. Search terms were identified in relevant studies during the planning of the systematic review. The terms were selected from qualitative studies of patient experiences in mental health care and from qualitative studies of suicidal patients' experiences. All identified search terms were included to increase search sensitivity. The full electronic search strategy for PsycINFO is outlined in Additional file [1](#MOESM1){ref-type="media"}. We also screened reference lists and conducted author searches in EMBASE and Google Scholar. We systematically searched all of the above databases using the following terms: patient\* satisfaction\, patient\ preference\, in-patient\ experience\, patient\ experience\, patient\ perception\, patient\ view\, patient\ perspective\, patient\ opinion\, user\ experience\, consumer\ experience\, consumer participation, suicide, suicidal, feeling safe and feeling unsafe. The study selection process was conducted according to the eligibility criteria displayed in the flow diagram in Fig. [1](#Fig1){ref-type="fig"}. First, all titles were screened, and the abstracts were read by one author (SHB). Ineligible studies were excluded. Full-text articles were obtained for the eligible studies. Two authors (SHB and KR) independently assessed the full-text articles for eligibility in a standardized manner. A third author (KAA) validated the assessments. The level of agreement was generally high; however, setting was often discussed, as the studies were conducted in mixed settings. Agreement was reached by re-reading the articles to determine whether information on patient experiences with psychiatric in-patient care could be extracted from the studies in question. All authors were in agreement regarding the final inclusion and exclusion of all articles. A data extraction sheet was developed to guide study selection. Information from all full-text articles was added to the sheet. All studies were assessed based on the abovementioned eligibility criteria and colour-coded as red (no), orange (maybe) or green (yes).Fig. 1PRISMA (2009) flow diagram Synthesis of results {#Sec6} -------------------- Thematic analysis, as proposed by Thomas and Harden \[[@CR21]\] and Braun and Clarke \[[@CR23]\], was used to facilitate the synthesis of the results of the included studies. The thematic synthesis consisted of two stages. The first stage entailed coding the text "line by line", condensing the meaning units and developing descriptive themes. An inductive approach was used in which the descriptive themes remained close to the original findings of the studies \[[@CR23]\]. The second stage developed relationships between the descriptive themes and patient safety to generate analytical themes \[[@CR21]\]. Thematic mapping was used to identify relationships between meaning units, descriptive themes and analytical themes \[[@CR21], [@CR23]\]. Connections between patients' needs, expectations, experiences, reported outcomes (such as experiencing increased or decreased suicidal behaviour) and use of the term 'safety' were studied in the analytical stage. Coding and preliminary theme development were conducted by one author (SHB) and reviewed by all three authors. The analysis yielded 83 meaning units, nine descriptive themes and three analytical themes ("Connection", "Protection" and "Control"). Forty-nine of the 83 meaning units were found in the "Connection" theme, which was thus considered the most comprehensive theme. An example of theme condensation is presented in Table [1](#Tab1){ref-type="table"}. A full overview of the meaning units and themes is provided in Additional file [2](#MOESM2){ref-type="media"}.Table 1Example of theme condensationExample of extracted dataMeaning unitsDescriptive themeAnalytical themeLack of acknowledgment from observers; these perceptions sometimes overlapped with perceptions of a lack of empathy. Such behaviors included observers' reading books, appearing distracted or uninterested in the participant, and acting like the participant was a burden \[[@CR40]\].61. Lack of observer support manifests as lack of empathy and acknowledgementReceiving support from the observersProtectionFeelings of objectifications in formal observation without interpersonal engagement...It's a scary thing going somewhere where you feel like you're isolated and locked away. (Claire)...Being watched like that; it's freaky...a bit invasive...that separation, that 'us and them'. It's a bit tricky. (Kate) \[[@CR26]\].62. Feeling objectified and detached without observer support"They don't care. You get that feeling quite often. It just kind of supports that hopeless kind of feeling that life isn't worth living and nobody cares about anything." Such encounters did little to alleviate hopelessness, and six participants noted that they increased their anxiety or aggravated their dysphoria \[[@CR40]\].63. Feeling objectified increases stress and hopelessness Two authors (SHB and KR) independently assessed the methodological quality of the included studies and rated the studies based on Malterud's \[[@CR22]\] checklist for qualitative research. Malterud's guidelines for assessing qualitative studies and an example of a scored article are provided in Additional file [3](#MOESM3){ref-type="media"}. An overview of the quality assessment of the included papers is presented in Additional file [4](#MOESM4){ref-type="media"}. Results {#Sec7} ======= Study selection {#Sec8} --------------- The study selection process utilized the PRISMA guidelines \[[@CR20]\] (Fig. [1](#Fig1){ref-type="fig"}) and identified a total of 1,097 records through database searches. Additional searches yielded 29 records. After removing duplicates, the remaining 984 records were screened. Title screening and abstract reading resulted in the exclusion of 904 records that did not meet the eligibility criteria. Eighty full-text articles were read, and relevant information was extracted and entered into the information sheet, assessed according to the inclusion criteria and coded (yes/maybe/no). Sixty records were excluded for not meeting the inclusion criteria, and we ultimately included the remaining 20 studies in the systematic review (Table [2](#Tab2){ref-type="table"}).Table 2The list of included studiesAuthor/year/originAimSampleSettingData collection methods and analysisKey points related to in-patient careVatne & Nåden, 2016 \[[@CR38]\]. NorwayTo develop a deeper understanding of suicidal patients in the aftermath of suicide attemptsTen patients considered with serious suicidality after a suicide attempt. Non-psychotic. Interviewed after suicide attempt.Two emergency psychiatric wards and one crisis resolution team.Semi-structured interviews. Analysed using thematic analysis inspired by Braun and Clarke. Gadamerian hermeneutic approach.• Connectedness, someone who cares\ • Hospital admission important for staying alive\ • Support from family and friendsLees, Procter and Fassett, 2014 \[[@CR26]\]. AustraliaTo explore the experiences and needs of mental health-care consumers who had a suicidal crisis (shortened).Nine patients recovered from a recent suicidal crisis where they received mental health in-patient care.Setting not specified. Experiences of psychiatric in-patient care are described.In-depth, semi-structured interviews collected as part of a larger multi-method study. Analysed with a constant comparative method and classical content analysis.• Therapeutic engagement central to quality of care\ • Isolation, loss of control, objectificationMontross Thomas et al., 2014 \[[@CR28]\]. USATo better understand suicide experiences from the perspective of patients diagnosed with serious mental illness.23 patients hospitalized after a suicide attempt. Diagnosed with serious mental illness. Interviewed after discharge.Veterans Affairs Hospital, mental health program.Qualitative interviews with audio/videotaping. Analysed using van Manen's phenomenological framework.• Need for clinicians' empathy, compassion and listening skills\ • Addressing problems underlying suicide attemptVatne & Nåden, 2014 \[[@CR32]\]. NorwayTo explore the experiences of being suicidal and encounters with health care personnel.Ten patients considered seriously suicidal. Psychosis excluded. Interviewed after suicide attempt.Psychiatric emergency ward, sub-emergency psychiatric wards and one crisis resolution team.Semi-structured interviews. Analysed using thematic analysis inspired by Braun and Clarke. Gadamerian hermeneutic approach.• Openness and trust\ • Someone who addresses the matter\ • Being met on equal terms, humiliatedCutcliffe et al., 2012a \[[@CR41]\]. Unknown origin.To better understand the observed increased risk for suicide following discharge from an in-patient psychiatric service. Key theme one.20 patients admitted to the hospital with suicidal ideation and/or a lifetime history of suicidal behaviour. Interviewed after discharge.In-patient psychiatric service.Hermeneutic interviews. Analysed using van Manen's phenomenology.• Anxiety to go back to life without having a sense of control\ • Need to be involved in discharge planningCutcliffe et al., 2012b \[[@CR42]\]. Unknown origin.To better understand the observed increased risk for suicide following discharge from an in-patient psychiatric service. Key theme two.20 patients admitted to the hospital with suicidal ideation and/or a lifetime history of suicidal behaviour. Interviewed after discharge.In-patient psychiatric service.Hermeneutic interviews. Analysed using van Manen's phenomenology.• Patients still suicidal at discharge\ • Disorientation concerning what to do with their life\ • Need for post-discharge supportPavulans et al. 2012 \[[@CR27]\]. SwedenTo explore the experience of being suicidal, including a suicide attempt, and identify possible implications for health care professionals.Ten patients interviewed after a suicide attempt while hospitalized in a psychiatric ward.Psychiatric in-patient care at one university hospital.Semi-structured interviews. Analysed using van Manen's phenomenology and qualitative content analysis.• Being in need of control\ • Re-establish control before the point of no return\ • Control related to problem-solving and insightVatne & Nåden, 2012 \[[@CR29]\]. NorwayTo explore experiences of persons after a suicide crisis or a recent suicide attempt.Ten patients considered seriously suicidal. Psychosis excluded. Interviewed after suicide attempt.Psychiatric emergency ward, sub-emergency psychiatric wards and one crisis resolution team.Qualitative interviews. Analysed using thematic analysis. Gadamerian hermeneutic approach.• Losing touch with the world\ • Someone to see, listen and understand\ • Desperation increases with involuntary hospitalizationHolm & Severinsson, 2011 \[[@CR31]\]. NorwayTo explore how recovery processes facilitate changes in suicidal behaviour in women with borderline personality disorder.13 patients with suicidal behaviour. Borderline personality disorder.Recruited from different settings within mental health. Experiences of psychiatric in-patient care were described.In-depth interviews. Data analysed with thematic analysis.• Changing suicidal behavior by feeling confirmed, safe, and trusted.Cutcliffe et al, 2006 \[[@CR36]\]. EnglandTo determine if psychiatric/mental health nurses provide meaningful caring responses to suicidal people, and if so, how was it achieved.20 patients with experiences from a serious suicide attempt.Crisis care in emergency psychiatric services.Semi-structured interview. Data analysed with constant comparative method. Glaserian grounded theory approach.• Reconnecting the person with humanity\ • Guiding the individual back to humanity, learning to liveSun, et al 2006b \[[@CR25]\]. TaiwanPresentation of a nursing care theory developed to guide the care given to people with suicidal ideas and those with a previous suicide attempt.15 patients with either suicidal ideas or attempted suicide. Interviewed while hospitalized.Psychiatric hospital ward.Semi-structured interviews and participant observation. A grounded theory approach.• Safe and compassionate care giving via the therapeutic relationshipSun et al, 2006a \[[@CR24]\]. TaiwanTo investigate nurses' and patients' perceptions of psychiatric wards (the context of care) and the professionals' response (the intervening conditions) that may impact the delivery of suicidal nursing care.15 patients with either suicidal ideas or attempted suicide. Interviewed while hospitalized.Psychiatric hospital ward.Semi-structured interviews and participant observation. A grounded theory approach.• Protective environment\ • Access to lethal items\ • Group support, spiritual supportTalseth, Gilje & Nordberg, 2003 \[[@CR30]\]. NorwayTo describe a process of consolation revealed by two suicidal patients' experiences.Two patients. Interviewed after a suicide attempt (from the Talseth et al., 1999 \[[@CR34]\] study).Psychiatric hospital ward.Qualitative interviews. Phenomenological hermeneutic study inspired by Ricoeur's philosophy.• Vulnerability and deep despair\ • Closeness\ • Connection\ • The dialogue with HCPsWiklander, Samuelsson, & Åsberg, 2003 \[[@CR33]\]. SwedenTo extract and analyse the interview data concerning experiences of shame.13 patients with experiences from attempted suicide. Interviewed after discharge.Specialized psychiatric in-patient care.Qualitative semi-structured interviews. Transcripts analysed using qualitative methods (not specified).• Sensitive to attitudes and behaviours of HCPs\ • Shame reactions related to aspects of careTalseth, Jacobsson & Nordberg, 2001 \[[@CR39]\]. NorwayTo illuminate the experience of being treated by physicians.21 patients expressing the wish to die or attempted to commit suicide. Interviewed while hospitalized.Psychiatric emergency wards, psychiatric sub-emergency wards and one psycho-geriatric ward.Qualitative interviews interpreted using a phenomenological hermeneutic approach inspired by Ricoeur's philosophy.• Need for confirmation in interactions with physiciansSamuelsson et al., 2000 \[[@CR35]\]. SwedenTo describe the attempted suicide patients' perceptions of receiving specialized in-patient psychiatric care.18 patients. Interviewed after a suicide attempt near the time of discharge.Specialized psychiatric in-patient care.Qualitative interviews. Analysed for qualitative content using methods inspired by Burnard.• Perception of care and caregivers, a sense of security\ • Confirmation and lack of confirmation\ • Commitment and respectCardell & Pitula, 1999 \[[@CR40]\]. USATo explore patients' experience of constant observation to determine whether they derived any therapeutic benefits beyond the intended protective benefit.20 patients placed under constant observation for suicidality.Psychiatric hospital ward and a general medical centre with a psychiatric in-patient unit.Extensive in-depth interviews. Analysis of themes consistent with Hutchinson's recommended management of grounded theory data.• Constant observation not merely a protective intervention, but with therapeutic potential.\ • Need for engaged and supportive observersFletcher, 1999 \[[@CR43]\]. UKTo explore the perceptions of staff regarding the constant observation of a suicidal patient in mental health settings.24 patients at risk for suicide, constantly observed for at least 48 h.Acute psychiatric hospital.Ethnographic study with participant observation and semi-structured interviews. Data transcribed onto cards and subjected to content analysis.• Patients' negative feelings of being under constant observation related to staff actionsMcLaughlin, 1999 \[[@CR37]\]. UKTo explore psychiatric nurses' and patients' opinions regarding the care offered to suicidal patients and how the care for suicidal patients could be improved.17 patients admitted for depression, suicidal ideation or overt suicidal behaviour.Three psychiatric hospital wards.Observation and semi-structured interview. Data analysed using content analysis by Field and Morse.• The need to address difficulties\ • Help with problem-solvingTalseth et al., 1999 \[[@CR34]\]. NorwayTo illuminate the meaning of suicidal psychiatric in-patients' experiences of being cared for by mental health nurses.21 patients admitted with suicidal ideations or after a suicide attempt.Psychiatric emergency wards, psychiatric sub-emergency wards and one psycho-geriatric ward.Qualitative narrative interviews. A phenomenological--hermeneutic method inspired by Ricoeur used in the data analysis.• Being confirmed\ • Lack of confirmationAbbreviations: HCP* health care professional Study characteristics {#Sec9} --------------------- The review consisted of 20 articles published between 1999 and 2016. The patients' ages ranged from 16 to 63 years. The most frequently occurring diagnoses in the sample were affective disorders, of which major depression was the most prevalent, followed by schizophrenia spectrum diagnoses and personality disorders. Patients reported different experiences and needs depending on their symptoms and level of functioning; however, these parameters could not be analysed because of the presence of mixed samples. All patients had experienced suicidal crises, and the majority had attempted suicide prior to hospitalization. The studies originated primarily from Western mental health care settings, with the exception of studies by Sun et al. \[[@CR24], [@CR25]\], which were conducted in Taiwan. Themes representing patients' experiences regarding safety {#Sec10} ---------------------------------------------------------- The results of the 20 studies were synthesized and organized under analytical and descriptive themes (Table [3](#Tab3){ref-type="table"}). The results of this synthesis are described in greater detail in the following text.Table 3Analytical and descriptive themesAnalytical themeDescriptive themeConnectionMeeting someone who cares\ Receiving a confirmation of feelings\ Being acknowledged as a human beingProtectionBeing protected from death\ Receiving support from the observersControlGaining insight\ Coping with difficulties and symptoms\ Attaining discharge readiness ### Connection {#Sec11} The "Connection" theme illustrates how connections with health care professionals were vital for patient recovery and feelings of safety. A lack of connection was also experienced by the patients and had potentially fatal consequences. The sample of suicidal patients included in this review reported multiple and diverse causes of their suicidal crises \[[@CR26], [@CR27]\], but all patients experienced feelings of overwhelming suffering and increased vulnerability \[[@CR27]--[@CR31]\]. Patients experienced increased emotional sensitivity regarding how they were perceived and approached by health care professionals, and this sensitivity affected their perceptions of themselves, their recent suicide attempt, their therapeutic relationships \[[@CR26], [@CR32], [@CR33]\] and their feelings of safety in the hospital \[[@CR31], [@CR34], [@CR35]\]. Patients' connections with health care professionals enabled them to feel valued as human beings by meeting someone who cares; to feel understood by receiving a confirmation of feelings; and to feel respected and trusted by being acknowledged as a human being. Meeting someone who cares {#Sec12} ------------------------- Suicidal patients expressed feeling lonely, being alone with their despair, being separated from the external world and feeling a need to be connected with others \[[@CR28]--[@CR30], [@CR34]\]. A sense of being cared for could be achieved by meeting the patient's basic needs, such as bodily contact, fresh air, food, hygiene, sleep and rest \[[@CR34]\]. Patients also felt cared for when they engaged with health care professionals who were active and empathetic listeners, who spent time with them, and who showed interest in them as well as compassion for their situation \[[@CR26], [@CR28], [@CR34], [@CR36]--[@CR38]\]. These interpersonal interactions and the physical presence of the health care professionals helped patients feel that they were valuable \[[@CR30], [@CR34], [@CR39]\] and that they mattered and belonged in the world \[[@CR30], [@CR36]\]; these feelings reduced their suicidal ideations \[[@CR36]\] and made them feel safe in the psychiatric ward \[[@CR34], [@CR35]\]. Cutcliffe (\[[@CR36]\], s. 797) described this recovery process as a "re-connection with humanity" driven by connecting with and feeling cared for by nurses. Some patients felt that their health care providers had neither time nor compassion for them \[[@CR25], [@CR34], [@CR37]\], and these feelings had potentially fatal consequences. These patients experienced that their health care providers spent little time with them because the providers were busy performing other tasks or were interrupted during patient visits. Some patients experienced having no one to talk to, feeling ignored or feeling that they were being stored away as though they were an object \[[@CR34], [@CR39]\]. When met with a lack of interest and disengagement from health care professionals, patients lost confidence in their providers \[[@CR34]\], refrained from seeking help and felt unsafe in the ward \[[@CR35]\]. The experience of being isolated and alone on the ward raised feelings of hopelessness and worthlessness \[[@CR39]\]. Some patients felt redundant and started to plan ways to take their lives on the ward \[[@CR34]\]. Receiving a confirmation of feelings {#Sec13} ------------------------------------ Patients indicated that they needed someone who could listen to and understand their story and situation \[[@CR29], [@CR32], [@CR34]\] and provide confirmation of their feelings \[[@CR24], [@CR34], [@CR36]\]. They also expressed a need to be taken seriously in their suffering, to be allowed to express their feelings \[[@CR33]--[@CR35]\] and to be able to talk about their suicidality \[[@CR28], [@CR32]\]. The patients positively described their experiences being asked directly about their suicidal thoughts and plans, as they longed for opportunities to talk about difficult questions \[[@CR32]\]. Patients felt confirmed when they perceived that their mental health providers understood their situation and their need to step away from the demands of their lives \[[@CR33]\] and supported their need for hospitalization \[[@CR35]\]. The quality of the patient-physician relationship depended on patients' experience of this confirmation, as it enabled them to feel safe and understood \[[@CR34], [@CR36]\] and mitigated the despair and shame elicited by their suicide attempts \[[@CR30], [@CR33], [@CR34]\]. Patients experienced a lack of confirmation when health care professionals denied their feelings, neglected their illness, diverged from topics that the patients wanted to address, did not address difficult feelings \[[@CR33]--[@CR35]\], merely emphasized their positive resources \[[@CR32]\], or did not provide adequate or empathetic responses when they disclosed sensitive issues \[[@CR33]\]. Some patients reported that their health care professionals did not spend sufficient time with them to properly understand the reasons for their suicide attempts or that the professionals avoided talking about their suicide attempt \[[@CR34], [@CR39]\]. Other patients felt that their nurses were concerned only about their symptoms or the effects of their medications and thus did not allow them opportunities to share their thoughts and feelings \[[@CR34], [@CR39]\]. Patients perceiving these types of non-responsive attitudes with respect to sensitive or important topics experienced worsening feelings of shame and humiliation \[[@CR32], [@CR33]\] that exacerbated their suicidal ideations and, in some cases, resulted in subsequent suicide attempts \[[@CR32], [@CR35]\]. Being acknowledged as a human being {#Sec14} ----------------------------------- Patients stated that it was important for providers to meet them on equal ground in order for them to feel acknowledged as a human being \[[@CR26], [@CR33], [@CR34]\]. This meant being treated non-judgementally \[[@CR24], [@CR28], [@CR33], [@CR36]\] -- being empowered and understood as individuals rather than as objects, cases or diagnoses \[[@CR30], [@CR31], [@CR33]\]. When the patients felt that they were achnowledged as a human being, they were able to feel trusted, respected, and safe in the ward and were thus receptive to help \[[@CR26], [@CR30], [@CR31], [@CR35]\]. Through these feelings, patients regained their sense of human dignity and thereby felt that it was worthwhile to be alive \[[@CR26], [@CR33], [@CR36]\]. Not being seen as a human being was related to feelings of inequality \[[@CR32], [@CR34]\], e.g., patients whose providers overused medical jargon or limited their visits to discussions about medications and diagnoses \[[@CR31], [@CR34]\], as well as the feeling of being punished by health care professionals through the use of ward rules, verbal expressions or body language to exert their power \[[@CR33]\]. Not being seen as a human being was also related to feelings of disempowerment, e.g., being talked about when they were present \[[@CR32]\], not being informed about ward routines \[[@CR33]\] or who their primary nurse was \[[@CR25], [@CR37]\], not being informed about their own arrangements \[[@CR35]\], or experiencing that their opinions, information or histories were not considered important \[[@CR32], [@CR39]\]. Suicidal patients with borderline personality disorder experienced that they were able to recover by experiencing feelings of safety and trust during their encounters with health care professionals. However, when treated as inferior, the patients did not feel safe in the hospital \[[@CR31]\]. ### Protection {#Sec15} The "Protection" theme pertained to patients' experiences when under constant observation and their struggles to feel safe from themselves and their invasive suicidal impulses \[[@CR31], [@CR40]\]. Patients felt safe from themselves and their suicidal impulses and protected from death during constant observation. Receiving support from the observers was the most important aspect during constant observation, as patients lacking these relationships felt detached and objectified, and their anxiety and symptoms worsened \[[@CR26], [@CR40]\]. Being protected from death {#Sec16} -------------------------- During constant observation, some patients experienced a state of mind in which they continually searched for available means to attempt suicide. Some experienced feeling powerless against their suicidal thoughts, whereas others experienced command hallucinations related to suicide \[[@CR40]\]. Patients perceived constant observation as a means of altering their suicidal ideations and self-destructive behaviour. Patients considered this practice life-saving because of the presence of vigilant observers, the limited availability of objects to use for suicide attempts, the passage of time \[[@CR40]\] and the distraction and escape from the outside world \[[@CR24]\]. Patients struggled to feel safe from themselves and to assume responsibility for their own lives when they lacked protection during acute suicidal crises \[[@CR31], [@CR41]\]. Adequate protection was also related to their perceptions of the hospital as a safe place \[[@CR41], [@CR42]\]. Accordingly, patients who easily found ways to attempt suicide in the ward and those who did not receive safety searches or monitoring often felt unsafe in the hospital \[[@CR25]\]. However, one patient explained that not being able to end his life actually increased his suffering, as he believed that being able to end his suffering in the event that it became unbearable was a source of comfort that helped him cope with his situation \[[@CR29]\]. Patients experienced a lack of freedom and privacy under constant observation \[[@CR25], [@CR40], [@CR41]\], and most were happy when it was discontinued because of its invasiveness. Some patients even lied about their suicidality to discontinue their observation \[[@CR40]\]. Receiving support from observers {#Sec17} -------------------------------- Cardell and Pitula \[[@CR40]\] concluded that the relationship with care providers was at the heart of constant observation and highlighted the importance of patients having supportive observers as opposed to impersonal and detached observers. Patients experienced observer support as vital for decreasing their suicidality during constant observation \[[@CR36], [@CR40]\], as these relationships facilitated reduced suicidality. It was important for the observers to have an optimistic attitude, encourage problem-solving, enable patients to gain self-esteem, acknowledge patients as unique and meaningful human beings \[[@CR40]\], and try to understand patients by talking with them about their feelings \[[@CR43]\]. By interacting with supportive observers, the patients internalized what the observers projected and felt worthy as human beings and thus worthy of being alive \[[@CR40]\]. Some patients experienced a lack of acknowledgement and a lack of interpersonal engagement under constant observation, in which the observers appeared disinterested or distant or behaved as though their patients were a burden \[[@CR26], [@CR40]\]. When attempting to start a conversation, the observers either did not respond or displayed hostile facial expressions, which was perceived as a lack of empathy \[[@CR40]\]. Lees \[[@CR26]\] observed that having minimal interpersonal engagements limits the therapeutic potential of interventions, such as formal observation and medications. Patients deprived of interpersonal engagement felt objectified and separated from their health care professionals \[[@CR26]\] or that nobody was there for them or acknowledged their existence \[[@CR40]\]; these experiences exacerbated their feelings of anxiety and hopelessness and supported their perceptions that nobody cared about them and that their lives were not worth living \[[@CR40]\]. ### Control {#Sec18} The "Control" theme involved patients' need to re-establish a feeling of control over their lives \[[@CR27]\]. Suicidal patients experienced a sense of not being in control, a desire to regain control and a sense of losing control during suicidal crises \[[@CR26], [@CR27]\], which they often described as periods of overwhelming emotional suffering that left them unable to cope with life \[[@CR27]--[@CR31]\]. Patients whose health care professionals enabled them to gain insight and cope with difficulties and symptoms were able to regain control of themselves. This sense of control was important for attaining discharge readiness and feeling safe from themselves. Patients without this sense of control experienced increased suicidal thoughts. Gaining insight {#Sec19} --------------- Gaining insight into their illnesses enabled patients to regain control after their suicide attempt \[[@CR27], [@CR31], [@CR41]\]; patients who understood themselves were able to address the difficulties in their life without attempting suicide \[[@CR27]\] and also felt safer from themselves \[[@CR31]\], which helped them feel in control of their lives \[[@CR41]\]. Coping with difficulties and symptoms {#Sec20} ------------------------------------- Patients felt that a sense of control could be achieved by being able to manage difficulties and by learning new problem-solving and help-seeking skills, as well as by receiving adequate treatment for mental health problems and obtaining assistance for social and economic problems. Patients who were able to manage difficulties were able to visualize a way back to their lives \[[@CR27], [@CR36], [@CR37]\]. Variations in coping strategies related to different support and independence needs were described, as some patients expressed a need for others to "fix" their problems, some expressed a need for a break from any type of demand, and others emphasized a need to strengthen their self-efficacy to more effectively cope with their life situations \[[@CR27], [@CR33], [@CR41]\]. Some patients experienced that their problems were best addressed through one-on-one conversations with health care professionals \[[@CR35], [@CR37]\], whereas others preferred group support \[[@CR25], [@CR28], [@CR37]\], spiritual support \[[@CR25]\], or family or friend support \[[@CR25], [@CR38]\]. Patients needed health care professionals who could adapt to their needs and coping strategies \[[@CR38]\]. Attaining discharge readiness {#Sec21} ----------------------------- Patients expressed the expectation that their admission would result in a cure for or solution to their problems; this belief represented a major disconnect between patients' expectations and the treatment provided during short-term hospitalization \[[@CR41]\]. At discharge, some patients felt that their problems were unsolved \[[@CR37]\] and that they lacked the skills and tools for coping with their problems and their unchanged circumstances; this feeling resulted in increased distress and suicidal thoughts \[[@CR41], [@CR42]\]. At discharge, patients experienced unaddressed problems related to their suicidality \[[@CR32], [@CR37], [@CR41]\]. Thus, they did not feel prepared for discharge and feared that leaving the hospital would lead to subsequent suicide attempts \[[@CR41]\]. These patients experienced the feeling that the system was failing them and indicated that they did not know where to seek support in the event that formal mental health services could not help \[[@CR41]\]. Patients' sense of control was strengthened by having a post-discharge support plan and by being able to contact the ward after discharge if necessary \[[@CR27], [@CR35], [@CR41]\], as well as by being prepared for the upcoming change in their freedom by feeling empowered and supported prior to discharge \[[@CR31], [@CR41]\]. Thus, it was important for patients to be allowed to participate in decision making regarding their post-discharge support, as this reduced their fears and anxieties at discharge when being sent "back to the lion's den" (\[[@CR41]\], s. 24). Discussion {#Sec22} ========== This paper posed the following review question: "How can we describe suicidal patients' experiences regarding safety during psychiatric in-patient care?" Suicidal patients' experiences with safety during psychiatric in-patient care were described in 20 studies that addressed whether their needs were met during their hospitalization. This review argues that patients define safety in terms of "feeling safe" and that connection, protection and control play vital roles in their safety-related experiences. Fulfilment of these needs are experienced as essential for recovery from their suicidal crises, in addition to the ability to feel safe during their encounters with health care professionals and to feel safe from their suicidal impulses. When experiencing unmet needs, the patients not only felt unsafe but also exhibited increased suicidal thoughts and feelings. For some patients, these experiences were characterized as triggers for another suicide attempt. The patient experiences discussed in our review are related to the relational and emotional aspects of hospital care and are consistent with the findings of other studies regarding patient experiences \[[@CR10], [@CR13]\]. Our findings also resonate with those of psychiatric in-patient care studies, in which patients identified psychological safety as the most common safety issue \[[@CR44]\]. The connection and protection components discussed herein emphasize the importance of the therapeutic relationship in not only establishing feelings of safety but also optimizing patient outcomes, such as those related to increases or decreases in patient suicidality. The suicidal patients in this review addressed the vital importance of the therapeutic relationship in helping patients both feel safe and be safe. These findings are consistent with those of studies highlighting the therapeutic alliance in effective suicidal patient assessments and management \[[@CR45]--[@CR47]\] and studies identifying the staff--patient relationship as important to patients' feelings of safety \[[@CR44], [@CR48], [@CR49]\]. Poor staff-patient relationships were found to play key roles in preventable suicides and were attributed to poor communication and relationship quality \[[@CR50]\]. This review highlights the importance of addressing the control component to enable suicidal patients to feel and be safe after discharge from the hospital ward. The control component demonstrates the importance of supporting external and internal processes that help suicidal patients feel a sense of control and of understanding the individual from an ideographic point of view. Consistent the results of this review, Connell \[[@CR51]\] found that, for mental health patients, a sense of control was linked to feelings of safety. The level of desired dependence or independence varied according to each patient's current circumstances and differed over time. Undrill \[[@CR52]\] stated that psychiatric risks should be perceived as manifestations of suffering. Thus, maintaining high-quality core activities during care and acknowledging suicidal patients' suffering through trust and therapeutic closeness should be the primary methods of addressing patients' suicide risk and improving their safety. In accordance with Undrill's \[[@CR52]\] findings, our review indicates that ensuring patient safety entails addressing patients' therapeutic needs and psychological safety in addition to their physical safety. Although integrating relational and technical patient safety measures into psychiatric care is challenging \[[@CR53], [@CR54]\], safety is dependent on this integration. The link between feeling safe and being safe is vital for suicidal patients; suicidal patients' physical safety cannot be ensured if they do not feel safe. A system that is designed to physically prevent patients from committing suicide but that neglects their need for a connection with health care professionals may not be successful, as patients may exhibit increased suicidality despite the implementation of procedures to prevent this outcome. Furthermore, patients may not only feel unsafe, but they may also be unsafe because of an increased suicide risk imposed by the complex dynamics between emotionally vulnerable patients and their health care professionals. A broader perspective regarding patient safety that integrates therapeutic needs, psychological safety and physical safety is therefore needed. Limitations {#Sec23} ----------- There were a few limitations to this review. There is a risk of missed studies due to a lack of common nomenclature. To address this limitation, the search terms and strategy were designed to increase the sensitivity to relevant literature. Furthermore, the systematic search included only published peer-reviewed studies, resulting in the exclusion of possibly valuable grey literature and unpublished papers. Although there is a risk of reviewer bias, efforts were made to minimize this bias by applying systematic search methods and by following the PRISMA guidelines for systematic reviews. The review was limited to studies regarding psychiatric in-patient care. Studies examining the experiences of suicidal patients when receiving emergency care and outpatient treatment were excluded, as were studies regarding the experiences of patients without access to psychiatric care. These types of studies should be included in future reviews that aim to explore patient pathways and continuity of care, as poor continuity of psychiatric care has been associated with preventable suicides \[[@CR50]\]. Implications for research and practice {#Sec24} -------------------------------------- The literature included a diverse group of patients characterized by suicidal behaviour. These different patient groups may present distinct experiences, thus limiting the general understanding of suicidal patients as a group. To account for the diversity of patients in suicidal crises, more studies involving the elderly, youths, low-income countries and non-Western health care settings are necessary. There is also a need to explore the experiences of suicidal patients in different diagnostic groups, such as suicidal in-patients with/without psychotic symptoms and patients with/without chronic suicidality or borderline personality disorder. The similarities and differences between the experiences of suicidal patients and non-suicidal patients must be elucidated to identify the generic versus group-specific characteristics that determine patient safety in psychiatric care. Additionally, patients may have different needs during different stages of their suicidal crises. For example, Rise et al. \[[@CR55]\] observed that patients indicated different safety-related needs depending on their symptoms. However, this distinction was not addressed in the studies included in our review and represents a direction for further research. We recommend the following changes regarding in-patient care practices for suicidal patients based on the results of our systematic review:Patient experiences should be considered an integral part of suicidal patients' safety to guide clinical practice and the design of patient safety measures.Suicidal patients' need for connection with health care personnel indicate that the relational component of patient safety is considered the most vital aspect of care and should thus be integrated into measures such as constant observation, suicide risk assessments, clinical supervision, ward therapeutic environments and encounters with health care personnel groups.Suicidal patients' need for protection highlights the importance of constant observation in suicidal crises and the need for skilled professionals in close proximity to patients.Suicidal patients' need for control emphasizes the need for therapeutic interventions that increases the patient's insight and problem-solving skills as well as shared decision making regarding treatment plans, crisis plans, support systems and post-discharge follow-up activities. Conclusion {#Sec25} ========== Our review addresses the importance of having a broader view of safety for suicidal patients rather than merely understanding safety in technical terms. When considering suicidal patients' experiences, safety appears to be related to more than the absence of suicide risk and the need for physical protection. Safety for the suicidal patient is highly dependent on patients' perceptions of their connections with health care professionals, the fulfilment of their needs during care and their psychological safety. To be safe, patients must feel safe through their connections with health care professionals; they must be protected against their suicidal impulses and they must have a sense of control over their lives. These components should serve as the basis of future efforts designed to understand the ontology of safety for suicidal patients during in-patient psychiatric care. Additional files {#Sec26} ================ Additional file 1:Search strategy for PsychINFO. (DOCX 12 kb) Additional file 2:Table of themes and meaning units. (DOCX 24 kb) Additional file 3:Malterud's \[[@CR22]\] "Guidelines for authors and reviewers of qualitative studies -- an example of checks and scores". (DOCX 15 kb) Additional file 4:An overview of the included studies' scores: high, middle or low. (DOCX 16 kb) We thank all the psychologists, doctors, nurses and patients in the Division of Psychiatry at Stavanger University Hospital who provided their reflections during the manuscript preparation. Funding {#FPar1} ======= The Western Norway Regional Health Authority funded this project under grant agreement no. 911846. Availability of data and materials {#FPar2} ================================== All data generated or analysed during this study are included in this article. The additional files include supplementary information. The protocol and a matrix of the data analysis of the excluded studies are available from the corresponding author upon reasonable request. Authors' contributions {#FPar3} ====================== All three authors (SHB, KR and KAA) devised the search strategy and eligibility criteria. SHB conducted the database searches and the primary exclusion of studies. SHB and KR performed the eligibility assessments of the full-text articles. SHB completed the synthesis of the results, and KR and SHB conducted the quality appraisal. All three authors (SHB, KR and KAA) validated the results synthesis and quality appraisal. SHB drafted the manuscript, and KR and KAA provided critical revision of intellectual content. All authors approved the final manuscript. Authors' information {#FPar4} ==================== SHB is a Ph.D. scholar in risk management and societal safety at the University of Stavanger and a psychologist at Stavanger University Hospital. KR has a Ph.D. and is a researcher in Psychiatric Nursing at Stavanger University Hospital. KAA is a professor and head of the research group "Quality and safety in health care systems" in the Department of Health Studies at the University of Stavanger. Competing interests {#FPar5} =================== The authors declare that they have no competing interests. Consent for publication {#FPar6} ======================= Not applicable. Ethics approval and consent to participate {#FPar7} ========================================== Not applicable.	{ "pile_set_name": "PubMed Central" }
Introduction ============ Aneurysmal subarachnoid hemorrhage (SAH) is a devastating disease with substantial morbidity and mortality. Prognostic modeling is an important instrument to identify high-risk patients in both clinical practice and research settings. Recently, a prognostic model to predict 60-day case fatality after aneurysmal SAH was developed with data from the International Subarachnoid Aneurysm Trial (ISAT) \[[@B1]\]. Our aim was to externally validate this model in a retrospective cohort of consecutive SAH patients. Methods ======= We included consecutive aneurysmal SAH patients admitted to one university hospital between October 2007 and October 2011. Exclusion criteria were: age \<18 years, hospital admission \>28 days after SAH, nonaneurysmal SAH, explicit objection by the patient to view the medical data and missing data on 60-day case fatality. The model\'s predictors were age, maximum lumen size of the aneurysm, Fisher grade and World Federation of Neurological Surgeons (WFNS) grade. Two versions of the model were validated: one with WFNS grade scored on admission and the other with WFNS grade assessed at the time of treatment decision, as a proxy to WFNS grade at randomization used in the ISAT. The outcome was 60-day case fatality. Model performance was assessed by studying discrimination, expressed by the area under the receiver operating characteristic curve (AUC), and calibration. Results ======= A total of 307 patients were included in the validation cohort. The observed 60-day case fatality rate was 30.6%. Discrimination was good, and was considerably better for the model with WFNS grade at treatment decision (AUC = 0.89) compared with the model with WFNS grade on admission (AUC = 0.82). Calibration was poor, with mean predicted probabilities of 17.0% for the model with WFNS grade on admission and 17.7% for the model with WFNS grade at the time of treatment decision. Conclusion ========== Our results indicate that the ISAT prediction model is generalizable, since the model showed adequate performance in an independent, unselected cohort of aneurysmal SAH patients. The model discriminated well between patients who died and those who survived the first 60 days after SAH. Additionally, use of WFNS grade at the time of treatment decision of the ruptured aneurysm improved model performance. However, since predicted probabilities were lower than observed probabilities, the ISAT prediction model needs to be adapted before use in clinical practice.	{ "pile_set_name": "PubMed Central" }
Introduction ============ The atherogenicity of type 1 diabetes has been increasingly recognized \[[@B1]\]. Patients with diabetes show a 2- to 10-fold risk for developing atherosclerotic lesions compared with the normal population \[[@B2],[@B3]\]. Even if these complications become manifest in the adult diabetic patient, the process of vascular changes starts much earlier \[[@B4],[@B5]\]. The most significant changes in early subclinical period of atherosclerotic disease are endothelial dysfunction and increase in intima-media thickness observed in all arterial beds \[[@B6]\]. Common carotid artery intima-media thickness (CIMT), measured by high-resolution B-mode ultrasonography, is a noninvasive marker of subclinical atherosclerosis \[[@B7]-[@B9]\]. Normative values for CIMT are available for subjects aged 10 to 20 years \[[@B10],[@B11]\]. Mechanisms involved in the increased oxidative stress in diabetes include not only oxygen free radical generation due to nonenzymatic glycosylation (glycation), autooxidation of glycation products, but also changes in the tissue content and activity of antioxidant defense systems. Increased levels of the products of oxidative damage to lipids have been detected in serum of diabetic patients, and their presence correlates with the development of complications \[[@B12]-[@B18]\]. Oxidative stress plays a pivotal role in the development of diabetes microvascular and cardiovascular complications \[[@B19]\]. In the diabetic macrovascular and in the heart, this appears to be a consequence of increased oxidation of fatty acids, resulting in part from pathway-specific insulin resistance \[[@B19]\]. Nitric oxide has cellular antioxidant and pro-oxidant actions \[[@B20]\]. The endothelial nitric oxide (NO) system plays a pivotal role in vascular physiology and pathology. NO is a potent vasodilator agent with anti-hypertensive, anti-thrombotic, anti-atherogenic, and anti-smooth muscle proliferative properties \[[@B21]\]. However, high amounts of NO produced by inducible NO synthase (iNOS) and/or peroxynitrite (ONOO-), a reactive intermediate of NO with superoxide anion are involved in pro-inflammatory reactions and tissue damage as well \[[@B22]\]. Total Antioxidant Capacity (TAC) is capable of serving as a parameter to monitor diabetes in patients with type 1 DM \[[@B23]\]. A depletion of the total antioxidant capacity is associated with a higher incidence of diabetic complications \[[@B24]\]. The relation between oxidative stress markers and carotid intima media thickness in type 1 diabetic children and adolescents was not extensively studied. Hence, this study aimed at assessment of carotid intima media thickness in children and adolescents with type1 diabetes mellitus in relation to plasma nitric oxide and plasma total antioxidant capacity levels and with diabetes duration, glycemic control and microvascular complications. Patients and methods ==================== This case control study was conducted at the Pediatric Diabetes Clinic, Children's Hospital, Ain Shams University, Cairo, Egypt in the period from April 2011 to February 2012.It included 50 patients with type 1 diabetes mellitus(DM) regularly attending the clinic. They were 37 females (74%) and 13 males (36%). Their ages ranged from 6--16 years with a mean age of 9.7 ± 3.4 years. Their duration of illness ranged from 1--13 years with mean diabetes duration of 4.5 ± 3.5 years. Patients were further subdivided into two groups according to the duration of diabetes: (Group I) included 22 children and adolescents with diabetes duration of 5 years or more, (Group II) included 28 children and adolescents with diabetes duration of less than 5 years. Fifty age and sex matched healthy individuals were included as a control group. They were 14 males and 36 females. Their age ranged from 2--14 years with mean age of 9.8 ± 3.14 years. The study was approved by the Ethical Committee of Ain Shams University Faculty of Medicine. Inclusion criteria ------------------ Patients were included in the study only if they have type 1 diabetes mellitus on regular insulin therapy regularly visiting the Clinic. Exclusion criteria ------------------ Included Type 2 diabetes mellitus, hypertension, patients with malignancy, connective tissue diseases, liver dysfunction, renal dysfunction (serum creatinine \> 1.2 mg/dl), congenital or acquired cardiovascular disorders, administration of drugs other than insulin (such as oral hypoglycemics, antihypertensives, antiplatelets or lipid lowering medications, aspirin, or vitamin supplements) at the time of the study and none of them was cigarette smoker. All patients were subjected to detailed history taking, thorough clinical examination, measurement of Glycosylated Hb (HbA1c) by HPLC (high performance liquid Chromatography). Patients were considered under optimal glycemic control when their HbA1c was \< 7.5% \[[@B25]\]. Microalbuminuria was assayed using SERA-PAK immuno-microalbumin Kit (Bayer Corporation, Benedict Ave, Tarry town, NY, USA). Persistent microalbuminuria was defined when two of three samples showed urinary albumin excretion rate of 30--300 μg/mg creatinine \[[@B26]\] fasting serum triglycerides, serum cholesterol using Synchron CX7 (Brea, California, USA) \[[@B27]\]. Serum total antioxidant capacity by colorimetric method \[[@B28]\], serum NO by colorimetric method \[[@B29]\] and Carotid intimal --media thickness (CIMT) assessment using B mode ultrasonography \[[@B30]\]. Measurment of serum Total Antioxidant Capacity (TAC) by colorimetric method. The determination of antioxidative capacity was performed by the reaction of antioxidants in the sample with a defined amount of exogenously provide hydrogen peroxide (H2O2) the antioxidants in the sample eliminate a certain amount of the provided hydrogen peroxide. The residual H2O2 was determined clorometrically by an enzymatic reaction which Involves the conversion of 3,5,dichloro \_2\_ hydroxyl benzensulphonate to a colored product \[[@B28]\]. Measurement of serum Nitric oxide (NO) was done using colorimetric method. This assay determines nitric oxide concentrations based on the enzymatic conversion of nitrate to nitrite by nitrate reductase. The reaction is followed by colorimetric detection of nitrite as an azo dye product of the Griess Reaction. The Griess Reaction is based on the two-step diazotization reaction in which acidified NO2^-^ produces a nitrosating agent, which reacts with sulfanilic acid to produce the diazonium ion. This ion is then coupled to N-(1-naphthyl) ethylenediamine to form the chromophoric azo-dervative which absorbs light at 540--570 nm \[[@B29]\]. Statistical analysis -------------------- Demographic and clinical data are presented as means ± SD or proportions. Differences in continuous variables between males and females were tested with the Student t test for normally distributed data and the Mann--Whitney U test for non-normally distributed data. The χ2 test for contingency tables with different degrees of freedom was obtained to detect associations between categorical independent variables. Adjustment for multiple confounding was done using linear regression analysis with a manual backward procedure. Multivariate analyses were preceded by estimation of the correlation between potential confounders. A significance level of \< 0.05 was used. All statistical analysis was done using the SPSS software package for Windows, version 15.0 (SPSS, Chicago, IL). Results ======= -- Patients mean systolic blood pressure (SBP) was103.500 ± 18.077 mmHg and mean diastolic blood pressure (DBP) was 68 ± 10.302 mmHg. Three (6%) of our patients had SBP and DBP \> 95% percentile while 97% were within the normal ranges. -- Diabetic patients showed significant increase in nitric oxide and decrease in total antioxidant capacity level than controls (P value \< 0.001)\] (Table [1](#T1){ref-type="table"}), Figure [1](#F1){ref-type="fig"}. ![Comparison between patients and controls regarding values of nitric oxide and total antioxidant capacity.](2251-6581-12-50-1){#F1} -- NO level was higher in patients with suboptimal glycemic control than those with optimal glycemic control \[(23.38 ± 6.7 μmol/L) vs (16.33 ± 5.6 μmol/L)\] (P value = 0.017) and in patients with nephropathy than patients without nephropathy \[(22.00 ± 4.14 μmol/L) vs (16.67 ± 6.5 μmol/L)\] (P value = 0.033). -- No statistically significant difference was found in TAC level between diabetic patients with and without nephropathy or between patients with suboptimal and optimal glycemic control (p value \>0.05) -- Serum total cholesterol and triglycerides were significantly higher in studied patients than control group, although within normal range. -- Carotid intima media thickness was significantly increased in diabetic patients compared to normal controls \[(mean 0.46 ± 0.04 mm) vs (mean 0.39 ±0.02 mm)\] (P \< 0.001) (Table [1](#T1){ref-type="table"}), Figure [1](#F1){ref-type="fig"}. -- Carotid intima media thickness was significantly increased in diabetic patients with suboptimal than those with optimal glycemic control \[(0.46 ± 0.04 cm) vs (0.40 ± 0.02 cm)\] (P value 0.04), in patients with diabetes duration \> 5 years than those with diabetes duration \< 5 years \[(0.50 ± 0.03 mm) vs (0.43 ± 0.02 mm)\] (P value \<0.001)\] (Table [2](#T2){ref-type="table"}) and in diabetic patients with than patients without nephropathy \[ (0.49 ± 0.02) vs (0.40 ± 0.04)\] (P value = 0.009). -- A statistically significant positive correlation was found between carotid intima media thickness and age, duration of diabetes mellitus, systolic and diastolic blood pressure and nitric oxide (P value \< 0.001 and \<0.001, 0.002, 0.007 and 0.01 respectively) and a significant negative correlation was found between carotid intima media thickness and total antioxidant capacity (P value = 0.02) (Table [3](#T3){ref-type="table"}). No significant correlation was found between carotid intima media thickness and mean HbA1c, cholesterol or triglycerides (P value \>0.05). ###### Comparison between diabetic patients and control group regarding studied parameters Groups t-test ----------------------- ---------------- ---------------- ---------- ----------- Age (years) 9.78 ± 3.45 9.86 ± 3.14 -0.12 0.90 BMI(kg/m2) 20.120 ± 4.155 18.740 ± 3.87 t = 1.72 0.09 Cholesterol (mg/dl) 147.35 ± 38.92 120.68 ± 35.26 t = 3.5 0.0005\* Triglycerides (mg/dl) 80.86 ± 25.10 62.40 ± 20.50 t = 4.03 0.0001\* NO (μmol/L) 17.07 ± 6.36 12.57 ± 4.74 3.98 \<0.001\* TAC (mmol/L) 0.41 ± 0.29 0.87 ± 0.23 -8.87 \<0.001\* CIMT (mm) 0.47 ± 0.04 0.40 ± 0.02 9.66 \<0.001\* Gender 37/13 36/14 0.437\# 0.509 NO nitric oxide, TAC Total antioxidant capacity, CIMT carotid intima media thickness; \#= Chi-Square. \Indicate statistical significance. ###### Comparison between patients with diabetes duration \<5 years and \>5 years regarding studied parameters Patients with diabetes duration ≤ 5 years(n = 28)* Patients with diabetes duration \> 5 years(n = 22) T-test ----------------------- ------------------------------------------------------- -------------------------------------------------------- ------------ ----------- Age (years) 7.64 ± 2.80 12.50 ± 1.946 -6.92 \<0.001\* Mean HbA~1~c% 8.47 ± 1.98 9.73 ± 2.07 -2.19 0.03\* Cholesterol (mg/dl) 134.36 ± 38.34 162.50 ± 35.23 -1.94 0.07 Triglycerides (mg/dl) 79.13 ± 24.74 81.86 ± 25.54 0.38 0.7 NO (μmol/L) 16.27 ± 6.70 18.09 ± 5.89 -1.01 0.32 TAC (mmol/L) 0.45 ± 0.31 0.36 ± 0.25 1.18 0.25 CIMT (mm) 0.44 ± 0.03 0.50 ± 0.03 -7.09 \<0.001\* NO nitric oxide, TAC Total antioxidant capacity, CIMT carotid intima media thickness. \Indicate statistical significance. ###### Correlation between CIMT, serum NO and serum TAC in diabetic patients CIMT (mm)* NO (μmol/L) ------------------------------ ----------- --------------- ----------------- Age (years) r 0.741 P-value \<0.001\* Duration of diabetes (years) r 0.656 P-value \<0.001\* SBP (mmHg) r 0.126 P-value 0.002\* DBP (mmHg) r 0.379 P-value 0.007\* NO (μmol/L) r 0.40 P-value 0.01\* TAC(mmol/L) r -0.34 -0.53 P-value 0.02\* \<0.001\* CIMT carotid intima media thickness, SBP systolic blood pressure, DBP diastolic blood pressure, NO nitric oxide, TAC Total antioxidant capacity. \*Indicate statistical significance. Discussion ========== The present study showed a significant increase in serum nitric oxide (NO) level in diabetic patients than controls which came in agreement with many studies \[[@B31]-[@B33]\]. Pitocco et al. \[[@B34]\] suggested a reduced asymmetric dimethyl arginine (ADMA) inhibition of NOS as possible mechanism involved in the pathogenesis of oxidative stress in female subjects with a short duration and uncomplicated type 1 diabetes. Serum NO level was significantly increased in diabetic patients with than those without nephropathy which came in accordance to previous studies \[[@B35],[@B36]\]. Prabhakar, \[[@B37]\] reported that the enhanced NO production may contribute to hyperfiltration and microalbuminuia at early diabetic nephropathy. NO level was significantly higher in patients with suboptimal than optimal glycemic control which came in agreement with others \[[@B38]\]. A significant decrease in total antioxidant capacity level was found in patients than controls which came in agreement with several studies \[[@B39]-[@B42]\]. Lack of difference in TAC level between our studied patients with and without nephropathy is concordant with El-desoky et al. \[[@B43]\] who found that plasma TAC in patients of diabetes mellitus type I with nephropathy was not significantly decreased as compared with diabetes mellitus type 1 with no nephropathy and they concluded that TAC cannot be used to differentiate between diabetes mellitus type 1 with and without nephropathy. Serum total cholesterol and triglycerides were significantly higher in studied diabetic patients compared to control group although within the normal range. This agreed with Margeisdottir et al. \[[@B44]\] who attributed those findings to the young age of the patients. Diabetic patients showed a significant increase in CIMT mean values compared to controls. Those results came in accordance to several studies \[[@B45]-[@B47]\]. A significant increase in CIMT in patients with nephropathy compared to patients without nephropathy came in concordance with Gul et al. \[[@B47]\] suggesting that diabetic microangiopathy is related with macroangiopathy and in patients with suboptimal glycemic control compared to patients with optimal glycemic control which came in concordance with Abdelghaffar et al. \[[@B46]\]. CIMT was directly correlated with age in studied diabetic patients which agreed with many studies \[[@B46]-[@B48]\]. A strong direct correlation was found between mean CIMT and patients systolic and diastolic blood pressure which agreed with several studies \[[@B46],[@B48],[@B49]\]. The relationship between increased CIMT and blood pressure suggests that smooth muscle proliferation plays a role in the early diffuse thickening of the arterial wall \[[@B49]\]. No significant correlation was found between mean CIMT and metabolic control parameter HbA1c, this agreed with many studies \[[@B47],[@B50]\]. Data from the literature indicated that, in contrast to the functional impairment of the endothelium, structural changes are not correlated to single parameter such as the HbA1c at a young age \[[@B50]\]. A statistically significant positive correlation was found between carotid intima media thickness and nitric oxide level which agreed with Dursun et al. \[[@B51]\] who explored the relation between CIMT and oxidative stress markers in type 2 diabetic patients on maintenance hemodialysis. Zineh et al. \[[@B52]\] reported the association between NOS3 polymorphisms and arterial stiffness in children with type 1 diabetes. Insulin acts by modulating the release of vasodilator substances, such as nitric oxide and prostaglandins, from vascular endothelium, by both stimulating and inhibiting the sympathetic nervous system and by protecting smooth muscle cells in blood vessel from apoptosis induced by oxidative stress \[[@B53]\]. Thus the vasodilatory and antioxidant effects of insulin are depressed in case of insulin deficiency (i.e. type 1 diabetes) \[[@B54]\]. A negative correlation was found between CIMT and total antioxidant capacity level. Lower serum TAC levels were observed in atherosclerotic coronary artery disease patients with an inverse association with number of damaged coronary vessels \[[@B55]\]. Other research reported increased DNA damage in the nucleus of coronary cells and decreased plasma TAC level in coronary artery disease patients \[[@B56]\]. Conclusion ========== The significant elevation in nitric oxide and reduction in total antioxidant capacity in children and adolescents with type 1 diabetes mellitus together with their correlation with carotid intima media thickness may reflect the role of oxidative stress in the development of atherosclerosis in young type 1 diabetic subject. Further studies to measure other antioxidant markers like glutathione or antioxidants or markers of oxidative stress like malondialdehyde in relation to carotid intima media thickness are warranted. Abbreviations ============= CIMT: Carotid intima media thickness; DM: Diabetes mellitus; HbA1c: Glycosylated hemoglobin; NO: Nitric oxide; TAC: Total antioxidant capacity; SBP: Systolic blood pressure; DBP: Diastolic blood pressure; ELISA: Enzyme linked immune sorbent assay. Competing interests =================== The authors declare that they have no competing interests. Authors' contributions ====================== MHE supervised the work and reviewed the manuscript. RMM designed the study, analyzed the data and drafted the manuscript. OIY carried out the carotid intima media thickness assessment, contributed in data analysis and drafting the manuscript MAS carried out the laboratory studies. NAK collected the data. All authors approved the manuscript. Acknowledgements ================ We are grateful to the staff of the Diabetes Clinic and echocardiography unit, Children\'s Hospital, Ain Shams University, Cairo Egypt.	{ "pile_set_name": "PubMed Central" }
Taking into account the SARS-coronavirus 2 (SARS-CoV-2 also know as COVID-19) spreading and invasiveness in this critic period and the lack of therapeutic drugs and antiviral vaccines, it is necessary for all of us researchers to discuss about new therapeutic devices and quickly define if agents, other than that conventionally used, could offer true clinical benefits. The interest in statins as an inexpensive, readily available and easily tolerated drugs useful to reduce the morbidity and mortality caused by influenza and viral diseases, has been provided from numerous scientific publications over the past 15 years \[[@bib0005],[@bib0010]\] but for some reason it has never gone deep. Statins, inhibitors of the HMG-CoA reductase an enzyme that limits the pathway of mevalonate and cholesterol, have pleiotropic effects, as we described in several publications and especially in a Thematic Issue \"Statin: new life for an old drug\" published in 2014 on Pharmacol Res in collaboration with the discoverer of these compounds, prof. A. Endo \[[@bib0015]\]. These are substances with anti-inflammatory and immuno-modulatory effects, which could certainly benefit patients with influenza and viral pathologies. Experimental evidence suggested the effectiveness of statins in viral infections and their potential mechanisms of action ([Table 1](#tbl0005){ref-type="table"} ) \[[@bib0010]\]. In some viruses, statin-induced reduction of cholesterol in the plasma membrane results in lower viral titres and failure to internalize the virus. Such data suggest that in some viruses, cholesterol affects the early stages of infection. In particular, in the initial stage of infection, viruses bind specific receptors concentrated within lipid rafts, areas of plasma membrane rich in cholesterol. Statins, by reducing the percentage of cholesterol present in the membrane, alter the assembly of the receptors and dramatically reduce the possibility of adhesion of the viral agent to the host. Lipid rafts are also involved in the viral replication phases, as they constitute packets of vesicles capable of concentrating virus replication factors. These processes are also highly destabilized by statins \[[@bib0010]\].Table 1Potential mechanisms triggered by statins as anti-virus compounds (see reference \[[@bib0010]\] for detailed studies).Table 1StatinsExperimental ModelsMechanism of action and observed effectsAtorvastatin, RusovastatinMadin-Darby Canine Kidney (MDCK) cells infected with influenza A (strains- H3N2 and H1N1) viruses-down-regulation of Rho/Rho kinase pathway;\ -inhibition of the virus proliferationAtorvastatinC57BL/6 mice infected with influenza A virus (strains H3N2 and H1N1)-reduced lung virus titers;\ -reduced mortality rates in infected miceAtorvastatin, SimvastatinCrandell Feline Kidney (CrFK) cells infected with influenza A virus (strain H1N1)-reduced TNF-α and IL-6 in supernatants of infected cellsSimvastatinPrimary normal human bronchial epithelial (NHBE) cells stimulated with synthetic dsRNA viral analogue;\ Human lung epithelial cell line A549 stimulated with synthetic dsRNA viral analogue-suppressed dsRNA-induced STAT3 activation;\ -inhibition of RANTES expressionSimvastatinsynthetic dsRNA-induced pneumonia in BALB/c mice-reduced STAT3 activation;\ -reduced RANTES release;\ -reduced neutrophilia in the lungsAtorvastatinMadin-Darby Canine Kidney (MDCK) cells infected with influenza A virus (strain H1N1)-inhibition of the early stage of virus multiplication;\ -reduced virus infectivity by decreasing the virus titer from infected cells;\ -increased cell viability of virus infected cellsSimvastatinMadin Darby canine kidney (MDCK) cells infected with Influenza A virus (strain H1N1)-inhibition of the early stage of virus multiplication;\ -decreased virus-induced cytotoxicity in infected cells;\ 2-fold decrease of secreted pro-inflammatory cytokines (TNF-α, IL-6, INF-γ);\ decrease of virus replication through inhibition of Rab/RhoA GTPase activity and LC3 membrane localizationLovastatinHEp-2 cells infected with Respiratory Syncytial Virus (RSV, strain A2)\ C57BL/6 and BALB/c mice infected with RSV-reduced RSV replication in HEp-2 cells;\ -reduced cell-to-cell fusion in cell culture (through inhibition of RhoA);\ -decreased RSV virus replication in mice;\ -decreased virus-induced weight loss and illness in miceSimvastatinBALB/c mice infected with Influenza A virus (strain H5N1)-decrease of secreted pro-inflammatory cytokines and chemokines (IFNγ, IL-10, TNFα);SimvastatinLeukocytes from HIV-infected patients-depletion of cell membrane cholesterol and dissociation of lipid-rafts;\ -decrease of subpopulations and macrophages acting as Antigen Presenting Cells\ -decrease of highly efficient HIV-1 infection transferred by macrophages to CD4^+^T cellsMevastatin, SimvastatinHuman hepatocarcinoma (Huh7) cells containing subgenomic HCV replicons-dose-dependent inhibition of HCV replicon replication (measured as luciferase signal);\ -additive antiviral activity in short-term used in combination with IFNα or HCV nonstructural (NS)5B polymerase or NS3 protease inhibitors Many viral pathogens encode substrates for mammalian prenylation pathway (mevalonate pathway). It is hypothesized that the pre-dilation of viral proteins is beneficial for the propagation of the virus and that in some viral infections the statins, by inhibiting HMG-CoAR and reducing the expression of prenylated proteins, carry out an evident antiviral activity \[[@bib0010]\]. A recent study has shown that the influenza A virus induces the formation of numerous lipid droplets in infected cells. In the experimental model used, Madin-Darby dog kidney cells were treated with atorvastatin (ATV, 5microM) before inducing the infection. Pretreatment reduced the infectious capacity and, consequently, the production of new influenza A viruses, with a percentage higher than 95 % \[[@bib0020]\]. Studies performed in other viral strains have shown that different statins perform antiviral activity with variable efficacy, probably due to the specific pharmacological and biochemical properties of the statins used in the analysis and, therefore, to the different antiviral power observed \[[@bib0025]\]. Furthermore besides statins there are also a series of mevalonate pathway inhibitors that could then be used in the next future with a more specific target and effectiveness. It is certainly worth studying in detail the mechanism by which statins inhibit virus replication. Although unsuitable for prophylaxis in all patients \[[@bib0030]\], to study their efficacy as antivirals in vitro could provide a valuable tool to identify their mechanism of action in preventing virus reproduction and, therefore, pandemic spread. These hypotheses are strongly supported by substantial meta-analyzes and retrospective studies showing reduced hospitalization rates and mortality in subjects already users of statins affected by flu diseases \[[@bib0035]\]. Of course, studies are needed to evaluate, in cellular and animal models, the direct effect of statins on different viral strains, and more than anything on coronavirus, obviously followed by clinical studies, largely dictated by the current clinical emergency. The analysis of the interaction with statins of viruses at very high costs, health and social, such as SARS-CoV-2, represents a valuable possibility. There is also to consider that the effects of statins on the family members of Coronaviridiae has never been investigated. Positive results from these studies could provide the Health Services with an effective, safe and low-cost tool to improve the prognosis of this viral pathology, prevent pandemics and improve antiviral prophylaxis in "frail patients". In conclusion with this letter, we want draw the attention from all researchers and physicians to this therapeutic perspective not yet and fully investigated. Declaration of Competing Interest ================================= The authors declare no conflicts of interest.	{ "pile_set_name": "PubMed Central" }
Gastrointestinal (GI) disorders are among the most frequent complaints in patients with HIV disease. Before combination antiretroviral therapy, it was suggested that 50-93% of all patients with HIV disease had marked GI symptoms during the course of their illness \[[@ref1],[@ref2]\]. Inflammatory changes, direct mucosal invasion of the GI tract by HIV, opportunistic infections and neoplasms are well recognized causative factors \[[@ref3]\]. GI manifestations of HIV disease may be debilitating and include anorexia, weight loss, diarrhea, nausea, vomiting, dysphagia and odynophagia, abdominal pain, anorectal disease, GI bleeding, jaundice and hepatomegaly, and GI tumors \[[@ref4]\]. Progressive immunosuppression is associated with increasing prevalence of GI symptoms \[[@ref5],[@ref6]\] and thus the evaluation of specific GI complaints must be based on an assessment of the degree of immunosuppression. However, antiretroviral therapy may change the nature of HIV disease and can lead to improvement in gastrointestinal symptoms for patients with advanced immunosuppression \[[@ref5],[@ref6]\]. Appropriate symptom management requires improved symptom recognition and should be an essential component of patient care at all stages of any disease worldwide \[[@ref7]\]. However, limited data exist regarding the incidence of various GI manifestations in HIV-infected patients in certain areas of the world including the Caribbean. The paper by Thompson et al in this issue of Annals of Gastroenterology \[[@ref8]\] investigated the prevalence of common GI symptoms in the ambulatory Jamaican HIV infected population compared to those seen in a healthy population of blood donors not known to have HIV. This study revealed that in HIV-infected patients the presence of symptoms was directly related to CD4 count similarly to previous studies \[[@ref9]\] and that GI symptoms were common in ambulatory HIV-infected patients and controls. Heartburn, belching and nausea were more common in the control patients whereas early satiety was the only symptom significantly more common in the HIV-infected patients. However, pain on swallowing, and vomiting were significantly more common in patients with CD4 \<350 cells/μL than in controls. The authors conclude that the relatively low incidence of significant GI symptoms in HIV-infected patients may reflect the changing manifestations of HIV infection due to the success of antiretroviral therapy. However, although the incidence of certain GI symptoms such as dysphagia and weight loss in the Jamaican population was similar to the incidence of GI symptoms reported in previous studies \[[@ref10]\], it was surprising that 35% of control patients had weight loss. Indeed, this present study has several limitations. Firstly, the weight loss as reported in this study was subjective and thus the estimated incidence may be inaccurate. Similarly, the subjective reporting of other GI symptoms such as belching and nausea may also explain why these symptoms were more common in the control patients. Secondly, the used questionnaire did not seek to determine if weight loss was intentional. Thirdly, the number of patients and controls was relatively small. Finally, the sample of HIV-infected patients represented only the ambulatory patients and was not entirely representative of the general population of HIV-infected patients. It would also be of interest to look at the possible association of specific antiretrovirals with the incidence of GI symptoms, similarly to previous studies \[[@ref5],[@ref10]\]. In conclusion, improved symptom recognition of GI symptoms in HIV patients should be an essential component of patient care worldwide and an astute clinical acumen is often required to avoid subjective interpretation of certain GI symptoms. Conflict of Interest: None David Geffen School of Medicine, University of California, Los Angeles, U.S.A.	{ "pile_set_name": "PubMed Central" }
Background {#Sec1} ========== Introduction {#Sec2} ------------ With the Sustainable Development Goals (SDGs), the global community sees greater than ever consensus about the crucial role that the health workforce plays in realizing goals to achieve universal health coverage \[[@CR1], [@CR2]\]. Qualified health workers, trained to work in effective teams within and across professional cadres to address the biomedical and social determinants of health, are critical to achieving health goals. While great progress has been made in maternal and child health and HIV/AIDS, continued shortages of adequately trained health workers raise the question of whether health professional education (HPE) systems are producing the health workforce needed to meet outstanding and emerging global health challenges. In 2010, the independent Commission on Education of Health Professionals for the 21st Century (henceforth "the independent Commission") released a comprehensive report on HPE. The Commission's report spurred a growing movement for HPE reform in many regions, including Africa, Asia, and the Americas \[[@CR3]\]. The report calls for institutional and instructional HPE reforms to target a variety of systemic problems \[[@CR3]\], including:Outdated, fragmented, and content-oriented curricula that produce graduates with narrow contextual understanding and insufficient knowledge, skills, and competencies to understand social and other determinants of health and diseasePoor teamwork and inadequate collaboration within and across health professional cadresEpisodic encounters with patient illnesses rather than continuous and holistic health careA predominant hospital orientation at the expense of primary careAn imbalance between health workforces and health needsWeak leadership in improving health system performance The independent Commission's report also mentions persistent gender stratification of professional status as a systemic deficiency and puts forth as one of its proposed reforms the need to pay particular attention to ensuring equal opportunities through more flexible working arrangements and career paths that accommodate temporary breaks, actively addressing gender discrimination and subordination \[[@CR3]\]. Yet gender-related deficiencies in HPE are not limited to gender stratification and include other forms of often invisible gender discrimination and inequalities as described in Table [1](#Tab1){ref-type="table"}. This paper identifies how particular gender-related deficiencies in HPE can be addressed as part of both instructional and institutional governance reforms. The definitions of the terms used in the paper are in Table [2](#Tab2){ref-type="table"}.Table 1Gender discrimination in health workforce systems \[[@CR4]\]Form of discriminationDescriptionSexual harassmentUnwanted, unwelcome, or offensive conduct that changes the terms and conditions of school or work, where either a person's rejection of, or submission to, such conduct is used explicitly or implicitly as a basis for a decision that affects that person's education or career (quid pro quo), or conduct that creates an intimidating, hostile, or humiliating work environment for the recipient (hostile environment). A form of violence as well as discrimination.Pregnancy discriminationExclusions, restrictions, or distinctions made on the basis of pregnancy, childbirth, or related conditions, such as unwillingness to hire, promote, or retain female students or workers who may get pregnant and leave school or the workforce or who require maternity leave and benefits. This type of discrimination is related to:Family responsibility discriminationExclusions, restrictions, or distinctions against individuals (such as pregnant women, mothers and fathers of young children, parents of disabled children, and individuals who care for their aging parents or sick spouses/partners) based on their responsibilities to care for family members.\ Pregnancy and family responsibility discrimination are related forms that target a broad range of reproductive functions and may be viewed together as caregiver (or reproductive role) discrimination. These related forms of discrimination usually target women of childbearing age who are not able to equally access opportunities for education, hiring, or promotion.Occupational gender segregationConcentration of men and women in different jobs (horizontal) or at different levels (vertical) in a job hierarchy. What has been called gender stratification \[[@CR3]\] may refer to vertical or horizontal segregation or both.Gender stereotypingOvergeneralized characterizations of persons in a particular group, occurring when the personal characteristics deemed necessary for a job are inconsistent with characteristics generally associated with a particular sex.Table 2Key definitionsBias \[[@CR43]\]\ An inclination or prejudice for or against one person or group, especially in a way considered to be unfair, that often results in discrimination.Discrimination in employment and occupation \[[@CR33]\]\ Practices that place individuals in a subordinate or disadvantaged position in school, the workplace, or the labor market because of characteristics (e.g., race, sex, age, religion, or other attribute) that bear no relation to the person's competencies or the inherent requirements of the job. Discrimination occurs when bias is enacted.Equal opportunity and non-discrimination \[[@CR44]\]\ The offering of employment, pay, or promotion to all, without discrimination as to sex, race, color, disability, and so forth.Gender-blind \[[@CR23]\]\ Gender-blind policies and programs that are designed without prior analysis of culturally defined economic, social, and political roles, responsibilities, rights, entitlements, obligations, and power relations associated with being female and male and the dynamics between and among men and women, boys and girls. Gender-blind policies and programs ignore gender considerations altogether.Gender discrimination \[[@CR45]\]\ Any distinction, exclusion, or restriction made on the basis of socially constructed gender roles and norms that prevents a person from enjoying full human rights.Gender equality (in the health workforce) \[[@CR46]\]\ A condition where women and men can enter the health occupation of their choice, develop the requisite skills and knowledge, be fairly paid, enjoy fair and safe working conditions, and advance in a career, without reference to gender. It implies that health professional education schools and workplaces are structured to integrate family and work to reflect the value of caregiving for women and men.Gender inequality\ Denotes the gender-based differences that result from gender discrimination and serve to diminish or enhance individuals' opportunities, access, power, conditions, and/or income.Gender transformative \[[@CR23]\]\ Policies and programs that seek to transform gender relations to promote equality and achieve program objectives. This approach attempts to promote gender equality by (1) fostering critical examination of inequalities and gender roles, norms, and dynamics; (2) recognizing and strengthening positive norms that support equality and an enabling environment; and (3) promoting the relative position of women, girls, and marginalized groups and transforming the underlying social structures, policies, and broadly held social norms that perpetuate gender inequalities.Special measures \[[@CR19]\]\ Programs, policies, and laws that seek to neutralize and redress embedded structures of discrimination and preferences for privileged groups that are already built into social institutions. Such affirmative measures place women or other marginalized groups in a situation of comparative advantage for a limited period, with the aim of achieving substantive equality in the long term.Substantive equality \[[@CR19]\]\ Takes into account the effects of past discrimination and recognizes that rights, entitlements, opportunities, and access are not equally distributed throughout society and therefore the need to sometimes treat people differently (through special measures) to achieve equal results.Systemic structural discrimination \[[@CR19]\]\ Patterns of behavior, policies or practices, and social, economic, or cultural background conditions that are part of the structures of institutions, which create or perpetuate disadvantage for members of a marginalized group relative to other groups in society or organizations.Intersectionality\ A feminist theory and analytical tool for understanding and responding to the ways in which gender intersects with other identities. The experiences of marginalization and privilege are not only defined by gender but by other identity factors, such as race, class, and sexual orientation to name a few---all of which are determined, shaped by, and imbedded in social systems of power. Intersectional paradigms view race, class, gender sexuality, and ethnicity among others as mutually constructing systems of power (\[[@CR47], [@CR48]\]; references [@CR36], [@CR49]--[@CR52] present the theoretical and methodological issues and opportunities related to this theory). Gender discrimination---whether culturally, socially, or structurally driven---is sometimes evident in health worker education and employment systems and sometimes so normative as to be invisible. Where gender discrimination and inequality exist, they can hinder production of the robust and competent health workforces needed to achieve health and development goals, by subjecting students and faculty to gender-based exclusions or restrictions and diminishing access and opportunities. As systemic problems that stem from (among other things) institutional cultures, norms, and policies, gender discrimination and inequality should be addressed in HPE reforms. To date, however, the HPE reform movement has paid scant attention to gender discrimination toward and among students and faculty. The new SDGs include a standalone goal that aims for achieving gender equality and empowering all women and girls, including a target to end all forms of discrimination against all women and girls \[[@CR1]\]. This gives gender discrimination and inequality a new primacy in development policy. As the independent Commission's report emphasizes, a key outcome of educational reform should be transformative learning \[[@CR3]\], which can cultivate a "new professionalism" and establish "enlightened change agents" who have the status, authority, ability, and willingness to challenge the numerous structural and cultural factors that keep gender-based discrimination in place. Transformative learning to address gender discrimination must necessarily be gender transformative. Equal opportunity, non-discrimination, gender equality, and respect for human rights should be core health professional values and competencies, promoted and enacted by HPE leaders, embedded in curricula, and enacted through HPE instruction, professional socialization, and institutional governance as foundations for the professional practice of future health workforce managers and frontline service providers. This paper summarizes the findings of an expert review that sought to assess interventions to combat gender discrimination and inequality in HPE settings and rank them according to whether they counter two broad types of gender discrimination in transformative ways: discrimination based on pregnancy and on family responsibilities (hereinafter called "caregiver discrimination") and sexual harassment. We also discuss challenges to implementing the identified interventions, recommendations for addressing the challenges, and implications for HPE reform. Gender discrimination and inequality {#Sec3} ------------------------------------ The global literature on gender and human resources for health (HRH) has demonstrated that gender discrimination and inequality are key barriers to entry, reentry, and retention in employment systems, especially for female health workers \[[@CR4]--[@CR6]\]. The common forms of gender discrimination documented in health workforce employment systems (Table [1](#Tab1){ref-type="table"}) also appear to operate in HPE settings, affecting HPE students' opportunities, treatment, and ability to complete their studies (Table [3](#Tab3){ref-type="table"}), and limiting faculty members' career satisfaction, advancement, and economic opportunities (Table [4](#Tab4){ref-type="table"}). We focus on sexual harassment and caregiver discrimination because they were apparent from the interventions described in the literature and because they are relevant to female HPE students and faculty \[[@CR7]\].Table 3Students' experience of gender discrimination and inequalitiesPhase of academic life cycle (students)Examples and resultsCareer selection• Gender stereotypes and segregation in health professional cadres such as nursing and nutrition (Kenya) \[[@CR10]\]Admission and entry• Negative attitudes against girls and women pursuing training and scholarship opportunities \[[@CR6]\]\ • Lower admission rates of female students at tertiary education institutions (Rwanda) \[[@CR53]\]Course participation and completion• Threats of failing grades against female students made by male faculty if students refuse sexual advances, leading to difficulty concentrating on studies or failed courses (Ghana, Kenya, Nigeria, Uganda, Zimbabwe) \[[@CR10], [@CR20], [@CR54]--[@CR56]\]\ • Demotion fees levied against students for taking time off for pregnancy and falling behind in their programs (Kenya) \[[@CR10]\]Career progression• Sexual harassment during medical training, affecting selection of medical specialty and residency programs (Japan, Sweden, USA) \[[@CR5], [@CR57]--[@CR60]\]\ • Attitudes discouraging female medical residents from becoming pregnant (USA) \[[@CR8]\]Retention and graduation• Unsafe living conditions, limiting students' ability to safely access university facilities and contributing to dropouts of female students \[[@CR11]\]\ • Threats of failing grades against female students by male faculty if students refuse sexual advances, leading to graduation delays or dropouts (Ghana, Kenya, Nigeria, Uganda, Zimbabwe) \[[@CR10], [@CR20], [@CR53]--[@CR56]\]\ • Required suspension or termination of studies when female students become pregnant (Namibia, Kenya) \[[@CR9], [@CR10]\]\ • Insufficient time for students with family and domestic responsibilities to participate in educational opportunities, contributing to dropouts (Tanzania, Uganda, UK, USA) \[[@CR11]--[@CR13]\]\ • Higher dropout rates of female students at all educational levels (Rwanda) \[[@CR53]\]Table 4Faculty members' experience of gender discrimination and inequalitiesPhase of academic life cycle (faculty)Examples and resultsRecruitment• Challenges in balancing work and family obligations, contributing to faculty leaving their positions or turning down employment offers (USA) \[[@CR31]\]\ • Hiring preferences for male faculty due to belief that female faculty taking maternity leave will be disruptive (Kenya) \[[@CR10]\]Career advancement opportunities• Discriminatory promotion decisions against female staff due to resistance to sexual advances (Nigeria) \[[@CR20]\]\ • Lower number of publications by female medical faculty with children (USA) \[[@CR31]\]\ • Taking reduced workload or time off for family responsibilities seen as a reflection of lower commitment to work (Australia, USA) \[[@CR16], [@CR61]\]Leadership• Higher numbers of male faculty in senior leadership, even in cadres traditionally considered female occupations, such as nursing (Kenya) \[[@CR10]\]Satisfaction and retention• Lower rates of career satisfaction among female medical faculty with children than among male medical faculty with children (USA) \[[@CR15], [@CR17]\] ### Students {#Sec4} For HPE students, gender discrimination and inequalities are apparent at different points throughout educational careers, from admission to career track designation to graduation. At the outset of the educational continuum, cultural and gender norms and stereotypes related to childbearing and childrearing can discourage girls and women from pursuing HPE training and scholarship opportunities \[[@CR6]\]. Once students are enrolled, gender-blind institutional policies and practices (i.e., policies and practices that do not take gender considerations into account) may prevent female HPE students from participating in classes, practicums, and other curricular offerings by failing to consider potential conflicts between educational requirements and students' caregiving responsibilities. Such gender blindness typically results in a lack of instrumental support for students, creating barriers to students' ability to equally access education or remain enrolled. Some professional programs also discourage women from becoming pregnant while they are students \[[@CR8]\]. In several countries, pregnant students are required to take mandatory time off before returning to school or may even face expulsion \[[@CR9]\]. Caregiver discrimination also may play out in the form of demotion fees for pregnant students who take time off and fall behind in their courses and practicums \[[@CR10]\]. Caregiving responsibilities have been shown to play a major role in attrition rates in countries such as Kenya, Tanzania, Uganda, the UK, and the USA \[[@CR10]--[@CR13]\]. In higher-education systems as a whole, insufficient and sometimes insecure living conditions can limit female students' ability to safely access university facilities and further contribute to decisions to drop out \[[@CR11]\]. Sexual harassment and assault have been documented in primary and secondary schools and universities in high- and low-resource settings \[[@CR14]\]. For HPE students, sexual harassment, threats, or assault by other students or teachers (whether quid pro quo or in a hostile environment) can make it difficult for the targeted student to concentrate on or complete coursework \[[@CR10]\]. Sexual harassment can fundamentally change students' educational environment and opportunities and may contribute to a student's decision not to pursue a particular career track. ### Faculty {#Sec5} For faculty, gender discrimination and associated inequalities often relate to conditions that structurally disadvantage members of one sex (typically women) in the academic system, such as requiring training that involves travel in order to obtain promotions \[[@CR6]\]. An academic culture of long working hours and implicit biases against faculty with family responsibilities can affect promotion and tenure decisions in both HPE and general higher-education institutions \[[@CR15]--[@CR17]\]. Recent organizational research found that pregnant women were perceived as less competent, less committed to their jobs, and furthest from meeting male "ideal worker" norms \[[@CR18]\]. Such forms of invisible bias and structures of discrimination can be embedded in social institutions over time and become culturally normative \[[@CR19]\]. Sexual harassment of faculty members can also affect career advancement. For example, Nigerian female academic staff reported that their refusal of university officials' sexual advances led to exclusion from promotion and other benefits \[[@CR20]\]. This type of discrimination lowers faculty self-confidence, career satisfaction, and retention, which in turn can affect the quality of education being provided at HPE institutions and contribute to faculty attrition \[[@CR5]\]. HPE institutions may enable (and fail to regulate) sexual harassment due to cultural norms that tolerate harassment and gender-blind policies \[[@CR21]\]. The failure to counter sexual harassment in HPE systems can have a serious and detrimental impact on both student and faculty experiences, as seen by system dysfunctions such as sex in exchange for grades or academic career advancement \[[@CR20]\]. The following sections describe our review of interventions to counter caregiver discrimination and sexual harassment in high- and low-resource school-based settings. Methods {#Sec6} ======= Review and assessment of gender-transformative potential {#Sec7} -------------------------------------------------------- A panel of experts with expertise in gender and health system strengthening reviewed and rated 52 distinct school-related interventions to address sexual harassment and caregiver discrimination in high- and low-resource settings (Fig. [1](#Fig1){ref-type="fig"}).Fig. 1Stages in the review, rating, and ranking of interventions There were 87 distinct educational institutions/organizations/programs that were included in the literature review, most of which were universities. Of the 87 institutions, the geographical breakdown was:North America and the Caribbean, 33 (38 %)West, East, and Southern Africa, 32 (36.8 %)Asia, 10 (11.5 %)Middle East and North Africa, 5 (5.75 %)Europe, 4 (4.6 %)Australia, 2 (2.3 %)South America, 1 (1.2 %) Interventions that had been previously identified in a review of 379 articles from peer-review and non-peer-review literature on HPE and general tertiary education \[[@CR22]\] provided the initial pool of interventions. The experts then assessed interventions' gender-transformative potential using criteria described in Table [5](#Tab5){ref-type="table"} by a given type of discrimination, focusing on key aspects such as location(s) where implemented, intervention features, and results of any formal evaluations or informal assessments. One practice was subsequently removed from the analysis because there was not enough information about the intervention itself, resulting in a final count of 51 interventions: 18 to counter sexual harassment, 27 to counter caregiver discrimination, and 6 to address gender equality more generally. Detailed, contextual information related to the review and assessment of interventions is available \[[@CR22]\].Table 5Criteria used to rate gender-transformative interventions in health professional educationCriteriaSexual harassment Take measures to end impunity for perpetrators of sexual harassment and other forms of gender discrimination Top critical criterion to counter sexual harassment Introduce, make use of, or further legal protections against gender discrimination Second critical criterion to counter sexual harassment Provide information and education about discrimination or rights Third critical criterion to counter sexual harassmentCriteriaCaregiver Discrimination Transform family, school, and/or work arrangements so that women are not penalized or disadvantaged for caregiving Top critical criterion to counter caregiver discrimination Challenge and change common discriminatory gender beliefs or norms Second critical criterion to counter caregiver discrimination Attempt to change imbalance of power or otherwise level the playing field Third critical criterion to counter caregiver discrimination Most interventions lacked the type of evaluation data related to inputs, processes, and outcomes that would allow the reviewers to determine their effectiveness, feasibility, or sustainability. Thus, reviewers rated the interventions' potential to counter gender discrimination and inequality by applying the six gender-transformative criteria in Table [5](#Tab5){ref-type="table"}. These criteria were formulated by the review team based on the USAID-supported Interagency Gender Working Group definition of gender-transformative policies and programs (see Table [2](#Tab2){ref-type="table"}) \[[@CR23], [@CR24]\]. The criteria were considered to be minimum standards for classification of a gender-transformative intervention in HPE settings. The reviewers discussed the meaning of the criteria and marked "Yes" or "No" for each criterion for each intervention to be rated. Intervention rating and ranking {#Sec8} ------------------------------- For interventions targeting the two types of discrimination (i.e., sexual harassment and caregiver discrimination), each reviewer was assigned a weight based on the reviewer's area of expertise, with assigned weights totaling 100 %. For each of the 51 practices, reviewers rated interventions applying the six criteria, with 0 representing "No" and 1 representing "Yes." The researchers then derived a weighted average for each intervention. The expert panel decided that an intervention had at least some gender-transformative potential if its weighted average was 0.5 or above. To rank interventions, one that had a 0.5 rating would be ranked higher than one with a 0 rating. Interventions were ranked by taking inventory of the reviewers' ratings using these critical criteria \[[@CR22]\]. The reviewers met to discuss the ranking process and final rankings. During these meetings, they refined their application of the six gender-transformative criteria, developed recommendations about core sets of interventions, and developed cross-cutting recommendations related to countering the two types of discrimination. Results {#Sec9} ======= Core sets of intervention {#Sec10} ------------------------- Of the 51 interventions reviewed, 13 interventions were identified as having significant stand-alone transformative potential in terms of the transformative criteria (see Table [5](#Tab5){ref-type="table"}) and more so when implemented in combination, that is, in "core sets." The reviewers formulated these core sets to include those interventions that were necessary (though not sufficient) to counter a particular form of discrimination in HPE settings. The core sets of priority interventions to counter sexual harassment and caregiver discrimination were those that met the critical criteria such that an intervention that met, for example, the top two critical criteria was ranked higher than an intervention that met only the top critical criterion. The panel also identified implementation challenges for each set where such information was available and formulated recommendations for addressing those challenges. ### Core set to counter sexual harassment {#Sec11} The expert reviewers identified a core set of three gender-transformative interventions with the potential to counter sexual harassment in HPE, shown in Table [6](#Tab6){ref-type="table"}. Establishing a sexual harassment policy and a grievance procedure appears to be feasible across high- and low-resource settings, as evidenced by implementation of the two practices in a number of African and North American universities \[[@CR22]\].Table 6Interventions included in the core set to counter sexual harassment in health professional educationInterventionDescriptionSexual harassment policy• Includes a single code of conduct for students, faculty, and staffGrievance or reporting procedure• Is confidential, outlines consequences for perpetrators, and takes concrete action to end impunity and reduce victims' fear of or vulnerability to retributionEducation and awareness raising• For students, faculty, and staff ### Key implementation challenges for interventions to counter sexual harassment {#Sec12} Although many HPE and other higher-education institutions included in the review had implemented one or more of the interventions in the core set to counter sexual harassment, the review identified a number of challenges:Sexual harassment policies may outline strong principles and institutional responsibilities, but the practical implementation of such policies can differ widely from their intentions. For example, the University of Stellenbosch's (South Africa) policy mandates a sexual harassment advisory and disciplinary committee, yet an assessment found that not only were many managers unaware of the policy but committee members' workloads made trainings on the policy difficult to schedule and implement \[[@CR25]\].The lack of awareness of grievance procedures (and of sexual harassment policies), along with inadequate individual and institutional training, can contribute to anemic use of grievance procedures in settings where sexual harassment is normative.Most policies explicitly prohibit retaliation against victims who report sexual harassment, but flawed grievance procedures and prevailing environments of intimidation or impunity can render anti-retaliation policies ineffective.Fear of retribution and lack of accountability discourage many victims of sexual harassment from using grievance procedures \[[@CR25]--[@CR28]\]. Assessments of Chancellor College in Malawi \[[@CR26]\] and the University of Botswana \[[@CR27]\] noted that when cases were reported, significant errors occurred in handling investigations, maintaining confidentiality, assuring that alleged harassers showed up, coordinating with responsible agencies, and even following the prescribed procedures, which caused students to lose confidence in the process. Given the possibility of culturally normative and unregulated sexual harassment in HPE settings, grievance procedures are an important intervention. To address the identified challenges, grievance procedures should pay special attention to confidentiality, guidance for documenting and reporting, clearly outline consequences for the perpetration of sexual harassment and retaliation, avoid an inadvertent chilling effect on reporting that may result from an overemphasis on false reporting, and take concrete action to both decrease and eliminate fear of retribution. Equally important, strategies must be implemented and enforced through strong institutional leadership, vigilant oversight, and timely follow-up and resolution. ### Core set to counter discrimination based on caregiver responsibilities {#Sec13} The reviewers also identified two core sets of interventions for students and faculty to counter caregiver discrimination (Table [7](#Tab7){ref-type="table"}). Practices included in these core sets have been shown to be feasible in some settings, with institutions in South Africa, Tanzania, and other countries offering child care \[[@CR29], [@CR30]\]. However, of the institutions reviewed, only the University of California and the University of Michigan, both in North America, offered the full set of interventions comprising the core set for faculty, and no institutions were identified that offered the full core set for students \[[@CR22]\].Table 7Interventions included in the core sets for HPE students and faculty to counter caregiver discriminationCore set for studentsCore set for facultyPregnancy• Pregnancy/maternity and parental leave• Pregnancy/maternity and parental leave (paid)• Continuation and reentry policies that do not require pregnant students to terminate their education• Pregnancy/maternity leave replacement funding to hire temporary replacements for employees on pregnancy/maternity leave to ensure continuity of instructionPostpartum• Lactation breaks and spaces• Lactation breaks (paid) and spaces• Parental leave• Parental leave• Child care (daily and emergency)• Child care (daily and emergency)• Child care financial assistance (or at low cost)• Child care financial assistance (or at low cost)• Flexible training schedules, such as part-time schedules and reduced workloads• Flexible working hours\ • Flexible tenure ### Key implementation challenges for interventions to counter caregiver discrimination {#Sec14} Interventions to counter discrimination based on caregiver responsibilities also face implementation challenges, although these, too, can be met by strong HPE leadership commitment:Adverse consequences---or fear thereof---are a significant barrier associated with some interventions. For example, faculty who opt for reduced duty leave or flexible training programs may experience resentment from colleagues. Moreover, HPE faculty may not always take advantage of interventions for fear that others will perceive them as uncommitted or that their careers will be negatively affected.Work-life integration is a key concern for many current and prospective HPE faculty (both women and men) \[[@CR31]\]. Institutions with family-friendly policies may, therefore, have a competitive edge in recruitment. Indeed, outside of the HPE sector, the University of Washington law school has used its family-friendly environment as a student- and faculty-recruiting tool \[[@CR32]\], and the University of California and University of Michigan both highlight their family-friendly initiatives to faculty candidates.Families and communities may resist some of the changes required to address discrimination based on caregiver responsibilities, because the interventions challenge longstanding gender norms, expectations, and divisions of labor. Girls and women who go to school likely need a reduced workload at home, potentially adding to their families' workload. Communication of policy and education of faculty and students, as well as ongoing public support of faculty who use such flexible policies, is key to preventing adverse consequences. A complementary strategy is to proactively plan for pregnancy coverage and flexible scheduling. HPE planners must also anticipate the different levels of resistance that may arise in recruitment and retention efforts and deliberately mobilize communities around reducing women's and girls' housework, preventing early marriage and pregnancy and sharing responsibility for caregiving. This implies a long-term, multidimensional, and multisectoral strategy to keep girls in school from the primary through tertiary levels. This might include provision of reproductive health services (including family planning) through HPE institutions. ### Interventions that address multiple forms of gender discrimination and inequality {#Sec15} In addition to selecting the core sets of interventions targeting sexual harassment and caregiver discrimination, the reviewers identified gender centers and equal employment opportunity units as having significant gender-transformative potential. These are institutional structures that advocate for, coordinate, oversee, implement, and evaluate multilevel strategies. These entities generally work to:Develop gender equality, equal opportunity, or affirmative action policiesEngage in awareness raising and information sharingServe advocacy and accountability functionsConduct gender sensitization workshops or sexual harassment training for women and menConduct research and university assessmentsProvide financial assistance to female studentsOffer mentoring and faculty career and leadership programs to women The last two objectives are examples of special measures to counter systemic structural discrimination and promote substantive equality (see Table [2](#Tab2){ref-type="table"}). These special measures counter the discrimination that may occur when poor families allocate scarce financial resources to fund boys' education and to compensate for the career barriers faced by women in HPE institutions characterized by high concentrations of men in top faculty and administrative positions. Whereas equal employment opportunity units which aim to counter discrimination in employment and occupation \[[@CR33]\] are often backed by national equal opportunity laws, gender centers face the challenge of not necessarily being backed by law. In addition, implementation challenges may arise from funding or staffing constraints. Leaders in HPE reform should educate stakeholders and advocate for the need for resource allocations to fund special measures to counter systemic structural discrimination to achieve substantive equality. Research limitations {#Sec16} -------------------- The relative lack of descriptive contextual and evaluation data for the 51 interventions limited the expert reviewers' ability to determine the exact nature, feasibility, sustainability, or effectiveness of the various interventions and, therefore, constrained their capacity to make recommendations for specific contexts such as low-resource settings. Overall, more information was available for institutions in high-resource than low-resource settings. The research team invited a variety of published experts representing different sectors and countries to participate as reviewers, but most experts were unavailable. A larger expert review group (including stakeholders such as students and faculty), or one with a more diverse range of expertise, would likely strengthen future reviews. Discussion {#Sec17} ========== A comprehensive HPE reform agenda aiming to produce a robust and competent health workforce should consider core sets of interventions to counter gender discrimination and inequality---even when discrimination is not overtly recognized by perpetrators or victims. Female health workers already constitute a large proportion of many countries' health workforces (in both the professional and non-professionalized cadres), and there is a growing focus on the role of women in the health workforce in the emerging human resources for health agenda \[[@CR5]\]. Failure to address gender discrimination and inequality in HPE can jeopardize broader health workforce and health system reform. The review findings can serve as a basis for evidence-based decisionmaking in planning and implementing appropriate gender-transformative interventions. The review demonstrated that interventions can address some of the gender-related obstacles to entry, retention, career progression, and graduation in HPE, in particular those related to sexual harassment and caregiver discrimination. Although many HPE institutions may not have the resources to implement all the gender-transformative interventions identified by the reviewers, the "core sets" provide a basis for establishing priorities and taking steps to counter gender discrimination. Potential implementation challenges spanning financial, informational, and cultural barriers need to be considered, however. Financing considerations are particularly relevant when interventions entail restructuring physical resources and human resources arrangements (such as child care, lactation spaces, or reduced duties leave) or require human resources (e.g., education and awareness-raising activities or adequately staffed and trained sexual harassment committees). Merely offering interventions does not guarantee that interventions are used or that equal opportunities and gender equality in HPE will result. Ensuring that the institutional community and especially the intended beneficiaries are aware of, committed to, and actually use the interventions is equally important. However, because many of the interventions challenge social and institutional norms and cultural stereotypes, some community members (inside or outside the institution) may resist their implementation. Increasing the actual use of new policies or procedures requires information and advocacy, anticipating potential resistance, providing incentives, and ensuring that HPE leaders are accountable for work-life integration and the speedy and effective handling of sexual harassment so that institutional efforts actually prevent and end impunity for discriminatory organizational structures and behaviors. What are some implications for HPE reform? First, gender-transformational HPE reform will in some cases entail both institutional and instructional changes \[[@CR3]\] sometimes in combination, since some reforms to institutional governance may require, or can be enhanced by, instructional reforms. For example, the gendering of the health workforce involves the distinction between occupations, as well as the relations between occupations (for instance, between medicine, nursing, and allied health professions) \[[@CR34]\]. These processes of gender segregation or stratification could be mitigated by both equal opportunity policy (an institutional governance reform) and transprofessional education (an instructional reform). Institutional reform, such as introducing a new policy, could also be combined with professional instruction to raise students' and faculty awareness of gender discrimination as unethical professional conduct, as well as a human rights violation. For example, the problem of quid pro quo sexual harassment (i.e., sex in exchange for grades, see Table [3](#Tab3){ref-type="table"}) is a form of unethical professional conduct within an HPE, an abuse of power by faculty which impacts educational or occupational opportunity, which has both instructional and institutional consequences. Heads of HPE institutions can work with an equal opportunity unit to introduce a code of conduct to faculty and students and include the subject in a course in ethics or human rights \[[@CR35]\], thus playing a dynamic role in inculcating shared attitudes, values, and behaviors related to respect for the dignity and rights of clients, students, and colleagues. Instructors can be trained to model ethical one-on-one and team interactions and conducting fair assessments free from the taint of sexual harassment, stereotyping, and other forms of discrimination. Third, the review highlights the importance of a gender-relational perspective in HPE reform. Gender relations shape health systems---including HPE systems---through their effects on the occupational segregation and stratification among health providers, the conditions of work, and processes of regulation, supervision, and management of health labor forces \[[@CR36]\]. The theoretical foundations of the approach in this paper include gender-relational theory, social dominance theory, and the sociology of patriarchy \[[@CR34], [@CR37], [@CR38]\]. This perspective gives a central place to the patterned relations between women and men (and among women and among men) that constitute gender as a social structure \[[@CR39]\]. The structure of gender relations (including power relations and hierarchies) in a given institution is its "gender regime" \[[@CR34]\]. The institutions through which health care is delivered (e.g., hospitals, clinics, private practices) have well-defined gender regimes \[[@CR34]\], and this applies to HPE institutions. If so, heads of HPE schools would benefit from a gender analysis approach to shed light on the "gender regime" in their institutions and possible impacts on faculty and students at various points in the academic career. A practical first step would be for heads of schools, equal opportunity staff, or human resources managers to conduct a gender audit of its institutional "gender regime," including policies and practices in relation to pregnancy, family responsibilities and sexual harassment, instructional content, socialization processes, and the socioeconomic characteristics of its student body and faculty. Ending all forms of discrimination against all women and girls everywhere (i.e., SDG 5.1) does not speak to the multiple or intersecting axes of gender discrimination, bias, and marginalization (e.g.,, economic class, region, race, or caste) which likely impact present and future student bodies. Such analysis could result in changes in recruitment, admissions, and financing \[[@CR3], [@CR40], [@CR41]\]. Operationally, this is a "field waiting for an analytical breakthrough" (\[[@CR36], p. 9\]) and should be part of an HPE-learning agenda. Fourth, institutional governance reforms suggest the utility of gender centers or equal employment opportunity units to drive policy and accountability. This unit could coordinate the development and enforcement of supportive, evidence-based policies to promote equal opportunity, non-harassment, and social equality. As a part of policies such as flexible working arrangements and career paths that accommodate temporary breaks, HPE leaders and HR managers should embrace gender equality in social roles and promote the value of caring and work-life integration for both women and men in educational and employment systems \[[@CR3], [@CR42]\]. Finally, the review highlighted the striking lack of contextual and evaluation documentation for gender-transformative interventions in HPE and the need to invest in a stronger learning agenda in relation to sexual harassment, caregiver discrimination, and, in fact, in all forms of discrimination already documented in health employment systems. We recommend building a rigorous knowledge base as well as evaluating the feasibility, sustainability, and effectiveness of gender-transformative HPE reforms. Table [8](#Tab8){ref-type="table"} outlines suggested areas of focus for this learning agenda.Table 8A learning agenda for gender-transformative health professional education reformForm of discriminationSuggested elements of learning agendaSexual harassment• Elements and features of grievance procedures that effectively resolve sexual harassment cases and end impunity\ • Effectiveness of various (HPE) institutional and non-institutional actors in investigating sexual harassment\ • Role of extra-institutional legal advocacy and redress\ • Desirability of separate grievance procedures for students, faculty, and staff versus a single procedureCaregiver discrimination• Impact of basic bundle on student retention, performance, and graduation\ • Impact of basic bundle on faculty recruitment, retention, and development\ • Feasibility and effectiveness of offering free or low-cost family planning for men and women in the basic bundle\ • Effective community messaging related to the need to reduce women's and girls' family and domestic burdens\ • Factors that contribute to use of family-friendly services by both women and menCross-cutting• Approaches for inter- and transprofessional education to promote equality and teamwork and effectively eliminate gender stereotyping and stratification (e.g., siloes, hierarchical chains of command)\ • HPE curricula that transmit core professional values and competencies (including human rights, social justice, gender equality, professional ethics and conduct, respectful care, and critical inquiry) and foster health workers as change agents in inequitable systems\ • Test new and efficient gender audit methodology and information systems to support improved institutional governance policies and monitor progress of reforms, targeting the analysis of gender regimes and intersecting discriminations, including the collection of data for indicators of social position (e.g., gender, income, geographical region, race, caste, disability)\ • New approaches to teach human rights and ethical professional conduct in combination with HPE governance.\ • Coordinated efforts across universities to implement special/affirmative measures and enabling conditions such as work and learning design to take parenting and caregiving responsibilities into account for all students and faculty, universal flexible working/study arrangements.\ • Structural and resource requirements for and impact of equal opportunity units in HPE institutions Conclusions {#Sec18} =========== The paper's broad aim is to encourage HPE leaders---heads of schools, human resources administrators, donors, and other human resources for health (HRH) stakeholders---to make gender-transformative changes in the current way of doing business and commit themselves to countering gender discrimination and inequality in HPE. These affect the functioning, quality, outputs, and outcomes of health professional education. Gender-transformative HPE reforms that eliminate impunity for sexual harassment, promote and protect the educational and labor rights of students and faculty, and develop student and faculty understanding of professional ethics will require changes in mindsets, institutional cultures, and leadership capability and the reframing of health workers as change agents committed to gender equality in often inequitable systems. Greater focus on the application of the core sets of interventions identified in this paper and a strong learning agenda should be part of ongoing HPE reform efforts. Competing interests The authors declare that they have no competing interests. Authors' contributions CoN, CrN, and SPM were responsible for the conception and design, acquisition of data, and analysis and interpretation of data for the systematic review on which the article was based and for drafting and revising the manuscript and graphics. DF was involved in interpreting findings from the systematic review and critically revising the manuscript. All authors gave final approval of the version to be published. Authors' information Authors Crystal Ng and Sara Pacqué-Margolis are former employees of IntraHealth International. The research described in this paper was funded through the United States Agency for International Development's (USAID's) CapacityPlus Project, under Grant GPO-A-00-09-00006-00. The views expressed in this paper are solely those of the authors and do not necessarily reflect the views of USAID or the United States government. The authors are grateful to Mesrak Belatchew, formerly of IntraHealth International; Asha George at the Johns Hopkins Bloomberg School of Public Health; Rebecca Bailey, formerly of IntraHealth International; and Nonie Hamilton, Lois Schaefer, and Temitayo Ifafore of USAID. Special thanks to Claire Viadro for editorial support.	{ "pile_set_name": "PubMed Central" }
Introduction {#S0001} ============ Acne is a common, multifactorial, inflammatory skin disease affecting the pilosebaceous unit resulting from altered keratinization, androgen-induced increased sebum production, inflammation involving both innate and acquired immunity and microbial colonization by Propionibacterium (P.) acnes.[@CIT0001]--[@CIT0003] Genetic predisposition, anxiety and depression, hyperglycemic diet, sunlight or artificial light, are additional factors known to be involved in the pathogenesis of acne.[@CIT0004] Acne is clinically characterized by different clinical aspects and severity, spanning from mild forms, mainly characterized by retentional lesions (RTL) (open and closed comedones) with or without few inflammatory lesions (IFL) (papules/pustules), to moderate or severe forms, with multiple IFL that seriously impact on patient's quality of life.[@CIT0005],[@CIT0006] The treatment vary according to disease severity. For mild acne, topical pharmacological treatments are generally recommended and include the use of benzoyl peroxide (BP), retinoid, and antibiotics.[@CIT0005]--[@CIT0010] The clinical efficacy of these agents is widely demonstrated, and it is well known that their use may cause some adverse effects, such as hypersensitivity reaction, erythema, burning/stinging in case of BP/retinoids, and limitations for the possible development of resistant bacterial strains for antibiotics.[@CIT0005]--[@CIT0010] Recent evidence suggests that some cosmetic agents if correctly prescribed may improve the therapeutic outcome with the advantage of minimizing the side effects and enhance the efficacy of prescription drugs.[@CIT0011] The aim of this study was to assess the efficacy and tolerability of a new cosmetic regimen consisting in the daily use of an anti-inflammatory/corneolytic/sebum controlling fluid in the morning and of an anti-inflammatory/corneolytic cream at bedtime in the treatments of mild facial acne by clinical and instrumental evaluations for 8 weeks duration study. Materials and Methods {#S0002} ===================== Study Design {#S0002-S2001} ------------ This was a multicenter, prospective, observational clinical study. Setting and Study Period {#S0002-S2002} ------------------------ From October 2017 to May 2018 ninety-one adult patients (71F/20M) with mild acne were enrolled at the Dermatology Department of the University of Catania (Italy) and of University "Federico II" of Naples (Italy). Study duration was 8 weeks. The study was performed in accordance with the ethical principles originating from the Declaration of Helsinki 1996 and Good Clinical Practices. The protocol was approved by an internal review board. A written informed consent was obtained from each patient before study procedures were started, which included consent to use their images. Inclusion/Exclusion Criteria {#S0002-S2003} ---------------------------- Inclusion criteria were: adult patients with mild facial acne (open and closed comedones and up to 10 papules), a wash out period of at least 2 weeks for any topical acne treatments, and 1 month for oral antibiotics and 2 months for hormonal therapy and/or oral isotretinoin. Exclusion criteria were: presence of pustules or nodules lesions, trunk localization, severe underlying diseases, concurrent exposure to sunlight and/or artificial ultraviolet sources, pregnancy and breastfeeding. No other topical products or drugs were allowed, except for a mild cleanser, make-up and oil-free SPF 50+ sunscreens. Methodology {#S0002-S2004} ----------- Patients were instructed to apply a fluid containing Licochalcone A combined with Salicylic acid, and L-Carnitine in the morning and a cream based on Licochalcone A and Hydroxy-complex 10% at bedtime for 8 weeks. A mild, perfume-free cleaning gel principally based on Ampho-Tenside 6% was suggested for daily care. In order to reduce potential evaluator bias, all subjects were assessed by an investigator not directly involved in the study at baseline (T0), and at 4 weeks (T1), and 8 weeks (T2). Clinical and Instrumental Evaluation Criteria {#S0002-S2005} --------------------------------------------- Clinical and instrumental evaluation were carried out at baseline, and at 4 and 8 weeks. Acne severity was rated by clinical evaluation using: 1) the Global Acne Grading System (GAGS) score consisting in a subjective assessment of acne severity proposed by Doshi et al in 1997 that divides the face into six areas, forehead, nose, chin, right cheek, left cheek, chest and upper back and assigns a factor (from 1 to 3) to each area (2=forehead; 1=nose; 1=chin; 2= right cheek; 2= left cheek; 3=chest and upper back): the local score is calculated using the formula factor x grade (depending on the most severe lesion type; 0= no lesions; 1= comedones; 2=papules; 3=pustules; 4=nodules); the global score is the sum of all global scores: a score of 1--18 is considered mild; 19--30 moderate; 31--38 severe and \>39 very severe and by 2) lesions count, method based on count of the total number of retentional and/or inflammatory lesions.[@CIT0012],[@CIT0013] Instrumental assessment included: 1) measurement of sebum by Sebutape^TM^ strips (CuDerm Corp., Dallas, TX, USA) placed on the patient's forehead, nose, cheeks, and chin and subsequently checked against a black background of a score card; sebum spots are scored on a scale ranging from 1 to 5, where 1 indicates a dry skin without sebum, and 5 identifies a very oily skin, and 2) facial imaging performed by high-tech facial photography characterized by 15 mega pixel resolution and flash cross-polarized Light by Reveal Photo Imager (Canfield Scientific Inc., Parsippany, NJ, USA) at Naples site and by VISIA-CR™ imaging system (Canfield Scientific Inc., Fairfield, NJ, USA) at Catania site.[@CIT0014]--[@CIT0016] Additionally, evaluation of product tolerability by a self-administered questionnaire based on 5 parameters (erythema, scaling, dryness, stinging/burning and itch) from 0 to 3 (0 =none; 1=mild; 2=moderate; 3=severe) was carried out. Study Endpoints {#S0002-S2006} --------------- Primary endpoint was the clinical efficacy at week 4 and 8 evaluated by reduction of GAGS score and lesion count; secondary endpoint was the evaluation of tolerability and cosmetic acceptability at the end of the study. Statistical Analyses {#S0002-S2007} -------------------- The characteristics of the patients were reported as absolute number and percentage for the categorical variables and as mean and standard deviation (SD) for the continuous variables. The analysis of longitudinal data was performed using a mixed effect regression model, which considered the correlation between measurements carried out over time in the same subject. In the mixed effect regression models, time, age, gender and center were considered as covariates. Interaction between center and time was also tested. For time covariate Dunnett post-test was used for multiple comparisons vs baseline. A two-tailed P value \<0.05 was considered significant. All statistical analyses were performed using SAS version 9.4 (SAS, Inc., Cary, NC). Results {#S0003} ======= Ninety-one subjects (71F/20M; age range: 18--30 years; mean age 21.5±3.8 years; mean GAGS: 11.6±2.4) with mild acne were enrolled. Eighty-eight (70F/18M) completed the study whereas three subjects were lost to follow-up for personal reasons. At 4 weeks a statistically significant reduction from baseline of GAGS (mean from 11.6±2.4 to 10.3±3.0; p\<0.001) was observed. In addition, the mean total count of RTL and IFL was significantly reduced of 41% and 45%, respectively, from baseline (comedones: mean from 42.3±14.3 to 24.8±12.7; p\<0.001; papules: mean from 5.5±5.0 to 3.0±3.8; p\<0.001) along with a significant reduction of mean sebum of about 47% (mean from 2.13±0.67 to 1.13±0.38; p\<0.001) ([Table 1](#T0001){ref-type="table"}, [Figure 1](#F0001){ref-type="fig"}).Table 1Results from Clinical and Instrumental Evaluation of GAG Score, Total Lesion Count of Comedones/Papules from Baseline to 8 WeeksEvaluations (Mean Value)Baseline4 Weeks8 WeeksP ValueMeanDSMeanDSMeanDSGAGSCatania11.22.29.63.47.73.3\<0.001Naples12.02.610.92.68.72.4\<0.001Overall11.62.410.33.08.22.9\<0.001Total Count ComedonesCatania39.311.920.813.612.610.9\<0.001Naples45.116.028.710.517.27.6\<0.001Overall42.314.324.812.715.09.6\<0.001Total Lesions PapulesCatania4.23.32.32.41.32.0\<0.001Naples6.76.03.74.81.83.7\<0.001Overall5.55.03.03.81.63.0\<0.001SebutapeNaples2.130.671.130.381.030.37\<0.001 Figure 1Advanced digital photography of two patients affected by mild acne at baseline (A,C) and after 4 weeks of treatment (B,D). At 8 weeks a further statistically significant reduction from baseline of GAGS (8.2±2.9; p\<0.001) was also recorded; similarly the mean total count of RTL and IFL was significantly reduced of 64% and 71%, respectively, from baseline (comedones: mean from 42.3±14.3 to 15.0±9.6; papules: mean from 5.5±5.0 to 1.6±3.0; p\<0.001), along with an additional mean sebum reduction of about 52% (p\<0.001) ([Table 1](#T0001){ref-type="table"}, [Figure 2](#F0002){ref-type="fig"}). For all considered outcomes, Dunnett multiple comparison post-tests versus baseline showed a significant reduction (p\<0.001) of 4-weeks and 8-weeks values. The two centers showed similar results in the reduction of GAGS (interaction center\time, p=0.655) and total count of comedones (interaction center\time, p=0.192), while for total count of papules a significant interaction center\time was found (p=0,007) showing a greater reduction in the Naples center in comparison to Catania center ([Table 1](#T0001){ref-type="table"}).Figure 2Advanced digital photography of two patients affected by mild acne at baseline (A,C) and after 8 weeks of treatment (B,D). Local side effects were documented in 3 cases only (severe erythema: 1 case; severe hitch: 2 cases) and product tolerability was rated as excellent in 90% of patients. Discussion {#S0004} ========== Cosmetic treatment for acne, if correctly prescribed and used, may improve outcomes in acne therapy.[@CIT0011] The topical cosmetic approach for acne treatment includes: hygiene and cleansing, ideally with a non-comedogenic, non acnegenic, non-irritating and non-allergenic cleanser; sebum controlling agents, generally based on the so-called "mattifyng" ingredients able to absorb sebum from skin surface, in case of hyperseborrea; corneolytics agents, particularly indicated for comedonic acne as able to induce a comedolytic effect, making comedones more superficial and at the same time may also facilitate skin absorption of other products; antimicrobial agents that have been proven to effectively reduce P.acnes growth both in vitro and in vivo clinical study; anti-inflammatory agents indicate to control the pivotal role of inflammatory events in the development of acne lesions; moisturizers product specifically designed for acne patients that improve adherence to therapy by alleviating the discomfort and feeling of dryness following topical and systemic retinoids or topical BP.[@CIT0011] The use of specific photoprotective agents and shaving products should also be encouraged. Finally, tailored camouflage make-up technique using non-comedogenic products may minimize the esthetic problem associated with acne and improve patients' self-esteem and adherence to treatment.[@CIT0011] The results of our study suggest that the tested products are effective in the treatment of mild acne achieving a good resolution in cure rates due to a synergic action of its agents concurrently able to control different factors involved in the pathogenesis of acne. In detail, licochalcone A is a natural flavonoid substance isolated from Glycyrrhiza glabra* and Glycyrrhizainflata (licorice root) with anti-inflammatory and anti-microbial effects, as confirmed by inhibition in vitro and in vivo of pro-inflammatory cytokines (PGE2, LTB4, IL-6 and TNF-α),[@CIT0014]--[@CIT0015] as well as anti ROS-induced cell damage action by activating of the expression of antioxidant transcription factor (Nf-E2-related factor 2/Nrf2) and of detoxifying enzymes (Heme Oxygenase 1/HO-1, Glutamate-Cysteine Ligase Modifier subunit/GCLM).[@CIT0017]--[@CIT0019] Hydroxy Complex is a combination of three exfoliating agents (Glycolic, Salicylic and Polyhydroxy Acids) able to cause intercorneocyte cell detachment so as to induce a comedolytic effect.[@CIT0020]--[@CIT0022] Finally, L-Carnitine is a sebum controlling agent by β-oxidation of intracellular lipid content as demonstrated in vitro and in vivo clinical study.[@CIT0023] Conclusions {#S0005} =========== Our results suggest that the daily regimen-based Licochalcone A in combination with Salicylic acid/L-carnitine as fluid or with Hydroxy-Complex 10% as cream represent an interesting approach for treating mild acne that undoubtedly deserves attention. It may represent a significant adjunct to the therapeutic armamentarium for the management of acne patients without any significant side effects. We are aware that our study has some restrictions being a non-controlled trial but it was a multicenter study with a considerable number of patients enrolled. Further studies on larger series of adolescent acne patients are necessary to confirm our results. Disclosure {#S0006} ========== The study was supported financially by Beiersdorf SpA (Milano, Italy), who manufacture the product. Paolo Chiodini received an honorarium for a consultancy from Beiersdorf SpA (Milano, Italy). The other authors declare no other conflicts of interest in this work.	{ "pile_set_name": "PubMed Central" }
Introduction ============ Chronic recurrent multifocal osteomyelitis (CRMO) is a rare, noninfectious, inflammatory bone disease, which occurs mainly in childhood \[[@b1-dp0804a09]\]. We present a case of CRMO and palmoplantar psoriatic skin lesions in a 12-year-old girl. Case Presentation ================= A 12-year-old girl presented with recurrent erythematous palmoplantar plaques and pustules. She also complained about pain in her left ankle that started 7 months earlier. Previous magnetic resonance imaging (MRI) had consistently revealed a multifocal bone edema of the left foot. Symptoms of weakness, fever, and morning stiffness were absent. The family history was unremarkable. Physical examination revealed well-demarcated erythematous plaques with remnants of dried pustules in a palmoplantar distribution ([Figure 1](#f1-dp0804a09){ref-type="fig"}). The active range of motion of the left upper ankle joint was painfully decreased by 50%. Laboratory results showed a slight increase of inflammation parameters, including c-reactive protein level and erythrocyte sedimentation rate. Antinuclear antibody level, rheumatoid factor, HLA-B27, and Lyme disease testing were negative. Serial MRIs revealed fluctuating T2 hyperintensities and T1 hypointensities involving the left talus, calcaneus, and metatarsal bones with undulating discrete joint effusion in the left upper ankle joint ([Figure 2a--d](#f2-dp0804a09){ref-type="fig"}). Additionally, synovial thickening of the left talocalcaneonavicular joint was noticed. T1-weighted images after contrast application were acquired at several MR-measurements over the course of 1 year, mainly reflecting the edema seen as T2 hyperintensities. Based on these results, the diagnosis of CRMO accompanied by palmoplantar pustular psoriasis (PPPP) was made. Treatment with oral nonsteroidal anti-inflammatory drugs (NSAIDs) and topical mometasone furoate 0.1% cream was initiated. CRMO was first described by Giedion et al in 1972. It primarily occurs in the distal metaphyses of long tubular bones \[[@b1-dp0804a09]\]. The involvement of the calcaneus, as described herein, was rarely reported. PPPP is found in approximately 15% of CRMO patients. The pathophysiology of CRMO is not well understood. Recent studies of CRMO patients described a reduced production of interleukin (IL) 10 by monocytes. This impairment may result in an increased activation of the Nod-like receptor family pyrin domains containing protein 3 inflammasome (NLRP3) leading to an enhanced expression of IL-1 β, which has a role in osteoclast activation via receptor activator of nuclear factor kappa-B ligand (RANKL) stimulation. Bissonnette et al \[[@b2-dp0804a09]\] described high levels of IL-1β and IL-17A in patients with PPPP leading to a secondary chemokine production of keratinocytes with accumulation of neutrophils. Thus IL-1β seems to play a key role in both CRMO and PPPP. Conclusion ========== CRMO should be treated interdisciplinarily, and NSAIDs should be the medication of first choice. Skin lesions may be alleviated by topical steroids. Moreover, bisphosphonates, TNF antagonists, IL-1-inhibitors, sulfasalazine or methotrexate have been described as effective. Fortunately, our patient showed a complete remission 2 years after the onset of symptoms, which is also observed in 30% to 40% of reported cases. In summary, we describe a rare case of CRMO initially presenting as PPPP with joint pain. Dermatologists and pediatricians should be familiar with the association of CRMO and PPPP in children to lead the way to correct diagnosis and treatment. Funding: None. Competing interests: The authors have no conflicts of interest to disclose. ![Well-demarcated erythematous scaly plaques with remnants of dried pustules on the left sole of the patient. \[Copyright: ©2018 Epple et al.\]](dp0804a09g001){#f1-dp0804a09} ![Serial MRIs over the course of 12 months (a--d) with fat suppressed T2-weighted images showing the fluctuating hyperintensities of the left ankle involving the calcaneus, talus, and metatarsal bones with an undulating joint effusion in the left upper ankle joint. The date of each MRI is indicated in the top row of images. \[Copyright: ©2018 Epple et al.\]](dp0804a09g002){#f2-dp0804a09} [^1]: All authors have contributed significantly to this publication.	{ "pile_set_name": "PubMed Central" }

Data sets that are subject to our analysis are available online. Data set B: <http://www.fc.up.pt/addi/ph2%20database.html>, Data Set A: <https://doi.org/10.17026/dans-zue-zz2y>. Because of concerns related to patient privacy and permissions set by the clinic that collected the data, some access restrictions apply to Data Set A. Introduction {#sec001} ============ Melanoma, the neoplasm of the pigment cells of the skin, is still a challenge both for clinicians and CAD (Computer Aided Diagnosis) specialists. Observations with the naked eye and with instruments, especially with popular optical or even with digital dermoscopes require long-term experience which is hard to achieve not only for general practitioners but also for dermatologists. Diagnosis of mature melanoma moles, due to asymmetry, variety in colors or border irregularity, may not be difficult, which is quite the contrary to early melanoma lesions that lack those indications. Effective treatment of melanoma i.e. a high (\>95%) 5 or 10-year survival rate consists in early detection and resection of the malignant skin lesion \[[@pone.0211318.ref001]\]. When not excited at an early stage, melanoma penetrates deep from epidermis to the skin and finally transfers to the lymph nodes and other internal organs by metastasis. At this stage the mortality rate is extremely high and especially for at least a decade has become a medical problem. This problem refers to all countries but particularly these where melanoma morbidity rate is elevated. Statistics says that women got melanoma moles on the legs and men on the back. Crucial for improving the patient survival rate is early detection and for that the only reliable method is biopsy. For obvious reasons (surgical complications, ANS-Atypical Nevus Syndrome, economic reasons) excision cannot be a diagnostic tool or standard mass treatment. The key role is precise detection of the tumor. Clinical diagnosis of early or micro-melanoma is extremely difficult even for experts. A differentiation between pigment and non-pigment skin lesions may be a challenging task but classification of different forms of pigment (melanocytic) lesions can be very complex. From the clinical point of view three forms of pigment lesions can be misdiagnosed: benign nevus, dysplastic (atypical) nevus and melanoma. The dysplastic nevus may be a precursor to malignancy and often exhibits features visually identical or close to melanoma \[[@pone.0211318.ref002]\]. On that account melanoma CAD systems are very popular and of key importance. They help in the early diagnosis of the skin lesions and support non-invasive methods in dermatology. The CAD methods implement machine learning paradigms and are based on data sets of (histopatology or expert ground truth) known cases. Unfortunately such data sets are sparse and with limited statistics of events and perhaps limited or unknown quality (e.g. image resolution, compression). Recognition of melanoma with such computer systems can be divided into: i) direct machine learning (ML) to recognize melanoma from its features, and ii) ML of specific skin lesion patterns/structures that are indicative of melanoma (lesion border, blue veil, geometric asymmetry etc.). The main feature types classified in the melanoma recognition systems include color and/or geometry-based features, histogram-like features, wavelet-based features and other less common (e.g. Gabor etc.). Literature on the melanoma CAD is ample and deal with all the steps: accuracy of clinical and histopathological examinations (i.e. the ground truth), dermoscopy and digital acquisition of lesion images, preprocessing (removal of artifacts, filtering), and finally pattern recognition and machine learning of melanoma. The majority of articles refer to dermoscopy and/or digital imaging of melanoma, since spectral, trans-illumination, ultrasonography or tomography systems, although promising, are not mass skin diagnostic systems. In 2013 Masood et al. \[[@pone.0211318.ref003]\] described the state of the art of the melanoma classification of dermoscopic images with statistics, plus they compared and discussed the results and conditions that affected the analyzed techniques. Later on we refer to this work to scrutinize our results. The latest review on computational methods and their applications as well as trends for automatic melanoma diagnosis was published in 2016 \[[@pone.0211318.ref004]\]. It was focused on feature extraction, feature selection, classification algorithms and evaluation procedures. Overview of statistics and results from the melanoma CAD and segmentation issues can be also found in \[[@pone.0211318.ref005], [@pone.0211318.ref006]\]. The main achievement of this work is first quantification of the C-SVM classification performance of melanoma (in terms of AUC) over a broad range of wavelet bases and SVM kernels. This article is a continuation of our latest research on wavelet-based features of melanoma, derived from dermoscopy images, and classified with various ML paradigms \[[@pone.0211318.ref007], [@pone.0211318.ref008]\]. Our motivation is driven by a systematic search for resolution invariant features and learning methods contributing to the computer pattern recognition of cutaneous melanoma. Motivation {#sec002} ---------- The first and still important approaches to the melanoma detection were segmentation techniques \[[@pone.0211318.ref004]--[@pone.0211318.ref006], [@pone.0211318.ref009]\]. Since dermoscopy images can be taken under different conditions, such techniques can suffer from unstable image illumination, different optical magnification, different skin complexions and presence of artifacts (hairs, reflections, bubbles of immersion fluid etc.). Under such circumstances successful pattern recognition is still possible, but may be limited in performance or strongly sensitive to datasets. Identification methods based on spatial and frequency information found in the skin texture can be a promising alternative. Application of the wavelet- and wavelet-packet- transform to human skin data was first applied in \[[@pone.0211318.ref010]\]. Patwardhan et al. \[[@pone.0211318.ref011]\] studied wavelet packets to decompose the sub-bands of the pigmented skin texture, which proved to derive features sensitive to the class of the dermoscopy image \[[@pone.0211318.ref007], [@pone.0211318.ref008], [@pone.0211318.ref012]\] and then be subject to appropriate machine learning methods. Some WPT attempts are also shown here \[[@pone.0211318.ref013]--[@pone.0211318.ref015]\]. Most of the melanoma wavelet-based classifiers quoted in the literature used only a single wavelet base (usually Daubechies 3 or 4 \[[@pone.0211318.ref011], [@pone.0211318.ref016]\]) to build classification models. In our latest research all popular and well defined wavelets are used. We showed that some selected wavelet bases prove to be more efficient and robust for the machine learning of dermoscopic images than the others. They also keep high classification efficiency in the downgraded image resolutions, that is exhibit resolution invariance. In \[[@pone.0211318.ref007]\] this was analyzed in the ensemble of different model types, and in \[[@pone.0211318.ref008]\] in simple back-propagated neural classifier setups. In this work we specifically study SVM classifiers of the wavelet-based features for the melanoma detection to analyze how wavelet bases can affect the quality of the margin-maximization learning paradigm. The choice of SVM is due to the inherent advantages of the method: i) the ability to generalize well as it maximizes the margin between the classes, ii) no local minima, and iii) flexibility of kernel selection within a unified architecture. Although in this article we want to study and compare with the previous works and literature the SVM classifiers (our work is the first quantification of the C-SVM classification performance of melanoma over a broad range of wavelet bases and SVM kernels), our primary objective is to contribute to optimal feature extraction methods. We believe that adequate wavelet base(s) representation can both yield high classification efficiency and be robust and efficient in worse or reduced resolution environments. For that objective we study one by one different learning paradigms, SVM included. Computational (memory, time) aspects of any machine learning experiment (not particularly SVM) are important factors which should be carefully planned. This is not only the algorithm convergence and regularness that matters, but if the developed classifier is to be used e.g. to support medical diagnosis (in CAD this is the ultimate goal), practical aspects play a role. Computer parts and systems constantly evolve. Nowadays personal smartphones with ARM-based processors are widely used as medical diagnostic tools, the melanoma CAD included (e.g. \[[@pone.0211318.ref017]\], see \[[@pone.0211318.ref018]\] for a review). Since (usually) ML algorithms for image recognition demonstrate high complexity, and small hand-held devices have limited processing power and memory (and the battery capacity), it is of great importance both to: efficiently probe the parameter space of the solution, and to look for features and learning paradigms that preserve high efficiency also in lower image resolutions. In this work our objective is to select the best wavelet bases in terms of AUC classification performance of the SVM binary classifiers of melanoma/dysplastic lesions and analyze how those wavelet bases perform with the reduced image resolutions. Materials and methods {#sec003} ===================== Experimental data {#sec004} ----------------- Data sets that are subject to our analysis are available on-line: 1. Data set A: <https://doi.org/10.17026/dans-zue-zz2y>, 2. Data set B: <http://www.fc.up.pt/addi/ph2%20database.html>. Because of concerns related to patient privacy and permissions set by the clinic that collected the data, some access restrictions apply to data set *A*. Data set *A* was collected in Poland from anonymous dermoscopy patients in 2012-2013 \[[@pone.0211318.ref019]\]. There were 102 malignant melanoma (M) and 83 dysplastic nevus (D) cases (altogether 185 images). Most (not all) of the JPEG images were annotated with some patients data like: sex, age, and location of the lesion (this kind of information is not used in this study). All of the images referred to lesions that were resected and examined under microscope in a histopathological lab. All the histopathological examinations were performed by one experienced team of pathologists, which could prevent from contingent false positives in the sample (we investigated this problem earlier and found few misdiagnosed cases by inexperienced specialists). This is important, since the histopathological examination is the ground truth which affects assignment of the class labels. The images were collected with resolution of 2272x1704 by Minolta Z5 digital camera with an extra dermoscopy extension and then JPEG-compressed as 3x8-bit RGB components. No magnification details were available. On the dermoscopic images no pre-processing tasks took place. This was due to the lack of rough artifacts (black borders, hairs, droplets of immersion fluid, etc.), but also to eliminate any bias on the final wavelet base selection. To be capable of performing the wavelet transform, the RGB representation (three integers per pixel) had to be translated into an indexed representation. This was done by linear, monotonic color mapping. Data set *A* was used to prepare two derived data sets of the same size by (recursively) averaging neighbor pixel values in 2x2 blocks: Images in the data set *A*2 had resolution of 1136x852. Images in the data set *A*4 had resolution of 568x426. Since each iteration of the wavelet decomposition downscales the input image by a factor of 2 both in rows and columns, to allow for three wavelet iterations the 568x426 set was padded with zeros (rows 427 and 428). Data set *B* was downloaded from the public database *PH*^2^ of the Automatic computer-based Diagnosis system for Dermoscopy Images (ADDI) in Portugal \[[@pone.0211318.ref020]\]. The *PH*^2^ database contains 200 dermoscopic images of melanocytic lesions (80 common nevi, 80 atypical nevi, and 40 melanomas) with medical annotations of expert dermatologists. The annotations refer to the dermoscopic criteria (Asymmetry: 0 = Fully Symmetry, 1 = Asymmetry in One Axis, 2-Fully Asymmetry; Pigment Network: AT = Atypical, T = Typical; Dots/Globules: A = Absent, AT = Atypical, T = Typical; Streaks: A = Absent, P = Present; Regression Areas: A = Absent, P = Present; Blue Whitish Veil: A = Absent, P = Present; Colors: 1 = White, 2 = Red, 3 = Light-Brown, 4 = Dark-Brown, 5 = Blue-Gray, 6 = Black), and clinical/histological diagnosis (0 = Common Nevus, 1 = Atypical Nevus, 2 = Melanoma). For most of the images binary masks from lesion segmentation is available. The *PH*^2^ images were taken during follow-up examinations at the Dermatology Service of Hospital Pedro Hispano (Matosinhos, Portugal) through Tuebinger Mole Analyzer with magnification of 20x. They are 3x8-bit RGB color images with a resolution of 768x560 pixels. The image size was cropped centrally to 760x552 pixels for the sake of the aforementioned wavelet recursive transformation. Since data set *A* contained only melanoma and displastic (atypical) lesions, we limited data set *B* to similar (not identical) medical cases. Finally data set *B* had the following structure: 40 melanomas (33 from clinial + histopathological, 7 lesions from only clinical examination), 77 common/dysplastic nevi (5 from clinial + histopathological, 72 lesions from only clinical examination). Altogether in *B* 113 lesion images were analyzed. In order to directly compare the efficiency of the SVM classifiers between data set A and B, the latter was additionally resized to resolutions A/A2/A4 and a series of experiments with the mixed data set A + B was done. Clinical statistics show that the incidence rate of melanoma is up to 5% (on average) within all 'suspicious' pigment lesions \[[@pone.0211318.ref021]\]. This means that the melanoma class is the minority class and is under-represented compared to the benign class. This status holds in the clinic, whereas in the melanoma CAD scientists attempt either to balance the number of cases from both classes (data set *A* is about this status) or to introduce sub- or over-sampling techniques \[[@pone.0211318.ref022]\] (this is required in *B*). In this work we took for both the data sets *A* and *B* the whole statistics for the minority class (melanoma) and experimented with different sub-sampling ratios \[[@pone.0211318.ref023]\] of the benign lesions. As in \[[@pone.0211318.ref008]\] for the data set *A* no major change in the classification performance (\<5% for *B*) was observed. This estimation allows us to conclude that the data imbalance problem in our experiment does not impact noticeably. In fact, in *B* the class ratio is 1:2, which is not a dramatic imbalance. Disregarding the way how imbalance is removed/controlled in this or (generally) other ML experiments, the problem in the clinic still holds, because a question arises how majority cases are picked out for the data set. It follows from our experience that usually those images are selected which visually fits in very well with the set of benign (dysplatic) cases. This may, however, seriously impact 'production' classifiers. We conclude that bias comes preferably from the 'clinical' source and not our data statistics or procedures. Wavelet features {#sec005} ---------------- Skin is texture that manifests repetitive patterns, pigment network and different structures (globules, streaks, etc.). This reality can be analyzed in a range of frequencies and spatial scales, which goes beyond pure Fourier analysis. In this section we introduce wavelet features that are used in this work. The Discrete Wavelet Transform (DWT) decomposes a signal into a coarse (average) signal and a signal of details. This can be interpreted, after Mallat \[[@pone.0211318.ref024]\], in terms of lowpass (L) and highpass filters (H). The subsequent levels of DWT operate recursively on the lowpass (scaling) part of the output. Products of the transform are downsampled (decimated) by two at each level. In the wavelet packet transform (WPT), the filtering operations are also applied to the signal of details. This results in a substantial amount of subbands in the output and makes the time (space) vs. frequency analysis much finer than in DWT. Multiresolution analysis with WPT can reveal scale-based features of the signal, which provides more precise information than other signal analysis techniques. Our working hypothesis is that WPT suits well to determine distinctive features characterizing the class of the dermoscopy image. Images are two-dimensional signals so one iteration of the Mallat filtering algorithm produces four sub-signals which can be considered as LL, LH, HL and HH filters after 1D wavelet transform on the rows and then on the columns. Those filters can be interpreted as decimated sub-images. We used three iterations of WPT. Since all wavelet filters are downsampled (by a factor of 2) after each run of the wavelet transform (in rows and columns separately), four subimages were produced. In [Fig 1](#pone.0211318.g001){ref-type="fig"} a summary image with all the used stages of decomposition is shown. It is evident that all channels of decomposition are used (WPT). ![The wavelet packet transform (WPT) decomposition.](pone.0211318.g001){#pone.0211318.g001} In one iteration four filters were produced, so in three turns we had 1 + 4 + 16 = 21 different transformation branches. In each branch we produced a list of twelve features. As features we took simple measures based on the energy content of the (sub)images \[[@pone.0211318.ref011], [@pone.0211318.ref012]\]: 1. *e*~*i*~---energies, 2. *e*~*i*~/*e*~*max*~---maximum energy ratios, 3. *e*~*i*~/Σ*e*~*k*~, *k* ≠ *i*---fractional energy ratios, where energy *e*~*i*~, *i* = 1, 2, 3, 4 was defined as a sum of absolute values of the pixels. Our feature vector was composed of 21 × 12 = 252 components. Such feature extraction was carried out in different wavelet bases: orthogonal wavelets: Haar, Daubechies, Symlets, Coiflets, bi-orthogonal wavelets, and reverse bi-orthogonal wavelets. Those wavelet bases possess certain characteristics which however is not discussed in this article. SVM learning {#sec006} ------------ The Support Vector Machine (SVM) learning method finds a hyperplane that separates data points from two distinct classes in the way that it maximizes the margin between the two classes \[[@pone.0211318.ref025], [@pone.0211318.ref026]\]. The mathematical formulation of the problem \[[@pone.0211318.ref027]\] involves finding *β* and *b* such that for all the data points (*x*~*i*~, *y*~*i*~) it holds: $$\begin{array}{r} {f\left( x \right) = x^{\prime}\beta + b = 0,\; y_{i}f\left( x_{i} \right) \geq 1} \\ \end{array}$$ where the latter equality is accomplished for those data points that are (called) support vectors (*y*~*i*~ *f*(*x*~*i*~) = 1). In a case the two classes are not separable by a hyperplane SVM introduces a soft margin i.e. it releases the condition *y*~*i*~ *f*(*x*~*i*~) ≥ 1 − *ξ*~*i*~. The optimization problem formulated by Lagrangian (in the simplest form of the *L*^1^ norm): $$\begin{array}{r} {L = \frac{1}{2}\beta^{\prime}\beta + C\sum\limits_{i}\xi_{i} - \sum\limits_{i}\alpha_{i}\left( y_{i}\left( x_{i}^{\prime}\beta + b \right) - \left( 1 - \xi_{i} \right) \right) - \sum\limits_{i}\mu_{i}\xi_{i}} \\ \end{array}$$ with the bounds: $$\begin{array}{r} {\beta = \sum\limits_{i}\alpha_{i}y_{i}x_{i},\;\sum\limits_{i}\alpha_{i}y_{i} = 0,\;\alpha_{i} = C - \mu_{i},\;\alpha_{i},\mu_{i},\xi_{i} \geq 0} \\ \end{array}$$ is usually translated into an expression with the *α* coefficients only: $$\begin{array}{r} {max_{\alpha}\sum\limits_{i}\alpha_{i} - \frac{1}{2}\sum\limits_{i}\sum\limits_{j}\alpha_{i}\alpha_{j}y_{i}y_{j}x_{i}^{\prime}x_{j}} \\ \end{array}$$ with ∑~*i*~ *α*~*i*~ *y*~*i*~ = 0 and 0 ≤ *α*~*i*~ ≤ *C*, C being the regularization parameter. This concept is very successful in machine learning and data mining applications \[[@pone.0211318.ref028]\], especially when non-linear transformations of the primary data space (kernels) are introduced. The appropriate choice of kernel can map primary features to a larger dimensional space where the classes are (better) separable. Four kernels are well known and widely used: the linear kernel $x_{i}^{T}x_{j}$, which is a linear combination of features, the polynomial kernel $\left( x_{i}^{T}x_{j} + r \right)^{p}$ with the parameters: p-degree of polynomial and r-offset, the gaussian (radial basis) kernel *exp*(−*γ* \* \|*x*~*i*~ − *x*~*j*~\|^2^), *γ* \> 0, and sigmoid (MLP) kernel. Quadratic programming which is required to solve the optimization problem defined above is broadly implemented in various statistical toolboxes and professional computer libraries e.g. in \[[@pone.0211318.ref029]\]. In this study we take advantage of both the LIBSVM library \[[@pone.0211318.ref029]\] and Matlab Statistical Toolbox \[[@pone.0211318.ref030]\]. We take into account the binary SVM classifiers with the linear, polynomial, and gaussian kernel. The SVM paradigm, as compared to standard back-propagated neural networks, does not have hidden parameters, and is a global approach not prone to stuck at local minima. Different bounds set on the margin can affect the (upper bound of) generalization error and can control overfitting. Bayesian optimization {#sec007} --------------------- The objective of every learning method is to minimize a cost function, which is usually the cross-validated loss. In the learning paradigm exploited in this work, i.e. C-SVM, we want to search for the best parameters of the kernel and the domain size C, as well as the number of support vectors in the resulting model. One framework to minimize a real-valued objective function within a certain box is Bayesian Optimization (BO) \[[@pone.0211318.ref031]\]. BO is a global optimization technique (starting points are of no importance) but there are cases where it does not always secure accurate results. If no a priori knowledge about parameters is available, the widest finite range should be covered. BO algorithm is time-consuming due to large computational burden every iteration, therefore for the sake of experimental requirements it should stop if the cross-validated error rate (objective function) drops below a certain level or (at least) after reaching a fixed number of iterations, or maximum run time. BO introduces the so called acquisition function, that determines which points of the space should be covered. 'Goodness' of the point is evaluated upon the 'amount' of improvement it provides for the optimization. There can be different methods to achieve this and our model of choice is based on the posterior distribution function *Q*. If at *x*~*best*~ the lowest value of the posterior mean is *μ*~*Q*~(*x*~*best*~) then the 'expected improvement' is: $$\begin{array}{r} {EI\left( x,Q \right) = E_{Q}\left\lbrack max\left( 0,\mu_{Q}\left( x_{best} \right) - f\left( x \right) \right) \right\rbrack} \\ \end{array}$$ To escape from local minima of the objective function, in order to balance sampling at points with high and low efficiency of the model, the acquisition function checks after each iteration the standard deviation of the posterior distribution function $\sigma_{Q}^{2}\left( x \right)$. If it drops below a defined level, the area is recognized as 'overexploited' and the algorithm modifies the kernel to raise the variance *σ*~*Q*~. This offsets the solution to a new not yet examined point. Such strategy both covers the whole space and concentrates on the best feasible points. Results and discussion {#sec008} ====================== Our setup was coded first with LIBSVM \[[@pone.0211318.ref029]\] and then recoded to Matlab \[[@pone.0211318.ref030]\]. The experiments ran on a 4-core i7 workstation with 32GB RAM. We cross-validated (10-fold CV) our SVM classifiers on the standardized (*N*(0, 1)) predictor vectors. Search for the best subset of features is usually a tradeoff between reducing the bias and computational cost, and preserving the classification performance. In this work we did not reduce our feature vector and kept all the wavelet features in the analysis. Our objective was to finally compare different learning methods, so we had to remove any bias coming from feature selection methods, which can affect the analyzed learning paradigms in a different way. This was discussed in \[[@pone.0211318.ref032]\] where feature selection algorithms (CFS, PCA, GSFS) reduced the complexity of the classification, but performance was highly dependent upon the classifier. Analysis and discussion of our results is divided into several stages: 1. Analysis of the absolute SVM classification performance for all the kernel functions for data set *A* (*A*/*A*2/*A*4) and data set *B* (original resolution), which includes Figs [2](#pone.0211318.g002){ref-type="fig"}--[5](#pone.0211318.g005){ref-type="fig"}. Our objective is to analyze an overall classification performance of the kernel functions. In [Fig 6](#pone.0211318.g006){ref-type="fig"} we analyze average AUC levels for the gaussian, linear, and polynomial kernels. We also present classification performance for all the kernel functions for data set *B* resized to resolutions *A*/*A*2/*A*4 (Figs [7](#pone.0211318.g007){ref-type="fig"}--[9](#pone.0211318.g009){ref-type="fig"}). Again, in [Fig 10](#pone.0211318.g010){ref-type="fig"} we analyze average AUC levels for the three kernels. 2. Analysis of the classification performance for a fixed kernel function for three image resolutions for data set *A* (Figs [11](#pone.0211318.g011){ref-type="fig"}--[13](#pone.0211318.g013){ref-type="fig"}) plus for the combined data set *A* + *B*(*resized*) (Figs [14](#pone.0211318.g014){ref-type="fig"}--[16](#pone.0211318.g016){ref-type="fig"}). Our objective here is to pick up those wavelet bases that show outstanding classification performance for individual data sets with specific image resolutions. Our second objective is to select those wavelet bases that preserve high classification efficiency through the presented spectrum of resolutions (resolution invariance). Analysis of the joint data set *A* + *B* cast light both on compatibility of the data sets and stability of the learning algorithm (independence on the data set). In [Fig 17](#pone.0211318.g017){ref-type="fig"} we show the number of support vectors for the three kernel functions used in the SVM classifier of data set A. This is also a test for quality and stability of features built upon certain wavelet numbers. 3. Analysis of our SVM results against our previous experiments with the wavelet features: ensembling (Figs [18](#pone.0211318.g018){ref-type="fig"}--[20](#pone.0211318.g020){ref-type="fig"}) and feed-forward neural networks classifiers (Figs [21](#pone.0211318.g021){ref-type="fig"}--[23](#pone.0211318.g023){ref-type="fig"}). Our objective is to show the property of the resolution invariant wavelet features for melanoma detection in the background of other tested classification paradigms. 4. Analysis of our SVM results with help of the melanoma CAD summary articles (Masood \[[@pone.0211318.ref003]\], Oliveira \[[@pone.0211318.ref004]\]) and some individual studies from the literature. ![AUC as a function of the wavelet number (see text) for data set *A* for the three SVM kernels optimized by the Bayesian search.\ Resolution is (2272x1704). To easily follow the trend the points belonging to the same kernel function are connected with a line.](pone.0211318.g002){#pone.0211318.g002} ![AUC as a function of the wavelet number for data set *A* for the three SVM kernels optimized by the Bayesian search.\ Resolution is A2(1136x852).](pone.0211318.g003){#pone.0211318.g003} ![AUC as a function of the wavelet number for data set *A* for the three SVM kernels optimized by the Bayesian search.\ Resolution is A4(568x426).](pone.0211318.g004){#pone.0211318.g004} ![AUC as a function of the wavelet number for data set *B* (original resolution 760x552) for the three SVM kernels optimized by the Bayesian search.](pone.0211318.g005){#pone.0211318.g005} ![Average AUC levels for the three kernel functions, for data set *A* (resolutions *A*, *A*2, *A*4).](pone.0211318.g006){#pone.0211318.g006} ![AUC as a function of the wavelet number for data set *B* for the three SVM kernels optimized by the Bayesian search.\ Resolution is resized to (2272x1704).](pone.0211318.g007){#pone.0211318.g007} ![AUC as a function of the wavelet number for data set *B* for the three SVM kernels optimized by the Bayesian search.\ Resolution is resized to (1136x852).](pone.0211318.g008){#pone.0211318.g008} ![AUC as a function of the wavelet number for data set *B* for the three SVM kernels optimized by the Bayesian search.\ Resolution is resized to (568x426).](pone.0211318.g009){#pone.0211318.g009} ![**Average AUC levels for the three kernel functions, for data set *B* resized to resolutions:** *A*, *A*2, *A*4.](pone.0211318.g010){#pone.0211318.g010} ![AUC for the SVM gaussian kernel as a function of the wavelet number for the data set *A* (resolutions: *A*, *A*2, *A*4).](pone.0211318.g011){#pone.0211318.g011} ![AUC for the SVM polynomial kernel as a function of the wavelet number for the data set *A* (resolutions: *A*, *A*2, *A*4).](pone.0211318.g012){#pone.0211318.g012} ![AUC for the SVM linear kernel as a function of the wavelet number for the data set *A* (resolutions: *A*, *A*2, *A*4).](pone.0211318.g013){#pone.0211318.g013} ![AUC for the SVM gaussian kernel as a function of the wavelet number for the combined data set *A* + *B* (*B* is resized to *A*, *A*2, *A*4).](pone.0211318.g014){#pone.0211318.g014} ![AUC for the SVM polynomial kernel as a function of the wavelet number for the combined data set *A* + *B* (*B* is resized to *A*, *A*2, *A*4).](pone.0211318.g015){#pone.0211318.g015} ![AUC for the SVM linear kernel as a function of the wavelet number for the data set *A* + *B* (*B* is resized to *A*, *A*2, *A*4).](pone.0211318.g016){#pone.0211318.g016} ![Number of support vectors for the three kernel functions used in the SVM classifier of data set *A*.](pone.0211318.g017){#pone.0211318.g017} ![AUC for the ensemble learning optimized by the 'recall' measure as a function of the wavelet number for data set *A* (resolutions: *A*, *A*2, *A*4).](pone.0211318.g018){#pone.0211318.g018} ![AUC for the ensembling learning as a function of the wavelet number for data set *B* resized to resolutions *A*, *A*2, *A*4.](pone.0211318.g019){#pone.0211318.g019} ![AUC for the ensembling learning optimized by the 'recall' measure as a function of the wavelet number for data set *A* + *B* (*B* is resized to *A*, *A*2, *A*4).](pone.0211318.g020){#pone.0211318.g020} ![AUC for the neural feed-forward classifier as a function of the topology and learning function (see text for the details).\ Data set is *A* in the native *A* and reduced resolutions *A*2 and *A*4.](pone.0211318.g021){#pone.0211318.g021} ![AUC for the neural feed-forward classifier as a function of the topology and learning function (see text for the details).\ Data base is *B* in the native resolution 760x552.](pone.0211318.g022){#pone.0211318.g022} ![AUC for the neural feed-forward classifier as a function of the topology and learning function.\ Data base is *A* + *B* (*B* is resized to the native and reduced resolutions of the data base *A*).](pone.0211318.g023){#pone.0211318.g023} Ad.1 {#sec009} ---- Full results of our experiments on data set *A*/*A*2/*A*4 are presented in Figs [2](#pone.0211318.g002){ref-type="fig"}--[4](#pone.0211318.g004){ref-type="fig"} for the three kernel functions and the three resolutions. In these and the follow-up figures the wavelet number denotes: 1 = Haar, 2-11 = Daubechies (db1-db10), 12-18 = symlets (sym2-sym8), 19-23 = coiflets (coif1-coif5), 24-38 = bi-orthogonal wavelets (Bior1.1, 1.3, 1.5, 2.2, 2.4, 2.6, 2.8, 3.1, 3.3, 3.5, 3.7, 3.9, 4.4, 5.5, 6.8), and 39-53 = reverse bi-orthogonal wavelets (Rbio1.1, 1.3, 1.5, 2.2, 2.4, 2.6, 2.8, 3.1, 3.3, 3.5, 3.7, 3.9, 4.4, 5.5, 6.8). The way how images, so 2D signals, are analized by the filter bank is another factor. In this work 1Dx1D filtering (separately along rows and columns) was performed, which followed Mallat \[[@pone.0211318.ref024]\] (in other approaches 2D filtering elements are potentially possible). For the SVM optimization we used the Sequential Minimal Optimizer \[[@pone.0211318.ref031]\] and initiated the alpha coefficients with zeros. To check for the optimization convergence the gradient tolerance was set for 10^−3^. The AUC values were averaged over 10-fold cross-validation (10 turns, 9 training subsets and 1 hold-out testing subset in each turn). The optimizatoin space that was checked included: - kernel functions: gaussian, polynomial, linear, - parameters of the kernel functions: for 'gaussian' it is the inverse variance parameter (*γ*), for 'polynomial' it is the order of the polynomial (d), - the maximum penalty imposed on margin-violating data points (*C*). We scanned the above-mentioned model parameters in a log step between 1e-3 and 1e6. The maximum number of iterations was fixed for 60 and the run time was left infinite. Figs [2](#pone.0211318.g002){ref-type="fig"}--[4](#pone.0211318.g004){ref-type="fig"} show AUC (the Area Under the ROC Curve) vs. the wavelet number for data sets *A*(2272x1704), *A*2(1136x852), and *A*4(568x426) for the binary SVM classifiers of melanoma vs. displastic lesion. This is a Bayesian search over an available space of the gaussian (marker is square), linear (triangle), and polynomial (circle) kernel. Clearly visible is an inphase sinusoidal fluctuation of AUC through all the resolutions. Since all the kernel functions fluctuate over the wavelet number in a similar way (the trend, not the absolute level) this must be the feature set that is responsible for the phenomenon, not the learning paradigm. Indeed, variability in the classification performance comes both from the individual wavelet base, and is also present within each wavelet family. The latter is obvious since wavelets from one wavelet family differ in scale and offset. The skin texture apparently 'prefers' some scales. The wavelets that were used in the analysis have different properties like smoothness, symmetry, or support. The wavelet base complies with some model constraints and has a diverse impact on the texture classification. Haar, Daubechies, symlets, and coiflets are orthogonal wavelets. Orthogonal wavelets are energy preserving and non-redundant, which follows the formula *X*^2^ = *A*^2^ + *D*^2^ (*X*-image, *A*-reconstructed image of approximation, and *D*-reconstructed image of details). After Mallat \[[@pone.0211318.ref024]\] wavelets can be represented by a filter bank. In this case this is a pair of a single scaling function and a wavelet. The bi-orthogonal wavelet filters, on the other hand, generate one scaling function and a wavelet for decomposition, and another pair for reconstruction. They have the property of perfect reconstruction i.e. *X* = *A* + *D* and a linear phase, which makes them more symmetric and compact (these properties are useful in signal analysis). For the sake of symmetry we can divide the selected wavelet bases into: symmetric functions (Haar and (reverse)-biorthogonal wavelets), nearly symmetric functions (symlets, coiflets), and asymetric functions (Daubechies). Taking into account this variety of properties we conclude that they are individual wavelets, not wavelet families that make the SVM classification of melanoma dermoscopy images more (or less) efficient, although certain wavelet families have a higher number of 'efficient' wavelets than the others. Performance of the SVM classification on data set *B* for the three kernels is shown in [Fig 5](#pone.0211318.g005){ref-type="fig"}. Images from this data set have rather low resolution but (unexpectedly) they yield (average) AUC values about 10% higher. Two factors may play a role: Compression: Data set A is a collection of JPEG images compressed by a lossy algorithm. Rough estimation on the information content of an image from data set A is possible from its JPEG file size (assumed we skip all the headers from different stages of the compression). We have the following (average) file sizes: 1.35MB (*A*, 2272x1704), 0.59MB (*A*2, 1136x852), and 0.17MB (*A*4, 568x426). This yields: 2.79 bit/pix, 4.89 bit/pix, and 5.45 bit/pix from the highest to the poorest resolution, respectively. Data set B, in turn, has (standardized) resolution of 760x552 and the average file size is 1.33MB. This yields information capacity of about 24.26 bit/pix. Such an extremely high information content is due to its lossless file format, which is BMP. Even if we compare data set *B* only to the lowest resolution data set *A*4, the factor between them is about five. We know that the JPEG algorithm has a quantization stage which 'flattens' pixels values based on the JPEG quality/compression ratio. This may affect the image structure which translates to decreased sensitivity for patterns in the compressed image. Quantitative estimations between data sets *A* and *B* are not possible as far as (pre)processing steps in *B* are unknown. Class content: The existing melanoma pattern recognition methods (a brief comparison is done later in this article) are rather 'advanced' systems so are 'costly' in terms of feature extraction/selection and classification. This reflects the 'clinical' difficulty in discriminating melanoma from displastic (anomal) skin lesions. Data set *A* has two distinct classes: melanoma and dysplastic nevus while data set *B*, in turn, has class melanoma and a mixture of dysplastic (1/3) and common (2/3) nevus. While the benign class in *B* (common + dysplastic nevus) must be more widely distributed in the feature space than dysplastic cases alone (*A*), the 'distance' between the melanoma and this (mixed) class in *B* should be larger i.e. more efficient for classification, than in *A*. Those two factors may be, though qualitative, but reasonable explanation for the difference in the average AUC level between the data sets *A* and *B*. Since there are a few variables in our SVM experiments (the wavelet bases, different resolutions, and finally the three kernel functions) in order to interpret the results and draw straight-forward conclusions we will limit the number of factors and answer the question which kernel function produces the best average classification performance in terms of AUC. [Fig 6](#pone.0211318.g006){ref-type="fig"} has three sections in which we show an average level of AUC produced by the SVM classifiers with the appropriate kernel functions: gaussian (dotted line), polynomial (dash-dotted line), and linear (solid line) for the nominal *A* and two downgraded resolutions *A*2 and *A*4. The polynomial kernel function keeps its average level in the reduced image resolutions while the liner one looses classification efficiency considerably. An interesting finding is that the gaussian kernel rises continuedly towards lower resolutions. Beyond this point (568x426), when we go further with the resolution reduction the classification performance drops rapidly to a level of about 70%. For the polynomial and linear kernel we can see a drop in performance for the intermediate resolution (1136x852) and then they recover loss and even outrun a little bit in 568x426. This behavior is independent of the absolute AUC level, this happens also for the linear kernel whose performance beyond the nominal 2272x1704 is poor. In order to check whether those average levels are data set specific (*A*) or general, we resized images from the data set *B* to the resolutions of the data set *A*. This can be found in Figs [7](#pone.0211318.g007){ref-type="fig"}--[9](#pone.0211318.g009){ref-type="fig"}. Based on the numerical results of Figs [7](#pone.0211318.g007){ref-type="fig"}--[9](#pone.0211318.g009){ref-type="fig"} we could depict in [Fig 10](#pone.0211318.g010){ref-type="fig"} an average level of AUC produced by the SVM classifiers with the same kernel functions (gaussian, polynomial, and linear) for data set *B* resized to resultions *A*, *A*2, and *A*4. Again, the linear kernel has a decreased (average) level of AUC along with the reduced resolution. As in *A* the gaussian kernel performs better for (568x426), but slightly worse than the polynomial kernel for (1136x852). Taking into account statistical aspects of learning and possible performance drops for (1136x852) apparent in *A*, both Figs [6](#pone.0211318.g006){ref-type="fig"} and [10](#pone.0211318.g010){ref-type="fig"} look similar. From this analysis we can conclude the following: 1. kernel functions play a key role in the performance of the SVM classification. The poorest linear kernel (dot product of the feature vector components) cannot appropriately map features-related (noisy) classes of melanoma and benign cases. The polynomial kernel, which can translate the primary feature vector into more dimensional space (degree of the polynomial is a parameter of the model) performs better, but the gaussian kernel outperforms all the other. This is because it provides the feature vector mapping in a more nonlinear fashion through introducing cluster-like subspaces. Apparently this is suitable for the SVM classifiers for the melanoma detection. 2. wavelet features classified by SVM with different kernel functions have 'preferential' resolutions where they reach the most optimal classification efficiency. Our analysis on two data sets (*A*, *B*) show that for the gaussian kernel the most efficient space for the SVM classifier is this of 568x426. This observation is important since can affect the computational aspects of the melanoma classification. Low(er)-resolution image analysis would be potentially resource-saving without (noticeable) degradation of the classification performance. Ad.2 {#sec010} ---- In stage 1 we performed an analysis of the average SVM classification performance for all the kernel functions for two data sets: *A* and *B*. We concluded that tha gaussian kernel outperforms the polynomial and linear kernel especially towards lower image resolutions and that both data sets yield compatible results. In stage 2 we focus on the gaussian kernel function and analyze the absolute classification performance (in terms of AUC) for the three image resolutions (2272x1704, 1136x852, 568x426) in *A* and the combined data set *A* + *B*. Our first objective here is to pick up the wavelet bases that show outstanding classification performance for the individual data sets with specific image resolutions. The second objective is to select those wavelet bases that preserve high classification efficiency through the presented spectrum of resolutions (resolution invariance). Analysis of the joint data set *A* + *B* should prove both compatibility of the data sets and stability of the learning algorithm (independence on the data set). Although our analysis is focused on the gaussian kernel (*A*: [Fig 11](#pone.0211318.g011){ref-type="fig"}, *A* + *B*: [Fig 14](#pone.0211318.g014){ref-type="fig"}), we present results for the other two kernel functions as well (*A*: Figs [12](#pone.0211318.g012){ref-type="fig"} and [13](#pone.0211318.g013){ref-type="fig"}, *A* + *B*: Figs [15](#pone.0211318.g015){ref-type="fig"} and [16](#pone.0211318.g016){ref-type="fig"}). This should cast light on robustness and stability of the wavelet features. Below we present for the individual kernel functions some consolidated figures with the three image resolutions. In each resolution we select a set (we arbitrarily choose n = 6) of wavelet bases that show an outstanding AUC classification performance. In [Table 1](#pone.0211318.t001){ref-type="table"} we call it 'Condition 1'. In 'Condition 2' we require that the wavelet number has the highest AUC. 10.1371/journal.pone.0211318.t001 ###### Wavelet numbers collected for the kernel functions (gaussian, polynomial, linear) satisfying some conditions for the AUC values (numbers in parentheses have equal AUC values). Condition 1: outstanding SVM classification efficiency. Condition 2: highest classification efficiency. Condition 3: high classification efficiency in all the resolutions (order of priority in A). Condition 4: high classification efficiency in all the resolutions (order of priority in A2). Condition 5: high classification efficiency in all the resolutions (order of priority in A4). ![](pone.0211318.t001){#pone.0211318.t001g} Data Set Resolution Condition 1 Condition 2 Condition 3 Condition 4 Condition 5 ------------ ------------ ----------------------------- ------------- ------------------------- ------------------------ ------------------------- gaussian A 2272x1704 (**46** 48) *26* (39 2 23) **46** **46** 48 *26* 2 39 12 A2 1136x852 *26* 43 (2 39 **46** 3) *26* *26* 2 **46** 39 19 30 A4 568x426 (**46** *26*) 25 47 48 42 **46** **46** *26* 48 2 19 12 B 760x552 3 (8 10 13 15 *26*) 3 \- \- \- A + B 2272x1704 41 (**46** 48) 45 40 23 41 41 **46** 48 40 6 9 A2 + B2 1136x852 24 (*26* 2) 39 25 42 24 24 2 *26* 25 12 47 A4 + B4 568x426 *26* (**46** 25) 12 (47 48) *26* *26* **46** 25 12 48 2 polynomial A 2272x1704 48 **46** 41 39 12 2 48 **46** 39 2 25 *26* 19 A2 1136x852 *26* 43 **46** 2 24 19 *26* *26* **46** 2 25 19 39 A4 568x426 *26* **46** 25 47 48 42 *26* *26* **46** 2 19 39 25 B 760x552 *26* (40 44 10 3 8) *26* \- \- \- A + B 2272x1704 **46** 45 41 48 53 15 **46** **46** 24 9 *26* 20 25 A2 + B2 1136x852 2 39 *26* 24 12 25 2 24 *26* 25 19 2 9 A4 + B4 568x426 **46** 25 42 49 *26* 24 **46** **46** 25 24 *26* 2 9 linear A 2272x1704 **46** 48 41 23 *26* 2 **46** **46** *26* 12 2 39 19 A2 1136x852 **46** 2 *26* 25 39 1 **46** **46** 2 *26* 39 19 18 A4 568x426 **46** *26* 47 25 14 12 **46** **46** *26* 19 18 12 39 B 760x552 (40 8) (25 37 47 13) 40 \- \- \- A + B 2272x1704 41 **46** 45 48 40 14 41 **46** 48 41 *26* 15 25 A2 + B2 1136x852 47 12 52 **46** 25 2 47 **46** 24 *26* 3 2 39 A4 + B4 568x426 **46** 25 *26* 43 48 1 **46** **46** *26* 2 3 48 43 As a next point we analyze high-AUC wavelet numbers almost constant in all the three resolutions. Our 'algorithm' to choose such cases is: 1. step 1: for *i* ∈ {*wavelet numbers*} calculate *SUM*(*i*) = ∑~*j*=1,2,3~ *AUC*~*i*~(*resolution*~*j*~) 2. step 2: choose n (n = 6) consecutive numbers: $i = \text{argmax}_{i}SUM_{i}$ Conditions 3,4,5 in [Table 1](#pone.0211318.t001){ref-type="table"} select wavelet numbers that show high efficiency in all the resolutions (assumed a data set and a kernel function). In such a set the order of priority of the wavelet numbers depends on the resolution preference of the analyzed dermoscopy images. In 'Condition 3' preferential is resolution 2272x1704, in 'Condition 4' 1136x852, and in 'Condition 5' 568x426. [Table 1](#pone.0211318.t001){ref-type="table"} shows the following: 1. Two most frequent wavelet numbers are marked: 46-in bold face, 26-in italic. 2. Wavelet numbers producing high AUC ('Condition 1), the best one included ('Condition 2'), belong for the most part to the Bi-orthogonal (24-38) and Reverse Bi-orthogonal (39-53) wavelet families, 3. wavelet numbers exhibiting good resolution invariance (keeping high AUC results in the three resolutions) come out in all the kernel functions ('Condition 3,4,5'). Some auxiliar information about how different wavelets 'fit' to the classifier can also be found in [Fig 17](#pone.0211318.g017){ref-type="fig"}. 'Good' wavelets from [Table 1](#pone.0211318.t001){ref-type="table"} (from the classification performance point of view) have a (preferably) small number of the support vectors. They are also less dispersed both between different kernel functions and different image resolutions. The perfect case *waveletnumber* = 46 shows very small values of the support vectors and almost perfect independence of the kernel function. This can be attributed to good quality and stability of features built upon wavelet number 46. Collective behavior of all the wavelet bases measured both in AUC and in the number of support vectors ([Fig 17](#pone.0211318.g017){ref-type="fig"}) casts light on how the SVM classifier 'prefers' the studied resolutions. It seems that the nominal resolution 2272x1704 and 568x426 are comparable in terms of the classification performance and stability, but 1136x852 is apparently worse (in [Fig 17](#pone.0211318.g017){ref-type="fig"} see the dispersion in the number of support vectors between the kernel functions). After analysis of [Table 1](#pone.0211318.t001){ref-type="table"} we come to some conclusions regarding the SVM classification of the dermoscopy images (class melanoma vs. class displastic/common nevus) with the presented wavelet features. First, we can see that efficient SVM learning of dermoscopy images is possible with some selected wavelet bases. Most of the high-performance wavelet bases belong to the symmetrical, bi-orthogonal or reverse bi-orthogonal wavelet families. Second, there are some wavelet numbers which keep the SVM classification performance on an almost stable level through a range of resolutions. The two wavelet bases that show best performance in all the studied resolutions are: 46 (RBio3.1), and 26 (Bior1.5). Ad.3 {#sec011} ---- In this section we compare our SVM results for data set *A* and *A* + *B* with the previous experiments with data set *A* concerning ensembling \[[@pone.0211318.ref007]\] and artificial neural networks \[[@pone.0211318.ref008]\]. Ensemble learning consists in learning of single weak and diverse learners \[[@pone.0211318.ref033]\]. To build an ensemble of models we started from an empty ensemble and added step-by-step the best model in the so called Out-of-Train procedure (OOT) (after Breiman's Out-Of-Bag technique \[[@pone.0211318.ref034]\]). Several models were trained on the training set and these models were compared by evaluating the prediction errors on the validation set. The best models became ensemble members until the ensemble got the desired size. Six model families were used for the training: LDA, Kernel Ridge Regression with a polynomial kernel *k*(*x*, *x*′) = (*a* + *x*.*x*′)^*b*^, MLP trained with the RPROP descent (with the changeable number of nodes), Perceptron trained with a second order gradient descent, C4.5 decision trees, and Matlab data trees (dtree). The wavelet features were learnt by an ensemble of models in this way, that the ensemble optimized (one by one) seven different quality measures: accuracy, precision, F1-score, fp rate, specificity, ber and recall. The final model was tested on a separate unseen set of data and the area under the ROC curve (AUC) was calculated. In those experiments we measured AUC and its error bar to draw some conclusions about stability of the models. Accuracy and recall were concluded to have the highest AUC levels plus modest error bars. In [Fig 18](#pone.0211318.g018){ref-type="fig"} we plot classification performance of the ensemble built on data set *A* by optimizing the recall quality measure. As we can see there are similar fluctuations of AUC over the wavelet number as in SVM ([Fig 11](#pone.0211318.g011){ref-type="fig"}) and the average level of AUC in ensembling is higher than this in SVM by about 6%. This difference in performance between ensembling and SVM should be attributed both to the different learning paradigms (ensembles may be more 'flexible' for optimization as far as high dimensional feature space is concerned), and (less) to the Bayesian optimization process, which is a statistical procedure. The most optimal wavelet base in our ensembling experiments (\[[@pone.0211318.ref007]\]) was reported wavelet number 46 i.e. reverse bi-orthogonal wavelet base Rbio3.1 and 26 i.e. bi-orthogonal wavelet base Bior1.5. The mixture set *A* + *B* ([Fig 20](#pone.0211318.g020){ref-type="fig"}) yields similar results, except for more wavelet numbers that satisfy the (almost) constant performance in all the three resolutions i.e. 46, 26, 24, 25, 39. We remember that those wavelet numbers are present in [Table 1](#pone.0211318.t001){ref-type="table"}. Artificial neural networks (ANN), in turn, is a black-box approach to the knowledge acquisition but can model complex relationships between inputs and outputs due to the nonlinear processing capabilities of its constituent neurons. ANN are widely used as classification systems for the melanoma CAD. In \[[@pone.0211318.ref008]\] we briefly commented on the existing works on ANN and contributed to the automatic classification of melanoma from dermoscopy images by ANN. In our study on data set *A* we used 252 input neurons that represented the same set of wavelet features as in this work, extracted for one wavelet base, which was RBio3.1 (wavelet number = 46). There are no no methodical studies in the literature how the ANN structure (hidden layers) affect the melanoma classification performance. After careful considerations and computational attempts on both CPU- and GPU-based parallel computing platforms we arbitrarily limited ourselves to a few combinations of hidden neurons grouped into one or two hidden layers: \[10\], \[20\], \[10-10\], \[20-10\], \[20-20\], which is not a considerably worse setup compared to the literature. The validation data was used to optimize (by mean square error) and update the weights in the backpropagation phase where all the layers had sigmoid activation functions. The testing set was used to calculate AUC on the unknown subset of dermoscopy images from data set *A*. We used several training algorithms with 'standard' base parameters: Levenberg-Marquardt *μ* = 0.001 (L1), Bayesian Regularization *μ* = 0.005 (L2), Broyden-Fletcher-Goldfarb-Shanno (L3), Conjugate Gradient with Powell-Beale restarts (L4), Fletcher-Powell Conjugate Gradient (L5), Polak-Ribiére Conjugate Gradient (L6), Gradient Descent *learn*\_*rate* = 0.01 (L7), Gradient Descent with Adaptive Learning *learn*\_*rate* = 0.01 (L8), Gradient Descent with Momentum *learn*\_*rate* = 0.01, *momentum* = 0.9 (L9), Variable Learning Rate Gradient Descent *learn*\_*rate* = 0.01, *momentum* = 0.9 (LA), One Step Secant (LB), Resilient backpropagation *learn*\_*rate* = 0.01, Δ = 0.07 (LC), and Scaled Conjugate Gradient (LD). Our objective was to find the best performing ANN for the classification of melanoma dermoscopy images under the assumption that only those aforementioned limited topologies are taken into account (starting with \[10\] hidden neuron on one hidden layer up to \[20 20\] hidden neurons on two layers). Beside classification performance our experiments had to show the computational efficiency (in terms of learning epochs required to gain a defined error level). For that we grouped results of our ANN setups in terms of pairs (*number*\_*ofepochs*, *setups*\_*finished*). As a necessary efficiency condition for further analysis we took the median of the number of epochs for the finished setups, which was 20 epochs. In [Fig 21](#pone.0211318.g021){ref-type="fig"} we plot numerical results of AUC for the resolutions 2272x1704, 1136x852, 568x426, for the five setups of the hidden layers and for the thirteen different back-propagation learning algorithms assumed that the number of epochs is below 20. Absence of L7-LA proves that methods based on (variations of) gradient decent converges very slowly (maximum number of epochs was sometimes even above 1000). Among the remaining cases there are no single winners in our optimization space, different combinations of the learning algorithms and topologies reach high AUC values. Each topology has a preferable learning algorithm assumed a resolution, and in reverse each learning algorithm has the most efficient topology. In any case one can optimally select a triple (*resolution*, *learning*\_*alg*, *hidden*\_*layers*) to reach a high AUC and match the experimental setup. In search for resolution invariance we scan through all the resolutions and see that especially L1 (Levenberg-Marquardt) shows extreme robustness for almost all the topologies e.g. (*A*, *L*1, \[10 10\]) = 0.97, (*A*2, *L*1, \[20\]) = 1.0, and (*A*4, *L*1, \[20\]) = 0.98. L6 (Polak-Ribiére Conjugate Gradient) proves to be the second best (*A*, *L*6, \[10 10\]) = 0.96, (*A*2, *L*6, \[20 10\]) = 0.99, and (*A*4, *L*6, \[10\]) = 0.93. The map of high AUC levels through all the learning functions shows us that images of 1136x852 pixels are the most suitable for classification with ANN (for SVM the preferable resolution was 568x426). Higher absolute results (although more sparse in the optimization space) are reached for data set *B* ([Fig 22](#pone.0211318.g022){ref-type="fig"}). The same applies to [Fig 23](#pone.0211318.g023){ref-type="fig"} for the joint data set *A* + *B*. Our ANN experiments for one particular wavelet base (wavelet number 46) agree qualitatively with the SVM results and the ensembling results in high classification efficiency and that the wavelet feature (46) is resolution-proof. Unfortunately no more wavelet bases were studied with ANN due to considerable computational burden. Ad.4 {#sec012} ---- A detailed/comprehensive comparison between various methods on the classification of melanoma and benign lesions cannot be done. While the reviews show the techniques and results of the published works, they cannot explain why and to which extent the classification performance varies with the data set, and with the method (the pre-processing step, feature extraction and selection included). This is because not enough information about all the crucial steps of the data collection and analysis are available from the papers, which was criticized by the reviews' authors. Sometimes even the presented methods have apparent drawbacks. In fact most of the results falling to the same machine learning paradigm seem to be comparable. This is because we usually compare some absolute performance measures i.e. levels (assumed the same measures) with no variance/dispersion of the results. There are only few works applying a method or the whole pattern recognition system to more than one data set. When different features are extracted/used, comparing performance of even identical machine learning methods is controversial. There is a lot of melanoma decision support systems, each having its own paradigm, a feature set, and machine learning approaches. Here we want to show our results against the background of existing experiments, based on the review of Masood \[[@pone.0211318.ref003]\], and more contemporary Oliveira \[[@pone.0211318.ref004]\] (crucial bibliography can be found there). Image acqusition, pre-processing, segmentation, feature extraction (and selection), and finally classification are fundamental steps in the non-invasive computer diagnosis of melanoma. Dermoscopic images, we use in this work, are the predominant part of the existing visual material recording skin lesions. This technique was briefly characterized in the introduction. The most important factor that makes this technique most useful in the clinical diagnosis is illumination and magnification of the moles. There are also drawbacks because artefacts such as hair or light reflections and liquid bubbles get magnified as well. For all that dermoscopy remains the most sensitive visual examinations that outperforms plain macroscopic images. Some more advanced techniques like trans-illumination, laser-based methods, ultrasonography or spectral systems are not widely used and are not significant in the mass melanoma screening. Melanoma image recognition can be reliable when trained on robust and comprehensive data. Unfortunately the only few existing standard data sets that can be used for training and testing in various melanoma CAD systems are relatively fresh and even if some statistics amount to some thousands of images, usually only a fraction of cases can be used. This is due to the variety of classes (non-pigment/pigment lesions, dermal/junctional/compound nevi, typical/dysplastic lesions, melanoma/other skin cancer cases etc.). For the melanoma binary classifiers, which have clinical significance, only melanoma and dysplastic cases should be selected. The most important existing ML experiments for melanoma usually operate on proprietary small data sets where the melanoma class has 20, 40, sometimes up to 100 cases. They are usually unavailable in public and those experiments are not compared against the standard data sets. Data set A, which we use, is a medium-size data set for melanoma classification, and its advantage is high resolution, which is not the case of most existing data sets. High resolution images allow for experiments with reduced, downgraded resolutions. In this context data set A is unique. To check performance of the trained SVM model one of the aforementioned public data sets was chosen---data set B described in the section 'Experimental Data'. Prior to segmentation some pre-processing steps can be done, if required. They can remove artifacts and enhance contrast. As it was explained in Motivation data sets A and B are taken without any methods enhancing image quality. Segmentation is an important step to extract regions of interest (ROI) in the image. Lesion border detection methods play an important role and this step has its own vast methodology and classification approaches (SVM included). As it is clearly visible from the literature, segmentation is rather unstable method, which can yield quite good results when fined tuned to the image local conditions. In fact even different images from the same data set can have slightly different magnification and/or illumination parameters (not to mention about the different artifacts), different skin complexions etc. so adjustment to one global (data set wide) parameter set is hard. Feature extraction methods for melanoma discrimination can be based on diverse grounds. They include clinical approaches (based on asymmetry, border, color, diameter, evolution or elevation of features), widely known as ABCD(E), 7-Point Checklist, Menzies etc., pattern analysis methods, where specific global or local patterns are visible in the dermoscopy image (homogeneous, starburst, globular, etc.), or various shape or color representation/variation features. Finally texture based features may be used, which determine some statistical texture descriptors e.g. the gray-level co-occurence matrix, (dis)similarity, entropy, momenta etc. Some popular methods to extract those aforementioned features is: thresholding, color discrimination, discontinuity-based segmentation (active contour), region-based segmentation (split and merge, morphological flooding), soft computing and fuzzy logic included. Other feature extraction methods found in the literature include filter-based approaches: Gabor (dyadic) filters, Fourier power spectrum, Gaussian derivative kernels or multichannel filtering. Some authors raise objections that a lot of different features are used to feed sophisticated classifiers of melanoma, but there is little discussion about the real meaning of those features and most of the studies do not report the details of their feature extraction procedures. Wavelet features which are exploited in this work probe the whole area of the image in various scales and frequency subsets, and are sensitive filters of localized frequencies in the skin texture. When this filtering method is done recursively through all the decomposition branches, the wavelet packet transform (WPT) is applied. Identification of malignant melanoma by wavelet analysis is less artifacts sensitive, independent of 'visual' conditions of the image thus less error-prone. References on experiments with the wavelet features are collected in the Introduction plus a brief summary is done in \[[@pone.0211318.ref003]\] and \[[@pone.0211318.ref004]\]. The wavelets features eliminate the segmentation step so are a promising alternative to most of the feature sets on the market for the melanoma detection. Results of our previous experiments and this work show that selected wavelet features are classification efficient and demonstrate robustness in a chosen spectrum of classifiers (ensembling, ANN, SVM). Feature selection is used to eliminate redundant, irrelevant or noisy features in order to reduce classifier complexity for better generalization and to make the learning process more time-, memory-, and storage-effective. From the known studies very few report the details of their feature selection procedures. The available ones treat feature selection as an optimization problem and apply greedy, heuristic, or genetic algorithms or some customized approaches. In our method we do not use any feature selection procedures, which was explained in the Motivation. The classification step to discriminate melanoma lesions vs. dysplastic or other lesions include different learning paradigms. This is usually divided into families of learning methods. They are: artificial neural networks (MLP, RBF), decision trees (CART, C4.5, etc.), Bayesyan networks, support vector machines (linear, polynomial and gaussian kernel) and ensembles (random forest, bagging, boosting, etc.). All those approaches are widely represented in the literature and summerized in the reviews. Below we restrict to SVM-based results. A brief theory to SVM methods were presented in the Introduction. This is a linear classifier, which can adapt to non-linearity of the model through kernel functions. Despite their inherent advantages, there exist relatively few studies investigating the utility of SVMs in melanoma recognition. SVMs are used as melanoma classifiers in two scenarios: - as an indirect classifier identifying one of the clinical (dermoscopic) features of melanoma (dots, streaks etc.), thus a feature extraction method, and - as a global image classifier based on various melanoma features (geometry, colors, wavelet filters etc.). The first function of SVM is represented in the five main articles. Celebi et al. \[[@pone.0211318.ref021]\] took a methodological approach and first determined the lesion border, then they extracted lesion shape and color/texture features. The feature set was optimized by various feature selection algorithms and the optimal feature subset size was ranked by an SVM classifier with the RBF kernel. Experiments on 564 images (15.6% melanoma, 84.4% benign) yielded a specificity of 92.34% and a sensitivity of 93.33% with a feature set of 18 out of all 437 features. The appropriate values of the kernel parameters, C (cost/penalty) and gamma (kernel width), were determined on a grid (*C* ∈ \[2^−5^, 2^15^\], and *γ* ∈ \[2^−15^, 2^3^\]). After the grid-search, the SVM classifier was trained with the optimal parameters (*C*, *γ*) = (2.0, 0.125). In \[[@pone.0211318.ref035]\] texture, border-based, and geometrical features were extracted from 289 dermoscopy images (114 malignant, 175 benign). The texture features were derived from a wavelet tranform, the border features were derived from spatial and frequency factors of the lesion border model, and the geometry features are derived from shape indexes. The most optimized features were selected by the gain-ratio method and the SVM classifier with the RGB kernel and optimized by SMO. Apart from SVM also random forest, logistic regression, and hidden naive Bayes models were tested. For the 23 features accuracy was 91.26% and ROC was 0.937. The authors concluded that the texture with (fewer) border and geometry features outperform pure texture information only. Based on color symmetry and texture analysis Abbas et al. \[[@pone.0211318.ref036]\] developed a system classifying the melanoma tumor patterns (reticular, globular, cobblestone, homogeneous, starburst, parallel ridge, multicomponent pattern). Such multicomponent patterns are well analyzed by learning algorithms that can assign each input pattern to multiple class labels simultaneously. In this work ML-SVM (RBF kernel), ML-kNN and multi-label ranking (AdaBoost.MC) were used, the latter being the best classifier for the problem. For the ML-SVM they reported classification sensitivity of 89.28%, specificity of 93.75% and AUC of 98.6% (as quoted by \[[@pone.0211318.ref004]\]) and concluded that the developed pattern classifier based on color--texture features agrees with dermatologists' perception). Mirzaalian \[[@pone.0211318.ref037]\] investigated visual streaks as one of the most important dermoscopic criteria for the diagnosis of malignant melanoma. The streaks in 99 dermoscopic images were identified by quaternion tubularness, which is sensitive to the radial components of streaks. Presence or absence of the regular and irregular streaks (measured by some flux-based descriptors for different number of bands, *K* ∈ \[5, 13\], and thicknesses Δ ∈ \[2, 8\]) was validated by the SVM classifier (RBF kernel), yielding AUC of 93% in the best model. Maglogiannis \[[@pone.0211318.ref038]\] was segmenting and counting dark dots and globules from dermoscopy images (108 benign + 100 melanoma lesions, 632 × 387 pixels) based on inverse non-linear diffusion. The optimal set of the dot features plus some region-based descriptors were classified (among others) by SVM with a polynomial kernel (degree = 5) in the malignant-non malignant lesion setup. For the best feature subset they achieved sensitivity of 88.46%, specificity of 92.31% and accuracy of 90.38%. The SVM classifier was the main supervised discriminator in the works mentioned below. They are widely commented in the reviews \[[@pone.0211318.ref003], [@pone.0211318.ref032], [@pone.0211318.ref039]\]. In \[[@pone.0211318.ref039]\] the data set contained 1619 lesion images: 600 common nevi, 144 dysplastic nevi, and 65 melanoma for the training set, and 690 common nevi, 80 dysplastic nevi, and 40 melanoma for the test set, all in resolution of 752x582 pixels. The authors analyzed basic, shape and color features (alltogether 107 features) with different normalization conditions and concluded that on both dichotomous (common nevi vs. dysplastic nevi + melanoma; melnoma vs. common + displastic nevi) and trichotomous tasks (correctly distinguishing all three classes) the ANNs performed similarly as logistic regression and SVMs, and better than the k-nearest neighbors and decision trees. For the dichotomous classification the optimal polynomial kernel was linear (*C* = 100) and yielded AUC of 0.92, and the optimal Gaussian RB kernel had inverse variance *γ* = 10^−4^ and *C* = 100 and yielded AUC of 0.97. \[[@pone.0211318.ref032]\] reviewed the state of the art of the visual features used for skin lesion classification in the stages: segmentation, border detection and color/texture processing. They also compared the performance of several classifiers and concluded that the SVM classifier (RBF kernel) seems to achieve higher performance in terms of sensitivity and specificity, followed by ADWAT and CART algorithms. Work \[[@pone.0211318.ref003]\] is a comparative assessment of skin cancer diagnostic models. They critically examined practices, problems, deficiencies and prospects from the image acquisition to classification of dermoscopic images. Some techniques are commented on the conditions that affect their feasibility. Single works that are referred to in the above mentioned reviews are briefly commented on below. Amico et al. \[[@pone.0211318.ref040]\] constructed various asymmetry measures (the so called Size Functions) to discriminate melanocytic lesions by SVM (kernel function was a third degree polynomial), which was implemented at a clinical level. They gained in the best tests sensitivity of 96.8% and specificity of 87.2%. \[[@pone.0211318.ref041]\] examined the melanoma discrimination capacity of some skin specialists (31 melanomas + 103 nevi) and the automatic data analysis for the melanoma early detection system (ADAM, 42 melanomas + 435 nevi). The ADAM system showed a slightly higher sensitivity (84%) and a lower specificity (72%), compared with the physicians. In \[[@pone.0211318.ref042]\] a local thresholding algorithm was proposed to extract separation, border, texture and color based features. Those features were used to construct a classifier based on SVM (malignant melanoma versus dysplastic nevus). The best accuracy of the RBF kernel was 91.84% (sensitivity = 91.30%, specificity = 91.87%). \[[@pone.0211318.ref043]\] developed a decision support system for the dermoscopic images by combining (by the Bayes theory) outputs from different classifiers (SVM, GML, kNN). On a collection of 358 dermoscopic images they used local color and texture-related features to achieve accuracy of 76%. In \[[@pone.0211318.ref044]\] a diagnostic system for dermatologists based on SVM models of melanoma was built. They investigated 14 geometry and color features. They tested four distinct kernels: polynomial, sigmoid, RBF and the k-MOD decreasing. The best SVM model with the k-MOD decreasing kernel function got 89% sensitivity and AUC = 76% using a set of 199 digital dermoscopic images (101 melanomas, 98 dysplastic). Work \[[@pone.0211318.ref045]\] aimed at determining the best system for the skin lesion classification out of one global (one of them was SVM, RBF kernel) and one bag-of-features classifier based on local features. The other objective was to compare the role of color and texture features in lesion classification and determine which set of features is more discriminative. They concluded that color features outperform texture features when used alone and that both methods achieve very good results on 176 dermoscopy images, i.e. sensitivity of 92% and specificity of 72% for the SVM global method against sensitivity = 100% and specificity = 75% for local methods. A classification system for four types of skin cancers: melanoma, basal cell carcinoma (BCC), actinic keratosis (AK), squamous cell carcinoma (SCC) was reported in \[[@pone.0211318.ref046]\]. The GLCM based texture features ware extracted from each of the four classes and given as input to a multi-class SVM. The accuracy of classification to the one of the four skin cancer classes was 81.43%. Masood et al. \[[@pone.0211318.ref047]\] analyzed 168 images (112 melanoma and 56 benign) rescaled to a resolution of 720x472. The lesion area was segmented using the Histogram based Fuzzy C Mean algorithm for Level Set initialization (H-FCM-LS). 45 features (15 GLCM, 5 GTDM-Gray Tone Difference Matrix, 15 FMI-WPT, and 10 Autoregressive) were extracted for each image The FMI-WPT features (Fuzzy Mutual Information based Wavelet Packet Transform (FMI-WPT) were extracted to level 3 and then some fuzzy sets were constructed. The number of features that maximized the classification performance was evaluated using fuzzy-set entropy based criterions. The constructed feature sets are used separately as well as in different combinations for feeding the classifier. Finally, Self-advisable Support Vector Machine (SA-SVM) was used for classification. SA-VM (both linear and kernel based) uses information generated from misclassified data in the training phase and thus, improves performance by transferring more information from training phase to the test phase. The results obtained by SA-SVM were significantly better than the results of traditional SVM. The SA-SVM diagnostic system achieved an overall accuracy of 90%, with sensitivity 91% and specificity 89%. In \[[@pone.0211318.ref049]\] features based on asymmetry, border irregularity, color variations and diameter were calculated from an illumination compensated segmented image after noise removal by iterative dilation. The SVM classifier was optimized by Sequential Minimal Optimization (SMO) for the parameters that varied in the ranges: *C* ∈ 0.1, 0.2, ..., 5 for soft margin and *σ* ∈ \[1, 11\]. The achieved sensitivity and specificity for the different sets of training and test data elements were respectively 87% and 94% (1-fold), and 90% and 75% (10-fold CV). Amerald et al. \[[@pone.0211318.ref050]\] proposed the high-level intuitive features (HLIFs) to model the ABCD criteria commonly used by dermatologists. They experimented with various data sets concluding that concatenating the proposed HLIFs with some low-level features increased classification accuracy. The SVM classifier with the linear kernel on the best data set achieved sensitivity 92.52%, specificity 96.22%, and accuracy 96.64%. Choudhury et al. \[[@pone.0211318.ref051]\] proposed a multilayer decomposition based textural and color features for SCC, BCC, melanoma and actinic keratosis. The normalized GLCM and histogram of oriented gradients (HOG) were used as textural feature descriptors, and were extracted from different layers of the image than color histograms. The base and detailed layers were decomposed by the weighted least squares (WLS) edge-preserving decomposition. These features were fed to multi-class SVM (MSVM) and extreme learning machine (ELM) for classification. An average accuracy of 94.18% for MSVM was better than 90.5% with ELM. Alquran et al. \[[@pone.0211318.ref052]\] collected asymmetry, border, color and diameter (ABCD) features extracted using the GLCM method after segmentation using thresholding. Those features were selected by principal component analysis (PCA), than the Total Dermoscopy Score (TDS)then the SVM classification was done. The achieved classification accuracy was 92.1%. The most recent work on SVM classified malicious and benign skin lesions by Ashtami \[[@pone.0211318.ref053]\] proposed new features characterizing border irregularities on both complete and incomplete lesions. Those features plus color and texture features (GLCM) were classified by a Support Vector Machine (SVM) model on two dermoscopy databases with images of two human races: caucasian an xanthous race. Their results, sensitivity 97.82%, specificity 75%, and accuracy 88.46% are slightly better than for their ANN classifier. There are few direct SVM systems based on wavelet features (and they are not fully comparable with our experiments): Surowka et al. \[[@pone.0211318.ref014]\] tested performance of several machine learning paradigms and methods: neural networks, support vector machines, and Attributional Calculus, applied to dermatoscopic images of potentially malignant pigmented lesions. The features were obtained using the multiresolution wavelet-based decomposition of the image. The SVM AUC was 0.937. In \[[@pone.0211318.ref048]\] 255 features were extracted from Wavelet Packet Transform (WPT) and reduced by Particle Swarm Optimization (PSO) to optimize the SVM classifier. In the pre-processing step the Wiener2, Gabor, median filtering and histogram equalization was introduced to the images. Each image was further segmented by edge detection and thresholding. All the tasks yielded classification sensitivity of 94.1%, specificity of 80.22%, and accuracy of 87.13% Takruri et al. \[[@pone.0211318.ref016]\] analyzed 448 digital and dermoscopic images (benign and malignant) from two sources and segmented them by k-means clustering. They derived wavelet (db4), curvelet and color based features both from grayscale and color original images which resulted in a sensitivity of 86.4% and 76.9% and specificity of 88.1% and 85.4% for the wavelets and curvelets respectively. The obtained results were discussed to be comparable to those obtained by dermatologists. In [Table 2](#pone.0211318.t002){ref-type="table"} we show our SVM numerical results (best wavelet numbers in each data set plus the parameters) against all the quoted literature results (Tables [3](#pone.0211318.t003){ref-type="table"} and [4](#pone.0211318.t004){ref-type="table"}) for the SVM learning paradigm of non-wavelet and wavelet features. We conclude that our results are reliable and compatible to the rest of the works. Unfortunately, due to too few melanoma classification systems utilizing both the wavelet features and the SVM classifiers plus unique conditions of the existing experiments, and last but not least, lack of details from the groups, makes a full quantitative comparison still unfeasible. 10.1371/journal.pone.0211318.t002 ###### Results of the melanoma SVM learning (this work). Legend: dn = dysplastic nevus, cn = common nevus. RBF kernel parameter: *γ*, polynomial kernel parameter: degree; \#features = 252, partition: 10-fold CV. Resolutions: *A* = *A* + *B* = 2272 x 1704, *B* = 760 x 552, *A*2 = *A*2 + *B*2 = 1136 x 852, *A*4 = *A*4 + *B*4 = 568 x 426. ![](pone.0211318.t002){#pone.0211318.t002g} Data set \#images \#class A (melanoma) \#class B wavelet number wavelet base kernel param. C AUC \#supp.vect. ------------ ---------- ---------------------- --------------- ---------------- -------------- --------------- ---------- ------- -------------- gaussian A 185 102 83 dn 46 Rbio3.1 4.68e-09 9.81e+06 0.870 72 A2 185 102 83 dn 26 Bior1.5 2.41e-04 9.69e+06 0.859 81 A4 185 102 83 dn 46 Rbio3.1 3.72e-06 4.76e+04 0.859 60 B 113 40 73 (cn + dn) 3 Db2 1.61e-03 1.54e+06 0.932 68 A + B 298 142 156 (cn + dn) 41 Rbio1.5 6.98e-04 6.96e+02 0.881 120 A2 + B2 298 142 156 (cn + dn) 24 Bior1.1 4.25e-03 2.36e+04 0.877 177 A4 + B4 298 142 156 (cn + dn) 26 Bior1.5 5.03e-03 1.64e+04 0.854 218 polynomial A 185 102 83 dn 48 Rbio3.5 3 9.53e+06 0.870 73 A2 185 102 83 dn 26 Bior1.5 3 5.78e+05 0.854 83 A4 185 102 83 dn 26 Bior1.5 4 6.50e+05 0.859 102 B 113 40 73 (cn + dn) 26 Bior1.5 3 1.53e+00 0.940 50 A + B 298 142 156 (cn + dn) 46 Rbio3.1 6 7.21e+03 0.868 92 A2 + B2 298 142 156 (cn + dn) 2 Db1 4 2.56e+05 0.871 125 A4 + B4 298 142 156 (cn + dn) 46 Rbio3.1 2 2.33e-01 0.868 138 linear A 185 102 83 dn 46 Rbio3.1 \- 2.63e+04 0.870 73 A2 185 102 83 dn 46 Rbio3.1 \- 5.48e+01 0.843 76 A4 185 102 83 dn 46 Rbio3.1 \- 8.28e+03 0.854 62 B 113 40 73 (cn + dn) 40 Rbio1.3 \- 5.58e+00 0.923 41 A + B 298 142 156 (cn + dn) 41 Rbio1.5 \- 9.39e+06 0.871 106 A2 + B2 298 142 156 (cn + dn) 47 Rbio3.3 \- 1.09e-05 0.858 118 A4 + B4 298 142 156 (cn + dn) 46 Rbio3.1 \- 1.16e-05 0.854 123 10.1371/journal.pone.0211318.t003 ###### The SVM-based melanoma machine learning experiments. Legend: n = nevus, cn = common nevus, dn = dysplastic nevus, otherwise = benign; parameters for the RBF kernel: (*C*, *γ*); parameter for the polynomial kernel: (degree).?/blank space = data unavailable. ![](pone.0211318.t003){#pone.0211318.t003g} ----------------------------------------------------------------------------------------------------------------------------------------------------------------- Ref\ \#images \#class A\ \#class B \#class C partition \#features\ kernel\ sens spec acc AUC (resolution) melanoma (best) (params) ----------------------------- ---------- ------------ ------------ ----------- -------------- ------------- ------------------- -------- -------- ------- ------- \[[@pone.0211318.ref039]\]\ 1619 65/40 600/690 cn 144/80 dn train/test 107 poly (1)\ 84,5%\ 88,5%\ 92%\ (752x582) RBF (100, 10^−4^) 92,1% 95% 97% \[[@pone.0211318.ref040]\]\ 977 25/25 500/427 n train/test poly (3) 96,4% 87,2% (magnif 16x) \[[@pone.0211318.ref041]\]\ 477 22/20/\ 218/217/\ train/test/\ 84% 72% (magnif 16x) 31 of 42 103 of 435 test doctors \[[@pone.0211318.ref042]\] 1041 69 972 n RBF (?, 7) 91.3% 91.9% 91.9% \[[@pone.0211318.ref014]\]\ 39 19 20 dn 231(10) RBF\ 97.4% (2272x1704) (512, 0.000244) \[[@pone.0211318.ref021]\] 564 88 476 10-fold CV 437(18) RBF (2, 0.125) 93.3% 92.3% \[[@pone.0211318.ref043]\] 358 134 106 118 dn RBF 75.7% \[[@pone.0211318.ref044]\] 199 101 98 dn 130train/\ 14 89% 64% 76% 69test \[[@pone.0211318.ref035]\] 289 114 175 $23$ RBF 91.3% 93.7% \[[@pone.0211318.ref037]\]\ 99 RBF 93% (768x512) \[[@pone.0211318.ref045]\] 176 RBF 92% 72% \[[@pone.0211318.ref036]\]\ 350 80/20%\ RBF (1,?) 89.3% 93.8% 98.6% (768x512) train/test ----------------------------------------------------------------------------------------------------------------------------------------------------------------- 10.1371/journal.pone.0211318.t004 ###### The SVM-based melanoma machine learning experiments (contd). Legend: dn = dysplastic nevus, otherwise = benign; parameters for the RBF kernel: (*C*, *γ*); BCC = Basal Cell Carcinoma, SCC = Squamos Cell Carcinoma, a.ker. = actinis keratosis. blank space = data unavailable. ![](pone.0211318.t004){#pone.0211318.t004g} -------------------------------------------------------------------------------------------------------------------------------------------------------------------- Ref\ \#images \#class A\ \#class B \#class C partition \#features\ kernel\ sens spec acc AUC (resolution) melanoma (best) (params) ----------------------------- ---------- ------------ -------------- ----------- ------------ ------------- ------------------------ -------- -------- ------- ----- \[[@pone.0211318.ref038]\]\ 208 100 108 poly (5) 88.5% 92.3% 90.4% (632×387) \[[@pone.0211318.ref046]\]\ 359 77 BCC(84)\ 81.4% (150x112) a.ker.(101)\ SCC(101) \[[@pone.0211318.ref047]\]\ 168 112 56 84 train\ 45 91% 89% 90% (720x472) 84 test \[[@pone.0211318.ref048]\] 79 50 29 64 train/\ 255 RBF\ 94.1% 80.2% 87.1% 15 test\ (3525.0051, 0.0084732) 15 test \[[@pone.0211318.ref016]\] 448 93 + 142 121 + 92 0.8train/\ 52\ RBF 86.4%\ 88.1%\ 0.2test 55(25) 76.9% 85.4% \[[@pone.0211318.ref049]\] 146 108 38 1-fold CV\ RBF 87%\ 94%\ 10-fold CV 90% 75% \[[@pone.0211318.ref050]\] LOO linear 92.5% 96.2% 96.6% \[[@pone.0211318.ref051]\] 94.2% \[[@pone.0211318.ref052]\] 92.1% \[[@pone.0211318.ref053]\] 151 97.8% 75% 88.5% -------------------------------------------------------------------------------------------------------------------------------------------------------------------- Conclusion {#sec013} ========== Computer aided diagnostic systems are common. For the early detection of cutaneous melanoma they play a crucial role to support clinics and general practitioners. Our work contributes to this effort in the field of feature extraction. Appropriate features project the information from a (dermoscopy) image into a space where classes (malignant and benign lesions) are well separated. This step is critical for the performance of the classifier and the learning procedure. In this work we extracted features from the WPT wavelet transform for the sake of efficient SVM classification. We analyzed 53 wavelet bases to select those that are efficient for the SVM learning, and to probe the SVM parameter space. Our experiments were performed on two different data sets of unequal resolution, and biased with different illumination, representation, and compression. The SVM paradigm involves optimization of a convex cost function, therefore it is a 'global' classifier, less prone to overfitting. It is flexible in a high dimension feature space through an appropriate choice of kernel, is robust and simple. We tested the linear, polynomial and gaussian kernel and searched (Bayesian optimization) the kernel parameters (C, *γ*, degree of polynomial). From this analysis we concluded that: 1. Kernel functions play a key role in the performance of the SVM classification, the best kernel was the gaussian function. 2. Wavelet features classified by SVM with different kernel functions have 'preferential' resolutions where they reach the most optimal classification efficiency. Analysis on two data sets (A, B) shows that for the gaussian kernel such most efficient resolution is 568x426. For the melanoma classification lower image resolution would be potentially resource-saving without (noticeable) degradation of the classification performance. 3. We selected wavelet bases that performed the best for a certain resolution and SVM kernel function, and also such that keep high classification efficiency towards downgraded image resolutions (Rbio and Bior wavelet families, particularly Rbio3.1 and Bior1.5). Our findings are compatible with our previous machine learning experiments on the melanoma discrimination (ensembling and artificial neural networks) where high classification accuracy and stable behavior over a range of resolutions was observed as well. Regarding our SVM classification of melanoma against dysplastic nevus, our classification performance is in line with other research groups, however a detailed comparison with other experiments with different data sets, selection criteria, different pre-processing steps and finally different features is problematic. General outlook is that further analyses on various (large) public data sets of dermoscopic skin lesion images should be done to probe new learning paradigms and yield quantitative results contributing to the melanoma feature extraction. [^1]: **Competing Interests:**The authors have declared that no competing interests exist.

{ "pile_set_name": "PubMed Central" }

Introduction ============ Heat shock protein 90 (HSP90), a 90 kDa ATP-dependent molecular chaperone, is the most abundant intracellular protein in mammalian cells and is essential for protein folding, assembly, and degradation processes.[@b1-ott-8-1849],[@b2-ott-8-1849] It is overexpressed in response to a variety of physiological and environmental insults, allowing cells to survive potentially lethal conditions due to its cytoprotective functions.[@b3-ott-8-1849]--[@b5-ott-8-1849] In this regard, HSP90 has recently received much attention due to its overexpression in certain types of cancer,[@b6-ott-8-1849]--[@b9-ott-8-1849] where it is associated with a poor prognosis and contributes to resistance to chemotherapy and radiation.[@b10-ott-8-1849] Unlike other molecular chaperones that are involved in the primary folding of nascent polypeptides, HSP90 uses repeated cycles of client protein binding, hydrolysis of ATP, and interaction with its co-chaperones such as HSP70, Cdc37, HOP, p23, and Aha1 to control the stability and activity of hundreds of client proteins.[@b11-ott-8-1849],[@b12-ott-8-1849] Importantly, the majority of these client proteins are involved in critical signaling pathways necessary for cellular proliferation, cell cycle progression, and apoptosis, including steroid hormone receptors, kinases, transcriptions factors, human epidermal growth factor receptor 2 (HER2), Akt, mutant BRAF, and mutant p53 among others.[@b12-ott-8-1849]--[@b16-ott-8-1849] Consistent with this, HSP90 inhibition results in the simultaneous degradation of many of these clients (via the ubiquitin-proteasome pathway), leading to cell-specific growth arrest, apoptosis of cancer cells, and antitumor activity in preclinical models.[@b3-ott-8-1849],[@b17-ott-8-1849] Thus, HSP90 has evolved as an important molecular target in cancer therapy, and 16 different first- and second-generation HSP90 inhibitors have entered clinical testing ([Table 1](#t1-ott-8-1849){ref-type="table"}). The prototype HSP90 inhibitor geldanamycin provided proof-of-concept for HSP90 inhibition; however, geldanamycin and its derivatives (17-allylamino-17-demethoxygeldanamycin \[17-AAG\] and 17-dimethylaminoethylamino-17-demethoxygeldanamycin \[17-DMAG\]) could not be fully developed due to a number of safety and pharmacological limitations. Consequent efforts using a variety of different chemical scaffolds have led to the development of highly potent, second-generation, small molecule HSP90 inhibitors with improved pharmacological properties and safety profiles.[@b18-ott-8-1849] This review will summarize information on the pharmacology, preclinical activity, and current clinical development of ganetespib, a novel small molecule, second-generation HSP90 inhibitor developed by Synta Pharmaceuticals. Pharmacology and preclinical data ================================= The HSP90 chaperone consists of an amino (N) terminal, a carboxy (C) terminal, and a middle (M) domain. The N-terminal contains the ATP-binding pocket, and both the N and C termini have a drug-binding site. The M domain mostly participates in forming active ATPase and acts as a docking site for clients and co-chaperones.[@b19-ott-8-1849],[@b20-ott-8-1849] Similar to many other HSP90 inhibitors, ganetespib also acts by binding to the N-domain ATP-binding pocket of HSP90.[@b21-ott-8-1849] The first-generation inhibitors were a semi-synthetic derivative of the natural product, geldanamycin, and included 17-AAG (17-Allyl-17-Demethoxygeldanamycin \[Tanespimycin\]), 17-DMAG (17-desmethoxy-17-N,N-dimethylaminoethylaminogeldanamycin \[alvespimycin\]), IPI-504/17-allylamino-17-demethoxygeldanamycin hydroquinone hydrochloride (Retaspimycin), and IPI-493/17-desmethoxy-17-amino geldanamycin ([Table 1](#t1-ott-8-1849){ref-type="table"}).[@b22-ott-8-1849] While the first-generation inhibitors provided proof-of-concept, they failed to advance further clinically due to various issues including poor pharmacological properties, adverse toxicity profiles, or suboptimal efficacy. The most clinically significant off-target toxicity with the geldanamycin derivatives was hepatotoxicity, which has largely been attributed to the presence of a benzoquinone moiety.[@b22-ott-8-1849] Although 17-DMAG lacked the solubility issues observed with 17-AAG, the toxicities reported with 17-DMAG, including liver, ocular, and cardiac toxicities, ultimately led to the cessation of its further clinical development.[@b23-ott-8-1849],[@b24-ott-8-1849] IPI-504 was designed to be more water soluble than both 17-AAG and 17-DMAG. Although the clinical activity of IPI-504 was promising in both non-small cell lung cancer (NSCLC)[@b25-ott-8-1849] and gastrointestinal stromal tumors (GIST),[@b26-ott-8-1849] there were treatment-related deaths due to hepatotoxicity in the Phase III RING (Retaspimycin in GIST) trial of this HSP90 inhibitor.[@b27-ott-8-1849] This prompted mandatory dose reductions of IPI-504 in the ongoing HER2-positive metastatic breast cancer trial, which ultimately may have led to the lack of efficacy seen with this inhibitor in that trial.[@b27-ott-8-1849] In contrast, the second-generation HSP90 inhibitors generally encompass the resorcinol moiety of radicicol or are purine derivatives; the exception to this classification is the SNX-522 agent, which does not fit among these designations ([Table 1](#t1-ott-8-1849){ref-type="table"}).[@b12-ott-8-1849],[@b23-ott-8-1849] Notably, this group of inhibitors lack the hepatoxicity noted with the first-generation inhibitors due to the absence of the benzoquinone moiety. Ganetespib (\[5-\[2,4-Dihydroxy-5-(1-methylethyl)phenyl\]-4-(1-methyl-1H-indol-5-yl)-2,4-dihydro-\[1,2,4\]triazol-3-one\]) is a resorcinol-based, non-geldanamycin, synthetic small molecule second-generation HSP90 inhibitor (molecular weight =364.4), similar to NYP-AUY922, AT-13387, and KW-2478. In differentiation to the other HSP90 inhibitors within this subgroup, ganetespib contains a triazolone moiety ([Figure 1](#f1-ott-8-1849){ref-type="fig"}, [Table 1](#t1-ott-8-1849){ref-type="table"}).[@b28-ott-8-1849] In vitro, ganetespib exhibits potent cytotoxicity in a wide variety of hematological and solid tumor cell lines, including those that express mutant kinases (including BCR-ABL, FLT3, c-KIT, EGFR, and B-RAF) that confer resistance to small molecule tyrosine kinase inhibitors. The half maximum inhibitory concentration values calculated for ganetespib are in the low nanomolar range, and ganetespib is at least 20-fold more potent than 17-AAG.[@b28-ott-8-1849] Interestingly, ganetespib not only induces rapid degradation of known HSP90 client proteins but also exhibits sustained activity even with short exposure times.[@b29-ott-8-1849] The cytotoxicity of ganetespib in these cell lines is predominantly mediated via an irreversible commitment to apoptosis, likely following growth arrest and effects on the cell cycle.[@b29-ott-8-1849] Similarly, ganetespib has potent antitumor activity in vivo as demonstrated by significant growth inhibition and/or regression in solid tumor and hematological xenograft models.[@b29-ott-8-1849] Evaluation of the microregional activity of ganetespib demonstrates efficient distribution throughout the tumor tissue, including the hypoxic regions \>150 µm from the microvasculature, resulting in sustained inhibition of proliferation and induction of apoptosis throughout the tumors.[@b28-ott-8-1849] Ganetespib also exhibits preferential tumor retention compared with normal tissues with a half-life of over 58 hours in NSCLC xenograft models vs 3 hours in the plasma and 5--6 hours in normal liver and lung tissues.[@b30-ott-8-1849] With regards to its pharmacological profile, as discussed earlier, ganetespib lacks the dose-limiting hepatotoxicity reported with the geldanamycin analogs. Additionally, ganetespib lacks the ocular toxicity that has been reported with NVP-AUY922 and SNX-522. The mechanism of this ocular toxicity is hypothesized to be related to photoreceptor degeneration due to retinal drug distribution and retention, which has been seen with NVP-AUY922 exposure in rodent models. In contrast, ganetespib does not accumulate in the rat eye and is rapidly eliminated from the retinal tissues.[@b31-ott-8-1849],[@b32-ott-8-1849] Taken together, these promising results suggest that ganetespib may be more potent in its antitumor activity compared with first-generation inhibitors and has an optimal safety profile that predicts for a superior therapeutic index. This has provided a compelling rationale to further develop this agent clinically. Single-agent Phase I/II trials ============================== An open-label dose-escalation study of single-agent ganetespib given intravenously in advanced solid tumors established the maximum tolerated dose at 216 mg/m^2^ given weekly for 3 out of 4 weeks. The recommended Phase 2 dose was hence established as 200 mg/m^2^. Dose-limiting toxicities (DLTs) included one grade 3 amylase elevation at 150 mg/m^2^, one grade 3 diarrhea, and two cases of asthenia (grade 3 and grade 4) at 259 mg/m^2^. Overall, the most common adverse events (AEs) were gastrointestinal in nature and fatigue, predominantly of grade 1 and 2 severity, and easily manageable.[@b33-ott-8-1849] Partial response (PR) was noted in a patient with metastatic colorectal cancer, and disease stabilization (SD) was noted in 23 of the 44 evaluable patients, including 1 patient each with NSCLC and GIST, whose tumors harbored the BRAF G469A and PDGFRA^D842V^ exon 18 mutations, respectively. The most common AEs in another Phase I trial of ganetespib in advanced solid tumors, administered twice weekly for 3 out of 4 weeks, were also diarrhea and fatigue. Two DLTs of elevated transaminases were noted at 10 mg/m^2^ and 144 mg/m^2^, respectively.[@b34-ott-8-1849] In this trial, a durable PR was noted in a patient with metastatic melanoma, and durable SD was reported in two patients with NSCLC. Two Phase I trials have also been conducted with single-agent ganetespib in patients with hematologic malignancies. These studies determined the recommended Phase 2 dose to be 200 mg/m^2^ for the once weekly and 90 mg/m^2^ for the twice weekly regimens, respectively. Again, the most common AEs in these studies were also diarrhea and fatigue. With regards to DLTs, there were elevated liver enzymes in one patient in the once weekly trial, and hyperbilirubinemia, hyponatremia, QTc prolongation, and transaminitis for the twice weekly trial.[@b23-ott-8-1849],[@b35-ott-8-1849],[@b36-ott-8-1849] Notably, there were no reports of severe hepatotoxicity, ocular toxicity, or cardiac toxicity in these patients. Since many HSP90 client proteins are mutated, overexpressed and/or chimeric kinases important for tumor growth and survival, including HER2 in breast cancer, mutant EGFR or anaplastic lymphoma kinase (ALK) in lung cancer, mutant KIT in GIST, and mutated BRAF in melanoma, a number of Phase II trials of single-agent ganetespib in selected tumor types that express such oncoproteins have been undertaken to exploit the concept of "oncogene addiction". Supported by preclinical evidence that ganetespib has potent antitumor activity in different breast cancer subtypes, a Phase II trial of single-agent ganetespib at 200 mg/m^2^ intravenously (IV) weekly for 3 out of 4 weeks was conducted in patients with unselected metastatic breast cancer.[@b32-ott-8-1849] The most common toxicities in this trial were of grade 1 and 2 severity and included diarrhea, fatigue, nausea, and hypersensitivity reactions. While the study did not meet its prespecified criteria for overall response rate in this heavily pretreated group of patients, clinical activity was notable in patients with trastuzumab-refractory HER2+ and triple negative breast cancer (TNBC). Specifically, two patients achieved a PR, and four of the seven patients who achieved SD had estrogen receptor positive (ER+)/HER2+ disease.[@b32-ott-8-1849] Unlike HER2+ breast cancer or hormone receptor-positive breast cancer, TNBC lacks a unique molecular alteration that can be therapeutically targeted. Yet, clinical activity has been noted with ganetespib in this subset of patients with evidence of tumor shrinkage, specifically in patients with lung metastases.[@b32-ott-8-1849],[@b37-ott-8-1849] This latter observation is in line with findings from syngeneic mouse models of spontaneous and experimental metastases, where ganetespib suppressed lung colonization and tumor growth.[@b37-ott-8-1849] There are many molecular alterations for TNBC that have been identified but are not yet validated as therapeutic targets. Several of these abnormal proteins are known HSP90 client proteins, such as EGFR, hypoxia-inducible factor 1 (HIF-1α), and KIT among others, thus underscoring the therapeutic potential for ganetespib even in this aggressive subtype of breast cancer.[@b38-ott-8-1849] Another Phase II trial evaluated the activity and tolerability of ganetespib administered at 200 mg/m^2^ IV for 3 out of 4 weeks in previously treated patients with NSCLC enrolled in three cohorts: cohort A (mutant EGFR), cohort B (mutant KRAS), and cohort C (no EGFR or KRAS mutations). Of the 66 patients enrolled to cohort C, four had PR with all four cases being crizotinib-naive and harboring the ALK gene rearrangement. As expected, the most common AEs were diarrhea, fatigue, nausea, and anorexia.[@b39-ott-8-1849] A Phase II trial in patients with metastatic ocular melanoma is also evaluating progression-free survival with single-agent ganetespib administered at the same dosing schedule ([ClinicalTrials.gov](http://ClinicalTrials.gov) Identifier: NCT01200238). Another pilot window-of-opportunity study is evaluating changes in the expression of biomarkers due to ganetespib, which is administered twice weekly for 2 weeks prior to surgery in patients with head and neck cancers ([ClinicalTrials.gov](http://ClinicalTrials.gov) Identifier: NCT02334319). The biology behind combination trials ===================================== Contrary to expectations based on the presence of mutant KIT, an HSP90 client, and therefore amenable to HSP90 inhibitor therapy, a Phase II trial of single-agent weekly ganetespib reported limited efficacy in patients with GIST (12/23 evaluable patients had SD: 4 SD ≥16 weeks, 8 SD ≥8 weeks).[@b40-ott-8-1849] This negative result was considered to be related to a possible lack of a sufficient duration of inhibition of this client or its downstream pathways, which was observed both in preclinical models and in patient biopsies done as a part of this Phase II trial.[@b40-ott-8-1849] The limited success of HSP90 inhibitors in GIST and other client-protein-driven patient populations remains poorly understood but has certainly led to the emergence of the concept of rationally designed combination studies to enhance efficacy. These strategies include combinations with cytotoxics, other targeted therapies, and radiation and are briefly discussed in the following. HSP90 protects cells under conditions of stress, and, therefore, HSP90 inhibitors have the ability to sensitize cells to the toxic effects of chemotherapy and radiation therapy.[@b22-ott-8-1849] In support of this, preclinical data from various cell lines and xenograft models suggest additive or synergistic antitumor activity when HSP90 inhibitors are combined with various systemic cytotoxics including anthracyclines and taxanes.[@b22-ott-8-1849],[@b41-ott-8-1849],[@b42-ott-8-1849] The synergistic benefit observed by the addition of HSP90 inhibitors to taxanes is likely multifactorial, resulting from increased cytotoxicity and apoptosis, Akt inactivation and sensitization of the tumor cells to induction of apoptosis by a taxane, loss of pro-survival signaling, and exacerbation of mitotic catastrophe.[@b12-ott-8-1849],[@b41-ott-8-1849],[@b43-ott-8-1849],[@b44-ott-8-1849] Indeed, the combination of ganetespib with paclitaxel or docetaxel enhances antitumor growth and is synergistic in TNBC, NSCLC, and ovarian cancer models.[@b37-ott-8-1849],[@b45-ott-8-1849],[@b46-ott-8-1849] Most importantly, this combination has a nonoverlapping toxicity profile. Ganetespib can also potentiate the cytotoxic activity of doxorubicin via enhanced DNA damage and mitotic arrest and thus can confer superior efficacy to doxorubicin-containing regimens.[@b37-ott-8-1849] Approximately 15%--25% of lung adenocarcinomas have tumor-associated KRAS mutations.[@b47-ott-8-1849] These mutations are negative predictors of response to currently available EGFR tyrosine kinase inhibitors, and these patients have unfavorable clinical outcomes.[@b48-ott-8-1849] Currently, there are no approved anti-KRAS-directed therapies. While KRAS by itself is not a known client protein, its downstream effector pathways, namely, the PI3K/AKT/mTOR and the RAF/MEK/ERK pathways, are shown to be sensitive to HSP90 inhibition. In fact, the combination of ganetespib with a dual PI3K/mTOR inhibitor demonstrated superior antitumor efficacy in a xenograft model, supporting further investigation of this dual-targeted approach.[@b49-ott-8-1849] On the other hand, ALK is a sensitive client of HSP90 inhibition, and ganetespib can overcome acquired resistance to the ALK inhibitor, crizotinib in ALK + NSCLC, both in xenograft models and in patients.[@b50-ott-8-1849] Furthermore, the L1196M mutant form of EML4-ALK, considered to be an acquired mutation conferring resistance to crizotinib, continues to be sensitive to other structurally different ALK tyrosine kinase inhibitors and HSP90 inhibitors like 17AAG and ganetespib in both cell lines and the clinical setting.[@b50-ott-8-1849],[@b51-ott-8-1849] Similarly, mutant BRAF is a very sensitive HSP90 client, and ganetespib was more potent than single-agent vemurafenib in BRAF-driven melanoma cell lines.[@b52-ott-8-1849] Since BRAF mutations lead to dysregulation of the RAF/MEK/ERK signaling axis, a number of trials are evaluating the combination of BRAF inhibitors and MEK inhibitors not only to improve efficacy but also to overcome acquired resistance. Along the same lines, the combination of ganetespib with vemurafenib or selective MEK inhibitors has shown synergistic activity in xenograft models. In a preclinical study, ganetespib in combination with the MEK inhibitor TAK-733 also caused significant tumor regressions in vemurafenib-resistant xenograft tumors, thus providing a rationale for this combination to treat tumors that have developed acquired resistance to vemurafenib.[@b52-ott-8-1849] HSP90 inhibitors can also enhance tumor cell sensitivity to radiation. HSP90 clients such as Akt and ErbB2 are thought to be associated with radioresponse, which in turn can protect against radiation-induced cell death. Degradation of these proteins by HSP90 inhibition therefore enhances tumor cell death in many cell lines and xenograft models.[@b53-ott-8-1849] This combination can also increase apoptosis and enhance G2 arrest and hence is being evaluated in tumors such as rectal cancer.[@b54-ott-8-1849] HSP90 inhibitors may also have a significant role to play in hormone-receptor resistant breast cancer based on the role of HSP90 in regulating the post-translational folding of the estrogen and progesterone receptors.[@b55-ott-8-1849] In preclinical hormone receptor positive breast cancer models, ganetespib reverses endocrine resistance and reduces heterogeneity in the disease control achieved by hormonal therapies. This work has formed the basis for evaluating the combination of ganetespib with endocrine therapies such as fulvestrant, a complete estrogen receptor antagonist ([Table 2](#t2-ott-8-1849){ref-type="table"}). Bortezemib was the first therapeutic proteasome inhibitor that was approved in the United States by the Food and Drug Administration for the treatment of multiple myeloma and mantle cell lymphoma. Preclinical findings suggest that the combination of HSP90 inhibition with bortezemib can enhance bortezemib-triggered apoptosis and induce a prolonged intracellular accumulation of ubiquinated proteins credited to the synergistic suppression of chymotryptic activity of the 20S proteasome.[@b56-ott-8-1849],[@b57-ott-8-1849] This has been the basis for combination trials of various HSP90 inhibitors, (including the ongoing ganetespib studies) with bortezemib in patients with multiple myeloma.[@b58-ott-8-1849],[@b59-ott-8-1849] Together, these compelling preclinical studies have formed the rationale for many Phase I/II clinical trials in combinations with ganetespib as a way to optimize efficacy and delay or limit acquired resistance. Combination Phase I/II trials ============================= [Table 2](#t2-ott-8-1849){ref-type="table"} lists several ongoing combination trials of ganetespib with cytotoxic agents such as taxanes and doxorubicin, radiation, fulvestrant, and other targeted agents such as sirolimus, crizotinib, Ziv-Aflibercept, and bortezemib for several tumor types. Some of these combination studies have reported results and are discussed in detail here. The combination of docetaxel and ganetespib is in advanced clinical testing in NSCLC based on the results from the Phase IIb GALAXY-I (Study of Ganetespib + Docetaxel in Advanced NSCLC) trial.[@b60-ott-8-1849] This Phase II trial randomized 252 stage III/IV NSCLC patients to docetaxel alone (75 mg/m^2^ on day 1 of a 3 week cycle; n=127) or the same dose of docetaxel and ganetespib (150 mg/m^2^ on days 1 and 15; n=125). There was a nonsignificant improvement in overall survival (OS) in the ganetespib arm vs the control arm (9.8 months vs 7.4 months; hazard ratio \[HR\] =0.82; *P*=0.082); there was also a trend toward improved PFS (4.5 months for ganetespib vs 3.2 months for the control group; HR =0.84; *P*=0.038). Importantly, in a prespecified group of patients enrolled in the trial more than 6 months after diagnosis (n=176), there was a significant 4.3 month OS advantage with ganetespib compared with the control of docetaxel alone (10.7 months vs 6.4 months; HR= 0.61; *P*=0.0093); PFS was also significantly improved in the treatment arm (5.4 months vs 3.4 months; HR= 0.61; *P*=0.0041). For the combination, the most common AE across all grades was diarrhea (48%), which was well managed with the use of over-the-counter antidiarrheals. The most frequently reported grade 3--4 AEs for ganetespib vs the control group were neutropenia (37% vs 38%), febrile neutropenia (11% vs 2%), and anemia (8% vs 2%).[@b60-ott-8-1849] The Phase III GALAXY-2 trial will compare second-line ganetespib and docetaxel vs docetaxel alone in advanced adenocarcinoma of the lung of those who are more than 6 months past a diagnosis of advanced disease. The target enrollment is 500 patients, and the primary endpoint is OS. Key secondary endpoints include OS in three subpopulations (mKRAS, elevated lactate dehydrogenase, and lactate dehydrogenase-5), PFS, overall response rate, and duration of response. Tumor tissue and blood samples will be collected for planned correlative studies ([ClinicalTrials.gov](http://ClinicalTrials.gov) Identifier: NCT01798485). Preliminary results from a Phase I trial of ganetespib plus paclitaxel and trastuzumab, a humanized monoclonal antibody against HER2, were recently presented at the San Antonio Breast Cancer Symposium.[@b61-ott-8-1849] There were no DLTs or grade 3 toxicities attributable to ganetespib. In this group of patients heavily pretreated with prior pertuzumab and ado-trastuzumab emtansine, one of six evaluable patients achieved a PR and four others had SD for a CBR of 60%. This trial is ongoing.[@b61-ott-8-1849] Given the exciting preclinical data of doxorubicin and ganetespib in xenograft models, this combination is being studied to determine a signal of efficacy in relapsed/refractory NSCLC ([Table 2](#t2-ott-8-1849){ref-type="table"}). Limitations and future directions ================================= As is the case with all targeted therapies, there is a pressing need to develop companion diagnostics or predictive biomarkers to better select patients who might derive the most benefit from ganetespib therapy. Due to the prolonged tumor retention of this inhibitor, it is evident that to optimize treatment, target modulation within the tumor itself is necessary. This will help expand the role of ganetespib and this class of agents beyond the tumors that are addicted to a sensitive client protein. Non-invasive \[18F\]-Fluorodeoxyglucose Positron Emission Tomography imaging has most recently been utilized as a means for measuring early response to ganetespib therapy; however, the utility of this approach has yet to be validated in large prospective trials.[@b12-ott-8-1849],[@b22-ott-8-1849],[@b23-ott-8-1849] Ongoing studies such as the window of opportunity trial in head and neck cancer, a Phase I trial in rectal cancer, and the ongoing Phase III GALAXY-2 trial in NSCLC are collecting tumor tissues for biomarker analyses that may also yield valuable information and guide further development of this agent. Conclusion ========== Altogether, our knowledge of the HSP90-complex machinery continues to grow. While as a class, HSP90 inhibitors have certainly emerged as an exciting multifaceted anticancer strategy, there are no currently approved HSP90 inhibitors for any cancer indication. The improved safety profile and superior efficacy of ganetespib compared with the other HSP90 inhibitors is attributed to its small molecular weight, increased lipophilicity, and absence of the benzoquinone moiety, allowing for a greater therapeutic index. Clinically, ganetespib has been shown to be very well tolerated with a notable lack of significant cardiac, liver, and ocular toxicity, which have been the limiting factors for further development of many of the other HSP90 inhibitors. While efficacy has been promising in NSCLC and HER2-positive metastatic breast cancer, many Phase I/II clinical trials are utilizing combinatorial approaches with other therapeutic agents to further improve the outcomes with ganetespib and also to expand its role in other disease types. Data from the ongoing Phase III trial in NSCLC along with the concurrent development of predictive biomarkers in other key studies have the potential to make a significant impact on the subsequent applications of ganetespib therapy and perhaps may lead to a path of regulatory approval for this promising HSP90 inhibitor. Komal Jhaveri would like to thank the Terri Brodeur Breast Cancer Foundation and all the patients and their families who participated in the clinical trials. **Disclosure** The authors report no conflicts of interest in this work. ![Chemical structure of ganetespib.](ott-8-1849Fig1){#f1-ott-8-1849} ###### First- and second-generation HSP90 inhibitors HSP90 inhibitors Class Pharmaceutical company ---------------------------------------- ------------------------- ---------------------------------------- **First-generation HSP90 inhibitors** Tanespimycin (17-AAG, KOS-953) Geldanamycin derivative Kosan Biosciences/Bristol-Myers-Squibb Alvespimycin (17-DMAG) Geldanamycin derivative Kosan Biosciences/Bristol-Myers-Squibb Retaspimycin (IPI-504) Geldanamycin derivative Infinity Pharmaceuticals IPI-493 Geldanamycin derivative Infinity Pharmaceuticals **Second-generation HSP90 inhibitors** CNF2024/BIIB 021 Purine Biogen Idec MPC-3100 Purine Myriad Pharmaceuticals/Myrexis Debio 0932 (CUDC-305) Purine-like DebioPharm PU-H71 Purine Samus Therapeutics Ganetespib (STA-9090) Resorcinol--Triazole Synta Pharmaceuticals NVP-AUY922 (VER-52269) Resorcinol--Isoxazole Novartis NVP-HSP990 Not reported Novartis KW-2478 Resorcinol Kyowa Hakko Kirin Pharma AT13387 Resorcinol Astex SNX-5422 Indazol-4-one Serenex/Pfizer DS-2248 Not reported Daiichi Sankyo Inc XL888 Not reported Exelixis **Abbreviation:** HSP90, heat shock protein 90. ###### Ongoing combination trials of ganetespib with other therapeutic agents ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Phase Disease Combination Dose and schedule Primary endpoint [ClinicalTrials.gov](http://ClinicalTrials.gov) Identifier --------------------- --------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------ -------------------------------- ------------------------------------------------------------ I Multiple myeloma Ganetespib +/−bortezemib Ganetespib IV days 1, 4, 8, 11 every 3 weeks\ MTD NCT01485835 Bortezomib IV or subcutaneous days 1, 4, 8, 11 every 3 weeks I/II Phase 1 includes multiple sarcoma subtypes and Phase 2 MPNST Ganetespib plus sirolimus Ganetespib 150 mg/m^2^ IV on days 1, 8, and 15 IV over 1 hour\ Toxicity/clinical benefit NCT02008877 Sirolimus 4 mg taken orally once daily on a continuous dosing schedule I Rectal cancer Ganetespib, capecitabine, and radiation Ganetespib 60 mg/m^2^ once weekly ×2 weeks prior to starting XRT and then on days 1, 8, 15 for cycle 1 and then on days 29 and 36 for cycle 2\ Response rate NCT01554969 Capecitabine at 825 mg/m^2^ twice daily for the entire duration of radiation therapy I HER2+ metastatic breast cancer Ganetespib, paclitaxel, and trastuzumab Ganetespib IV over 1 hour on days 1, 8, and 15\ MTD NCT02060253 Trastuzumab IV over 30 minutes, and paclitaxel IV over 1 hour on days 1, 8, 15, and 22 II Neoadjuvant breast cancer (I-SPY 2) Ganetespib + paclitaxel Ganetespib 150 mg/m^2^ once weekly for 3 out of 4 weeks\ pCR NCT01042379 Paclitaxel: 80 mg/m^2^ IV weekly ×12 weeks Randomized Phase II HR+ breast cancer Arm A: Fulvestrant\ Fulvestrant I M on day 1 and 15 of cycle 1, day 1 of cycle 2 and each subsequent cycle\ PFS NCT01560416 Arm B: Fulvestrant + ganetespib\ Ganetespib IV over 1 hour on days 1, 8, and 15 of each cycle Arm C: Crossover from Arm A to Arm B at progression I ALK positive lung cancers Crizotinib and ganetespib Ganetespib IV over 1 hour on days 1 and 8 of a 21-day cycle\ MTD and PFS NCT01579994 Crizotinib 250 mg orally twice daily continuously I/II Solid tumors/refractory small-cell lung cancer Ganetespib + doxorubicin Ganetespib 100 mg/m^2^ or 150 mg/m^2^ IV on days 1 and 8 of a 21-day cycle\ MTD/ORR NCT02261805 Doxorubicin 50 mg/m^2^ IV on day 1 of a 21-day cycle Phase I/II Malignant pleural mesothelioma (MESO-02) For Phase II: Ganetespib + cisplatin/pemetrexed Or Ganetespib + carboplatin/pemetrexed Ganetespib IV on day 1 and day 15 of each cycle\ DLT, MTD/PFS NCT01590160 Cisplatin 75 mg/m^2^, day 1 every 21 days Pemetrexed 500 mg/m^2^, day 1 every 21 days\ Carboplatin AUC5, day 1 every 21 days Pemetrexed 500 mg/m^2^, day 1 every 21 days I/randomized II Metastatic, p53-mutant, platinum-resistant ovarian cancer (GANNET53) For Phase II: Arm A: ganetespib plus paclitaxel\ Ganetespib IV once weekly for 3 out of 4 weeks\ PFS NCT02012192 Arm B: paclitaxel Paclitaxel 80 mg/m^2^, given IV once weekly for 3 out of 4 weeks I/II Recurrent ovarian, fallopian tube, or primary peritoneal cancer Paclitaxel and ganetespib Paclitaxel and ganetespib IV over 1 hour on days 1, 8, and 15 DLT, RP2D, PFS at 6 months, RR NCT01962948 I Refractory gastrointestinal carcinomas, non-squamous non-small- cell lung carcinomas, urothelial carcinomas, and sarcomas Ganetespib and Ziv-Aflibercept Ganetespib IV weekly on days 1, 8, and 15 of a 28-day cycle\ MTD, safety NCT02192541 Ziv-Aflibercept IV every 2 weeks, on days 1 and 15, of a 28-day cycle ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- **Abbreviations:** MTD, maximum tolerated dose; IV, intravenous; MPNST, malignant peripheral nerve sheath tumors; HER2+, human epidermal growth factor receptor 2 positive; DLT, dose limiting toxicity; RP2D, recommended Phase 2 dose; PFS, progression-free survival; RR, response rate; HR+, hormone-receptor positive; IM, intramuscular; ORR, objective response rate; pCR, pathologic complete response; ALK, anaplastic lymphoma kinase.

{ "pile_set_name": "PubMed Central" }

Introduction {#s1} ============ In most epithelial tissues, the Na^+^-K^+^-ATPase is located at the basolateral membrane and provides the driving force for a variety of Na^+^-dependent transport processes across the plasma membrane, which are critical for sustaining animal homeostasis (Schuurmans-Stekhoven and Bonting, [@B40]; Skou and Esmann, [@B42]). In marine teleost fish, the coordinated physiology of gill and intestinal epithelia is responsible for the mobilization of ingested seawater toward the submucosal vascular beds, and the extrusion of the excess of NaCl from ingested seawater, leading to a net balance of water intake (Evans, [@B21]; Evans et al., [@B22]). The key molecular components in the machinery for the elimination of NaCl against a dramatic electrochemical gradient reside in a specialized cell type, namely chloride cells, which are located in the branchial epithelia (Evans, [@B21]; Evans et al., [@B22]). Despite the complexity of iono- and osmoregulatory processes in marine fish, involved mechanisms are ultimately dependent on the exergonic activity of the Na^+^-K^+^-ATPase (Evans, [@B21]; Evans et al., [@B22]). Previous studies performed in our laboratory have disclosed differences in the biochemical properties of the Na^+^-K^+^-ATPase activities along the intestinal tract of the gilthead seabream, *Sparus aurata* (Díaz et al., [@B13]; Almansa et al., [@B1]; Dópido et al., [@B16]). On the basis of the differential sensitivities to ouabain, calcium and to ionic strength, it was concluded that α-subunit of Na^+^-K^+^-ATPase expression varied along the intestinal tract with α~1~ being ubiquitously expressed, but α~3~-subunit only observed in distal regions (Almansa et al., [@B1]). Later on, we found that differences in thermodynamic properties existed between intestinal segments from gilthead seabream and that these differences were likely due to different phospholipid microenvironment rather than to differential expression of α-subunits (Almansa et al., [@B2]). The influence of the lipid microenvironment composition on the thermodynamic and kinetic properties of the Na^+^-K^+^-ATPase is pervasive and has been recognized for long (Wheeler and Whittam, [@B49]; Klingenberg, [@B30]; Brasitus, [@B6]; Cornelius and Skou, [@B11]; Yeagle et al., [@B51]; Muriana et al., [@B34]; Gerbi et al., [@B25], [@B26]; Ventrella et al., [@B47]). From these early studies it soon became clear that the degree of unsaturation of membrane phospholipids, as well as membrane cholesterol contents, played modulatory roles on the kinetic features of the Na^+^-K^+^-ATPase, which were initially interpreted as secondary to differences in membrane fluidity. However, the precise molecular association between degree of phospholipid unsaturation, type of phospholipid, cholesterol-to-phospholipid relationships, on the membrane biophysics (and geometry) and dynamics of membrane-bound integral proteins is just starting to be unraveled. Studies using membrane models have demonstrated direct interactions between Na^+^-K^+^-ATPase α-subunits and specific membrane phospholipids and fatty acids, and that there exit interaction sites at the hydrophobic surfaces of both α-subunits and regulatory FXYD protein in the buried helixes exposed to the phospholipid bilayer (Arora et al., [@B3]; Cohen et al., [@B8]; Esmann and Marsh, [@B20]). At present, there exist solid evidence for direct and specific interactions of different phospholipids and cholesterol which affect both the stability and molecular activity of the Na^+^-K^+^-ATPase, with essential roles in physiological regulation linked to membrane lipid composition (Gerbi et al., [@B25], [@B26]; Yeagle et al., [@B51]; Crockett and Hazel, [@B12]; Else and Wu, [@B18]; Cornelius, [@B9]; Almansa et al., [@B2]; Esmann and Marsh, [@B20]; Cornelius et al., [@B10]). Unlike model membranes, where lipid composition is set to specific compositions, usually containing few molecular lipid species, membranes from living cells display an enormous biochemical complexity with \~10,000 different molecular species, likely depending on the organism and cell type, which, in turn, are subjected to continuous remodeling in response to intracellular and extracellular signals (Dowhan, [@B17]; Ernst et al., [@B19]). Important (and abundant) lipid molecules tightly associated to membrane physical properties in living cells are polyunsaturated fatty acids (LCPUFA). These fatty acids generally esterify glycerol backbone at *sn*-2 position in membrane phospholipids. Unlike saturates and monoenes, which can be synthesized by most vertebrates, including marine fish, LCPUFA cannot be produced (or are produced in very limited amounts) and their needs are mostly covered by their food intake. Amongst most important LCPUFA fatty acids, those belonging to the n-3 series (n-3 LCPUFA) and n-6 series (n-6 LCPUFA) are considered essential (Sargent et al., [@B39]; Díaz and Marín, [@B14]; Spector and Kim, [@B43]), and their deficiency are associated with a number of pathophysiological conditions and with the failure in adaptive responses (Bell et al., [@B4]; Gerbi et al., [@B25], [@B26], [@B24]; Sargent et al., [@B39]; Bogdanov et al., [@B5]; Russo, [@B38]; Díaz and Marín, [@B14]; Matsunari et al., [@B33]). In the present study we have aimed to determine the influence of n-3 LCPUFA-deficient diets in the lipid composition of isolated intestinal and branchial epithelial cells of the gilthead seabream *in vivo*, and the extent to which such lipid modifications impact the thermodynamic properties of Na^+^-K^+^-ATPase from epithelial cells. The outcomes indicate that branchial epithelium is much more resistant to diet induced-LCPUFA depletion, in terms of phospholipid remodeling, than the intestinal counterparts. To the best of our knowledge, this is the first study demonstrating that the lipid composition of epithelial cells differs depending on their histological origin. Also, the results disclose severe changes in the thermodynamic behavior of the Na^+^-K^+^-ATPase that correlate with altered lipid environment-protein interactions *in vivo*. Materials and methods {#s2} ===================== Animals, diets, and cell preparation ------------------------------------ Gilthead seabream (*S. aurata*) (average weight 400 g) were initially reared at the National Institute of Oceanography of Tenerife (Spain) in seawater (35°/~oo~) at 20°C and fed commercial fish pellets (CONTROL), containing 2.07% n-3 LCPUFA (dry weight basis, DWB). A subset of specimens were then reared in the Atlantic Ocean (Los Gigantes, Southwest coast of Tenerife, Spain) in parallel tanks containing 10 individuals each, and fed once a day with an amount of pellets equivalent to 1% their biomass of the control diet (CONTROL) or a LCPUFA-deficient diet (DEFICIENT) for 6 months. The experimental diet containing no n-3 LCPUFA and very low amounts of n-3 PUFA was based on olive oil as lipid source, and obtained from Stirling University\'s Fish Nutrition facility (UK). Detailed composition of diets is shown in Table [1](#T1){ref-type="table"}. After decapitation, intestinal segments were isolated from pyloric caeca and from the proximal and distal portions of the intestine as described previously (Almansa et al., [@B1]). For branchial epithelia, animals were first perfused through continuous ventricular injection of Ringer physiological solution to remove blood cells from gills. All tissues were rinsed in ice-cold Ringer solution, and submitted to epithelial cells isolation following the procedures described in detail in Dópido et al. ([@B16]). Isolated cells were immediately placed in the homogenization solution containing 50 mM sucrose, 20 mM TRIS, 1 mM EDTA, 1 mM of the protease inhibitor, phenylmethylsulfonyl fluoride (PMSF), pH 7.5 (adjusted with TRIS/HCl) and kept at 4°C. Once homogenized, samples were stored in 1 ml aliquots at −80°C until analysis. All experimental manipulations were carried out following the procedures approved by the ethics committee (Comité de Ética de la Investigación y de Bienestar Animal: CEIBA) from Universidad de La Laguna. ###### **Composition of control and PUFA-deficient diets**. **Control** **Deficient** ---------------------- ------------- --------------- Protein 40 38 Lipids 21 19 Celulose 3 5 Humidity 9 10 Ash 7 9 **Data in % Weight** **Fatty acids** **Control** **Deficient** 14:0 6.8 tr 16:0 18.9 11.8 16:1 7.6 1.4 18:0 3.7 2.9 18:1 12.7 72.1 18:2 n-6 4.3 9.6 18:3 n-3 1.4 0.6 18:4 n-3 2.9 -- 20:1 2.4 0.4 20:2 n-6 0.2 tr 20:4 n-6 0.7 tr 20:4 n-3 0.7 tr 20:5 n-3 12.6 tr 22:1 2.2 0.4 22:5 n-6 0.4 tr 22:5 n-3 1.4 tr 22:6 n-3 13.2 tr Monoenes 26.3 74.3 Saturates 30.7 14.7 n-3 PUFA 32.2 0.9 n-6 PUFA 5.9 9.8 n-3 LCPUFA 27.9 tr n-6 LCPUFA 1.1 tr n-3/n-6 5.5 0.1 *Data in % area. PUFA: polyunsaturated fatty acids (≥2 double bonds). LCPUFA: Long-chain polyunsaturated fatty acids (≥18 carbons and ≥2 double bonds)*. ATPase assays ------------- Na^+^-K^+^-ATPase activities were measured in triplicate as the difference in inorganic phosphate (Pi) production from ATP in the presence or absence of 1 mM ouabain, under steady-state conditions, as described previously (Ventrella et al., [@B48], [@B47]; Díaz et al., [@B13]; Almansa et al., [@B1]). Briefly, 25 μl of protein suspension, containing 50--100 μg protein, were added to test tubes containing 1 ml of incubation media (25 mM HEPES, 200 mM NaCl, 10 mM KCl, 5 mM MgCl2, and adjusted to pH 7.4 with TRIS) and allowed to preincubate for 5 min in a temperature-controlled water-jacketed chamber set at the desired temperature. The reaction was started by the addition of 50 μl (5 mM final concentration) of vanadate-free ATP and incubated for 10 min at different temperatures ranging from 4 to 50°C under continuous agitation. The amount of Pi produced was determined by the method of Forbush ([@B23]). Accordingly, at the end of the incubation period, the reaction was stopped by adding 1 ml of an ice-cold solution containing 2.8% ascorbic acid, 0.48%, ammonium heptamolybdate, 2.8% SDS and 0.48 M HCl, and the tubes were placed at 4°C for 10 min. Afterwards, the reaction was developed by incorporating 1.5 ml of 2% sodium citrate, 2% sodium-m-arsenite and 2% acetic, incubated at 37°C for another 10 min, and the absorbance read at 705 nm. Na~2~HPO~4~ was used as Pi standard for calibration curves. Corrections for unspecific ATP hydrolysis were made by measuring the amount of Pi liberated in the absence of protein samples at each temperature tested. Specific Na^+^-K^+^-ATPase activities were expressed as μmol Pi/mg prot.hr. Lipid analysis -------------- Gill and intestinal epithelial homogenates were also submitted to lipid analyses. For these purposes, cell collections were homogenized in chloroform/methanol (2:1 v/v) containing 0.01% of butylated hydroxytoluene as antioxidant, and then the organic solvent was evaporated under a stream of nitrogen. Lipid contents were determined gravimetrically. Lipid extracts were re-dissolved in chloroform/methanol (2:1 v/v) and stored in 1 mL glass vials in a nitrogen atmosphere free of O~2~ at −20°C until analyses. Lipid classes were separated by one-dimensional double development high performance thin layer chromatography (HPTLC) as described elsewhere (Olsen and Henderson, [@B35]; Martín et al., [@B32]) and quantified by scanning densitometry using a Shimadzu CS-9001PC dual wavelength flying spot scanner. The polar lipid fraction (PL) was separated from neutral lipids by silica sep-pak cartridges and then subjected to acid-catalyzed transmethylation to yield fatty acid methyl esters (FAME) (Christie, [@B7]). FAME were separated and quantified using a Shimadzu GC-14A gas chromatograph equipped with a flame ionization detector, an integrator and a fused silica capillary column Supelcowax TM 10 (30 m × 0.32 mm I.D.). Individual FAME were identified by reference to a multi-standard mixture which included authentic standards from marine fish (Supelco, Bellefonte, USA) and further confirmation of identity carried out by mass spectrometry when necessary. Statistics and calculations --------------------------- All results are expressed as means ± SEM for, at least, four different determinations. Experimental data were submitted to one-way ANOVA followed by Tukey\'s test or to Kruskall-Wallis analysis followed by Mann-Withney *U* test, were appropriate. Correlation and determination coefficients were obtained by Pearson\'s approach. Multivariate analyses were performed using Principal Component Analyses (PCA). Statistical calculations were performed using SPSS (v.15.0 SPSS Inc., Chicago). Estimations of regression equations (lineal and exponential) and related parameters were performed by non-linear regression analysis tools using *Sigma Plot* software (Jandel Scientific, San Rafael, CA). A *p*-value below 0.05 was considered to achieve statistical significance. Materials --------- Ouabain, sodium dithionite, vanadate-free Na~2~ATP, HEPES, EDTA, TRIS, type IV Collagenase and PMSF were purchased from Sigma-Aldrich (Biosigma, Spain). Dimethylsulfoxide and HPTLC plates were obtained from Merck (Germany). Silica sep-pak cartridges were supplied by Millipore (Milford, MA). All reagents were analytical grade. Results {#s3} ======= Lipid profiles in isolated gill and intestinal epithelial cells from control fish --------------------------------------------------------------------------------- Lipid composition of gill epithelia and isolated enterocytes obtained from animals fed control diets are shown in Table [2](#T2){ref-type="table"}. It can be observed that neutral and polar total lipid contents as well as individual lipid classes are rather homogeneous between intestinal segments. Only phospahtidylglicerol (PG, lowest in posterior intestine) and free fatty acids (FFA, lowest in anterior intestine) significantly differ between regions. However, it turns out that lipid profiles in epithelial cells from gill origin notably differ from those of enterocytes. Thus, gill epithelial cells contain lower amounts of phosphatidylcholine (PC), phosphatidylinositol (PI), phospahtidylglycerol (PG), phosphatidylserine (PS), phosphatidylethanolamine (PE), and as a consequence, total polar lipids (TPL), as well as higher levels of cholesterol (CHO), sterol esters (SE), free fatty acids (FFA) and resultant total neutral lipids (TNL) compared to enterocytes (Table [2](#T2){ref-type="table"}). ###### **Lipid classes and total polar lipid fatty acid composition of isolated epithelia from gilthead seabream reared under standard conditions**. **Pyloric caeca** **Anterior intestine** **Posterior intestine** **Gill epithelia** ------------------------ ------------------- ------------------------ ------------------------- -------------------- **LIPID CLASSES** LPC 0.27 ± 0.46 0.18 ± 0.21 0.27 ± 0.65 0.20 ± 0.29 SM 1.85 ± 0.56 2.08 ± 0.70 2.10 ± 0.86 1.04 ± 0.44 PC 19.38 ± 2.46a 17.49 ± 2.13a 17.77 ± 1.87a 4.21 ± 1.50b PS 3.36 ± 0.80a 3.23 ± 0.56a 2.81 ± 0.89ab 1.08 ± 0.46b PI 4.66 ± 0.57a 4.30 ± 0.77a 3.72 ± 0.70a 1.03 ± 0.21b PG 5.31 ± 0.58a 4.31 ± 0.95a 3.11 ± 0.78b 1.02 ± 0.77c PE 12.13 ± 2.14a 10.97 ± 1.86a 10.27 ± 2.54a 3.03 ± 2.10b TPL 46.96 ± 6.48a 42.56 ± 5.46a 42.02 ± 6.56a 11.61 ± 4.91b MAG 1.93 ± 1.14b 1.92 ± 1.12b 1.41 ± 1.24b 7.39 ± 3.82a DAG 0.00 ± 0.00 0.00 ± 0.00 0.27 ± 0.67 0.52 ± 0.74 CHO 18.95 ± 2.24b 18.42 ± 1.64b 19.46 ± 0.57b 25.83 ± 2.11a FFA 1.78 ± 1.83b 5.33 ± 3.64a 7.39 ± 4.53a 14.44 ± 6.08a TAG 23.24 ± 5.16b 24.48 ± 2.50b 23.16 ± 4.14b 33.46 ± 1.24a SE 5.57 ± 1.66 5.61 ± 2.81 6.06 ± 2.00 3.22 ± 0.36 TNL 53.04 ± 6.48b 57.44 ± 5.46b 57.98 ± 6.56b 84.86 ± 6.72a LT (μg/mgprot) 554 ± 60a 375 ± 49b 345 ± 53b 560 ± 80a **FATTY ACIDS** 14:0 1.07 ± 0.53b 0.44 ± 0.24b 0.32 ± 0.21b 2.05 ± 0.11a 16:0 21.99 ± 1.38 21.34 ± 1.09 21.05 ± 0.92 22.08 ± 0.75 16:1 0.77 ± 0.34b 0.45 ± 0.17b 0.57 ± 0.22b 3.73 ± 0.64a 18:0 10.31 ± 0.79b 11.04 ± 0.84ab 12.11 ± 0.23a 8.56 ± 0.45b 18:1 n-9 9.84 ± 0.62b 10.70 ± 0.27b 10.10 ± 0.87b 14.85 ± 0.21a 18:2 n-6 2.49 ± 0.05 2.49 ± 0.03 3.21 ± 0.40 2.85 ± 0.01 18:3 n-3 0.00 ± 0.00b 0.00 ± 0.00b 0.00 ± 0.00b 0.17 ± 0.00a 18:4 n-3 0.09 ± 0.08 0.12 ± 0.10 0.13 ± 0.11 0.17 ± 0.02 20:1 1.41 ± 0.14b 1.98 ± 0.40b 1.49 ± 0.13b 0.45 ± 0.16a 20:2 n-6 0.36 ± 0.07 0.30 ± 0.03 0.20 ± 0.23 0.00 ± 0.00 20:4 n-6 2.59 ± 0.27b 2.41 ± 0.18b 2.62 ± 0.25b 5.10 ± 0.43a 20:4 n-3 0.48 ± 0.04a 0.53 ± 0.06a 0.54 ± 0.03a 0.39 ± 0.05b 20:5 n-3 6.15 ± 1.55b 6.32 ± 1.94b 6.14 ± 1.13b 9.15 ± 0.38a 22:1 0.25 ± 0.05a 0.40 ± 0.27a 0.30 ± 0.16a 0.00 ± 0.00b 22:5 n-6 0.46 ± 0.04 0.56 ± 0.07 0.60 ± 0.05 0.57 ± 0.10 22:5 n-3 2.25 ± 0.36a 2.47 ± 0.23a 2.55 ± 0.21 2.53 ± 0.32 22:6 n-3 34.13 ± 2.17a 35.04 ± 0.79a 32.70 ± 1.50a 21.69 ± 0.33b **TOTALS AND INDEXES** Monoenes 12.29 ± 0.90b 13.53 ± 0.53b 12.47 ± 0.58b 20.19 ± 0.21a Saturates 33.69 ± 1.71 33.35 ± 1.72 34.03 ± 1.14 34.06 ± 0.67 n-6 6.03 ± 0.38b 5.98 ± 0.09b 6.86 ± 0.53b 8.85 ± 0.57a n-3 LCPUFA 43.03 ± 1.93a 44.38 ± 1.67a 41.95 ± 1.33a 33.76 ± 1.08b 18:1/n-3 LCPUFA 0.16 ± 0.00b 0.17 ± 0.01b 0.17 ± 0.01b 0.44 ± 0.01a UI 248.96 256.74 244.40 194.35 *Different letters in the same row state for statistical differences with p-values below 0.05. LPC, Lyso-phosphatidylcholine; SM, Sphingolipids; PC, Phosphatidylcholine; PS, Phosphatidylserine; PI, Phosphatidylinositol; PG, Phosphatidylglycerol; PE, Phosphatidylethanolamine; TPL, Total Polar Lipids; MAG, Monoacylcerides; DAG, Diacylcerides; CHO, Cholesterol; TAG, Triacylcerides; FFA, Free fatty acids; SE, Sterol esters; TNL, Total Neutral Lipids; TL, Total Lipids; UI, Unsaturation Index. With the exception of 18:1/n-3 LCPUFA ratio, UI and TL, all data are expressed as % of total lipid. TL is expressed as μg lipids/mg total protein*. Regarding fatty acids from polar lipids, significant differences between enterocytes were observed for stearic acid (18:0), linoleic acid (18:2n-6) and docosapentaenoic acid (DPA, 22:5n-6), which were higher in posterior intestine. However, when compared with gill epithelium significant differences in most fatty acids were observed. Thus, levels of palmitoleic acid (16:1), linoleic acid (18:2n6), linolenic acid (18:3n-3), and total monoenoic and total n-6 fatty acids were higher than in enterocytes, while contents of eicosatetraenoic acid (20:4n-3), eicosapentaenoic acid (20:5n-3), docosahexaenoic acid (22:6n-3) and total n-3 LCPUFA were notably lower than in the enterocytes within the same animals. Consequently, the 18:1/n-3 LCPUFA relationship, was significantly higher in gill epithelial cells. The results presented so far reveal that lipid profiles in gill epithelia substantially differ from that in enterocytes. This was further confirmed by multivariate analyses using the principal component analyses (PCA). The results of the factor scores plot shown in Figure [1](#F1){ref-type="fig"} indicate that the three populations of enterocytes were clustered together while the gill population was clearly segregated (Figure [1A](#F1){ref-type="fig"}). The variables explaining most variance (overall variance 76.90%) were 18:1n-9, 20:4n-6 (ARA, arachidonic acid), linolenic acid (18:3n-3), myristic acid (14:0), which were positively related to PC1, and DHA, stearic acid (18:0) and 20:1 which were negatively related to PC1 (Figure [1A](#F1){ref-type="fig"}). PC2 was positively correlated to three n-6 fatty acids (18:2n-6 or linoleic acid, 20:3n-6, and 22:5n-6) and negatively to palmitic acid (16:0). Indeed, differences in PC2, allowed a certain differentiation between enterocytes, with caecal and posterior enterocytes appearing as opposite clouds, and anterior enterocytes positioned between them (Figure [1B](#F1){ref-type="fig"}). ![**Multivariate analyses using PCA of fatty acids from total polar lipids in gilthead seabream epithelia. (A)** Scatterplot for Factor scores 1 and 2, and grouped to identify different epithelial sources in control animals. **(B)** Factor loadings for Principal Components 1 and 2 (PC1 and PC2) obtained for fatty acids from TPL in control animals. The fraction of total variance explained by each principal component is indicated in parentheses. **(C)** Factor loadings for Principal Components 1 and 2 (PC1 and PC2) obtained for fatty acids from TPL in enterocytes from Control and LCPUFA-deficient animals. As in **(B)**, the fraction of total variance explained by each principal component is indicated in parentheses. **(D,E)** Scatterplots for Factor scores 1 and 2 grouped to identify different diets **(D)**, and enterocyte sources **(E)**. For details see corresponding results section.](fphys-07-00589-g0001){#F1} Effects of n-3 PUFA-deficient diets on lipid profiles of isolated gill and intestinal epithelial cells ------------------------------------------------------------------------------------------------------ In the next experiment we aimed to determine the impact of dietary n-3 PUFA deficiency in the lipid composition of the four populations of epithelial cells. The results in Table [3](#T3){ref-type="table"} indicate that TPL were reduced by the treatment in enterocytes (at least in caecal and anterior enterocytes). This effect was due to general reductions of PC, PS, PI, and PE in deficient animals although differences only scarcely reached statistical significance. However, none of these trends were observed in gill epithelia. Regarding total neutral lipids, a generalized increase in the contents of SE, triacylglycerides (TAG) and TNL was observed in enterocytes from deficient animals irrespective of the intestinal origin. As in the case of TPL, none of these effects were observed in branchial cells. ###### **Effects of experimental diets on lipid classes and total polar lipid fatty acid composition of isolated epithelia**. **Pyloric Caeca** **Anterior Intestine** **Posterior Intestine** **Gill epithelia** ------------------------ ------------------- ----------------------------------------------- ------------------------- ----------------------------------------------- -------------- ----------------------------------------------- -------------- ---------------------------------------------- **LIPID CLASSES** LPC 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 1.86 ± 0.12 0.68 ± 0.29[^\*^](#TN1){ref-type="table-fn"} SM 1.18 ± 0.36 1.08 ± 1.01 0.67 ± 0.06 0.61 ± 0.14 0.80 ± 0.04 0.41 ± 0.21[^\*^](#TN1){ref-type="table-fn"} 1.52 ± 0.29 1.02 ± 0.38 PC 16.21 ± 5.36 10.31 ± 2.75 16.61 ± 0.10 11.30 ± 1.89[^\*^](#TN1){ref-type="table-fn"} 13.61 ± 3.50 11.40 ± 3.23 8.23 ± 0.70 8.71 ± 3.06 PS 2.12 ± 0.25 1.58 ± 0.31 2.24 ± 0.49 2.04 ± 0.10 3.07 ± 1.00 2.04 ± 0.34 3.16 ± 0.31 3.79 ± 0.38 PI 4.50 ± 1.18 3.32 ± 0.70 3.82 ± 0.12 2.85 ± 0.73 3.28 ± 0.64 3.36 ± 0.61 2.49 ± 0.24 2.13 ± 0.41 PG 5.06 ± 0.51 3.89 ± 0.67 4.59 ± 0.02 3.49 ± 1.05 2.27 ± 1.64 3.29 ± 0.65 2.54 ± 0.112 2.44 ± 0.23 PE 12.14 ± 1.57 7.43 ± 1.04[^\*^](#TN1){ref-type="table-fn"} 13.12 ± 0.30 10.97 ± 1.93 12.39 ± 2.51 13.30 ± 1.42 6.81 ± 0.43 6.11 ± 1.80 TPL 41.22 ± 8.18 27.63 ± 4.20[^\*^](#TN1){ref-type="table-fn"} 41.07 ± 0.21 31.28 ± 5.21[^\*^](#TN1){ref-type="table-fn"} 35.44 ± 7.40 33.84 ± 5.02 26.64 ± 0.46 24.88 ± 5.57 MAG 1.10 ± 0.82 1.32 ± 1.01 0.56 ± 0.05 0.15 ± 0.11[^\*^](#TN1){ref-type="table-fn"} 1.05 ± 0.43 0.69 ± 0.08 1.33 ± 0.44 1.38 ± 2.23 DAG 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.02 ± 0.03 0.00 ± 0.00 0.00 ± 0.00 0.35 ± 0.34 0.28 ± 0.66 CHO 28.90 ± 3.38 28.25 ± 3.18 21.37 ± 0.53 20.77 ± 1.62 26.46 ± 1.72 28.45 ± 0.84 28.52 ± 1.02 27.38 ± 3.78 FFA 8.72 ± 5.87 5.26 ± 2.75 12.66 ± 1.35 10.20 ± 3.03 17.25 ± 8.64 9.22 ± 1.23 10.21 ± 0.79 10.72 ± 3.20 TAG 17.18 ± 5.57 31.89 ± 5.62[^\*^](#TN1){ref-type="table-fn"} 21.06 ± 1.45 33.51 ± 3.00[^\*^](#TN1){ref-type="table-fn"} 16.64 ± 2.21 23.79 ± 4.27[^\*^](#TN1){ref-type="table-fn"} 21.78 ± 1.33 21.07 ± 3.65 SE 2.85 ± 0.08 5.62 ± 0.93[^\*^](#TN1){ref-type="table-fn"} 2.44 ± 0.13 4.03 ± 0.83[^\*^](#TN1){ref-type="table-fn"} 2.91 ± 1.41 3.82 ± 0.67 11.13 ± 0.31 11.29 ± 0.70 TNL 58.77 ± 8.18 72.36 ± 4.20[^\*^](#TN1){ref-type="table-fn"} 58.10 ± 0.63 68.71 ± 5.21[^\*^](#TN1){ref-type="table-fn"} 65.04 ± 7.29 65.98 ± 5.03 73.35 ± 0.46 72.12 ± 5.57 PHO/CHO 1.43 0.98 1.92 1.51 1.34 1.19 0.87 0.88 LT (μg/mgprot) 767 ± 90 1090 ± 65[^\*^](#TN1){ref-type="table-fn"} 865 ± 60 1230 ± 180[^\*^](#TN1){ref-type="table-fn"} 761 ± 90 998 ± 190[^\*^](#TN1){ref-type="table-fn"} 520 ± 20 518 ± 70 **FATTY ACIDS** 14:0 0.42 ± 0.12 0.60 ± 0.24 1.45 ± 1.45 0.44 ± 0.28 0.59 ± 0.08 0.41 ± 0.20 0.96 ± 0.39 0.68 ± 0.23 16:0 16.88 ± 0.55 14.59 ± 2.57 18.61 ± 5.12 12.10 ± 2.68 17.36 ± 1.19 11.63 ± 0.67[^\*^](#TN1){ref-type="table-fn"} 22.85 ± 3.92 22.65 ± 5.23 16:1 1.60 ± 0.32 1.78 ± 0.24 2.19 ± 1.03 1.19 ± 0.48 1.61 ± 0.23 1.18 ± 0.23 3.76 ± 0.50 2.19 ± 0.16[^\*^](#TN1){ref-type="table-fn"} 18:0 11.22 ± 1.25 9.88 ± 1.80 13.11 ± 0.64 10.63 ± 2.72 13.47 ± 1.18 10.15 ± 1.69[^\*^](#TN1){ref-type="table-fn"} 11.81 ± 2.01 12.11 ± 4.14 18:1n-9 9.07 ± 2.85 21.41 ± 6.72[^\*^](#TN1){ref-type="table-fn"} 9.98 ± 3.34 23.84 ± 9.39[^\*^](#TN1){ref-type="table-fn"} 6.86 ± 1.52 22.92 ± 5.35[^\*^](#TN1){ref-type="table-fn"} 14.29 ± 3.22 20.51 ± 4.55 18:1n-7 2.73 ± 0.17 2.71 ± 0.10 2.76 ± 0.11 2.72 ± 0.20 2.31 ± 0.22 2.60 ± 0.20 3.06 ± 0.19 2.74 ± 0.30 18:1n-5 0.10 ± 0.09 0.11 ± 0.10 0.00 ± 0.00 0.00 ± 0.00 0.06 ± 0.06 0.06 ± 0.11 0.00 ± 0.00 0.09 ± 0.16 18:2n-6 5.29 ± 0.28 7.43 ± 1.97 4.87 ± 0.78 8.32 ± 3.42 4.95 ± 0.68 7.37 ± 1.50 5.81 ± 0.58 6.87 ± 2.92 18:3n-3 0.32 ± 0.06 0.24 ± 0.04[^\*^](#TN1){ref-type="table-fn"} 0.44 ± 0.07 0.34 ± 0.05[^\*^](#TN1){ref-type="table-fn"} 0.30 ± 0.08 0.27 ± 0.07 0.38 ± 0.04 0.11 ± 0.18[^\*^](#TN1){ref-type="table-fn"} 18:4n-3 0.16 ± 0.08 0.20 ± 0.17 0.17 ± 0.07 0.10 ± 0.11 0.07 ± 0.06 0.14 ± 0.02 0.06 ± 0.11 0.05 ± 0.09 20:1 0.72 ± 0.06 1.05 ± 0.17 0.99 ± 0.13 0.92 ± 0.80 0.77 ± 0.06 1.52 ± 0.28[^\*^](#TN1){ref-type="table-fn"} 1.47 ± 0.65 1.23 ± 0.24 20:2n-6 0.47 ± 0.04 0.53 ± 0.14 0.43 ± 0.09 0.58 ± 0.24 0.20 ± 0.34 0.28 ± 0.49 0.49 ± 0.07 0.64 ± 0.16 20:4n-6 2.68 ± 0.48 1.33 ± 1.18 1.56 ± 1.36 1.48 ± 0.42 2.16 ± 0.15 1.52 ± 0.21[^\*^](#TN1){ref-type="table-fn"} 2.11 ± 0.96 1.78 ± 0.62 20:4n-3 0.63 ± 0.04 0.54 ± 0.16 0.60 ± 0.26 0.50 ± 0.24 0.56 ± 0.18 0.25 ± 0.26 0.10 ± 0.17 0.08 ± 0.13 20:5n-3 7.71 ± 0.65 6.10 ± 3.00 6.62 ± 2.28 5.92 ± 2.75 4.84 ± 1.56 4.37 ± 1.96 2.88 ± 1.08 1.37 ± 1.20 22:1 0.05 ± 0.04 0.10 ± 0.09 0.09 ± 0.09 0.14 ± 0.14 0.02 ± 0.04 0.05 ± 0.08 0.41 ± 0.72 0.00 ± 0.00 22:5n-6 0.70 ± 0.05 0.46 ± 0.15 0.79 ± 0.11 0.68 ± 0.40 0.96 ± 0.10 0.57 ± 0.22[^\*^](#TN1){ref-type="table-fn"} 0.55 ± 0.14 0.24 ± 0.21[^\*^](#TN1){ref-type="table-fn"} 22:5n-3 3.08 ± 0.50 2.32 ± 0.54 2.31 ± 0.74 2.26 ± 0.54 3.19 ± 0.51 2.50 ± 0.53 1.31 ± 0.32 1.23 ± 0.22 22:6n-3 27.12 ± 1.09 18.18 ± 2.10[^\*^](#TN1){ref-type="table-fn"} 23.57 ± 3.39 18.17 ± 3.07[^\*^](#TN1){ref-type="table-fn"} 28.83 ± 1.50 21.72 ± 2.32[^\*^](#TN1){ref-type="table-fn"} 13.46 ± 4.99 11.87 ± 2.96 **TOTALS AND INDEXES** Monoenes 15.73 ± 2.42 28.17 ± 6.45[^\*^](#TN1){ref-type="table-fn"} 17.31 ± 2.67 29.86 ± 9.76[^\*^](#TN1){ref-type="table-fn"} 12.93 ± 0.98 29.55 ± 5.13[^\*^](#TN1){ref-type="table-fn"} 25.54 ± 5.32 29.79 ± 4.23 Saturates 30.02 ± 1.25 26.56 ± 4.66 35.17 ± 7.79 24.39 ± 6.12 33.11 ± 2.76 23.37 ± 2.60[^\*^](#TN1){ref-type="table-fn"} 37.46 ± 6.44 36.58 ± 9.60 n-3 39.20 ± 0.45 27.74 ± 5.67[^\*^](#TN1){ref-type="table-fn"} 33.88 ± 6.20 27.32 ± 6.13 38.05 ± 3.35 29.60 ± 5.06 18.19 ± 6.56 14.71 ± 3.83 n-6 9.46 ± 0.29 10.40 ± 1.24 7.81 ± 2.14 11.20 ± 3.05 8.60 ± 0.46 10.14 ± 1.54 9.16 ± 1.70 9.82 ± 3.82 n-3LCPUFA 38.61 ± 0.43 27.17 ± 5.62[^\*^](#TN1){ref-type="table-fn"} 33.10 ± 6.27 26.85 ± 6.13 37.51 ± 3.31 28.92 ± 4.74 17.75 ± 6.49 14.55 ± 3.56 n-3/n-6 4.15 ± 0.17 2.73 ± 0.89 4.44 ± 0.56 2.70 ± 1.42 4.44 ± 0.51 3.02 ± 0.97 1.95 ± 0.34 1.56 ± 0.29 18:1/n-3LCPUFA 0.23 ± 0.07 0.85 ± 0.43 0.30 ± 0.08 0.98 ± 0.61 0.18 ± 0.03 0.83 ± 0.34[^\*^](#TN1){ref-type="table-fn"} 0.88 ± 0.35 1.42 ± 0.19[^\*^](#TN1){ref-type="table-fn"} UI 261.62 206.09 227.69 209.64 252.62 220.28 152.06 137.98 *p \< 0.05 compared to controls. Abbreviations as in Table [2](#T2){ref-type="table"}*. Profiles of fatty acids in polar lipids shown in Table [3](#T3){ref-type="table"}, also indicate severe changes in enterocytes from deficient animals. Thus, a dramatic reduction of DHA occurs in all enterocyte preparations, reflecting the composition of deficient diets. Furthermore, the depletion of DHA in enterocytes was accompanied by pervasive increase in oleic acid (18:1n-9), linoleic acid and total monoenoic fatty acids. Strikingly, none of these changes appear to affect gill epithelium, as only few minor fatty acids were significantly changed. Given that main differences between enterocytes from LCPUFA deficient animals were observed for fatty acids, we next performed multivariate analyses using PCA. The results illustrated on Figure [1C](#F1){ref-type="fig"} revealed that two principal components explained 64.1% of overall variance, with PC1 being positively related to saturates (mainly 14:0, 16:0, and 18:0) and negatively to polyunsaturated fatty acids (mainly 20:4n-6, 20:5n-3, and 22:6n-3). Interestingly, PC2 was positively related to 18:1n-9, but PC1 was insensitive to this fatty acid, despite large differences between groups. Plotted factor scores revealed that deficiency-induced changes in fatty acids composition of phospholipids was so dramatic that two clusters corresponding to control and deficient enterocytes were plainly segregated (Figure [1D](#F1){ref-type="fig"}). Further, when factor scores were plotted for each intestinal segment, it was observed that irrespective of the intestinal origin, the three groups of enterocytes are represented in both CONTROL and DEFICIENT clouds (Figures [1D,E](#F1){ref-type="fig"}). Thermodynamic properties of Na^+^-K^+^-ATPase in isolated enterocytes and branchial cells ----------------------------------------------------------------------------------------- We next examined the temperature-dependence curves for the Na^+^-K^+^-ATPase reaction rates from the three groups of enterocytes and branchial cells under control conditions. Arrhenius plots for the Na^+^-K^+^-ATPase from intestinal and branchial preparations along with the Eyring plots (in the insets) are shown in Figure [2](#F2){ref-type="fig"}. Arrhenius plots indicate that the overall reactions of ATP hydrolysis for the Na^+^-K^+^-ATPase activities from all sources proceeded with temperature discontinuity points (*Td*) and two activation energies, i.e., above (*Ea*~1~) and below (*Ea*~2~) the breaking temperature. The calculated values for *Td* were lowest for posterior enterocytes (15.6°C) and highest for gill epithelia (22.3°C), being *Td* for pyloric caeca (17.2°C) and anterior intestine (16.4°C) enterocytes very close each together. Regarding *Ea*~1~ it is noticeable that values of enterocytes were similar between segments (in the range 12.33 kcal/mol -- 15.58 kcal/mol), but notably lower than the value obtained in gill epithelia (17.98 kcal/mol). Conversely, *Ea*~2~ were similar between epithelial cells from intestine and gills (in the range 5.46 kcal/mol--7.72 kcal/mol). Of note, it can also be observed that maximal activities are attained in the range 30--35°C, well beyond the standard rearing (and environmental) temperature of this species, and that increasing the assay temperature above 45°C (*I*~*t*~, inactivation temperature) leads to a pronounced decrease in ATPase activity, which has been interpreted as result of progressive enzyme denaturation and/or destabilization of membrane microenvironment as thermal stress increases. This thermosensitive behavior of Na^+^-K^+^-ATPase is common between ectotherms (Else and Wu, [@B18]). ![**Thermodynamic features of Na^**+**^-K^**+**^-ATPase from gilthead seabream epithelia reared under standard conditions**. Illustrated correspond to Arrhenius plots (Log\[activity\] vs. 1/T) and Eyring plots (insets, Ln\[k/T\] vs. 1/T) for the hydrolysis of ATP by Na^+^-K^+^-ATPase in each epithelial preparation. *Td*: Arrhenius breaking (or discontinuity) temperature. *Ea*~1~ and *Ea*~2~: activation energies below and above the point of discontinuity, respectively. *I*~*t*~: discontinuity at inactivation temperature. Each point corresponds to the average of four different experiments.](fphys-07-00589-g0002){#F2} By applying the transition states theory using Eyring plots we estimated activation enthalpy (Δ*H*^‡^) and activation entropy (Δ*S*^‡^). Results are illustrated in the insets of Figure [2](#F2){ref-type="fig"}. From the slopes and intercepts of ln(k/T) vs. 10^3^/T plots below and above *Td*, it was calculated the activation enthalpy and entropy values. In general, lower enthalpy and entropy values are observed above *Td*. In enterocytes, Δ*H*^‡^ values were similar between sections and ranged between 11.78 kcal/mol and 12.50 kcal/ mol below *Td*, and between 4.92 kcal/mol and 7.77 kcal/mol above *Td*. These values were considerably higher for the case of gill epithelia, which were 18.53 kcal/mol and 10.06 kcal/mol as calculated below and above *Td*, respectively. Regarding activation entropy, highest entropy values were observed for gill epithelia both above (36.86 kcal/°K/mol) and below (65.95 kcal/°K/mol) *Td*, while in enterocytes, Δ*S*^‡^ remained similar and ranging between 48.49 kcal/°K/mol (caeca) and 46.01 kcal/°K/mol (anterior intestine) below *Td*, and 31.78 kcal/°K/mol (anterior intestine) and 22.07 kcal/°K/mol (caeca) above *Td*. Effects of alterations in lipid profiles on the thermodynamic properties of Na^+^-K^+^-ATPase --------------------------------------------------------------------------------------------- We next assessed the potential effects of lipid alterations induced by n-3 PUFA deficient diets on the thermodynamic properties of Na^+^-K^+^-ATPase from intestinal and branchial epithelial cells. The results are illustrated in Figure [3](#F3){ref-type="fig"}. The first obvious differences when compared to control cells were the disappearance of *Td* and the increase in activation energy values in the enzyme from enterocytes. Indeed, *Ea*~1~ values ranged from 50.13 kcal/mol in caecal enterocytes to 24.9 kcal/mol in the case of anterior enterocytes. The second most dramatic change was the increased thermosensitivity displayed by the enzyme from enterocytes, which started to inactivate (*I*~*t*~) around 26°C, nearly 20° lower than in control cells. Consequently, *Ea*~2~ could not be estimated in enterocyte populations. These deleterious effects of n-3 PUFA deficiency in enterocytes are indicative of membrane instability and protein inactivation as a consequence of environment destabilization. Strikingly, gill epithelium demonstrates, once again, a considerable resistance to LCPUFA depletion and its effects were much less prominent in *Td, Ea*~1~, and *Ea*~2~, though slight differences compared to control animals were detected for *Td* and *Ea*~2~, but not for *Ea*~1~. ![**Effects of PUFA-deficient diets on thermodynamic properties of Na^**+**^-K^**+**^-ATPase from gilthead seabream enterocytes from anterior intestine, posterior intestine, pyloric caeca, as well as from gill epithelium**. Illustrated correspond to Arrhenius and Eyring plots for the hydrolysis of ATP by Na^+^-K^+^-ATPase in each epithelial preparation as indicated in Figure [2](#F2){ref-type="fig"}. Each point corresponds to the average of four different experiments.](fphys-07-00589-g0003){#F3} Changes in activation enthalpy (Δ*H*^‡^) and activation entropy (Δ*S*^‡^) for Na^+^-K^+^-ATPase in enterocytes below (*I*~*t*~) were dramatic, with a generalized increase in Δ*H*^‡^ (24.36 kcal/mol, 33.14 kcal/mol and 44.61 kcal/mol in anterior, posterior and caecal enterocytes, respectively) and Δ*S*^‡^ (84.26 kcal/°K/mol, 113.94 kcal/°K/mol and 171.27 kcal/°K/mol in anterior, posterior and caecal enterocytes, respectively). However, for gill epithelia, Δ*H*^‡^ values remained similar to controls both above (10.06 kcal/mol) and below (18.53 kcal/mol) *Td*. Likewise, Δ*S*^‡^ changes closely resembled those of controls with values of 36.68 kcal/°K/mol and 65.95 kcal/°K/mol, above and below discontinuity point, respectively. As n-3 LCPUFA deficiency affected only fatty acid composition of membrane phospholipids, but not lipid classes, levels of most relevant fatty acids were compared with the Arrhenius plot parameters (*Td, Ea*~1~, *Ea*~2~) and activation enthalpy (Δ*H*^‡^) and entropy (Δ*S*^‡^) for enterocytes belonging to the three regions as well as for gill epithelia (Figure [4](#F4){ref-type="fig"}). Initially, we obtained the Pearson\'s correlation coefficient matrixes for all fatty acids and ratios which were differentially affected by diets against thermodynamic parameters. First, we found significant relationships between levels of 18:1n-9 (*Td* = 9.46 + 0.83^\*^\[18:1n-9\], *r* = 0.96, *p* \< 0.001), total monoenes (*Td* = 6.76 + 0.63^\*^\[monoenes\], *r* = 0.96, *p* \< 0.001), DHA (*Td* = 32.72 − 0.61^\*^\[DHA\], *r* = 0.93, *p* \< 0.05) or n-3 LCPUFA (*Td* = 3.04 − 0.38^\*^\[n-3 LCPUFA\], *r* = 0.91, *p* \< 0.05), and *Td* (Figure [4A](#F4){ref-type="fig"}, left panel). Noticeably, it was observed an opposite influence of monoenes (positively related) and n-3 polyunsaturated fatty acids (negatively related) and *Td*. The positive effect of monoenes was also related, on one side, to saturated fatty acids as revealed by the negative relationships between saturates/18:1n-9 (*Td* = 34.48 − 8.012^\*^\[saturates/18:1n-9\], *r* = 0.85, *p* \< 0.05) and *Td*, and to n-3 LCPUFA, on the other side, as indicated by the negative correlation between n-3 LCPUFA/monoenes (*Td* = 26.30 − 3.93^\*^\[n-3 LCPUFA/monoenes\], *r* = 0.88 *p* \< 0.05) and *Td* (Figure [4A](#F4){ref-type="fig"}, middle panel). More precisely, this later relationship could be defined as a very significant negative exponential relationship between the ratio DHA/18:1n-9 (*Td* = 14.69 + 20.66^\*^e^−0.81\*\[DHA/18:1n-9\]^, *r* = 0.96, *p* \< 0.005) and *Td* (Figure [4A](#F4){ref-type="fig"}, right panel). Since *Td* disappeared in n-3 LCPUFA deficient enterocytes (but not in gills) these epithelia were excluded from these regression analyses. ![**Relationships between thermodynamic signatures of Na^**+**^-K^**+**^-ATPase from gilthead seabream epithelia and their lipid features. (A)** Effects of individual, grouped or ratios fatty acids on Arrhenius (discontinuity) breakpoints (*Td*). **(B)** Effects of lipid parameters on activation energies below *Td or I*~*t*~*:* (*Ea*~1~). **(C)** Effects of lipid ratios and selected fatty acids on activation enthalpy (Δ*H*^‡^) and activation enthalpy (Δ*S*^‡^*)*.](fphys-07-00589-g0004){#F4} On the other hand, *Ea*~1~ was only related to total monoenes and particularly to 18:1n-9 (*Ea*~1~ = 14.43 + 0.086^\*^\[1.2^\[18:1n-9\]^\], *r* = 0.91, *p* \< 0.005) levels, but not obvious relationships were observed for saturated or polyunsaturated fatty acids individually (Figure [4B](#F4){ref-type="fig"}, left panel). However, these later two lipid groups appear to substantially affect *Ea*~1,~since their ratios to 18:1n-9 provided good estimations of *Ea*~1~(Figure [4B](#F4){ref-type="fig"}, right panel),revealing in both cases very significant exponential decays. Finally, we assessed the relationships between activation enthalpy (Δ*H*^‡^) and activation entropy (Δ*S*^‡^) as dependent variables and lipid parameters. We found that enthalpy of activation was closely related to the DHA/18:1n-9 ratio (Δ*H*^‡^ = 13.17 − 198.7^\*^e^−3.17\*\[DHA/18:1n-9\]^, *r* = 0.92, *p* \< 0.01) and to saturates/18:1n-9 (Δ*H*^‡^ = 11.43--94.57^\*^e^−2.11\*\[saturates/18:1n-9\]^, *r* = 0.98, *p* \< 0.001) (Figure [4C](#F4){ref-type="fig"}, left panel). Enthalpy (Δ*H*^‡^) of activation was also positively (and exponentially) related to monoenes, in particular to 18:1n-9 (Δ*H*^‡^ = 11.63 + 0.28^\*^e^−0.17\*\[/18:1n-9\]^, *r* = 0.96, *p* \< 0.005) but not to n-3 LCPUFA or DHA (Figure [4C](#F4){ref-type="fig"}, middle panel). In the case of activation entropy (Δ*S*^‡^), two relevant relationships were observed, i.e., for 18:1n-9 (Δ*S*^‡^ = 46.22 − 0.57^\*^e^−0.18^\*^\[18:1n-9\]^, *r* = 0.96, *p* \< 0.005) (shown in Figure [4C](#F4){ref-type="fig"}, right panel), and for the ratio DHA/18:1n-9 ratio (Δ*S*^‡^ = 50.38 + 778.31^\*^e^−3.46\*\[DHA/18:1n-9\]^, *r* = 0.92, *p* \< 0.01) (not shown). Discussion {#s4} ========== In the present study, we have assessed the potential involvement of membrane lipid profiles and thermodynamic properties of the Na^+^-K^+^-ATPase in four populations of epithelial cells in the gilthead seabream. First, we observed that fatty acid composition of membrane phospholipids slightly differ between enterocytes isolated from the different intestinal segments. However, most dramatic differences were observed between enterocytes and branchial epithelial cells. To the best of our knowledge this is the first study convincingly demonstrating such heterogeneities in membrane lipid composition of epithelial cells. Highlighting main differences, we observed that, under control conditions, enterocytes are more enriched in long-chain polyunsaturated fatty acids (mainly DHA), total polar lipids, PC, PE, PS, PI, and PG, compared to gill epithelial cells. Conversely, branchial cells contain higher levels of total neutral lipids (mainly CHO, FFA, MAG, and TAG), and 20-carbons polyunsaturated fatty acids (ARA and EPA), as well as monoenes. These features allowed a neat discrimination between epithelial cells in the multivariate analyses performed here, which allowed us to define a differential lipid fingerprint between epithelial cells of intestinal and branchial origins. As the biochemical structure of membrane bilayer is finely tuned to cope with the specific cellular functions, it can be speculated that in gill epithelia, membrane phospholipids are more adequate to provide an efficient permeability transmembrane barrier against severe osmo-ionic gradients (Wood and Shuttleworth, [@B50]; Evans et al., [@B22]). Another conclusion that can be extrapolated from these data, is that as membrane surface is greater in enterocytes than in branchial cells (due to the presence of an extensive brush border), it would be expected a higher level of TPL. On the other hand, epithelial branchial cells exhibit higher levels of FFA, MAG, and TAG, which is in consonance with the high density of mitochondria and higher metabolic rates, especially in chloride cells (Wood and Shuttleworth, [@B50]; Evans et al., [@B22]), which accomplish the major task of NaCl extrusion against a very large electrochemical gradient. The second goal of our study was the assessment of the extent to which lipid profiles in epithelial cells might be affected by dietary modifications. Our dietary manipulation consisted in providing an n-3 PUFA-deficient diet which contained similar amount of total lipids compared to control standard diets, but changed proportions of PUFA, monoenoic and saturated fatty acids. The results showed that deficient diets have a profound impact on lipid profiles in enterocytes. With the sole exception of TAG which were significantly increased in all intestinal segments studied, lipid classes remained unaffected. However, most important changes were observed in fatty acid composition of membrane phospholipids. Thus, levels of total monoenes (mainly 18:1n-9) were significantly increased, while DHA, total n-3 LCPUFA and 18:3n-3, were reduced. All these findings revealed a direct influence of diet composition on the enterocyte membrane composition. However, it is worthwhile mentioning that the increase in 18:1n-9 in membrane phospholipids may also be part of a homeoviscous strategy to keep membrane fluidity in conditions of LCPUFA depletion. Strikingly, fatty acid composition of branchial cells from the same animals remained very similar between control and deficient diets, with only small changes in some minor fatty acids. These observations are very relevant from the homeostatic point of view, and pinpoints to the existence of powerful lipostatic mechanisms responsible for the biochemical stability of branchial cell membranes (including chloride cells, mucous cells, epithelial respiratory cells and pillar cells). This indeed leads us to hypothesize that the extremely relevant physiological functions of branchial cells (Wood and Shuttleworth, [@B50]; Evans et al., [@B22]) are tightly linked to a precise lipid composition in cell membranes. The next goal in the present study was the assessment of how changes in lipid profiles of cell membranes might affect the thermodynamic properties of epithelial Na^+^-K^+^-ATPase. Results from Arrhenius plots indicate that the overall reactions of ATP hydrolysis for the Na^+^-K^+^-ATPase activities from all sources followed convex behaviors (Truhlar and Kohen, [@B46]) with temperature discontinuity points (*Td*) and two activation energies, above (*Ea*~1~) and below (*Ea*~2~) the discontinuity, with *Ea*~1~\>*Ea*~2~. Slight differences in these parameters were detected for control enterocytes, but values for control gill epithelia were substantially different, being *Td* and *Ea*~1~ higher than in enterocytes. Similar results have been observed in mucosal scrapings (Almansa et al., [@B2]) and branchial microsomal fractions from gilthead seabream (Ventrella et al., [@B48]) and in the sea bass *Dicentrarchus labrax* (Ventrella et al., [@B47]) reared under similar conditions. Thermodynamically, the lower *Td* values obtained for enterocytes imply lower values for the thermotropic transitions of protein-lipid and protein-protein interactions and, coherently, a temperature reduction in the activation energies-shift point (Truhlar and Kohen, [@B46]). Accordingly, the enzyme from pyloric caeca would display lower activation energy above 17.20°C, which would allow an increase in the enzyme activity at lower temperatures (close to *Td*), compared with anterior and posterior enterocytes, whose shift points were clearly lower. Conversely, the enzyme from gills would require substantially higher temperatures (above 22.26°C) to benefit from a lower activation energy (as inferred from *Ea*~2~). This apparent unfavorable condition may not be compensated by the Δ*H*^‡^ values (both above and below Td) found in gill Na^+^-K^+^-ATPase. These assertions have been demonstrated experimentally in enzyme activity assays performed under standard conditions, where branchial Na^+^-K^+^-ATPase specific activity measured at 20°C was found \~50% lower than in averaged enterocytes. Also, these results are in line with lower activation enthalpy for enterocytes compared to gill epithelia (Table [4](#T4){ref-type="table"}). Thus, it seems that there exist environmental restrictions leading to restricted degrees of freedom (i.e., translational, rotational, and vibrational) in the transition state for the hydrolysis of ATP by Na^+^-K^+^-ATPase in branchial membranes. This is perhaps the price must be paid in gill epithelia to cope with extreme osmotic and ionic gradients far from equilibrium. It is known that the Na^+^-K^+^-ATPase reaction cycle includes two major conformational intermediates, namely *E1* (with high affinity for ATP) to *E2* (with phosphatase activity) (Schuurmans-Stekhoven and Bonting, [@B40]; Skou and Esmann, [@B42]; Kaplan, [@B29]), and that the transition between them occurs with a large enthalpy and entropy change. Moreover, changes in membrane order (reduction) favors the *E1* conformation and ATP binding (i.e., by PUFA or ethanol) while the increase in membrane order by reduced temperature or PUFA depletion, favors the *E2* conformation with ATP hydrolysis and K^+^ binding (Swann, [@B45]; Skou and Esmann, [@B42]; Kaplan, [@B29]). Thus, the higher absolute values of Δ*H*^‡^ (and Δ*S*^‡^) of activation for the hydrolysis of ATP in gill Na^+^-K^+^-ATPase suggest more ordered membranes, where the stabilization forces required for the transition of *E1* to *E2* conformations for hydrolysis are higher than for enterocytes. Indeed, our analyses of membrane lipid composition (with significantly lower contents of total n-3 LCPUFA in gill epithelia) and the lower membrane unsaturation index found in gill cell membranes (194.35) compared to enterocytes\' (ranging 244.4--256.7) support the thermodynamic observations and may well explain the observed differences between intestinal and branchial epithelia. ###### **Summary of thermodynamic parameters calculated for epithelial cells from control individuals and animals fed LCPUFA deficient diets**. ***Ea~1~* (kcal/mol)** ***Ea~2~* (kcal/mol)** ***Td* (°C)** **Δ*H*^‡^ \< *Td* (kcal/mol)** **Δ*H*^‡^ \>*Td* (kcal/mol)** **Δ*S*^‡^ \< *Td* (kcal/°K/mol)** **Δ*S*^‡^ \> *Td* (kcal/°K/mol)** ---------------------- ---------------------------------------- ------------------------ ---------------------------------------- ---------------------------------------- ------------------------------- ----------------------------------------- ----------------------------------- **CONTROLS** Gill 17.98 7.72 22.26 18.53 10.06 65.96 36.87 Pyloric caeca 13.14 5.46 17.20 12.59 4.92 48.50 22.07 Anterior intestine 12.33 6.48 16.42 11.78 7.00 46.01 25.87 Posterior intestine 12.75 8.28 15.58 12.20 7.77 47.13 31.78 **LCPUFA DEFICIENT** Gill 17.40 4.40 26.24 16.84 3.83 59.41 15.92 Pyloric caeca 50.13[^\#^](#TN2){ref-type="table-fn"} 26.09[^\#^](#TN2){ref-type="table-fn"} 49.61[^\#^](#TN2){ref-type="table-fn"} 171.27[^\#^](#TN2){ref-type="table-fn"} Anterior intestine 24.91[^\#^](#TN2){ref-type="table-fn"} 27.22[^\#^](#TN2){ref-type="table-fn"} 24.36[^\#^](#TN2){ref-type="table-fn"} 84.26[^\#^](#TN2){ref-type="table-fn"} Posterior intestine 33.68[^\#^](#TN2){ref-type="table-fn"} 25.77[^\#^](#TN2){ref-type="table-fn"} 33.14[^\#^](#TN2){ref-type="table-fn"} 113.93[^\#^](#TN2){ref-type="table-fn"} *Values were obtained below I~t~*. The relationships between epithelial membrane lipids and thermodynamic properties of Na^+^-K^+^-ATPase was further demonstrated in the n-3 PUFA deficiency experiments. This maneuver dramatically affected lipid profiles in enterocytes. As mentioned before, the main effects being a considerable increase in oleic acid and a substantial DHA depletion. Paralleling these findings, *Td* disappeared and *Ea*~1~ increased by 2--3.8-fold compared to control animals. LCPUFA depletions also augmented the thermosensitivity of intestinal Na^+^-K^+^-ATPase and reduced the values for inactivation temperature (*I*~*t*~). Noticeably, none of these parameters were significantly affected in gill epithelia, which reinforce the hypothesis that branchial cells are endowed with a highly efficient lipostatic mechanism ensuring the biochemical stability of membrane lipids (and lipid-lipid and lipid-protein interactions), ultimately giving support to the splendid physiological complexity of fish gills. Severe changes in absolute values of Δ*H*^‡^ (and Δ*S*^‡^) of activation for the hydrolysis of ATP were observed in enterocytes (but again not in gill epithelia), which increased by 2--4-fold for Δ*H*^‡^ and 1.8--3.5 fold for Δ*S*^‡^ below inactivation temperature (*I*~*t*~). These observations suggest more ordered membranes and higher stabilization forces for the transition *E1* to *E2* conformations. In agreement, unsaturation indexes in enterocytes (but not in branchial cells) were reduced by LCPUFA depletion by 14% (index values ranging 261-227 in controls, and 206-227 in LCPUFA deficient cells). The deleterious effects of increased *Ea*~1~, Δ*H*^‡^ and Δ*S*^‡^ were readily demonstrated in the specific activity assays (as measured at 20°C, the rearing temperature of seabream). Indeed, unlike branchial preparations, in enterocytes\' homogenates, Na^+^-K^+^-ATPase specific activity fell by average 82.8% compared to controls. In order to explore more deeply the individual relationships between specific membrane lipids and lipid ratios, as independent variables, and thermodynamic parameters, we used non-linear regression analyses. We found that *Td* was significantly negatively related to DHA and positively to 18:1n-9 (and monoenes). Expectedly, the ratio n-3 LCPUFA/monoenes (and more precisely the ratio DHA/18:1n-9) was negatively related to both *Td*, following a negative exponential relationship. Overall, these data indicates that the relative amounts of 18:1n9 and DHA in membrane phospholipids are critical in determining the discontinuity point and the transition *Ea*~1~↔*Ea*~2~ and pinpointed that small changes in DHA and/or 18:1n-9, have a great impact in setting *Td* as per their exponential relationship. Moreover, the same ratio (DHA/18:1n-9) was also found to be a significant predictor of *Ea*~1~ and Δ*H*^‡^ values, observing in both cases negative exponential equations describing their relationships. In these cases, both total monoenes and 18:1n-9 (but not saturates or DHA alone) were significantly related to *Ea*~1~, Δ*H*^‡^ and Δ*S*^‡^ but following exponential relationships, which indicates a very relevant role of 18:1n-9. Overall, these data indicate the relevant effects of DHA and monoenoic fatty acids (and their relative proportions) in setting thermodynamic traits for Na^+^-K^+^-ATPase, at least for ATP hydrolysis. Finally, current available information based on reconstitution and crystallographic studies may provide, at least in part, a molecular interpretation in support of our observations. Thus, recent studies using purified detergent-soluble recombinant α/β or α/β/FXYD Na^+^-K^+^-ATPase complexes, have revealed three separate functional effects of phospholipids and cholesterol that are exerted at separate binding sites for phophatidylserine/cholesterol (site A), polyunsaturated phosphatidylethanolamine (site B), and saturated PC or sphingomyelin/cholesterol (site C) (Cornelius et al., [@B10]; Habeck et al., [@B27]). These binding sites likely correspond to three lipid-binding pockets identified in the crystal structures of the Na^+^-K^+^-ATPase (Shinoda et al., [@B41]; Laursen et al., [@B31]). The three sites appear to have different effects on Na^+^-K^+^-ATPase complexes, being site A stabilizing, site B stimulatory and site C inhibitory (Cornelius et al., [@B10]; Habeck et al., [@B27]). Therefore, it is expected that direct and specific interactions of different phospholipids and cholesterol within the protein molecular framework, will determine both the stability and molecular activity of Na^+^-K^+^-ATPase, eventually giving membrane lipid composition an essential role in its physiological regulation. As it has been mentioned before, the activating site is associated to polyunsaturated fatty acid binding, especially when esterifying the *sn*-2 position of neutral lipids such phosphatidylethanolamine or phosphatidylcholine. Indeed, these two phospholipids represent by average 66% of epithelial TPL (Tables [2](#T2){ref-type="table"}, [3](#T3){ref-type="table"}), and DHA is, by far, the most abundant fatty acid in epithelial polar lipids. Recently, the stimulation of Na^+^-K^+^-ATPase activity of the purified human α1β1 or α1β1FXYD1 complexes by neutral PUFA PC or PE has been reported, being the stimulation structurally selective for neutral phospholipids (Haviv et al., [@B28]). The structural selectivity for the neutral phospholipid and asymmetric saturated plus PUFA fatty acyl chain structure is actually a strong indication for a specific interaction with Na^+^-K^+^-ATPase. Indeed, molecular modeling of *E1*\~*P* and *E2*·*P* conformations bound to any of these polyunsaturated phospholipids suggests that their specific binding may facilitate the *E1P*↔*E2P* conformational transition (mainly by lowering the activation energy) (Cornelius et al., [@B10]). In addition, DHA (and other LCPUFA) impose a general physical effect on the bilayer physicochemical state as it has conformational properties that keep highly structured but fluid membrane bilayer capable to accommodate rapid protein conformational changes (Rabinovich and Ripatti, [@B36]; Stillwell and Wassall, [@B44]; Diaz et al., [@B15]). Presumably, adjustment of DHA levels within membrane phospholipids (mainly phosphoglycerides) would accomplish effective physicochemical changes over a wide temperature range, ensuring the adaptation of cell membranes to environmental fluctuations and metabolic activity (Sargent et al., [@B39]). The latter being considered an important mechanism involved in thermal acclimation in ectothermic organisms (Raynard and Cossins, [@B37]; Else and Wu, [@B18]; Ernst et al., [@B19]). Interestingly, the stimulatory effects of polyunsaturated PE or PC seem to be independent of cholesterol and the FXYD protein (Cornelius et al., [@B10]). This is relevant because cholesterol is quite abundant in epithelial lipid profiles (the second most abundant neutral lipid after TAG, and accounting for \~40% of TNL), and because cholesterol has been long recognized to be essential for optimal Na^+^-K^+^-ATPase activity (Crockett and Hazel, [@B12]; Cornelius, [@B9]; Cornelius et al., [@B10]). The effects of cholesterol on Na^+^-K^+^-ATPase are associated to its ability to impose a high degree of conformational order on the phospholipids acyl chains thereby stabilizing the liquid-ordered lipid phase, which is dominated by changes in activation entropy (Cornelius et al., [@B10]), but also to direct interaction with site A (Shinoda et al., [@B41]; Habeck et al., [@B27]). However, as we observed no changes in cholesterol contents, neither between tissues nor between diets (Table [3](#T3){ref-type="table"}), observed differences in activation Δ*H*^‡^ and Δ*S*^‡^ cannot be explained by distinct membrane cholesterol. Finally, we wish to introduce a word of caution on the extent to which our results may be interpreted. Indeed, although we have empirically studied the thermodynamic parameters of the epithelial Na^+^-K^+^-ATPase from intestinal and branchial origins, and correlated them with the membrane phospholipid composition, it is not possible to provide molecular details on the real enzyme-membrane interactions nor giving precise insights into the physical mechanisms underlying the observed responses. In conclusion, the results illustrated in the present study reveal that subtle differences in the lipid matrix of the lipid microenvironment surrounding the Na^+^-K^+^-ATPase may explain the heterogeneous thermodynamic features of the Na^+^-K^+^-ATPase in epithelia from the gilthead seabream reared under control conditions. In enterocyte preparations, these effects were exacerbated by induction of membrane n-3 LCPUFA deficiency. Noteworthy, epithelial cells from gill origin were notably resistant to diet-induced modifications of membrane lipid composition and, consequently, to alterations in the thermodynamic features of the Na^+^-K^+^-ATPase. We conclude that n-3 LCPUFA and 18:1n-9 (together with cholesterol) are critical elements for the fine tuning of the Na^+^-K^+^-ATPase activity within the context of the epithelial plasma membrane, which underlies the essential role of the Na^+^-K^+^-ATPase in the physiological regulation of osmo/ionoregulatory tasks in euryhaline teleost fish. Author contributions {#s5} ==================== RD and CR performed the lipid analyses and specific ATPase activity experiments. TG was in charge of temperature-dependence experiments. MD designed the study, analyzed the data (together with RD) and wrote the manuscript. All authors reviewed the drafted manuscript. Conflict of interest statement ------------------------------ The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. We are indebted to Lupe Acosta for invaluable assistance during the experimental stage of the study. We are grateful to Biosearch Life-Puleva Biotech for continuous support to our project at the Laboratory of Membrane Physiology and Biophysics. This research was partly supported by research grant SAF2014-52582-R from MINECO (Spain). [^1]: Edited by: Angel Nadal, Universidad Miguel Hernández de Elche, Spain [^2]: Reviewed by: Pasquale Stano, University of Salento, Italy; Angel Catala, National University of La Plata, Argentina [^3]: This article was submitted to Membrane Physiology and Membrane Biophysics, a section of the journal Frontiers in Physiology

{ "pile_set_name": "PubMed Central" }

Review of the literature {#sec0001} ======================== Background {#sec0002} ---------- Tuberculosis (TB) is an airborne infectious disease that most commonly affects the lungs. The causative organism is the obligate aerobic, acid-fast bacillus *Mycobacterium tuberculosis* [@bib0001], [@bib0002], [@bib0003]. There are numerous manifestations of extrapulmonary tuberculosis, many involving the eye [@bib0004]. This review will focus primarily on the intraocular forms of TB. A previous review summarized orbital TB [@bib0005]. Literature search strategy {#sec0003} -------------------------- The literature was reviewed using a PubMed search with both Medical Subject Headings (MeSH) and keywords. MeSH terms included tuberculosis, ocular tuberculosis, eye infections and visual acuity. Keywords included eye, intraocular, periocular, ocular, uveitis, sclerouveitis, panuveitis, choroiditis, retina, retinal, tuberculosis, and "ocular tuberculosis." Results were limited to available peer-reviewed, English-language journals published between 1930 and 2015. All papers were reviewed, (including single case reports), to determine if they should be referenced in this review. Epidemiology {#sec0004} ------------ The Centers for Disease Control and Prevention (CDC) estimates that one third of the world\'s population is infected with TB, but only ten percent of infected persons develop clinical manifestations of the disease [@bib0006]. Of the ten percent with detectable disease, sixteen to twenty-seven percent have extrapulmonary TB involvement, which includes those with intraocular findings [@bib0007]. The incidence of intraocular TB has been reported to range from 1.4 to 18 percent [@bib0004], [@bib0008], [@bib0009], [@bib0010], [@bib0011], [@bib0012], [@bib0013]. Age over forty, female gender, and HIV infection increase the risk of extrapulmonary TB, and individuals with HIV also have an increased risk of ocular TB [@bib0004]. Intraocular involvement {#sec0005} ----------------------- Hematogenous spread is the primary mechanism by which TB affects the eye [@bib0014], [@bib0015], [@bib0016]. However, direct local extension and hypersensitivity responses from infection elsewhere in the body can also result in intraocular findings [@bib0014]. Intraocular TB often affects the ciliary body and choroid due to the high regional oxygen tension of these tissues, and uveitis, especially posterior uveitis, is the most common form of intraocular TB [@bib0006], [@bib0014], [@bib0016]. Regardless of the clinical presentation, multiple recurrences of inflammation despite treatment should increase the level of suspicion for intraocular TB in a patient with TB risk factors. The wide variety of ways in which TB can affect the intraocular tissues are described below and summarized in [Table 1](#tbl0001){ref-type="table"} with a corresponding diagram outlining the relevant structures of the eye in [Fig. 1](#fig0001){ref-type="fig"}.Table 1Clinical presentations of intraocular tuberculosis.Table 1TissuePossible presentationsAnterior uvea/ Pars planaGranulomatous anterior uveitisIris nodulesIris atrophyIntermediate uveitis  LensCataract  ChoroidTuberclesTuberculomasAbscessesChoroiditis  RetinaMacular edemaIntra- or preretinal hemorrhageRetinitisVasculitisNeovascularizationNeuroretinitisEales disease  Optic nerveOptic neuritisRetrobulbar neuritisPapillitisPapilledemaTubercles  GlobePanuveitisEndophthalmitisPanophthalmitisGlobe ruptureFig. 1Anatomical diagram of the eye. The various anatomical structures of the eye that can be affected by tuberculosis are labeled. Labels correspond to the structures listed in [Table 1](#tbl0001){ref-type="table"}Fig 1 ### Anterior uveitis {#sec0006} TB can cause a granulomatous uveitis with iris and angle granulomas, mutton-fat keratic precipitates, posterior synechiae, and occasionally hypopyon [@bib0017], [@bib0018], [@bib0019], [@bib0020]. A pigmented hypopyon, iris nodules, and iris atrophy have also been reported [@bib0014], [@bib0021], [@bib0022]. Cataract can develop as a result of ongoing inflammation and steroid treatment, and extensive synechiae can lead to angle closure glaucoma [@bib0014]. Of patients with TB-related uveitis, anterior uveitis has been reported in 12 to 36 percent of the cases; these patients are more likely to have broad-based posterior synechiae and less likely to have filiform synechiae than patients with uveitis unrelated to TB [@bib0006], [@bib0023], [@bib0024]. ### Posterior and panuveitis {#sec0007} Panuveitis has been reported in 11 to 20 percent of patients with TB uveitis ([Fig. 2](#fig0002){ref-type="fig"}), whereas posterior uveitis accounts for 35 to 42 percent of intraocular TB [@bib0006], [@bib0023]. Multifocal choroiditis is the most common manifestation of posterior segment involvement ([Fig. 3](#fig0003){ref-type="fig"}) [@bib0006], [@bib0014], [@bib0015], [@bib0023]. Retinitis usually occurs in the setting of concomitant choroiditis rather than as an isolated syndrome [@bib0006]. TB-associated posterior uveitis can also take the form of serpiginous-like choroiditis ([Figs. 4](#fig0004){ref-type="fig"} and [5](#fig0005){ref-type="fig"}), which is hypothesized to be a hypersensitivity reaction that progresses relentlessly despite steroid treatment. [@bib0006], [@bib0025]. Serpinginous-like choroiditis may be an important marker for TB even in patients residing in non-endemic regions [@bib0026], [@bib0027]. In contrast to classic serpiginous choroiditis, the lesions in TB-associated serpiginous-like choroiditis are more pigmented, more likely to be multifocal, and often arise outside the peripapillary region [@bib0028]. However, both TB-associated choroiditis and true serpiginous choroiditis can be difficult to treat. Patients with serpiginous-like choroiditis are typically from TB-endemic regions and/or of Asian Indian ethnicity [@bib0029]. In one report, serpiginous-like choroiditis was found in 21.5 percent of patients with TB uveitis in Tunisia, North Africa, and another report found that patients with TB-associated uveitis are more likely to have serpiginous choroiditis than patients with uveitis unrelated to TB [@bib0023], [@bib0024]. Since chorioretinal inflammation may breach Bruchs membrane, all types of chorioretinal scars should be monitored for the development of neovascularization [@bib0014].Fig. 2Panuveitis secondary to tuberculosis. Optical coherence tomography of the right eye (A) reveals subfoveal fluid with intraretinal cysts and diffuse thickening. There is mild diffuse macular leakage on fluorescein angiography (B) with more distinct areas of leakage in the central and temporal macula. Indocyanine green demonstrates a vague area of hypercyanescence in the superior macula (C). The left eye looked similar.Fig 2Fig. 3Multifocal choroiditis secondary to tuberculosis. Montage color fundus photos of the right and left eyes (A and B) demonstrate multifocal punched out lesions in both eyes. The lesions are hyperfluorescent on mid-phase fluorescein angiography (C and D). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)Fig 3Fig. 4Serpiginous-like choroiditis secondary to tuberculosis. Montage color fundus photos of the right and left eyes (A and B) demonstrate extensive retinal pigment epithelial changes, atrophy, and scarring in both eyes. The lesions are hypofluorescent centrally with hyperfluorescent borders which do not leak on fluorescein angiography (C). The lesions are hypocyanescent on indocyanine green (D). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)Fig 4Fig. 5Serpiginous-like choroiditis secondary to tuberculosis. Montage color fundus photos of the right and left eyes (A and B) demonstrate multiple white, irregular lesions and mottling of the retinal pigment epithelium in both eyes as well as a peripapillary serpiginous-like lesion in the left eye. The lesions are hyperfluorescent on fluorescein angiography (C) and hypocyanescent on indocyanine green (D) as shown for the left eye. Optical coherence tomography reveals macular thinning with ratty outer segment changes in the right eye (E); the left eye has an epiretinal membrane with peripapillary subretinal material corresponding to a fibrotic scar as well as elevation of the inferotemporal macula due to a large choroidal lesion (F). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)Fig 5 ### Tubercles, tuberculomas, and subretinal abscesses {#sec0008} Another manifestation of TB-associated posterior uveitis is the tubercle, which is a small nodule that is usually located in the posterior pole. Tubercles may be unilateral or bilateral, typically lack associated vitreitis, and are usually multifocal in miliary TB [@bib0014], [@bib0030]. Tubercles are strongly associated with meningitis [@bib0031]. Large choroidal tuberculomas measuring up to 14 mm are encountered less commonly [@bib0032], [@bib0033]. Tuberculomas are yellow subretinal masses that mimic tumors and often have associated exudative retinal detachments [@bib0006]. They can occur in immunocompetent patients, those with disseminated TB, and those with central nervous system TB [@bib0034], [@bib0035], [@bib0036], [@bib0037], [@bib0038]. Sub-retinal abscesses occur as a result of liquefaction necrosis in caseating granulomas and can develop in patients with disseminated TB [@bib0014], [@bib0039]. Vitreitis and retinal hemorrhages can also be associated with these abscesses [@bib0006]. ### Endophthalmitis and panophthalmitis {#sec0009} If sub-retinal abscesses go untreated, they can rupture and result in endophthalmitis [@bib0015], [@bib0040], [@bib0041]. TB endophthalmitis is exceedingly rare, with only 18 cases published worldwide [@bib0042]. Panophthalmitis, which includes scleral involvement, can also occur, leading to globe rupture or scleral calcification in advanced cases [@bib0014], [@bib0043]. ### Retinal vasculitis and other retinovasculopathies {#sec0010} Retinal vasculitis, which is more common with TB-associated intraocular inflammation than non-TB associated uveitis, typically presents as a periphlebitis and very rarely involves the arterioles in intraocular TB [@bib0024], [@bib0044]. Periphlebitis is typically accompanied by vitreitis and is the second most common presentation of intraocular TB [@bib0045]. In a patient with TB risk factors, TB should be highly suspected in the setting of exudative hemorrhagic periphlebitis [@bib0046] Ischemic central retinal vein occlusion has also been reported in the setting of retinal vasculitis due to TB [@bib0047]. Eales disease, first described by Henry Eales in 1882, is characterized by peripheral capillary nonperfusion, neovascularization, recurrent vitreous hemorrhages, periphlebitis, and intraocular fibrovascular proliferation in a quiet eye ([Fig. 6](#fig0006){ref-type="fig"}). This entity typically affects otherwise healthy men in their third to fourth decade of life from TB-endemic countries. Associated systemic symptoms have been reported to include epistaxis, peripheral circulation disorders, headache, and constipation [@bib0048], [@bib0049], [@bib0050]. *M. tuberculosis* DNA has been detected in patients with Eales disease, but it remains unclear whether or not TB is directly causative of this entity, or if it is a separate retinal hypersensitivity reaction [@bib0019], [@bib0046], [@bib0051], [@bib0052], [@bib0053].Fig. 6Eales disease. Montage color fundus photo of the right eye (A) reveals a temporal frond of neovascularization in a patient with a positive Mantoux skin test. The corresponding fluorescein angiogram (B) shows leakage from the area of neovascularization with adjacent peripheral capillary nonperfusion. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)Fig 6 ### Neuroretinitis and optic neuropathy {#sec0011} Neuroretinitis typically occurs secondary to choroidal spread of a peripapillary infection [@bib0006], whereas optic nerve infiltration and subsequent optic neuropathy can occur secondary to hematogenous spread or choroidal extension [@bib0015]. Optic neuropathy can also be the result of a hypersensitivity response without direct infection of the nerve [@bib0014]. Manifestations of TB optic nerve involvement can also include papillitis, papilledema, optic neuritis, retrobulbar optic neuritis, and optic nerve tubercles [@bib0051], [@bib0054], [@bib0055]. ### Immunocompromised patients {#sec0012} Ocular involvement of TB is both more common and more severe in immunocompromised patients. There have been reports of extensive choroidal infiltrates and nodules, severe vitreitis, endophthalmitis, choroidal tuberculomas, choroiditis, chorioretinitis, and bilateral optic neuritis in the immunocompromised with TB [@bib0056], [@bib0057], [@bib0058], [@bib0059], [@bib0060], [@bib0061]. Immune recovery uveitis occurs in HIV positive patients with TB who have reconstitution of the immune system after starting antiretroviral therapy [@bib0046]. The findings in immune recovery uveitis can vary widely and include anterior uveitis, hypopyon, vitreitis, papillitis, panuveitis, retinal or optic disc neovascularization, retinal detachment, cystoid macular edema, epiretinal membrane formation, vitreomacular traction, and macular hole [@bib0062]. One reported case was so severe that it resulted in globe rupture [@bib0063]. It should be noted that immune recovery uveitis can also occur in treated HIV patients who do not have TB. Ophthalmic imaging characteristics of intraocular TB {#sec0013} ---------------------------------------------------- ### Fluorescein and indocyanine green angiography {#sec0014} On fluorescein angiography, active choroidal tubercles show early hypofluorescence and later hyperfluorescence [@bib0006]. Tuberculomas show early hyperfluorescence with late pooling in areas of exudative retinal detachment [@bib0006]. Serpiginous-like choroiditis demonstrates early hypofluorescence with late hyperfluorescence and leakage of the active borders of the lesion [@bib0015]. Inactive serpiginous-like choroiditis lesions have staining borders and less hyperfluorescent centers. Retinal vasculitis from TB will demonstrate leakage surrounding the involved vessels as well as peripheral capillary nonperfusion [@bib0006]. Indocyanine green angiography of choroidal lesions demonstrates early hypocyanescence followed by late hypercyanescence surrounding a central zone of hypocyanescence [@bib0015]. ### Optical coherence tomography {#sec0015} Optical coherence tomography (OCT) may be helpful in detecting and quantifying macular edema and other pathology such as chorioretinal lesions [@bib0014]. While a significant degree of intraocular inflammation could result in decreased image resolution, the OCT may still provide valuable information when clinical examination by slit lamp and ophthalmoscopy is very difficult [@bib0064]. ### Fundus autofluorescence {#sec0016} Fundus autofluorescence can be particularly useful in cases of serpiginous-like choroiditis [@bib0065]. Serpiginous-like choroiditis progresses from an ill-defined area of hyperautofluorescence in the acute stage, to a well-defined hypoautofluorescent halo surrounding an area of hyperautofluorescence in the subacute stage, and finally a uniformly hypoautofluorescent lesion after resolution [@bib0065]. ### Ocular ultrasound {#sec0017} B-scan ultrasonography is also useful when severe inflammation or cataract limits the view of the fundus. While mass-like tuberculomas can mimic tumors, A-scan shows low to medium internal reflectivity, which may help differentiate the tuberculoma from other entities [@bib0003], [@bib0015]. Diagnosis {#sec0018} --------- The myriad of intraocular TB manifestations can mimic other diseases, making diagnosis challenging. Delays in diagnosis of intraocular TB can lead to permanent vision loss and even loss of the affected eye(s) [@bib0066]. Without microbiologic confirmation, which can be challenging to obtain, a diagnosis of intraocular TB may be empiric; therefore it is necessary to maintain a high index of suspicion aided by a targeted review of systems and directed laboratory testing [@bib0014], [@bib0067], [@bib0068]. Response to anti-tuberculosis treatment (ATT) may be the only definitive evidence of intraocular TB, but patients should not be committed to long courses of systemic ATT without first undergoing a thorough evaluation to rule out other entities such as syphilis and toxoplasmosis that may have similar findings [@bib0006], [@bib0069]. ### Mantoux skin test {#sec0019} A Mantoux skin test involves the intradermal injection of a purified protein derivative followed by examination for cutaneous induration 48 to 72 hours after the injection [@bib0070]. A positive result is caused by a delayed-type hypersensitivity reaction and is defined as 1) induration measuring greater than 5 mm in HIV-positive patients; 2) induration greater than 10 mm in high-risk persons, including healthcare workers, nursing home patients, and those living in endemic areas; 3) induration greater than 15 mm in all other patients [@bib0006]. Cutaneous hypersensitivity has been shown to correlate directly with ocular hypersensitivity in rabbits, indicating this test may be useful in diagnosing ocular disease [@bib0071]. Unfortunately, Mantoux test interpretation is subjective, and the test has low sensitivity and specificity [@bib0006]. Anergy to the protein can lead to false negative results, especially in immunocompromised patients [@bib0006]. False positive reactions may be encountered in patients who have been exposed to different species of mycobacteria, those who have received multiple Mantoux tests in the past, or those who have previously received Bacillus Calmette-Guerin (BCG) vaccinations [@bib0006], [@bib0014], [@bib0072]. However, induration greater than 10 mm should not be attributed to prior BCG vaccination according to the United States Preventative Services Task Force [@bib0073]. ### Interferon gamma testing {#sec0020} The QuantiFERON TB-Gold test (QFTG: Qiagen, Germantown, Maryland, USA) is a blood test designed to specifically detect TB. One advantage over the Mantoux test is the fact that the QuantiFERON blood test does not require a return office visit for interpretation [@bib0074]. The blood test quantifies interferon gamma release by using *M. tuberculosis* antigens to stimulate sensitized T cells from TB-infected patients. Theoretically this test should not yield false positive results due to exposure to other mycobacteria or prior BCG vaccination [@bib0074], [@bib0075]. However, the false positive rate has been reported to be as high as 41% in healthcare workers and 80.5% in HIV-positive patients at low risk for TB [@bib0076], [@bib0077], [@bib0078], [@bib0079]. Nevertheless, while the QuantiFERON test may have the most utility in patients who have been vaccinated with BCG or for those who are unlikely to return for skin test reading, the CDC recommends the QuantiFERON test as a screening tool that may be used in place of a Mantoux skin test in all situations were a Mantoux test would be recommended [@bib0080]. ### Chest x-ray and computerized tomography {#sec0021} Chest x-ray or computerized tomography of the chest demonstrating cavities, consolidation, or lymph node enlargement can provide supporting evidence for TB [@bib0009]. Any positive chest imaging finding should prompt sputum evaluation to distinguish between healed TB lesions and active disease [@bib0081]. Ocular TB can occur in the absence of pulmonary TB, so a normal chest radiograph does not exclude the diagnosis of intraocular TB [@bib0003], [@bib0082]. ### Acid-fast bacteria and histology {#sec0022} The diagnosis of intraocular TB can be proven if acid-fast bacteria are demonstrated via microscopy or Lowenstein-Jensen media culture of intraocular fluids or tissue; however, the yield of this diagnostic test is typically low due to a paucity of organisms [@bib0014], [@bib0015]. Necrotizing granulomas with positively staining acid-fast bacilli by Ziehl-Neelsen stain are also compatible with TB [@bib0009]. ### Polymerase chain reaction {#sec0023} Polymerase chain reaction (PCR) amplification of mycobacterial DNA can be utilized in the evaluation of suspected intraocular TB, as the ocular specimen volume or size is often too small to detect the mycobacterium by microscopy or culture [@bib0083], [@bib0084]. PCR has the potential to be an important tool in the diagnosis of intraocular TB, and has proven useful in vitreous samples from patients with serpiginous-like choroiditis [@bib0083], [@bib0085], [@bib0086]. However, testing of aqueous and vitreous humor may be falsely negative due to a relative paucity of organisms in TB-associated uveitis [@bib0087]. Therefore, while it can be useful, positive *M. tuberculosis* PCR is not required to make the diagnosis of intraocular TB [@bib0015], [@bib0088], [@bib0089]. Treatment {#sec0024} --------- ### Medical treatment {#sec0025} Because systemic treatment is required to treat ocular TB, patients should be co-managed by an ophthalmologist and an infectious disease specialist. Medical ATT for ocular TB utilizes the same quadruple-drug regimen used to treat pulmonary TB. The CDC-recommended regimen consists of isoniazid, rifampin, pyrazinamide, and ethambutol for a total of 2 months, followed by an additional 4 to 7 months of dual therapy with isoniazid and rifampin [@bib0090]. Extended courses of therapy may be necessary due to multi-drug resistance and/or the typically slow and sometimes relapsing nature of ocular TB [@bib0014], [@bib0066]. Rifabutin, fluoroquinolones, interferon gamma, and linezolid may be utilized in cases of multi-drug resistance or patient intolerance to any of the four recommended agents [@bib0006]. Infectious disease specialists may choose to treat latent TB with isoniazid and/or rifampin rather than quadruple therapy, although it has become increasingly common to conclude that intraocular TB requires quadruple drug therapy [@bib0090]. Pyridoxine (vitamin B6) supplementation is necessary for patients taking isoniazid in order to prevent peripheral neuropathy [@bib0091]. ### Ocular side effects of medical therapy {#sec0026} A baseline eye exam must be performed prior to starting ATT, and patients should be monitored for medication side effects, which can sometimes be confused with worsening ocular TB [@bib0006]. Isoniazid and ethambutol can cause optic neuropathy, which presents as decreased visual acuity, decreased color vision, or cecocentral scotomas [@bib0092]. Ethambutol can also cause optic neuritis characterized by red-green dyschromatopsia and disc edema, which can lead to optic atrophy [@bib0006]. The offending medication should be promptly discontinued at the first sign of vision-threatening side effects, as these symptoms may be reversible, with vision often returning to normal 10 to 15 weeks after drug cessation [@bib0006]. Vitamin B12 can also occasionally help vision recovery, but in some cases vision loss may be permanent [@bib0006]. Rifabutin can cause anterior, intermediate, or panuveitis that is typically treatable with corticosteroids but may also necessitate discontinuation of the medication [@bib0003]. ### Use of steroids {#sec0027} While choroidal tubercles and tuberculomas often respond well to ATT alone, many other intraocular manifestations of TB require corticosteroids plus ATT to prevent damage to the ocular tissues secondary to the inflammatory response [@bib0015], [@bib0093]. Corticosteroid treatment can be critical for certain manifestations of ocular TB, such as choroiditis, which can actually undergo paradoxical worsening upon initiation of ATT [@bib0006]. Topical and locally injected corticosteroids are important in the management of anterior and intermediate uveitis, and systemic forms are an important component in retinal vasculitis treatment [@bib0003], [@bib0006], [@bib0017]. It is critical, however, that corticosteroids are never used in the absence of ATT, as this not only leads to more frequent recurrences of ocular inflammatory symptoms but can also cause significant worsening and potentially dissemination of the infection [@bib0006], [@bib0094]. ### Surgical treatment {#sec0028} Typically, ocular TB can be managed with medical therapy alone. However, in cases where retinal neovascularization develops, laser photocoagulation may be necessary to treat areas of retinal capillary nonperfusion and decrease the stimulation of neovascular growth [@bib0006]. Pars plana vitrectomy has also been reported as an adjunctive therapy for TB endophthalmitis, and full thickness eye wall resection with pars plana vitrectomy has been used to treat a tuberculous granuloma [@bib0042], [@bib0095]. Occasionally, patients may also require cataract surgery; however, this procedure should be postponed until all intraocular inflammation has been completely controlled for at least 3 months [@bib0014]. Elective surgical procedures such as cataract surgery should be delayed even longer in young or noncompliant patients and in those with severe ocular damage from intraocular TB [@bib0014]. Conclusion {#sec0029} ========== Intraocular TB is a difficult diagnosis, as it can mimic many other more common etiologies of intraocular inflammation or uveitis. A high index of suspicion for TB and cooperation with infectious disease specialists are paramount, because timely diagnosis and treatment may prevent irreversible vision loss [@bib0096]. Ann M. Farrell for assistance with literature search. Funding: Supported by Mayo Foundation, Rochester, MN; Research to Prevent Blindness, New York, NY; and the VitreoRetinal Surgery Foundation.

{ "pile_set_name": "PubMed Central" }

Introduction {#s1} ============ Dengue (DEN) and chikungunya (CHIK) are two mosquito-borne viral infections transmitted by mosquitoes of the genus *Aedes*. Dengue viruses (DENV-1, -2, -3, -4) belonging to the *Flaviviridae* family, *Flavivirus* genus [@pntd.0000706-Lindenbach1] are of primarily concern as they are responsible of the most important arboviral disease widely distributed in the tropical world [@pntd.0000706-Gubler1]. Dengue infection may be unapparent or induce an undifferentiated febrile illness, a classic dengue fever (DF), or a dengue hemorrhagic fever (DHF). The highest prevalence of DEN is observed in South-East Asia and South Americas with approximately 50--100 million cases and 25 000 deaths per year [@pntd.0000706-Gubler2]. The transmission is mainly ensured by the highly anthropophilic *Aedes aegypti* in urban areas [@pntd.0000706-Gubler3]. However, *Aedes albopictus* may act as a secondary vector in rural areas and even as the main vector when *Ae. aegypti* is not present or too scarce as observed in some localities in China, Japan, Hawaii, and Seychelles [@pntd.0000706-Gratz1]. *Ae. albopictus* was indeed the only vector in the recent dengue outbreaks observed in the Indian Ocean on La Reunion Island [@pntd.0000706-Pierre1] and Madagascar [@pntd.0000706-Ratsitorahina1]. Chikungunya virus (CHIKV), first isolated in Tanzania in 1953 [@pntd.0000706-Ross1], belongs to the *Togaviridae* family, *Alphavirus* genus [@pntd.0000706-Kuhn1] and is endemic to Africa, India and South-East Asia. In Africa, the virus is maintained within a sylvatic cycle with wild mosquitoes (*Aedes furcifer*, *Aedes luteocephalus*, *Aedes taylori*, *Aedes africanus*) feeding preferentially on primates [@pntd.0000706-Diallo1], [@pntd.0000706-Jupp1]. In Asia, CHIKV is mainly transmitted within an urban cycle in an inter-human transmission achieved essentially by the human-biting *Ae. aegypti* and the less anthropophilic *Ae. albopictus*, which prefers suburban and rural areas where it colonizes both artificial and natural containers [@pntd.0000706-Lam1], [@pntd.0000706-Laras1]. CHIKV mainly induces high fever and severe arthralgia, and had limited impacts on public health before its emergence in the Indian Ocean in 2005. This major epidemic started in the Comoro Islands in January 2005 then spread rapidly to the other islands of the region, Mayotte, Seychelles, La Reunion and Mauritius [@pntd.0000706-Schuffenecker1]. In April 2006, one third of the population in La Reunion Island had been in contact with the virus [@pntd.0000706-Renault1]. Surprisingly, the vector in this epidemic was not *Ae. aegypti*, only present as residual populations on the island, but *Ae. albopictus* [@pntd.0000706-Salvan1]--[@pntd.0000706-Delatte1]. This latter species was proved to be a very efficient vector of a mutated strain CHIKV harboring a switch from an alanine to a valine in the E1 glycoprotein, mutation that appeared in the course of the outbreak and was then selected as a major epidemic genotype [@pntd.0000706-Vazeille1], [@pntd.0000706-Schuffenecker1]. CHIK outbreaks spread rapidly and caused several million clinical cases in the Indian Ocean Islands and India, where outbreaks had been absent for 32 years [@pntd.0000706-Pialoux1], [@pntd.0000706-Santhosh1]. One consequence was an increasing overlapping in the distribution of the two arboviral diseases, DEN and CHIK leading to the detection of a higher number of co-infections in humans. Co-infection by DENV and CHIKV in patients has been known for a long time. First described in 1964 in South India [@pntd.0000706-Myers1], it has been more frequently reported since the re-emergence of CHIK: Sri-Lanka [@pntd.0000706-Hapuarachchi1], India [@pntd.0000706-Chahar1], [@pntd.0000706-Schilling1], Malaysia [@pntd.0000706-Nayar1] where the main vector involved is *Ae. aegypti* and in Gabon [@pntd.0000706-Leroy1], Madagascar [@pntd.0000706-Ratsitorahina1] and La Reunion Island [@pntd.0000706-World1] where viral transmission is achieved by *Ae. albopictus.* On La Reunion Island, the CHIK outbreak was preceded by a small outbreak due to DENV-1 with 228 cases reported between March and July 2004 [@pntd.0000706-Pierre1]. Despite the limited impact of the DEN outbreak, in January 2006, suspected cases of co-infection DENV-CHIKV in patients were reported [@pntd.0000706-World1] and the same phenomenon was also observed in Madagascar in January 2006 [@pntd.0000706-Ratsitorahina1]. Dual arboviral infections in humans can occur following the bite of two mosquitoes, each infected by one virus, or the bite of one mosquito infected by the two viruses. If isolations of either virus from mosquitoes collected in the course of an epidemic have been already noticed, to our knowledge, doubly infected mosquitoes have never been described. A recent study even concluded on the failure to prove co-infection of *Ae. aegypti* by these two viruses [@pntd.0000706-Rohani1]. Here, we orally infected in a single blood-meal, *Ae. albopictus* from La Reunion Island with autochthonous viral strains isolated during the 2004--2005 outbreak of DEN-1 and the 2005--2006 CHIK outbreak. Dissemination of both viruses within the vector was checked by immunofluorescence assay on female head squashes. Furthermore, saliva was collected from each female to check their ability to deliver simultaneously both DENV and CHIKV infectious particles. Additionally, we performed secondary infections by inoculating first DENV-1 then providing CHIKV in an infectious blood-meal. Methods {#s2} ======= Ethics statement {#s2a} ---------------- All experiments on live vertebrates were performed in compliance with French and European regulation and according to the Institut Pasteur guidelines for laboratory animal husbandry and care. Mosquitoes {#s2b} ---------- *Ae. albopictus* Providence (ALPROV) were collected as eggs in 2006 on La Reunion Island and provided by the DRASS (Direction Régionale des Affaires Sanitaires et Sociales). The F5 or F6 generation was used for experimental infections. Colonies were maintained at 28±1°C with a light:dark cycle of 16 h:8 h and a 80% relative humidity. Larvae were reared in pans containing 1 yeast tablet in 1 liter of tap water. Adults were provided with 10% sucrose solution *ad libitum* and fed three times a week on anaesthetized mice. Virus {#s2c} ----- The CHIKV 06.21 isolated in November 2005 from a new-born male from La Reunion Island presenting meningo-encephalitis symptoms was used for all experiments. This strain harbored the A-\>V mutation at the position 226 in the E1 glycoprotein (E1-226V) [@pntd.0000706-Schuffenecker1]. Viral stock used was a third passage on *Ae. albopictus* C6/36 [@pntd.0000706-Iragashi1] stored at −80°C in aliquots. Procedure for C6/36 cell infections and passages are described elsewhere [@pntd.0000706-Vazeille1]. The DENV-1 185/04 was isolated in May 2004 from the plasma of a patient in La Reunion Island. The strain belonged to the Brazil group of the Pacific genotype which was the main genotype isolated during the outbreak (GenBank: DQ285558.1). The virus was provided as a second passage on C6/36 cells. DENV-1 production on mosquito cell cultures being insufficient to allow mosquito oral infections, the virus stock was produced by inoculating intra-thoracically mosquitoes with the viral strain [@pntd.0000706-Rosen1]. Inoculated mosquitoes were incubated 10 days at 28°C and their infectious status checked by indirect immunofluorescent assay (IFA) on head squashes [@pntd.0000706-Kuberski1]. Bodies were then pooled and triturated in heated (56°C for 30 min) FCS (Fetal Calf Serum). The supernatant fluid recovered after low speed centrifugation was used as a source of virus in mosquito blood-meals. Both viruses were provided by the French National Reference Center for Arboviruses at the Institut Pasteur which had obtained the verbal consent from patients or parent\'s patients who provided blood sera. Intrathoracic inoculation of mosquitoes {#s2d} --------------------------------------- Adult females were inoculated using the protocol described by Rosen and Gubler [@pntd.0000706-Rosen1], each mosquito receiving 0.17 µL (i.e. 10^2.8^ PFU/mL) of the DENV-1 strain. Oral infection of mosquitoes {#s2e} ---------------------------- Infection assays were performed with 7 day-old females which were allowed to feed for 15 min through a chicken skin membrane covering the base of a glass feeder containing the blood-virus mixture maintained at 37°C. The infectious blood-meal was composed of a virus suspension diluted (1∶3) in washed rabbit erythrocytes isolated from arterial blood collected 24 h before the infectious blood-meal [@pntd.0000706-VazeilleFalcoz1]. A phagostimulant, ATP, was added at a final concentration of 5×10^−3^ M. Fully engorged females were transferred to small cardboard containers and maintained with 10% sucrose at 28±1°C for 14 days. Viral suspension provided in the blood-meal contained one or two viruses. For the secondary infection experiment, the blood-meal with CHIKV yielding 10^6^ FFU (foci forming unit)/mL was provided 7 or 13 days after inoculation with DENV-1. As control, females were inoculated with DENV-1 alone or orally infected with CHIKV alone. For the two trials where both viruses were provided by oral route, titers of the blood-meals were respectively: 10^6^ FFU/mL for CHIKV in both trials and 10^4,5^ FFU/mL for DENV-1 in the trial 1 and 10^5,9^ FFU/mL for DENV-1 in the trial 2. Saliva collection and virus detection/titration {#s2f} ----------------------------------------------- At day 14 post-infection, females were chilled, and their wings and legs removed, the stress inducing a forced salivation. Proboscis was inserted into 1 µL micropipette (microcaps®, Drummond Scientific Company, USA) filled with FCS. After 45 min, medium containing the saliva was expelled under pressure into 1.5 mL tubes containing Leibovitz L15 medium supplemented with 10% FBS. To allow a specific detection of both viruses, each sample was inoculated in two wells, one for the detection of CHIKV and one for the detection of DENV-1. 20 µL of each sample were added to monolayers of C6/36 cells in 24 wells plaque to detect infectious particles by the foci forming technique using an immunoperoxydase assay. Cells were incubated 3 days for CHIKV and 5 days for DENV-1 at 28°C under an overlay consisting of 50% of Leibovitz L-15 medium supplemented with 10% FBS and 50% of carboxyl methyl cellulose. Cells were then fixed with 3.6% formaldehyde at room temperature (RT) for 20 min and a immunoperoxydase assay staining was performed to detect foci. After a first incubation of 4 min with PBS 0.1% Triton X-100 (Sigma) at RT, cells were incubated 20 min at 37°C with a mouse ascitic fluid at a dilution of 1∶1000 for CHIKV and 1∶100 for DENV-1 (both ascetic fluids were provided by the French National Reference Center for Arbovirus at the Institut Pasteur). After a wash in PBS 1X, cells were incubated at 37°C for 45 min with peroxydase-conjugated goat anti-mouse IgG antibody (Pierce biotechnology, Rockford, USA) at a 1∶100 dilution in PBS 1X. After final wash in PBS 1X, Fast 3,3′ Diaminobenzidine (Sigma) was used to reveal the staining and foci were counted. The titer of infectious particles per saliva was expressed as FFU/mL (mean ± standard deviation). Female status analyzed by IFA {#s2g} ----------------------------- After salivation, females were tested for the presence of CHIKV and DENV-1 by IFA on their head squashes [@pntd.0000706-Kuberski1]. CHIKV and DENV-1 antigens were detected with the same mouse ascitic fluids used for saliva titration. Head squashes being performed between two slides, infection status of females fed with both viruses, could be checked for both antigens by using one slide for each IFA. Mosquitoes inoculated with CHIKV and DENV-1 were used as positive controls, negative controls were inoculated with cell culture media. Results {#s3} ======= Secondary infection with CHIKV of *Ae. albopictus* previously inoculated with DENV-1 {#s3a} ------------------------------------------------------------------------------------ As shown on [Table 1](#pntd-0000706-t001){ref-type="table"}, only few females inoculated with DENV-1 were willing to feed on an artificial blood-meal in the BSL-3 insectarium. When a blood-meal was proposed at day 7 post-inoculation, eight females out of 106 fed and among them, three survived until day 13 post-inoculation. By IFA on head squashes, we found that all 3 females had disseminated both viruses. Besides, when a blood-meal was offered at day 13 post-inoculation, eight females out of 54 females fed and four survived until day 20 post-inoculation. These four females had disseminated both viruses. Control females inoculated or orally infected by only one virus were all positive. 10.1371/journal.pntd.0000706.t001 ###### Superinfection with CHIKV of *Aedes albopictus* Providence previously inoculated with DENV-1. ![](pntd.0000706.t001){#pntd-0000706-t001-1} at day 7 post-inoculation at day 13 post-inoculation ---------------------------------- ------------------------------------ -------------------------------------- ----------------------------------- ------------------------------------ -------------------------------------- Engorged females at day 7 pi (N) Surviving females at day 13 pi (N) Co-infected females at day 13 pi (N) Engorged females at day 13 pi (N) Surviving females at day 20 pi (N) Co-infected females at day 20 pi (N) 8 (106) 3 (8) 3 (3) 8 (54) 4 (8) 4 (4) N, number of females tested. Oral co-infections {#s3b} ------------------ Females were exposed to both viruses in a same blood-meal and disseminated infection rates were estimated at day 14 post-infection (pi) ([Table 2](#pntd-0000706-t002){ref-type="table"}). In the trial 1, 71.6% of females have only disseminated CHIKV, 0% only DENV-1, 18.6% both viruses and 9.8% did not disseminate any virus. In the trial 2, 30.8% of females have only disseminated CHIKV, 7.7% only DENV-1, 50.8% both viruses and 10.7% did not disseminate any virus. When providing a higher titer of DENV-1 in the blood-meal (trial 2), a higher proportion of females were co-infected with both viruses. 10.1371/journal.pntd.0000706.t002 ###### Disseminated infection rates (%) of *Aedes albopictus* Providence 14 days after a blood-meal with both CHIKV and DENV-1. ![](pntd.0000706.t002){#pntd-0000706-t002-2} CHIKV DENV-1 CHIKV + DENV-1 Non-infected Total --------- ------- -------- ---------------- -------------- ------- Trial 1 71.6 0 18.6 9.8 102 Trial 2 30.8 7.7 50.8 10.7 65 Disseminated infection rates were estimated by IFA on head squashes. For CHIKV, both trials used a blood-meal at a titer of 10^6^ FFU/mL. For DENV-1, trial 1 corresponded to a blood-meal at a titer of 10^4.5^ FFU/mL and trial 2 to a titer of 10^5.9^ FFU/mL. Total corresponds to the total number of females tested. For females having disseminated both viruses, saliva was collected at day 14 pi and titrated. Relative transmission of the two viruses are shown in [Table 3](#pntd-0000706-t003){ref-type="table"}. In the trial 1, out of 19 saliva, 4 contained simultaneously CHIKV and DENV-1, 4 only CHIKV, 3 only DENV-1 and 8 no virus. In the trial 2, out of 33 saliva, 9 presented concomitantly CHIKV and DENV-1, 8 only CHIKV, 2 only DENV-1 and 14 no virus. Mean titers, expressed as FFU per saliva, and standard deviation are shown on [Table 4](#pntd-0000706-t004){ref-type="table"}. 10.1371/journal.pntd.0000706.t003 ###### Relative transmission of the two viruses by dually infected mosquitoes. ![](pntd.0000706.t003){#pntd-0000706-t003-3} DENV-1 ------- -------- --- --- --- ---- CHIKV Yes 4 4 9 8 No 3 8 2 14 Saliva from females having disseminated both viruses 14 days after an infectious blood-meal was tested for the presence of both viruses. For CHIKV, both trials used a blood-meal at a titer of 10^6^ FFU/mL. For DENV-1, trial 1 corresponded to a blood-meal at a titer of 10^4.5^ FFU/mL and trial 2 to a titer of 10^5.9^ FFU/mL. 10.1371/journal.pntd.0000706.t004 ###### Mean numbers ± standard deviations of infectious viral particles in saliva of *Aedes albopictus* Providence co-infected with CHIKV and DENV-1. ![](pntd.0000706.t004){#pntd-0000706-t004-4} Trial 1 (N) Trial 2 (N) ------------------------------------- -------- ------------- ------------- Saliva with both viruses detected CHIKV 16±10 (4) 44±75 (9) DENV-1 74±134 (4) 46±66 (9) Saliva with only one virus detected CHIKV 4±1 (4) 45±64 (8) DENV-1 4±1 (3) 17±0 (2) The saliva of females detected positive for both viruses by IFA on head squashes, 14 days after an infectious blood-meal were titrated. Titers were expressed in FFU (foci forming units). For CHIKV, both trials used a blood-meal at a titer of 10^6^ FFU/mL. For DENV-1, trial 1 corresponded to a blood-meal at a titer of 10^4.5^ FFU/mL and trial 2 to a titer of 10^5.9^ FFU/mL. N refers to the total number of females tested. Discussion {#s4} ========== We report here the ability of *Ae. albopictus* from La Reunion Island to replicate simultaneously autochthonous strains of DENV-1 and CHIKV provided in the same blood-meal and to deliver both infectious viral particles in their saliva. To our knowledge, such co-infection has never been shown neither under laboratory conditions nor in the field. Lastly, we succeeded in inducing a secondary infection with CHIKV 7 or 13 days after a first infection with DENV-1 virus. CHIKV and DENV are both transmitted by *Ae. aegypti* and *Ae. albopictus*, the former being considered the major vector and the latter, the secondary vector. However *Ae. albopictus* is able to sustain DEN outbreaks in the absence of *Ae. aegypti* [@pntd.0000706-Gratz1]. Indeed, in the Indian Ocean, *Ae. albopictus* was predominant in Seychelles and in La Reunion Island where the species took part of DEN outbreaks in 1976-77 [@pntd.0000706-Metselaar1], [@pntd.0000706-Kless1] and in 2005 [@pntd.0000706-Pierre1]. On La Reunion Island, *Ae. aegypti* populations are scarce and do not exhibit a high anthropophily [@pntd.0000706-Salvan1], [@pntd.0000706-Paupy1] while *Ae. albopictus* has favored the emergence of a new CHIKV strain harboring a substitution (alanine → valine) at the position 226 of the E1 glycoprotein during the 2005--2006 outbreak. This variant presents high levels of replication in *Ae. albopictus* [@pntd.0000706-Vazeille1] and a short extrinsic incubation period as the virus could be found in saliva as early as two days after infection [@pntd.0000706-Dubrulle1]. Subsequent outbreaks due to the new CHIKV variant were often related to transmission by *Ae. albopictus* corroborating an adaptative mutation in response to a requirement for transmission by this species. Moreover, co-infections with both DENV and CHIKV have been detected in some patients from La Reunion Island [@pntd.0000706-World1] and Madagascar in 2006 [@pntd.0000706-Ratsitorahina1]. Co-infections CHIKV-DENV in patients have been first described in 1967 and since the emergence of CHIKV in the Indian Ocean, reports of co-infections are increasing: in the Indian Ocean, as mentioned above, but also in Sri-Lanka [@pntd.0000706-Hapuarachchi1], Malaysia [@pntd.0000706-Nayar1], in India [@pntd.0000706-Chahar1], [@pntd.0000706-Schilling1] where the main vector involved is *Ae. aegypti* and in Gabon [@pntd.0000706-Leroy1]. Except for the Americas still free of CHIK infection, the geographic range of CHIK is now largely overlapping that of DEN. Furthermore, the emergence of CHIK outbreaks due to the new variant coincided with the recent establishment of *Ae. albopictus* in Central Africa, in Cameroon [@pntd.0000706-Fontenille1], [@pntd.0000706-Peyrefitte1] and Gabon [@pntd.0000706-Coffinet1], [@pntd.0000706-Peyrefitte2]. In 2007, patients with co-infections DENV-CHIKV were indeed detected for the first time in Africa in Gabon [@pntd.0000706-Leroy1]. Co-infection of a mosquito vector by two different viruses can occur by the way of two successive infectious blood-meals taken on two different viremic hosts or by one blood-meal taken on a co-infected host. We chose to mix both viruses in the same meal since *Ae. albopictus* females from La Reunion Island were reluctant to feed twice at 7 or 13 days interval on an infectious meal in the BSL-3 conditions. If *Ae. albopictus* females can be readily fed on an artificial meal in a BSL-3 laboratory it is very difficult to make them take a second artificial meal in the same conditions after 1 or 2 weeks of incubation and even to make then take a first meal if they have been kept more than 48 h in the depression conditions of the BSL-3. Temperature and humidity were optimal, females had the opportunity to lay their eggs if needed, were starved prior blood-feeding as usual. Therefore, we were unable to offer viruses in sequential meals and choose to test the possibility of a superinfection by first injecting one virus (DENV-1) then offering an infectious blood meal with the second virus (CHIKV) 7 or 13 days latter. Even then, we had very few individuals blood-fed but could however detect replication and dissemination of both viruses. *Ae. albopictus* from La Reunion Island were already known to be a very efficient vector for the new CHIKV variant and able to sustain DEN outbreaks. We demonstrated that replication of both viruses can take place simultaneously and vectors become able to transmit the two viruses in a single bite. In *Ae. albopictus*, CHIKV infectious particles can be found in saliva from two days after oral infection [@pntd.0000706-Dubrulle1] when this delay is much longer for DENV at least 10 days [@pntd.0000706-Gubler4]--[@pntd.0000706-Salazar1]. We performed the saliva analysis 14 days after feeding the females on a co-infected blood-meal but were not able to detect infectious viral particles in all saliva collected from females presenting a disseminated infection of both viruses. As CHIKV was not detected, when it should have been considering its rapid excretion in saliva [@pntd.0000706-Dubrulle1], females with negative saliva must have been females unable to salivate using our technique [@pntd.0000706-Dubrulle1] or excreting a very small amount of infectious particles that we could not detect due to the low sensitivity of the technique of viral detection. It would be interesting to perform a kinetic study of excretion of infectious particles from females exposed to one or both viruses while controlling the presence of saliva in the collected samples. We performed our study with *Ae. albopictus* from La Reunion Island with two viral strains isolated in the same area during the DENV-1 outbreak of 2005 and the CHIKV outbreak of 2005--2006. This combination vector/virus fits well with the natural context strengthening the findings of this work. Nevertheless, it seems difficult to infer to other epidemiological contexts as shown by the failure of Rohani *et al.* [@pntd.0000706-Rohani1] to prove co-infection of *Ae. aegypti* by CHIKV and DENV. The mechanism of co-infection in *Ae. albopictus* by two different arboviruses needs to be further investigated and tested using additional trials with different viral strains to determine if this phenomenon is an exception due to the particularly well adapted partners or a quite common mechanism. The still ongoing expansion of this species, particularly in Africa where numerous arboviruses are transmitted, in of particular concern and, if the co-infection was a quite usual phenomenon could have great implication on human health. It should be noted that superinfection is possible in mosquitoes as well as in cells infected with heterologous viruses (i.e., different genus) and not with homologous ones [@pntd.0000706-Beaty1]--[@pntd.0000706-Zou1]. Little is known about the molecular and cellular basis of co-infection which should be explored. Acknowledgments {#s5} =============== We are grateful to the DRASS in the Reunion for providing the ALPROV mosquito strain and to the French National Reference Center for Arboviruses for the E1-226V CHIKV and the DENV-1 strains. We also wish to thank Sara Moutailler for critical reading of the manuscript. We are grateful to Michèle Bouloy and Félix Rey for their support. The authors have declared that no competing interests exist. This work was funded by the ANR EntomoChik drive, the Fondation Duranton de Magny, and the Institut Pasteur (PTR 201). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: MV ABF. Performed the experiments: MV LM EM. Analyzed the data: MV ABF. Wrote the paper: MV ABF.

{ "pile_set_name": "PubMed Central" }

1. Introduction {#sec1} =============== Acute scrotal pain includes some urgent diagnoses, such as testicular torsion and incarcerated hernia or omentum. Traditionally, an in-depth interview and color Doppler ultrasonography (US) can lead to an acute diagnosis, although some cases exist where clinicians cannot eliminate emergency etiologies because the case history and physical examination are not performed accurately or the US revealed equivocal imaging. Here, we present a case where a presurgical diagnosis compromised incarcerated omentum, and the operation revealed that intense hydrocele inflammation was the cause of the symptoms. 2. Case Presentation {#sec2} ==================== A 14-month-old boy who had been diagnosed with a hydrocele in the right spermatic cord was referred to our department from the outpatient clinic with a right swollen scrotum. He presented to our hospital for the management of intermittent right scrotum pain for 4 days. He was in grumpy mood, his right scrotum was red and swollen, and the cremasteric reflux was absent ([Figure 1](#fig1){ref-type="fig"}). The urine test showed no white blood cells, and blood tests showed a slight increase in CRP (1.27 ng/mL). The US showed normal testes and epididymis with normal blood flow by the color Doppler US ([Figure 2(a)](#fig2){ref-type="fig"}). On the cranial of the right testis, low-echoic hydrocele with multiseptum was present, and its capsulizing wall was as thick as 3--5 mm ([Figure 2(b)](#fig2){ref-type="fig"}). In the right inguinal canal, omentum was present in the patient\'s processus vaginalis with blood flow by the color Doppler US ([Figure 2(c)](#fig2){ref-type="fig"}). However, we could not completely rule out the partial omental incarceration considering the sensitivity of US, and thus, emergent operative intervention was performed. Under inguinal exploration, processus vaginalis (PV) was present with the omentum inserted, but there was no evidence of necrotic or adherent tissue. PV was not communicating with hydrocele. Surgical findings did not indicate that the omentum itself was the cause of scrotal pain, and we next went on to the testis exploration. Upon opening the right tunica albuginea, the testis was normal colored, and the spermatic cord did not experience torsion. However, the hydrocele wall was ubiquitously adherent with the surrounding tissue and was much thicker and more solid than typical pediatric testicular hydrocele. The enlarged hydrocele wall was removed, and its edges were oversewn in the surgery. Pathology of the hydrocele wall showed edema in the wall, fibrin precipitation, and lymphocyte infiltration ([Figure 3(a)](#fig3){ref-type="fig"}), indicating that severe inflammation caused scrotal pain and redness. 3. Discussion {#sec3} ============= Acute scrotum is characterized by scrotal pain that has rapid or acute onset. The most serious condition in pediatrics is testicular torsion that requires emergent surgery to avoid testicular necrosis. As the sensitivity of the US against testicular torsion is as high as 88.9--100% \[[@B1], [@B2]\] and the US in this case showed clear blood flow into both testes, we had to rule out another disease that required emergent surgery---incarcerated intestine or omentum into the inguinal canal. Even the highest sensitivity of the US for incarcerated inguinal hernia is 91% \[[@B3]\], and thus, the possibility of an incarcerated hernia could not be eliminated. Therefore, we proceeded to emergent surgical intervention. Although the accurate etiology is uncertain, there are three reasons that suggest the painful red scrotum was derived from inflammation of hydrocele wall. First, the pathology findings revealed a thick hydrocele wall, a high number of inflammatory cells, and firm fibrin precipitation, those of which are not usually seen in typical pediatric hydroceles ([Figure 3(b)](#fig3){ref-type="fig"}). Second, there were no other differential diagnoses that could explain the cause of the pain. The other possible etiology was compression of the testicular vessels by a massive hydrocele or edematous incarcerated inguinal hernia \[[@B4]\]. These conditions are rare, with only five cases being reported to date \[[@B4]--[@B8]\]. In each of those cases, the preoperative US clearly showed reduction in testicular blood flow. Based on these findings, the effect of compression could be ruled out in our case. The third observation is the type of onset. Acute scrotum pain due to ischemia is caused by vessel obstruction; therefore, the onset should be sharp and sudden. In the present case, the patient was referred to our department after 4 days of pain initiation. Retrospectively, the time course suggests that the cause is not likely to be ischemic, although we could not be certain that the testis or omentum was intact. All these observations suggest that the cause of the pain was inflammatory changes in the hydrocele. The reason for hydrocele inflammation is uncertain. It is widely known that incarcerated hernias sometimes cause hydroceles. In our case, however, there was no ischemia of the omentum; nor was the hydrocele communicating with abdominal cavity, suggesting other reasons exist for hydrocele inflammation. Clinically, it was reported that omentum often form adhesions to inflamed regions such as the patient\'s PV \[[@B3]\]. In our case, possibly, omentum caused inflammatory changes in the PV and the inflammation spread to the hydrocele wall, although this is just a speculation. Pediatric hydroceles are common and often asymptomatic. To our knowledge, this is the first case of a pediatric hydrocele that caused acute scrotum pain. In retrospect, surgery was not necessary in this case; the intention was to ensure that we were ruling out tissue ischemia, a critical condition. Physicians should not hesitate to perform emergency surgery if the cause of scrotal pain cannot be determined with confidence. The authors appreciate Naomi Kasuga, who is a laboratory technician, for supporting tissue staining. Conflicts of Interest ===================== The authors declare that they have no conflicts of interest. ![The scrotum was red and swollen.](CRIM2018-2862514.001){#fig1} ![(a) Both testes were normal sized with normal blood flow. (b) The hydrocele wall was as thick as 3--5 mm (scale bar = 1 cm). (c) The omentum is inserted into the right inguinal canal. The blood flow was detected.](CRIM2018-2862514.002){#fig2} ![(a) The stroma was thick and edematous with fibrin precipitation and lymphatic cell infiltration. (b) Pediatric noncommunicating hydrocele aged at 1 year as a control. The tissue was not as edematous as in our case, and less lymphatic cells were infiltrated (scale bar = 100 *µ*m).](CRIM2018-2862514.003){#fig3} [^1]: Academic Editor: Bruno Megarbane

{ "pile_set_name": "PubMed Central" }

All relevant data are within the paper and its Supporting Information files. Introduction {#sec001} ============ In agriculture, insecticides are commonly used to control insect pest species. However, pest management is hindered by the increasing level of insecticide resistance in various pest species worldwide. One of the most serious insect pest species to evolve resistance against many insecticides from different chemical classes is the Old World cotton bollworm, *Helicoverpa armigera* (Hübner). Its pest status is further strengthened by its highly polyphagous nature and extremely wide geographical distribution, with the latter due to its ability to migrate long distances \[[@pone.0197760.ref001]--[@pone.0197760.ref004]\]. Recent work confirmed a major incursion of *H*. *armigera* into Brazil, as well as Argentina, Uruguay and Paraguay and it has been detected in Puerto Rico and Florida \[[@pone.0197760.ref005]--[@pone.0197760.ref008]\]. Pyrethroid resistance of *H*. *armigera* was first reported in Australian populations in 1983, six years after the introduction of these insecticides \[[@pone.0197760.ref009]\]. Since then, pyrethroid resistance has been reported from various populations globally but, in most cases, the molecular resistance mechanism in *H*. *armigera* is yet to be determined. The two major pyrethroid resistance mechanisms common to other species are target site insensitivity, and metabolic resistance by cytochrome P450 monooxygenases (P450s) and carboxylesterases \[[@pone.0197760.ref010], [@pone.0197760.ref011]\]. Target site resistance against pyrethroids, also known as knockdown resistance (*kdr*), is based on one or more point mutations in the insect sodium channel protein, which is the target of pyrethroid insecticides. In *H*. *armigera*, two mutations (D1549V and E1553G) in the sodium channel protein have been described \[[@pone.0197760.ref012], [@pone.0197760.ref013]\]. In the case of metabolic resistance, 18 P450s have been identified as capable of metabolizing one or more pyrethroid insecticides after heterologous expression \[[@pone.0197760.ref011], [@pone.0197760.ref014]--[@pone.0197760.ref019]\]. One of them is the chimeric *CYP337B3*, recently identified in Australian *H*. *armigera* and thought to have arisen through an unequal crossing-over between the parental genes, *CYP337B1* and *CYP337B2* \[[@pone.0197760.ref018]\]. Joussen et al. \[[@pone.0197760.ref018]\] demonstrated that, after heterologous expression, *CYP337B3* is capable of metabolizing fenvalerate, a type II pyrethroid ester, to the nontoxic 4\'-hydroxyfenvalerate. In contrast, the parental enzymes exhibit no detectable fenvalerate metabolism. *CYP337B3* has been shown to confer 42-fold resistance towards fenvalerate in Australian *H*. *armigera* lines differing solely in the presence of *CYP337B3* and its parental genes \[[@pone.0197760.ref018]\]. Similarly, a 49-fold resistance factor was reported by Forrester et al. \[[@pone.0197760.ref020]\] for an Australian field-collected population, indicating that metabolism of fenvalerate by *CYP337B3* is the detoxification mechanism *in vivo*. Recently, *CYP337B3* was also identified in field-collected populations from Pakistan \[[@pone.0197760.ref021]\] and China \[[@pone.0197760.ref022]\], confirming its presence outside Australia. Sequence analysis revealed a distinct *CYP337B3* allele (*CYP337B3v2*), in the Pakistani population and three distinct alleles in the Chinese populations (*CYP337B3v2*, *CYP337B3v3*, *CYP337B3v4*) that differ from the Australian allele (*CYP337B3v1*), by a number of synonymous and non-synonymous SNPs, in addition to variability of the intron sequence and size. This variation may result from different crossing-over positions during recombination of the *CYP337B1* and *CYP337B2* parental genes, with different alleles of *CYP337B1* and *CYP337B2* involved in the crossing-over. Such a pattern is indicative of independent origins of the various *CYP337B3* alleles, however further work would be required to determine whether this is accurate across a wide geographic distribution of samples. Given the high frequency of this allele around the world, it is possible that the population of *H*. *armigera* now established in South America \[[@pone.0197760.ref005], [@pone.0197760.ref023], [@pone.0197760.ref024]\] may be carrying this allele and that the possession of *CYP337B3* may confer a selective advantage over local species. Previous studies with allozymes and mitochondrial DNA indicated that very little population structure among *H*. *armigera* populations, though more recent work using many thousand genomic markers has shown that the population of *H*. *armigera* in Australia can be differentiated from the rest of the world \[[@pone.0197760.ref025], [@pone.0197760.ref026]\]. This lack of structure makes the identification of the *CYP337B3* chimeric origin difficult, as well as more generally hampering the identification of source populations in the case of incursions into the New World. However, examining patterns of distribution and diversity among recently selected genes can shed light on such processes. *CYP337B3* is a strong candidate for having undergone rapid selection due to its role in pyrethroid resistance, which has a short (approximately 50 year) history across much of the globe \[[@pone.0197760.ref027]\]. Here, we use field-collected populations from 17 countries distributed through Africa, Asia, Oceania, Europe, and South America to first examine the evolutionary history of the chimeric *CYP337B3* gene and the global spread in *H*. *armigera*. We use global allelic frequency data to show that *CYP337B3* is commonly found around the world, with at least six likely independent unequal crossing-over events giving rise to the same resistance phenotype in different regions. We also confirm recent data showing evidence of hybridisation between invasive *H*. *armigera* from Brazil and the local *Helicoverpa zea* population, including the exchange of the *CYP337B3* gene \[[@pone.0197760.ref026], [@pone.0197760.ref028]\]. Furthermore, we then use the different *CYP337B3* alleles to shed light on the nature of selection of *CYP337B3*, finding evidence to support recent selection operating on standing variation as opposed to novel mutations arising as a result of pyrethroid application. Materials and methods {#sec002} ===================== Population samples {#sec003} ------------------ A total of 1063 insect samples were used in the current study, including several specimens collected and used in previous work \[[@pone.0197760.ref005], [@pone.0197760.ref025], [@pone.0197760.ref026], [@pone.0197760.ref028], [@pone.0197760.ref029]\]. *H*. *armigera* was collected from 17 countries, *H*. *zea* from two ([Fig 1](#pone.0197760.g001){ref-type="fig"}; [S1 Table](#pone.0197760.s007){ref-type="supplementary-material"}). All insect samples were collected from agricultural areas with the permission of the landholder or by trapping using species-specific pheromones. No collecting permits were required because both species are common agricultural pests. No endangered or protected species were used in this work. Samples were usually collected as larvae from wild and crop host plants, as adult moths via light/pheromone traps or as larvae after bioassay. For one set of samples from China (n = 240), larvae were collected from colonies derived from eight recently collected field samples (established with 20 to 50 individuals). Collection dates range from 2002 to 2013. Samples were preserved in ethanol (\>95%), in RNAlater, or at -20°C prior to DNA extraction. DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen) or the Fast Tissue-to-PCR Kit (Fermentas, Thermo Scientific). ![Collection localities.\ *H*. *armigera* (n = 986) were sampled from 17 countries (black stars): Australia, New Zealand, China, Korea, India, Pakistan, Greece, Spain, France, Burkina Faso, Cameroon, Chad, Ghana, Madagascar, Senegal, Uganda, and Brazil. *H*. *zea* were sampled from Brazil (n = 43) and the United States of America (n = 16). Modified from public domain map from Natural Earth (<http://www.naturalearthdata.com>).](pone.0197760.g001){#pone.0197760.g001} The species status of several preserved specimens was confirmed by mitochondrial gene (*COI* and *Cytb*) sequencing, either from previous work \[[@pone.0197760.ref005], [@pone.0197760.ref029]\] or, where new samples were available, by amplifying and sequencing the same regions ([S2 Table](#pone.0197760.s008){ref-type="supplementary-material"}). PCR amplification followed the protocols of Behere et al. \[[@pone.0197760.ref029]\] and Tay et al. \[[@pone.0197760.ref005]\]. PCR products were sequenced at Macrogen (Seoul, Korea) and the Biological Resources Facility (Australian National University, Canberra, Australia). Assembly of DNA trace sequences was performed using CLC Genomics Workbench version 8.0. Screening for *CYP337B3* {#sec004} ------------------------ Following species identification, samples were screened for the presence of *CYP337B3* using the CYP337B2F and CYP337B1R primers described in Joussen et al. \[[@pone.0197760.ref018]\]. Heterozygote/homozygote status was determined through relevant band detection on 1.5--2% agarose gels containing 1% (w/v) of GelRed (Biotium) and visualised under UV light. Initial sequence was generated from these short fragments for a subset of samples following the PCR amplification protocol of Joussen et al. \[[@pone.0197760.ref018]\]. However, in order to generate more information, primers were designed to amplify the intron of the gene ([S2 Table](#pone.0197760.s008){ref-type="supplementary-material"}). In the chimeric *CYP337B3* gene, *CYP337B1* contributes the intron and several hundred base pairs of coding sequence depending on the allele, while *CYP337B2* contributes much of the 5\' end of the coding sequence. Thus, the *CYP337B3* intron corresponds to the *CYP337B1* intronic sequence. We also screened samples from Brazil (n = 181) and Australia (n = 97) for the reciprocal recombination event using CYP337B1F and CYP337B2R primers ([S2 Table](#pone.0197760.s008){ref-type="supplementary-material"}) but no products were detected with this primer combination. Further amplification of full-length genes and transcripts was performed for selected individuals with representative alleles. PCR products and cloned fragments were sequenced at several institutes, including: Macrogen (Seoul, Korea), Biological Resources Facility (Australian National University, Canberra, Australia), UC Davis Genome Center (University of California, Davis, US), StarSEQ (Mainz, Germany), Bioneer Corporation (Daejeon, Korea) and the Department of Entomology at the Max Planck Institute for Chemical Ecology (Jena, Germany). Assembly and sequence analysis of the PCR products was done using CLC genomics workbench v8.0 and T-Coffee (<http://www.ebi.ac.uk/Tools/msa/tcoffee/>). Resistance bioassays {#sec005} -------------------- To establish the role of *CYP337B3* in pythrethroid resistance, fenvalerate field resistant and susceptible individuals were identified by bioassay of field-collected material. Fenvalerate (95.3%) was provided by Sumitomo Chemical (Sydney, Australia) and dissolved in analytical grade acetone to produce a diagnostic concentration (0.125 μg/μL). A 1 μL volume of the discriminating concentration was applied to the dorsal thorax of late 3^rd^/early 4^th^ instar larvae within a weight range of 30-40mg using a 50 μL micro-syringe in a repeating dispenser (Hamilton Company, Reno, NV, USA). Bioassays were maintained for 7 d at 25°C with a 14:10 (L:D) h cycle. Relative humidity (RH) was not controlled. Dead and moribund larvae were collected and described as sensitive by using one or more of the following criteria: larvae unable to demonstrate coordinated movement when prodded; paralysis of prolegs; larvae very slow to right themselves (time exceeding 3 s). Larvae that were actively feeding and developing normally were described as resistant. Population genetic analysis {#sec006} --------------------------- The elimination of variation in regions linked to an adaptive allele on one or even a few haplotypes is referred to as a "selective sweep" \[[@pone.0197760.ref030]\]. This process will lead to low nucleotide diversity (π) across the region of the genome under selection as well as an excess of rare variants as highlighted by Tajima's D \[[@pone.0197760.ref031]\]. This statistic identifies deviation from a neutral model of evolution, whereby a negative score signifies purifying selection or population expansion that one would expect following a selection event. To provide evidence that the *CYP337B3* gene is under selection in *H*. *armigera*, we searched for genomic signals of a recent selection event using high-throughput sequencing data. DNA samples from 12 Australian *H*. *armigera* homozygous for *CYP337B3v1* were sequenced on an Illumina HiSeq. Briefly, Nextera libraries were produced following the manufacturer's instructions and sequence was generated as 100 bp PE reads (Illumina HiSeq 2000, Biological Resources Facility, Australian National University, Canberra, Australia). Raw reads were aligned to the bacterial artificial chromosome (BAC) 33J17 (JQ995292.1) sequence using BBMAP v33.43 (<http://sourceforge.net/projects/bbmap/>), trimming reads when quality in at least 2 bases fell below Q10. We did not align reads to the reference *H*. *armigera* genome sequence, which contains only *CYP337B1* and *CYP337B2* but not *CYP337B3* \[[@pone.0197760.ref032]\]. Only uniquely aligning reads were included in the analysis, to prevent spuriously inferring evolutionary processes occurring independently on each BAC. Outputted BAM files were sorted before duplicate reads were removed and files were annotated with read groups using Picard v1.138 (<http://picard.sourceforge.net>). The reference sequences were indexed using Samtools \[[@pone.0197760.ref033]\]. UnifiedGenotyper in GATK v3.3--0 \[[@pone.0197760.ref034]\] was used to estimate genotypes across all individuals simultaneously, implementing a heterozygosity value of 0.01. Genotypes annotated as "LowQual" were removed prior to subsequent analysis with VCFtools v0.1.12b \[[@pone.0197760.ref035]\], whereby π and Tajima's D were calculated in sliding windows of 2500 bp that progressed by 1250 bp across biallelic sites of 33J17. Results were plotted in R v3.1.2 using ggplot2 v1.0.1 \[[@pone.0197760.ref036]\], while gene annotations were derived via tblastx \[[@pone.0197760.ref037]\] and visualised with CLC Genomics workbench v8.0. To examine potential mechanisms underlying the origin of the *CYP337B3* alleles, an intron dataset containing sequences for representative *CYP337B1* and *CYP337B3* individuals (since the chimeric *CYP337B3* intron comes from the *CYP337B1* parental gene; see above) was created. MAFFT ver. 7.182 \[[@pone.0197760.ref038]\] was used to align these sequences using the -linsi option. Subsequently, IQ-TREE ver. 1.3.0 \[[@pone.0197760.ref039]\] was used, first in the -m TESTNEWONLY mode to determine the appropriate nucleotide substitution model, and then in full mode to generate a maximum likelihood tree. Subsequently 10,000 bootstrap replicates were carried out, and the bootstrap values transferred from the consensus tree to corresponding nodes of the maximum likelihood tree. BEAST ver. 2.3.0 \[[@pone.0197760.ref040]\] was used for the intron-only dataset to test whether the *CYP337B3* allele is monophyletic in origin (i.e., has arisen once from a *CYP337B1*/*CYP337B2* cross-over and subsequently diverged independently). Two analyses were performed--one, in which monophyletic taxon sets for each of the *CYP337B1* and *CYP337B3* introns was enforced, and the other in which no monophyly was assumed. In BEAST, all runs used the evolutionary model identified in IQ-TREE (HKY + G4), a constant population coalescent tree prior, and a chain length of 10 x 10^6^. Other coalescent tree priors were also tested, but these did not quantitatively impact the results. BEAST results were examined in Tracer to confirm convergence and check ESS values (all exceeded 1,500). The likelihood of each analysis was then compared using Bayes Factors and the ACIM model, with 1,000 replicates, in Tracer ver. 1.6 \[[@pone.0197760.ref041]\]. Results {#sec007} ======= A total of 1063 insect samples was collected from 18 countries around the world ([Fig 1](#pone.0197760.g001){ref-type="fig"}). The identification of *H*. *armigera* was initially based on morphological assessments by collectors and subsequently confirmed with the mitochondrial markers, *cytochrome oxidase I* and *cytochrome b* \[[@pone.0197760.ref029]\]. Any individuals not identified as *H*. *armigera* by mitochondrial DNA were excluded from subsequent analysis (*H*. *armigera* n = 999, *H*. *zea* n = 59, *Chloridea* (*Heliothis) virescens* n = 5). A PCR assay using one primer from each of the two parental genes *CYP337B1* and *CYP337B2* was used to identify the presence of the chimeric *CYP337B3* gene. *CYP337B3* was extremely common in field collected *H*. *armigera* throughout its native range ([Fig 2](#pone.0197760.g002){ref-type="fig"} and [S1](#pone.0197760.s001){ref-type="supplementary-material"}--[S5](#pone.0197760.s005){ref-type="supplementary-material"} Figs). From a total of 999 individuals tested, *CYP337B3* was present in 969 (97%) and only 30 (3%) individuals were homozygous for *CYP337B1* or *CYP337B2*. 878 individuals were *CYP337B3* homozygotes (88%) and 91 were heterozygotes, meaning they carried one chromosome carrying *CYP337B3* and the other chromosome with the two parental genes *CYP337B1* and *CYP337B2* (9%). In some countries, almost all individuals were at least heterozygous for the *CYP337B3* gene, but homozygous *CYP337B3* individuals were the most common everywhere. In India, Pakistan, Korea, China, and Australia, almost all individuals were homozygous for *CYP337B3*, while heterozygotes (*CYP337B1*, *CYP337B2* and *CYP337B3*) at the *CYP337B3* locus were more common in European and African samples ([Fig 2](#pone.0197760.g002){ref-type="fig"} and [S1](#pone.0197760.s001){ref-type="supplementary-material"}--[S5](#pone.0197760.s005){ref-type="supplementary-material"} Figs). ![The frequency of *CYP337B* genotypes in selected geographic regions.\ For a more detailed breakdown see [S1 Fig](#pone.0197760.s001){ref-type="supplementary-material"} through [S5 Fig](#pone.0197760.s005){ref-type="supplementary-material"}.](pone.0197760.g002){#pone.0197760.g002} The PCR product of *CYP337B3* used in the screening PCR reaction \[[@pone.0197760.ref018]\] was sequenced from positive individuals (alleles n = 460) and a pattern of single nucleotide polymorphisms was identified. By comparing this short (355 bp) sequence between individuals, 6 alleles (*CYP337B3v1-2*, *5--8*) were identified worldwide ([Fig 3](#pone.0197760.g003){ref-type="fig"}). An allele originally identified in an Australian (Toowoomba) strain \[[@pone.0197760.ref018]\] was found only in Australia in the current study, where it was the most common allele (*v1*: n = 91), and *v2* the next most common allele (*v2*: n = 23). In Asian samples, *v2* was the most common allele (n = 203) and two other rare alleles were identified (*v7* and *v8*; n = 1 each) in addition to *v3* and *v4* identified by Han et al. \[[@pone.0197760.ref022]\]. Three alleles were found in Africa including two that were new (*v2*: n = 11, *v5*: n = 79, *v6*: n = 8). In the European samples, we identified an approximately 50:50 ratio of *CYP337B3v2* and *v5* (v2: n = 20; v5 n = 17). In addition, a proportion of samples from Australia (n = 11), Africa (n = 2), and Europe (n = 7) appeared from the sequencing traces to be heterozygous for different *CYP337B3* alleles. Overall, the various *CYP337B3* alleles appeared to show distinct patterns for each region, suggesting different alleles are present for pyrethroid resistance around the world. ![Frequencies of the different alleles of *CYP337B3 (*B3vX*)* in different geographic regions.\ Africa = Burkina Faso, Cameroon, Chad, Ghana, Madagascar, Senegal, Uganda; Australasia = Australia and New Zealand; Europe = Greece, France and Spain. n = the number of alleles sequenced, homozygotes and heterozygote alleles summed.](pone.0197760.g003){#pone.0197760.g003} Sequencing of the entire coding sequences of all variants revealed a number of non-synonymous SNPs resulting in different predicted proteins when compared to *CYP337B1v1*: 3 in *v2* \[[@pone.0197760.ref021]\], 3 in *v5*, 8 in *v6*, 9 in *v7*, and 11 in *v8*. All variants of the chimeric gene appear to show a similar pattern with slightly different cross over points between *CYP337B1* and *CYP337B2* during the formation of *CYP337B3* and different intron sequences and sizes that derived from *CYP337B1* determined by sequencing genomic DNA ([Fig 4](#pone.0197760.g004){ref-type="fig"}; [S6 Fig](#pone.0197760.s006){ref-type="supplementary-material"}; [S3 Table](#pone.0197760.s009){ref-type="supplementary-material"}). This suggests that multiple origins of the same functional phenotype from different genetic backgrounds define the development of this chimeric gene. ![A schematic of the different crossing-over points and intron sizes observed in the chimeric *CYP337B3* gene.](pone.0197760.g004){#pone.0197760.g004} All of the 93 *H*. *armigera* individuals collected in Brazil were positive for *CYP337B3*, and all except one were homozygous. The majority of *CYP337B3* alleles detected in Brazilian *H*. *armigera* were identical to *CYP337B3v2*, which is predominantly found in Asia. However, the single heterozygous individual was carrying an African allele (*CYP337B3v5*). This suggests that Asia may be the most likely source for the Brazilian incursion of *H*. *armigera*, however, other source populations are also possible, as is the occurrence of multiple incursions. As well as examining *H*. *armigera*, we also examined *H*. *zea* from Brazil (n = 59) and we were able to amplify *CYP337B3* from one individual from Goias while from the others (n = 23) only *CYP337B1* provided consistent results. DNA from the suspect individual ([S1 Table](#pone.0197760.s007){ref-type="supplementary-material"}) was re-extracted twice in a different laboratory, using new reagents (extraction kit, primers, PCR reagents) and PCR was performed for the *CYP337B3* gene, intron and the mitochondrial marker COI. Extraction blanks were performed and were negative for both the COI sequence and the *CYP337B3*. Sequencing of both fragments from the three different extractions confirmed the presence of the *CYP337B3v2* allele (as found in the Brazilian *H*. *armigera*) in an individual with a mitochondrial sequence from *H*. *zea*. Bioassays for pyrethroid resistance on *H*. *armigera* from cotton growing areas in eastern Australia using the discriminating dose clearly showed a higher frequency of homozygotes for *CYP337B3* in survivors (100% *CYP337B3* homozygous; n = 26) than in dead or dying insects (9 *CYP337B3* homozygotes, 2 heterozygotes and 1 homozygous *CYP337B1*/*B2* individual; n = 12), although a subset of the dead or dying caterpillars was actually positive for *CYP337B3*. In addition to bioassayed individuals, samples were collected from crops in Kununurra, Ord River Irrigation Scheme, Western Australia seven days after a field application of fenvalerate on cotton. Survivors were found to possess *CYP337B3*, with a 100% homozygous frequency (n = 7). Insects collected from the same region on unsprayed crops (cotton, chickpea and chia) showed a lower frequency of *CYP337B3* (2 homozygote *CYP337B3*, 1 heterozygote and 3 homozygote *CYP337B1*/*B2* individuals; n = 6). Although these very small samples from field populations are insufficient for claims of statistical significance, they are consistent with the significant genotypic differences previously found in the Toowoomba laboratory population derived from Australia \[[@pone.0197760.ref018]\]. Using individuals from Australia that were homozygous for the *CYP337B3v1* allele, high throughput sequencing data was aligned to the BAC clone *33J17* which was the original source of the *CYP337B3v1* identification and originates from the Australian Toowoomba strain \[[@pone.0197760.ref018]\]. Of the 100,377 genotypes called across *33J17*, 32,381 remained after "LowQual" SNPs were removed. Both π and Tajima's D were variable, however values were decisively differentiated where *CYP337B3* lies; nucleotide diversity reaches its lowest point at the *CYP337B3* locus (0.00057) relative to the BAC-wide mean at 0.0179 ([Fig 5](#pone.0197760.g005){ref-type="fig"}; [S4 Table](#pone.0197760.s010){ref-type="supplementary-material"}). This is the expectation following a "hard" selective sweep, whereby adaptive alleles appear as a single haplotype before increasing in frequency \[[@pone.0197760.ref030]\]. Tajima's D is also lowest across the region where *CYP337B3* is located (-2.18) and points towards the process of purifying selection. It is worth noting that the BAC-wide mean value for Tajima's D is -0.43; such a low negative value may signify that other evolutionary processes have also acted upon this region \[[@pone.0197760.ref042]\]. While this variability is likely a result of small sample sizes and relatively small sliding window values, the small size (\~ 5 kbp) of the region around *CYP337B3v1* may signify ongoing high recombination rates and gene flow \[[@pone.0197760.ref043], [@pone.0197760.ref044]\]. ![Selective sweep among Australian *H*. *armigera* homozygous for *CYP337B3v1*.\ (a) Nucleotide diversity (π) and (b) Tajima's D calculated across sliding windows of the 33J17 BAC (JQ995292.1), with the red dashed line showing the average value of π (0.0179). Locations of gene bodies are indicated between the plots, with those in red identified as potential reverse transcriptases and those in yellow as exons of *CYP337B3v1*.](pone.0197760.g005){#pone.0197760.g005} To establish the potential mechanisms that underlie the creation of the chimeric gene and account for its global distribution, we sequenced the intron of *CYP337B3*. This region should contain more polymorphisms and allow a better diagnosis of the frequency of the different genotypes. As for the coding region data (above), we found a pattern of eight alleles in the intron data. Introns sequenced from *B3v3*, *v4* and *v8* are very similar and differ only by two SNPS whereas introns from *B3v1*, *v2*, *v5*, *v6* and *v7* show more significant differences in both length and sequence ([Fig 4](#pone.0197760.g004){ref-type="fig"}). Individuals that were homozygous for the *CYP337B3* gene were predominantly homozygous at the allele level, but as in the coding region, heterozygotes were again observed in the sequence traces of a small number of samples from Africa, Australia and Europe. Cloning and sequencing of the heterozygous PCR products showed that, in Australia, *v1/v2* (Australia/Asia) heterozygotes were present, while in Africa, there were heterozygotes of *v2/v5* (Asia/Africa) and *v5/v6* (both African). European heterozygotes possessed *v2/v5* (Asia/Africa). Phylogenetic relationships were inferred for the intronic region of a subset of representative *CYP337B1* and *CYP337B3* sequences. If the chimeric *CYP337B3* were due to a single cross-over event followed by subsequent divergence of allelic variations, then the phylogenetic estimates should show a monophyletic clade with the *CYP337B3* clade nested within the *CYP337B1 clade*. Our tests with BEAST indicated that this was not the case; specifically Bayes Factor (BF) analysis found overwhelming support for a paraphyletic origin of *CYP337B3* (BF score for paraphyly *vs*. monophyly = 494.412, with BF \> 10 indicating very strong evidence against the null model of monophyly). This does not rule out the possibility of some subsequent recombination between derived *CYP337B3* alleles and either parent gene. [Fig 6](#pone.0197760.g006){ref-type="fig"} presents the results of the maximum likelihood phylogenetic analysis, which also suggests multiple origins of the various *CYP337B3* alleles. Although bootstrap support values for some nodes of the tree are quite low, several well-supported relationships are present, and these are consistent with multiple geographic origins for the *CYP337B3* alleles. For example, *CYP337B3v1* (found only in Australia) is most closely related at the sequence level to a *CYP337B1v1* allele from Australia. Furthermore there is a cluster of closely related but rare *CYP337B3* alleles found in China (*CYP337B3v3*, *v4*, and *v7*) which could indicate local origins for those alleles. Thus, while some *CYP337B3* alleles may be recently derived from local *CYP337B1* alleles, the overall signal indicates that multiple unequal crossing-over events are responsible for the geographic distribution of chimeric *CYP337B3* while gene flow as well as some intergenic recombination may also have played a role producing the current geographic distribution patterns of *CYP337B3* alleles. ![Maximum likelihood phylogeny of CYP337B1 and CYP337B3 alleles.\ The maximum likelihood tree was generated in IQ-TREE with the HKY+G4 model of sequence evolution, using representative *CYP337B1* and *CYP337B3* intron sequences. Bootstrap values from 10,000 replicates are given on tree nodes. The eight *CYP337B3* alleles are indicated with labels (v1:v8), different CYP337B1 alleles are indicated by B1- followed by an allele ID and the country of origin. Sequences and related NCBI identifiers can be found in [S3 Table](#pone.0197760.s009){ref-type="supplementary-material"} and [S1 File](#pone.0197760.s011){ref-type="supplementary-material"}.](pone.0197760.g006){#pone.0197760.g006} We performed an additional calculation to determine the molecular rate that would be required to produce the measured *CYP337B3* allelic diversity if it had all arisen in the last \~50 years. Even using the rapidly evolving mitochondrial divergence rate of 1.5--2.5% divergence per million years as a baseline \[[@pone.0197760.ref045]\], we found that a 1000-fold increase in this rate would be required to place all *CYP337B3* divergence events inside a 50-year time window. This rate is so high, it suggests that rather than recently arising through *de novo* mutation in response to pyrethroid applications and then spreading around the world, cytochrome P450-mediated pyrethroid resistance is most likely the result of selection acting over the last 50 years on existing genetic variation present in the populations at low frequency. Discussion {#sec008} ========== *Helicoverpa armigera* is one of the world's most destructive pests and has developed resistance to numerous pesticides over the years \[[@pone.0197760.ref027]\]. Globally, pyrethroids are one of the most commonly used insecticides and resistance occurred rapidly at a global scale after the first use of pyrethroid insecticides in the late 1970s \[[@pone.0197760.ref027]\]. A number of mechanisms have been proposed for this resistance, including target site insensitivity \[[@pone.0197760.ref027], [@pone.0197760.ref046], [@pone.0197760.ref047]\], and metabolism by carboxylesterases \[[@pone.0197760.ref048], [@pone.0197760.ref049]\] and P450s \[[@pone.0197760.ref050]--[@pone.0197760.ref056]\]. Recent work suggests that a chimeric gene, *CYP337B3*, may play an important role in cytochrome P450-mediated pyrethroid resistance and we set out to examine the frequency of this gene worldwide \[[@pone.0197760.ref018], [@pone.0197760.ref021]\]. We detected the chimeric *CYP337B3* gene in every population examined and found a total of eight allelic variants of *CYP337B3* globally. Of the eight global alleles identified, one was found only in Australia (*CYP337B3v1*), one was dominant across Asia (India, Pakistan, China, Korea), and Brazil (*CYP337B3v2*), four were rare and uniquely found in China (*CYP337B3v3*, *v4*, *v7 and v8*) and two were predominantly African (*CYP337B3v5*, *v6*). European samples showed a mixture of alleles (*CYP337B3v2*, *v5*). *CYP337B3v2* was the most widespread allele, identified on every continent examined but was present at the highest frequencies in Asia and Brazil. One issue in using *CYP337B3* as a population marker is that diversity in *CYP337B3* is not new diversity, but rather a re-shuffling of diversity already present in the parental *CYP337B1* and *CYP337B2* genes. However, the presence of multiple alleles at the *CYP337B3* locus suggests that this chimeric gene has formed independently several times over the evolutionary history of the locus. In the overall patterns, we found that three alleles, *v1*, *v2* and *v5*, together account for \~98% of all detected *CYP337B3* alleles. The geographic distribution of these common alleles may suggest older recombination events in Australia, Asia, and Africa producing the common alleles and potentially more recent additional events in Asia and Africa producing the rare alleles. The alternative explanation that a single event has occurred, and then been distributed globally by migration and subjected to subsequent local recombination, is much less likely as it is not supported by the phylogeny or the molecular rate analysis. Indeed, our phylogenetic analysis found strong support for a paraphyletic origin of *CYP337B3* rather than the alternative of a single unequal cross-over event followed by subsequent allelic divergence. In addition, our calculations suggest that a 1000-fold increase in a commonly accepted baseline divergence rate (1.5--2.5% divergence per million years) would be required to place all *CYP337B3* divergence events inside a 50-year time window \[[@pone.0197760.ref045]\]. Taken together, the allele frequencies and phylogenetic analysis suggest that, rather than recently arising through *de novo* mutations in response to pyrethroid application, cytochrome P450-mediated pyrethroid resistance is most likely the result of recent selection acting on existing genetic variation maintained in the population at low frequency. Our selective sweep analysis supports this scenario, with patterns of estimated π and Tajima's D across the *33J17* BAC indicating that the *CYP337B3* locus has undergone recent selection, at least in Australia and with regard to *CYP337B3v1*. Recent work confirms that this can also be shown for the other alleles \[[@pone.0197760.ref026]\]. Further work will reveal how the diversity in this one region compares to the rest of the genome. There are a number of examples in the literature where the same insecticide resistance mechanisms have developed independently in different locations. In particular pyrethroid resistance (perhaps because of the short period between the introduction and the development of resistance) associated with identical point mutations in the voltage-gated sodium channel has developed in different haplotypes with different geographical origins in mosquitoes and houseflies \[[@pone.0197760.ref057], [@pone.0197760.ref058]\]. In the sheep blowfly, the same mutations in an esterase gene conferring organophosphate resistance were identified in very different haplotypes \[[@pone.0197760.ref059]\]. Although further work incorporating whole genome scale markers will be necessary to determine the origin of *H*. *armigera* in the New World and global patterns of movement in this species \[[@pone.0197760.ref026], [@pone.0197760.ref028]\], we gain insight into some of these processes here. The vast majority of the Brazilian insects examined possessed the allele *CYP337B3v2* found predominantly in Asia. However, we also detected the *CYP337B3v2* allele in other regions of the world, and so it is possible that the Brazilian populations of *H*. *armigera* could have come from localities where the *CYP337B3v2* allele also exists. For example, the closest native population of *H*. *armigera* to the New World is in West Africa and it is possible that the New World *H*. *armigera* are the result of a similar incursion from West Africa. However, only Asia has close to the same proportion of *CYP337B3v2* alleles. A recent survey \[[@pone.0197760.ref008]\] showed that the pattern of mitochondrial markers in *H*. *armigera* populations collected from southern Brazil, Argentina, Uruguay and Paraguay is very different from *H*. *armigera* populations from central and northern regions of Brazil, suggesting multiple incursions of this global pest into the New World \[[@pone.0197760.ref060]\]. It would be interesting to examine the *CYP337B3* locus in these individuals to see if they share the same pattern as found in Brazil. Recently, evidence for other pyrethroid detoxicative mechanisms in addition to *CYP337B3* has been found in China \[[@pone.0197760.ref022]\], and our own data showed some mortality in *CYP337B3-* homozygous individuals. This might suggest that possession of the *CYP337B3* is not enough for resistance alone and other mechanisms related to expression or other genes may exist. However, as the resistance based on *CYP337B3* is metabolic, it is not unusual that a high dose of insecticide will kill even homozygous individuals, for example if the enzyme is not abundant enough to decrease the amount of insecticide at its target site beneath the toxic threshold. Furthermore, we clearly see evidence of conservation of the *CYP337B3* alleles at this locus across the globe and we would not expect the observed level of conservation in the absence of selection, particularly in the non-coding regions. Indeed, when we examined individuals from Australia we found evidence of a selective sweep around the gene, indicating maintenance of the allele at this locus. Throughout its native range, phenotypic pyrethroid resistance is very common in *H*. *armigera* and it is of course possible that other or additive resistance mechanisms are at work in populations around the world. The arrival of *H*. *armigera* in the New World represents a significant expansion of its geographc range. It remains to be seen how large the problem *H*. *armigera* will become but, clearly, arriving with a resistance gene already in place would provide *H*. *armigera* a selective advantage over local susceptible populations of pests such as *H*. *zea* and *Chloridea* (*Heliothis*) *virescens*. One further complication is that *H*. *zea* has been shown to interbreed with *H*. *armigera* in the laboratory to produce viable offspring \[[@pone.0197760.ref061]--[@pone.0197760.ref063]\] and there is evidence of hybridisation in Brazil using genome wide markers \[[@pone.0197760.ref028]\]. In the current study, we applied the same PCR reaction conditions to individuals identified as *H*. *zea* from Brazil and we able to amplify *CYP337B3v2* from one individual. Clearly this one result is not enough to suggest that hybridisation is widespread between the two species but the detection of a presumably highly selectable gene indicates that this is something that should be considered a risk. While it is unclear what risk hybrids may pose, it is certain that an exchange of resistance and host association genes could lead to significant problems in control. The fact that the presence of the chimeric resistance gene *CYP337B3* can be ascertained by a simple screening PCR enables the rapid determination of susceptible *vs*. pyrethroid-resistant populations, which could lead to improved pest management strategies and the detection of hybrid individuals in the field. This work highlights a case of resistance alleles in *H*. *armigera* evolving on a local scale in a global context and provides an example of a species invading new continents, already able to overcome some of the control measures deployed for other species. We also show that resistance genes can arise locally and then mix on a global scale as a result of either natural movement or human-mediated transport. The implications for *H*. *armigera* and potentially *H*. *zea* as species are unclear but what is not in doubt is that the pesticide resistance selection pressure on *H*. *armigera*, in terms of numbers exposed, has dramatically increased with this incursion. This is true not just for conventional pesticide resistance but also for transgenic insecticidal traits. Transgenic cotton and especially corn and soybean which express Bt toxins are grown over a far larger area in the New World and the management of Bt and conventional pesticide resistance should be a priority. Ultimately, we grow the same crops around the world, control pests using the same chemistries, and transport commodities through global trade routes, so it is not a surprise that we are developing global, well-adapted pests. Our paper not only sets the scene for further work in this area but also highlights the value of considering biosecurity threats in terms of both the actual pest as well as its resistance profile (both phenotype and genotype). Supporting information {#sec009} ====================== ###### The frequency of *CYP337B3* on the Asian Sub-continent by region. Northern, Central and Southern represent general geographic regions of India. (TIF) ###### Click here for additional data file. ###### The frequency of *CYP337B3* in selected African countries. Homozygous and heterozygous frequencies are shown. (TIF) ###### Click here for additional data file. ###### The frequency of *CYP337B3* in China and Korea. \*Larvae from lab colonies derived from each field collection (20--50 individuals) and maintained at a population size of 300--500 adults per generation without exposure to any insecticide for 3--5 generations. (TIF) ###### Click here for additional data file. ###### The frequency of *CYP337B3* in Australia and New Zealand. Homozygous and heterozygous frequencies are shown. (TIF) ###### Click here for additional data file. ###### The frequency of *CYP337B3* in Europe. Homozygous and heterozygous frequencies are shown. (TIF) ###### Click here for additional data file. ###### Alignment of the crossover points for the various *CYP337B3* alleles. *CYP337B1* is in red, *CYP337B2* is green and the boxed sequence represents crossover point for each *CYP337B3* allele. (TIF) ###### Click here for additional data file. ###### Collection sites of *H. armigera* and *H. zea* from 18 different countries. (DOCX) ###### Click here for additional data file. ###### PCR primers used in this work. (DOCX) ###### Click here for additional data file. ###### Sequence identifiers, gene names, origin and NCBI numbers used for the maximum likelihood estimate of phylogenetic relationships. (DOCX) ###### Click here for additional data file. ###### Sequencing and alignment statistics for the 12 individuals used to demonstrate the selective sweep around *CYP337B3v1*. (DOCX) ###### Click here for additional data file. ###### Intron sequence data used for the maximum likelihood estimate of phylogenetic relationships. (DOCX) ###### Click here for additional data file. We thank George Mironidis (IMBB-FORTH, Greece), Athanase Badolo (University of Ouagadougou), Juil Kim (National Institute of Horticultural and Herbal Science, Korea), and Muhammad Ashfaq (National Institute for Biotechnology and Genetic Engineering, Pakistan; now: Biodiversity Institute of Ontario, Canada) for sampling and providing insects for this study from Greece, Burkina Faso, Korea, and Pakistan, respectively; Hae Kyung Kwon (National Institute of Horticultural and Herbal Science; Korea) for technical assistance; and Domenica Schnabelrauch (MPI for Chemical Ecology, Germany) for sequencing. In Brazil, we thank Miguel Soria and Cecelia Czepak for the collection of specimens. In Northern Australia, we also thank Penny Goldsmith from Ord Co. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Current address: Department of Crop and Soil Science, Oregon State University, Corvallis, Oregon, United States of America [^3]: Current address: Centre for Animal Sciences and Fisheries, University of Swat, Odigram, Khyber Pakhtunkhwa, Pakistan

{ "pile_set_name": "PubMed Central" }

Due to confidentiality restrictions and personal data protection, data are available upon request. Requests for data may be sent to: <mila.vulchanova@ntnu.no>/<sobh.chahboun@ntnu.no>. Introduction {#sec001} ============ Figurative language is a pervasive phenomenon in daily communication, and is reflected in a range of expressions, such as metaphors, idioms, clichés, irony, jokes. Such expressions vary in structure and degrees of transparency and familiarity, and can fluctuate from a single word to a complex sentence. A common feature that unites the categories of figurative language is the fact that reference in such expressions is indirect. More importantly, the interpretation of figurative expressions is non-compositional (non-literal). This deviation from compositional on-line computation and the heterogeneity of figurative uses makes figurative language more demanding in terms of processing \[[@pone.0168571.ref001]\]. In language development, the ability to interpret figurative language develops gradually, and comprehension of such expressions can be especially demanding in deficit populations where language is affected. Individuals with High Functioning Autism (HFA) are distinguished by relative preservation of language and cognitive skills. Intelligence in this group is at normal or above average levels, and quite often such individuals demonstrate specific advantages in certain cognitive areas. The term high functioning captures the level of adequate cognitive functioning in that group, while often, when referring to the level of language competence (i.e., intact structural language), they are described as highly verbal. However, problems with pragmatic language have been consistently reported across the autistic spectrum, even when structural language is intact \[[@pone.0168571.ref002]--[@pone.0168571.ref006]\]. In addition, some studies have found evidence of weaknesses in the figurative and idiomatic language domains, contrasting with clear strengths in areas of grammar \[[@pone.0168571.ref004], [@pone.0168571.ref006]\]. Several behavioral and imaging studies have shown that there are many issues to take into account regarding the interpretation of figurative language. It depends on the type of expression, its degree of compositionality/transparency, its linguistic structure, its source domain of knowledge, and not least, its novelty/conventionality. Thus, the main question is how these factors impact on the comprehension of the categories of interest (idioms, metaphors, irony), and whether the dissociations between structural language skills and figurative language competence reported in highly verbal individuals with autism are still maintained with a careful choice of experimental paradigm. The current study is designed to answer these questions by focusing on figurative language processing in highly verbal individuals with autism. The main goal was to investigate whether individuals with autism and preserved language competence understand idiomatic expressions and novel metaphors at the same level as carefully matched controls, and, specifically, whether in this process they benefit from visually presented context and can integrate information from multiple sources (e.g. visual modality and oral language presented simultaneously). Idiomatic expressions and their processing {#sec002} ------------------------------------------ The mechanisms underlying the processing of figurative language have been widely debated. Idiomatic expressions are one of the most common and popular forms of figurative language used in daily interaction. Idioms are multi-word expressions whose interpretation is non-literal, meaning it cannot be derived by using regular compositional processes. As such, they require the language user to go beyond what is stated literally in order to infer the communicative intent of the speaker \[[@pone.0168571.ref007]\]. In addition, idioms have attracted attention both on theoretical and empirical grounds. The reason is that they cannot be placed immediately within a grammar-lexicon dichotomy of language. On the one hand, it can be assumed that they belong in, and are retrieved from, the mental lexicon, following the assumption that they need to be acquired and stored in the same way as lexical items. On the other, their (complex) structure suggests that they may be processed like structures generated by grammar \[[@pone.0168571.ref008]--[@pone.0168571.ref009]\]. Idiom processing has been addressed extensively in recent research, but mainly, two different types of theoretical approaches have emerged regarding the processing and comprehension of idioms \[[@pone.0168571.ref010]\]. Representative of the first type of approach is the Lexical Representation Hypothesis. Originally proposed by Swinney & Cutler \[[@pone.0168571.ref011]\] (see also Bobrow & Bell \[[@pone.0168571.ref012]\] and Gibbs \[[@pone.0168571.ref013]\]), this hypothesis claims that idioms are stored as lexical items, and understanding an idiom involves two parallel processes, a retrieval process, and a literal compositional computation process based on decomposing every element separately \[[@pone.0168571.ref008]--[@pone.0168571.ref010]\]. Further, it is assumed that the retrieval process is faster, while the compositional process demands longer time. In a further development, Hamblin & Gibbs \[[@pone.0168571.ref014]\] suggest that idioms are essentially decomposable. For these authors, idiomatic interpretation rests on identifying the individual constituents, mainly because of the decomposability of the idiomatic expressions. Along similar lines, some authors have claimed that the processing and understanding of idioms cannot be reduced to lexical access or lexical retrieval only, and an additional interpretation process might be necessary \[[@pone.0168571.ref015]--[@pone.0168571.ref017]\]. Thus, Gibbs \[[@pone.0168571.ref016]\] suggests that idioms invoke complex meanings which are motivated by independently existing conceptual metaphors which are part of everyday thought. The second type of approach, labelled the Configuration Hypothesis, aims to capture the dynamic process involved in idiom interpretation. The authors in favour of this position argue that idioms are represented in a concrete and distributed way and are processed as complex expressions \[[@pone.0168571.ref015]\]. This latter approach can be described as a "compositional" approach whereby each constituent is assumed to contribute semantically to the idiom interpretation. In a comprehensive review of approaches to idioms, Titone & Connine \[[@pone.0168571.ref007]\] suggest that either view fails to account adequately for findings in research or capture the nature of idioms. These authors thus suggest a "hybrid" model, which acknowledges both the arbitrary "word" nature and the compositional aspect of idioms. Their model takes into account an important parameter on which idioms vary: degree of decomposability. A similar view, albeit from a slightly different perspective, is defended in Sanford \[[@pone.0168571.ref018]\] who describes idioms as the intersection of conceptual and syntactic schemas. Another factor which has been shown to impact on idiom processing is the mode of presentation. Tabossi, Fanari & Wolf \[[@pone.0168571.ref019]\] found that idiom identification differs depending on whether the expression was presented auditorily or visually (as text). This clearly implies that the different modalities in which the idioms are presented, demand different forms of processing. Another related figurative and formulaic language category are proverbs. Proverbs are fixed sentential expressions which invoke well-received wisdom, social or moral norms \[[@pone.0168571.ref020]--[@pone.0168571.ref021]\]. They reflect themes from everyday life and often refer to general principles of reasonable human behaviour and action \[[@pone.0168571.ref021]--[@pone.0168571.ref022]\], and, as such, may seem to express self-evident truths. However, like idioms, proverbs display a wide range of interpretations depending on the type of mechanism involved: they can be metaphorical, as *You can't judge a book by its cover*, or based on personification, hyperbole or a paradox (*No news is good news*). Unlike idioms, however, proverbs are typically based on some kind of cause-effect reasoning (*Once bitten*, *twice shy*) or reasoning by analogy, but may also reflect a certain type of cultural practice (*Never look a gift horse in the mouth*). Although proverbs may have prototypical figurative meanings, they are very much vital due to the necessity to draw precise inferences in each specific context of encounter \[[@pone.0168571.ref021]\]. Thus, the correct interpretation of a proverb requires the ability to navigate between their concrete, literal meaning and their figurative meaning by establishing a metaphorical connection between these two interpretations. These properties of proverbs make them an interesting category to study, especially in comparison to idioms. In the current study we refer to this type of figurative expression as "instructive expressions" reflecting their communicative role in language use. Unlike idioms and proverbs, metaphors are a category which is often less conventionalised and more transparent. It is often assumed that metaphors are a device which maps concepts from two distinct domains, which are not otherwise conceptually linked to each other \[[@pone.0168571.ref023]\]. Like idioms, however, metaphors vary in degree of novelty/conventionality which affects the way they are processed. In the current study we use novel metaphors as a category which does not depend on stored vocabulary knowledge, but rather invites active processing based on analogy. Thus, the processing of novel metaphors can act as a base-line for the processing of novel expressions, in contrast to all other idiomatic categories included in the study. Figurative language in High Functioning Autism (HFA) {#sec003} ---------------------------------------------------- Despite the long-standing interest in idioms, their nature and processing in typical populations, research on idiom comprehension in atypical populations overall, and in autism, in particular, has been surprisingly scanty. Mashal & Kasirer \[[@pone.0168571.ref024]\] conducted a study where individuals with HFA, children with learning disorder and typically developing controls were compared. The authors used 11 subtests, ranging from interpretation of figurative expressions, including visual metaphors, idioms, conventional metaphors, novel metaphors, to comprehension of homophones and semantic tests (synonymy, similarity). Their results showed that children with learning deficit scored significantly lower than typically developing peers on comprehension of conventional metaphors and idioms, but no significant differences were observed between the two groups on comprehension of visual and novel metaphors. Furthermore, both deficit groups displayed a dissociation between the visual modality and the auditory modality, supporting the assumption that modality is a salient factor in figurative language processing. In contrast, the control group demonstrated an association between idioms and conventional metaphors. This result supports the idea that idioms and conventional metaphors share a common ground, in that both are relatively frozen conventionalised expressions and most likely rest on similar mechanisms for their processing. This is consistent with findings in a recent study of metaphor processing in highly verbal individuals with autism \[[@pone.0168571.ref025]\]. Vogindroukas & Zikopoulou \[[@pone.0168571.ref026]\] conducted a study aimed at the developmental trajectory of idiom comprehension in Asperger's syndrome (AS) and HFA. The authors compared children with HFA/AS to their typically developing peers and to control adults. They found that children with AS/ HFA performed more poorly on the idiom task compared to the other two groups, while the two typically developing groups did not display any differences in terms of performance. In addition, the authors did not find a significant correlation between IQ scores and performance on the task for the HFA group, but IQ did correlate with idiomatic comprehension and age for the two control groups. It is important to mention that this study used idioms out of context. The fact that the idioms were presented without a supportive context means that the participants had to rely exclusively on a semantic analysis or retrieval. Thus, the study design might have introduced an additional problem and a potential confound between expression familiarity/frequency and decomposability. In that study, Vogindroukas & Zikopoulou \[[@pone.0168571.ref026]\] confirm that children with AS/HFA have a tendency to make literal interpretations, but that the difficulty in idiom comprehension is definitely not due to an intellectual deficit. Whyte et al \[[@pone.0168571.ref027]\] addressed what factors predict performance on idiom comprehension in highly-verbal children with autism in comparison to an age-matched control group and a control group matched on grammar-age equivalence. This study focused specifically on the contribution of syntax competence and advanced Theory of Mind (ToM) abilities to idiom comprehension in autism. The results document that children with autism indeed perform more poorly compared to age-matched controls, but that there is no significant difference with the syntax competence matched group. Furthermore, while advanced ToM ability did not predict performance in the control groups, it correlated with idiom comprehension in the participants with autism. Lee et al \[[@pone.0168571.ref028]\] tested Korean high-functioning children with autism compared to similarly aged children with Attention Deficit Hyperactivity Disorder (ADHD) and a control group of typically developing children. The test design included matching idioms to a picture, which either described the figurative meaning of the expression or the literal interpretation. Non-figurative sentences paired with their literal interpretation were used as a base-line condition. This study found that both the children with autism and the ADHD group performed significantly worse than controls on the two idiom conditions, and especially in the condition where the idiom was paired with its literal interpretation. No differences were observed on the non-figurative expressions. Norbury \[[@pone.0168571.ref029]\] investigated idiom comprehension in context in children with communication disorders and age-matched peers. Children belonging to the communication disorders group were selected to represent a spectrum of communication difficulty, including Specific Language Impairment (SLI), Pragmatic Language Impairment (PLI) (with or without presence of autistic symptoms), autistic disorder, and Asperger syndrome. An idiom definition task was chosen to measure idiom comprehension in context. The main finding of that study was that children with language and communication impairments benefit from context in interpreting unfamiliar idioms. However, children with deficits in structural language did not benefit from context as much as typically developing peers. Furthermore, children with autism spectrum disorder (ASD) symptoms and pragmatic impairments, but without concomitant structural language deficits, were not impaired to the same degree in their idiom understanding and were able to use context as effectively as their peers. Importantly, language ability was one of the most significant predictors of idiom understanding. These findings highlight the role of language skills as a key determinant in the use of context to aid idiom comprehension, and further suggest that contextual cues can provide sufficient information to derive at least a cursory understanding of new idiomatic expressions. Factors in idioms processing {#sec004} ---------------------------- Previous research has documented that idioms are more easily processed in the presence of supportive context (\[[@pone.0168571.ref016]\]; see Vulchanova et al \[[@pone.0168571.ref010]\] for a review). Findings from several studies suggest that the main role of context is to provide semantic support for decoding the target (appropriate) meaning of a sentence or an expression \[[@pone.0168571.ref017], [@pone.0168571.ref030]--[@pone.0168571.ref033]\]. In contrast, it is widely argued that individuals with autism are impaired in processing ambiguous linguistic information in context \[[@pone.0168571.ref034]--[@pone.0168571.ref036]\]. In addition, they often fail to attach context to their memories, and are specifically impaired in processing the social aspects of contextual information \[[@pone.0168571.ref037]\], while Ozonoff & Miller \[[@pone.0168571.ref038]\] demonstrate deviant use of context. However, the evidence concerning a contextual deficit is somewhat controversial. Thus, other studies have shown that competence at making use of context depends on language status, and children with higher verbal age can perform at the level of controls \[[@pone.0168571.ref039]\]. Also, success in that group may depend on the exact position of the ambiguous word (homograph) in the sentence. López & Leekam \[[@pone.0168571.ref036]\] found that, even though children with ASD performed worse than controls, they were more successful with homographs occurring later in the sentence (middle or end), suggesting sensitivity to context also in that group. A study by Brock, Norbury, Einav & Nation \[[@pone.0168571.ref040]\] showed that participants with autism display a tendency similar to controls to use predictions based on the meaning of the lexical verb, as reflected in an increase in looks to the picture matching the object of the verb. At the same time both groups were less likely to be distracted by a phonological competitor of the object word. This tendency, however, was mediated by language status in the autism group. In that study context was restricted to the level of the sentence, with a focus on the information encoded in the lexical verb and the ability to use that information to successfully orient to the possible object. Also, Frith \[[@pone.0168571.ref041]\] argues that individuals with autism need a concrete context to be able to understand words and expressions, and Saldaña & Frith \[[@pone.0168571.ref042]\] document that children with autism have a normal reduction in reading times for expressions which are congruent with previous events. Given this evidence, we expect that context will not place any additional burden on the interpretation of the stimuli used in the current design. Idiomatic expressions vary in degree of transparency/decomposability, and it seems clear that the more transparent the expression, the easier it is to understand, and conversely, the more opaque, the more difficult. A number of studies \[[@pone.0168571.ref043]--[@pone.0168571.ref044]\] demonstrate that the lexical makeup of idioms is a matter of degree and largely depends on the idiom's compositional properties, and that idioms are accessed differentially according to their degree of syntactic frozenness. These authors thus claim that idioms do not form a unique class of linguistic items, but rather share properties with "more" literal language. Following this tradition, and with Titone & Connine \[[@pone.0168571.ref007]\], we expect that comprehension difficulty will be a direct function of the transparency/decomposability of the idiom \[[@pone.0168571.ref031], [@pone.0168571.ref033]\]. It can be assumed that the ability to understand idioms relies on the competence to process these expressions beyond the literal interpretation of individual words. This means that understanding idiomatic expressions demands making inferences, as well as integrating contextual information from both verbal and non-verbal sources. In our daily life, our knowledge about the world around us is based on the semantic and conceptual information gathered through our senses, implying that we perceive the world through different modalities \[[@pone.0168571.ref045]\]. Consequently, it seems logical to expect that some of the difficulties that individuals with autism demonstrate in figurative language may be due to a more generic deficit in multimodal integration. Indeed, problems with multi-sensory information have been documented in autism, and specifically at speech-specific multisensory integration \[[@pone.0168571.ref046]\]. Irwin, Tornatore, Brancazio & Whalen suggest that there are fundamental differences in audio-visual integration in children and adolescents with autism in comparison to typical peers \[[@pone.0168571.ref047]\]. Furthemore, individuals with autism have been shown to present with a reduced McGurk effect in the processing of audio-visual speech \[[@pone.0168571.ref048]\]. It has also been suggested that the neurodevelopmental abnormalities in autism may be due to poor information integration across brain regions \[[@pone.0168571.ref049]\]. Van Der Smagt, Engeland & Kemner \[[@pone.0168571.ref050]\] further suggest that any problems arising from integrating auditory and visual information must stem from higher processing stages, as is the case in processing complex linguistic material, and not low-level integration processes (e.g., at the level of individual sound/visual stimulus). Hypotheses {#sec005} ========== In this study, we investigate the interaction of linguistic and visual context in the interpretation of figurative expressions in highly-verbal individuals with autistic spectrum disorder compared to typically developing peers. The aim of this research is to test a more homogeneous sample of participants within the broad group of individuals with autism by focusing on the higher end of the spectrum (high functioning individuals with intact structural language) in two homogeneous chronological age samples. Thus, we are interested in establishing a potential developmental trajectory, albeit through a cross-sectional design. Consistent with the evidence from previous research, we expected to find effects of age and group. In particular, we hypothesized that the two groups with autism will perform significantly worse than the age- and verbal abilities-matched control groups. Specifically, we expected to find higher reaction times (RTs) and error means in the participants with autism compared to their typically developing peers, despite an improvement with age, as observed in the case of metaphors in a recent study (Chahboun et al., \[[@pone.0168571.ref025]\]). Furthermore, it can be expected that, since individuals with HFA have a tendency for literal interpretation, they will have difficulties in appreciating the figurative nature of non-literal language and more often select interpretations that match the literal meaning of the expression. Given that figurative language skills develop constantly through childhood, we also expect that younger participants (the two children groups) will perform at a lower level than the young adult groups. While cross-modal tasks, which require the integration of information from different modalities may not present a problem for typically developing participants, they may tax processing in individuals with autism and make the task more demanding. Concerning the presence of visually supported meaning (pictures that matched the interpretation of the idioms), it can be expected that visual context may facilitate the processing of the figurative meaning, or, at least, will not present an additional problem, as suggested by López & Leekam \[[@pone.0168571.ref036]\] and indicated by the results in Brock et al \[[@pone.0168571.ref040]\], also consistent with the findings in Norbury about the facilitatory role of context \[[@pone.0168571.ref029]\]. Alternatively, it may be the case that the presence of both auditory and visual information might rather present a problem, as a result of the demand for information integration in the face of documented problems in that domain in autism. Thus, we anticipated higher error rates and RTs in the condition when stimuli were presented in a cross-modal fashion (auditory stimulus/visual target) in the groups with autism, and the absence of such an effect in the control groups. One of the most important criteria in the processing of figurative language is expression transparency. We expected that degree of transparency would impact both on reaction latencies and accuracy in both participants with autism and controls. More specifically, the most transparent category, the one labelled *biological idioms* would be easiest to process, while *cultural idioms*, which are more highly conventionalised and less transparent would, in comparison, present greater difficulty, especially for the child participants and participants with autism. We also expected novel metaphors and *instructive idioms* to be more easily processed than the less transparent *cultural idioms*. Method {#sec006} ====== Participants {#sec007} ------------ Two age groups of HFA (*n* = 45) and typically developing (TD) controls (*n* = 39) were recruited, all native speakers of Spanish. The first age group included children in the age range ten to twelve years. The second age group included young adults from sixteen to twenty-two years old (see [Table 1](#pone.0168571.t001){ref-type="table"} for the participants' descriptive and demographic data). Participants and their legal guardians (usually the parents) provided written informed consent for entry into the study according of the declaration of Helsinki principles. The study was approved by the Andalusian Biomedical Research Ethics Committee. The diagnosis of HFA was confirmed with the Autism Diagnostic Observation Schedule (ADOS) \[[@pone.0168571.ref051]\] (Children, $\overline{x}$ *=* 10,8; Young adults, $\overline{x}$ *=* 12,3). 10.1371/journal.pone.0168571.t001 ###### Mean (and standard deviation) of background measures for each age and group. HFA (high functioning autism), FSIQ (intelligence quotient, Full Scale score), VCI (verbal comprehension index), RG (receptive grammar scores), RC (reading comprehension scores). Note: *p* \< .05 for scores with the same superscript. ![](pone.0168571.t001){#pone.0168571.t001g} **Children** **Adults** ---------------------- -------------- -------------- --------------- ---------------- **n** 25 19 20 20 **Age (months, SD)** 11,3 ±0,96 11,9±0,89 18,1±1,65 18,3±1,92 **Gender (M)** 21 17 17 14 **FSIQ** 110,71±14,58 105,76±11,56 108,3±13,39 118,34±8,46 **VCI** 118,7±17,47 107,4±15,03 122,62±16,71 117,63±12,97 **RG** 72,72±5,56 74,31±4,30 74,40±3,53^a^ 78,42±1,74 ^a^ **RC** 0,58±0,17 0,59±0,16 0,63±0,15 0,72±0,12 The individuals with HFA and their typically developing peers were matched on age, gender and the verbal comprehension based on the Wechsler scale (WISC-IV \[[@pone.0168571.ref052]\] or WAIS-IV \[[@pone.0168571.ref053]\] were used, depending on the chronological age of the participant). The Verbal Comprehension Index (VCI) was used as an overall measure of verbal reasoning and includes measures of semantic/conceptual reasoning (similarities and vocabulary), and verbal comprehension. The descriptive statistics are given in [Table 1](#pone.0168571.t001){ref-type="table"}. The matching of the groups was based on the Wilcoxon Test and the smallest *p* value was *p* = .222 suggesting that there were no significant differences observed between the groups (See [Table 1](#pone.0168571.t001){ref-type="table"}). We also considered reading comprehension scores and receptive grammar (based on CEG, which is the Spanish adaptation of TROG \[[@pone.0168571.ref054]\]) as background measures to ensure our participants display scores within the norm (see [Table 1](#pone.0168571.t001){ref-type="table"}). No differences were observed between participants for reading comprehension. However, despite all our participants being within the norm for receptive grammar, differences were observed between the two groups of young adult participants. For this reason, we included receptive grammar as a co-variate in the analyses of the data. Apparatus and stimuli {#sec008} --------------------- In a pilot study with 50 adults, all native speakers of Spanish, we determined the degree of familiarity and frequency of the figurative expressions to be used in the main study. We selected 124 idioms and instructive expressions and 20 metaphors, for which participants in the pilot study had to determine: 1. "Do you know this expression?" 2. "If yes, do you know what it means?" 3. "Do you use this expression yourself?". Participants were asked to answer truthfully, and to take into account that there were no wrong answers. We established ratings for each expression by assigning numerical values to the answers (1 for *yes* and 0 for *no*) and averaging the scores from the 3 questions. The idiomatic expressions which received a rating of over 0.80 were selected for the study and classified into three main categories: a/ biological idioms; b/ cultural idioms; and c/ instructive expressions, as used in Vulchanova et al. \[[@pone.0168571.ref008]\]. For the first two categories they followed the typology adopted in Penttilä, Nenonen & Niemi \[[@pone.0168571.ref055]\], in turn inspired by Searle's idea of deep background (the human biological nature) vs. local background (local cultural practices), to account for the grounding of language in human experience and practice. The stimuli in these two categories were selected by 2 independent expert linguist raters (91% interrater agreement). Examples of biologically-based idioms are e.g., *estar con el agua hasta el cuello* (to be with water up to the neck); or *venir como anillo al dedo* (to come as ring to the finger), and of culturally-based idioms, *e*.*g estar como una cabra* (to be like a goat). The difference between the two categories is that, while biological idioms are derived from human (bodily) experience and interaction with the environment, and as such, tend to be more transparent and available to compositional parsing, cultural idioms are more idiosyncratic, tend to vary from one culture to another, and are, in general, less transparent. Our third category were instructive expressions (i.e., proverbs), whose meaning can be computed on-line, by drawing inferences on the basis of their individual constituent phrases (e.g., *Las mentiras tienen las patas cortas---*Lies have short legs---). The inclusion of this category was dictated by the need to compare such expressions, whose interpretation may be more directly available in appropriate context, to the comprehension of idioms. Finally, novel metaphors were included as a category that may be assumed to be less frozen/fixed, and as such more readily interpretable. From the pilot study, we selected for inclusion in the novel metaphors category those metaphors which earned a rating below .20 on the familiarity and frequency scale (e.g.: *Estas flotando en el aire*---You are floating in the air). The study was designed as a sentence-picture matching task (38 visual---presented as text---and 38 auditory expressions (a total of 76 expressions---20 biological idioms/20 cultural idioms/16 instructive expressions/20 metaphors) randomized and counterbalanced between groups and participants). Each expression was linked to four possible images: One reflecting the figurative target meaning of the expression, a second image reflecting the literal meaning, a third image corresponding to a figurative meaning, but not the target one, and finally, a distractor as forth image (Cf. [Fig 1](#pone.0168571.g001){ref-type="fig"}). Each expression was presented in a short appropriate context (See [S1 Appendix](#pone.0168571.s001){ref-type="supplementary-material"} for examples). Thus, the study design included both verbal (presented as text/discourse) and visual (pictures) context. ![Images for the idiom "*estar con el agua hasta el cuello*!*"* (lit.: be with water up to the neck/fig.: drown in work)](pone.0168571.g001){#pone.0168571.g001} Matlab (R2014b) (matrix laboratory) was used as a multiparadigm numerical programming language to design, build and run the experiment. Procedure {#sec009} --------- In this experiment, we used visual stimuli (images and sentences) that reflected literal meanings and idiomatic meanings. First, a fixation point (+) appeared for 500ms. Participants then saw on the screen the idiom or metaphor. Depending on the experimental block, the participant heard the expression in context via loud-speakers or read it in context in a short text presented on the computer screen orthographically. Thus, the condition in which stimuli were presented auditorily presented a cross-modal environment whereby participants had to select a visual target which matched a figurative expression they had heard via the loud-speakers. In the auditory experimental block, first the figurative expression appeared on the computer screen followed by a blank screen, while the expression in appropriate context were being presented auditorily. In the text (orthographic) condition both stimulus (expression and context) and target picture were all in the same (visual) modality. First the expression appeared on the screen, and was then followed by the context. Then, in both experimental conditions a fixation point (+) reappeared in the middle of the screen followed by four images in each corner, each one reflecting the figurative meaning (target) meaning, the literal meaning, figurative, but not target, meaning, and a distractor image. The position of presenting the images was counterbalanced between participants and between the modality of presenting the context and the order of the expressions. The participant's task was to click on the stimulus he/ she considered most appropriate for the meaning of the expression. Accuracy and reaction times were collected to determine the ease of processing. Results {#sec010} ======= The data of both the control and experimental groups (n = 84) were analysed with R \[[@pone.0168571.ref056], [@pone.0168571.ref057]\]; and the lme4, afex and lmerTest packages \[[@pone.0168571.ref058]\]. We used two performance measures in the study; reaction time which reveals the ease with which the figurative expression is processed, and accuracy which essentially measures the ability to link the figurative expression to its target image. We ran linear mixed models on reaction times and generalized linear mixed models on accuracy, including the variables of Group (ASD vs. control), Age (Children vs. Young Adults), Modality (Visual vs Auditory), and Type of Expression (Biological, Cultural, Proverbs, and Novel Metaphors). We report relevant F and p-values below (based on the Satterthwaite approximation for degrees of freedom) for reaction times and Chi square and p-values for accuracy. We ran four analyses. The first one, the reaction time analysis, provided us with a general overview of the ease with which the participants responded to the different experimental manipulations in the study. The next analyses explore the ability to identify the correct target compared to other options. Thus, a second model provided us with participants' accuracy and compared overall target and non-target responses. The third and fourth specifically explore correct to other specific non-appropriate responses. The third analysis, called "comparison between figurative target responses and literal responses", provides a comparison between responses to target images and images corresponding to the literal meaning. Finally, our fourth model is based on a comparison between figurative target responses and figurative non-target responses. This model only includes responses to images reflecting target meanings and responses to images which reflected a figurative albeit not target meaning. Reaction time analyses {#sec011} ---------------------- Reaction times were analysed using Log(RT) (ms) as a measure rather than reaction times to correct for right skewedness of the reaction time data. Only correct responses are included in the analyses. We also excluded extremely fast or slow responses. We excluded two trials with reaction times longer than 18s and we calculated the Z value of Speed (Log(RT) (ms)) for each participant (Z~p~) and for each item per age and group (Z~i~). If the sum of the squares of the Z values was smaller or equal than 8 (Z~p~^2^+Z~i~^2^ ≤ 8), the trial was included, and values bigger than 8 were excluded. An overall linear mixed model analysis on Log (RT) (ms) performance has been done with R, which provided us with the best fitting model by step-wise backward selection based on the Akaike Information Criterion (See [S2 Appendix](#pone.0168571.s002){ref-type="supplementary-material"}). We were left with a model including a three-way interaction of group, modality and type of expression, a two-way interaction of age and type of expression, and all lower effects. Also included in the model were the by-subject random intercepts and slopes for modality and by-item random intercepts and slopes for Age and Group. The overall model reveals that there are differences in reaction latency for Age (children/young adults) (*F* (1, 82.64) = 20.38, *p* \< .001) (Satterthwaite approximation for degrees of freedom) and Group (control/HFA) (*F* (1, 86.85) = 10.64, *p* = .001), with slower responses by children and individuals with autism (see [Fig 2](#pone.0168571.g002){ref-type="fig"}). Furthermore, an interaction between age and type of expression was found (*F* (3, 66.11) = 2.903, *p* = .041) (see [Fig 3](#pone.0168571.g003){ref-type="fig"}). A multiple comparison with Tukey contrasts showed that the interaction is due to differences between different types of expressions in children, but not adults. Here, a significant difference was observed between novel metaphors and instructive expressions (*p* = .0014). In addition, a marginally significant difference between cultural idioms and novel metaphors was observed (*p* = .07) showing that instructive expressions were the most demanding and difficult to process for the children. ![Average reaction times (ms) for each age (children/young adults) and group (control/HFA).\ Error bars denote one standard error of the mean.](pone.0168571.g002){#pone.0168571.g002} ![Average reaction times (RTs) for each age (children/young adults) and group (control/HFA) for each type of expression.\ Error bars denote one standard error of the mean.](pone.0168571.g003){#pone.0168571.g003} As we have pointed out, a three-way interaction between type of expression, group and modality was observed (*p* \< .001) (see [Fig 4A and 4B](#pone.0168571.g004){ref-type="fig"}). In order to establish the origin of the three-way interaction, additional multiple comparisons with Tukey contrasts were run. For the HFA group in the auditory modality, no significant differences were observed between different types of expressions. However, for the visual modality, an almost significant difference was observed between instructive expressions and novel metaphors (*p* = .0501), with novel metaphors processed faster than instructive expressions. For the control group in the auditory modality, significant differences were observed between instructive expressions and biological idioms (*p* = .013), with slower responses in the instructive expressions category, and between novel metaphors and cultural idioms (*p* = .033), with faster responses in the novel metaphor category, and between novel metaphors and instructive expressions (*p* \< .001), with faster responses for novel metaphors. No significant differences were observed in the visual modality in the control group between different types of expressions. These results suggest a different interaction pattern between participants with autism and controls. While for controls there is differential processing depending on type of expression in the oral language modality, with more transparent expressions processed faster, this pattern is absent in participants with autism. ![**4a, 4b. Average reaction times (ms) group (control/HFA) type of expression and modality**. Error bars denote one standard error of the mean.](pone.0168571.g004){#pone.0168571.g004} Moreover, we explored the differences between groups and modalities for each type of expression. For biological idioms, while the control group was faster in the auditory modality, the HFA group was faster in the visual modality (*p* = .026), but always significantly slower than controls (*p* = .001). For cultural idioms, the HFA group was significantly slower than controls (*p* = .001). In addition, no modality differences were observed in the HFA group for this type of expression, while the control group were faster in the visual modality (*p* = .02). For instructive expressions, the control group was significantly slower in the auditory modality (*p* = .01), while the HFA group showed no modality preference in this category and was slower in comparison to the control group overall in this type of expression (*p* = .015). Finally, for the novel metaphor category, again the HFA group was much slower overall (*p* \< .001), and showed no modality differences, while the control group showed faster responses in the auditory modality (*p* = .01). To sum up those results, with increase of expression non-transparency, the advantage of modality changes only for controls. Thus, for controls transparent expressions are processed faster in the oral modality and less transparent expressions are processed faster in the orthographic modality. No such trend is observed for the participants with autism. These results suggest a different processing pattern in autism. Comparison between figurative target responses and non-target responses {#sec012} ----------------------------------------------------------------------- An overall generalized linear mixed model analysis on ACC (mean errors) performance has been done with R, which provided us with the best fitting model by step-wise backward selection based on the Akaike Information Criterion. We were left with a model including a two-way interaction of age and group and main effects. Also included in the model were the by-subject random intercepts and slopes for modality and by-item random intercepts and slopes for Age and Group. The overall model thus reveals differences in accuracy in Age (children/young adults) (*χ^2^* (1, 23) = 5.73, *p* = .0016) and Group (control/HFA) (χ^2^ (1, 23) = 11.21, *p* \< .001), with more errors by children and individuals with autism (see [Fig 5](#pone.0168571.g005){ref-type="fig"}). In addition, a main effect of type of expression was observed as well (*χ^2^* (1, 21) = 8.49, *p* = .036). Multiple Comparisons of Means with Tukey contrasts (See [S3 Appendix](#pone.0168571.s003){ref-type="supplementary-material"}) revealed that this effect was due to marginally significant differences between instructive expressions and biological idioms (*p* = .07), and between instructive expressions and novel metaphors (*p* = .06), with more errors in the instructive expressions category. ![Average response accuracy (mean errors) for each age (children/young adults) and group (control/HFA).\ Error bars denote one standard error of the mean.](pone.0168571.g005){#pone.0168571.g005} Moreover, a two-way interaction between age and group was observed (*χ^2^* (1, 23) = 4.98, *p* = .025). Multiple Comparisons of Means with Tukey contrasts (See [S3 Appendix](#pone.0168571.s003){ref-type="supplementary-material"}) revealed that the interaction was due mainly to the differences between young adults (HFA and control) (*p* \< .001), and between the two control groups (children and young adults) (*p* = .01). No differences were observed between the two HFA groups (*p* = 0.99) or children (HFA and control) (*p* = 0.74). These results suggest that participants with autism perform at the same level regardless of age. Comparison between figurative target responses and literal responses {#sec013} -------------------------------------------------------------------- In order to establish responses to target figurative meaning contra literal interpretations, we re-ran the model, while excluding non-target figurative, and distractor responses (See [S4 Appendix](#pone.0168571.s004){ref-type="supplementary-material"}). In this model, Target responses were scored with 1 and literal responses were scored with 2. The model includes a significant three-way interaction between modality, age and group (control/HFA), and all lower interactions and main effects. Also included in the model were the by-subject random intercepts and slopes for modality and by-item random intercepts and slopes for Age and Group. This model reveals a main effect of group (control/HFA) ((*χ^2^* (1, 26) = 5.22, *p* = .022), with more literal responses by children and young adults with autism and a marginally significant difference in accuracy between Age (children/young adults) ((*χ^2^* (1, 26) = 3.51, *p* = .06) (see [Fig 6](#pone.0168571.g006){ref-type="fig"}). In addition, a main effect of type of expression was observed (*χ^2^* (1, 24) = 10.37, *p* = .015). Multiple Comparisons of Means with Tukey contrasts (See [S4 Appendix](#pone.0168571.s004){ref-type="supplementary-material"}) revealed that this effect was due to a difference between instructive expressions and novel metaphors (*p* = .02), with more literal responses in the instructive expressions category, and to a marginally significant difference between cultural idioms and novel metaphors, with more literal responses in the cultural idioms category. These results suggest that less transparent expressions are likely to trigger more literal responses. ![Average literality for each age (children/young adults) and group (control/HFA).\ Target responses were scored with 1 and literal responses were scored with 2. Higher means indicate more literal responses. Error bars denote one standard error of the mean.](pone.0168571.g006){#pone.0168571.g006} Furthermore, a two-way interaction between age and group was observed (*χ^2^* (1, 26) = 4.89, *p* = .02). We ran multiple comparisons of means with Tukey contrasts which revealed that this interaction was due to a significant difference between control young adults and HFA young adults (*p* = .015), with the young adults with autism giving more literal responses than their respective typically developing peers. Also, the difference between both ages of control participants was significant (*p* = .03), where the young adults detected the figurative target image more accurately. No differences were observed between the two HFA groups (*p* = 0.99) or children (HFA and control) (*p* = 0.99). These results suggest that children and participants with autism are more likely to provide literal responses. A three-way interaction between age, group and modality was observed (*χ^2^* (1, 26) = 3.97, *p* = 0.046) (see [Fig 7](#pone.0168571.g007){ref-type="fig"}). Again, a multiple comparison of means with Tukey contrasts showed no significant effects in the visual modality in any of the age groups, but clear differences in the auditory modality between the control young adults and the young adults with autism (*p* = 0.001), where the older groups of participants had inverted patters of responses. The young adults with HFA were less literal in the visual modality than the auditory modality, while the control young adults were better in the auditory modality, suggesting a differential pattern in response. Moreover, a marginally significant difference was observed between the control children and the control young adults (*p =* .*055)*, where the children did not show modality differences, but the older group was better in the auditory modality. ![Average literality for each age (children/young adults) and group (control/HFA) and modality (visual/auditory).\ Target responses were scored with 1 and literal responses were scored with 2. Higher means indicate more literal responses. Error bars denote one standard error of the mean.](pone.0168571.g007){#pone.0168571.g007} Comparison between figurative target responses and figurative non-target responses {#sec014} ---------------------------------------------------------------------------------- The final part of the analyses compared figurative target responses and figurative non-target ones. Therefore, we ran the model again without responses to the literal image and responses to the distractor (See [S5 Appendix](#pone.0168571.s005){ref-type="supplementary-material"}). From the overall model it is clear again, that there are differences in accuracy between groups (control/HFA) (*χ^2^* (1, 11) = 6.49, *p* = .01), with more errors by individuals with autism. Additionally, a main effect of age (children/young adults) was found as well (*χ^2^* (1, 11) = 12.92, *p* \< .001 and (see [Fig 8](#pone.0168571.g008){ref-type="fig"}), with more errors in the two younger groups. These results suggest that figurative language comprehension is not yet stabilized and still developing in the child participants. No other significant effects were found in this comparison. Despite this, figurative non-target responses only appeared in very few instances (4.5% of the responses in comparison with 81.8% for figurative target ones). Consequently, the comparison between figurative target and figurative non-target responses shows a very small error range. ![Average accuracy in figurative meanings for each age (children/young adults) and group (control/HFA) and modality (visual/auditory).\ Target responses were scored with 1 and literal responses were scored with 0. Higher means indicate more accurate responses. Error bars denote one standard error of the mean.](pone.0168571.g008){#pone.0168571.g008} General linear model---Receptive grammar as a covariate {#sec015} ------------------------------------------------------- We were also interested in the extent to which the difference between the two young adult groups on receptive grammar influenced these groups' performance on the tasks. Repeated measures ANOVA were run in SPSS \[[@pone.0168571.ref059]\], where we considered age and group as between subject factors, modality and type of expression as within subject variables and receptive grammar as a covariate. Bonferroni confidence interval adjustment was considered and observed power was calculated in this analysis (See [S6 Appendix](#pone.0168571.s006){ref-type="supplementary-material"}). We followed the same procedure for all the previously considered dependent variables (reaction time, overall accuracy, scores for target and literal responses, and scores for target and figurative non target responses). Receptive grammar as a covariate did not affect in any way any of the models and none of the main effects or interactions changed with the inclusion of receptive grammar scores as a covariate in the regression model. We report the power effects analyses in [S7 Appendix](#pone.0168571.s007){ref-type="supplementary-material"}. Discussion {#sec016} ========== Our main hypothesis was that we would find poorer performance on figurative language processing by children and the participants with autism compared to their peers, carefully matched on age, non-verbal IQ, and verbal comprehension. Consistent with our hypotheses, the analyses of reaction times in the current study found a main effect of age and group, with slower responses for children and participants with autism. While no significant difference was observed between the two children's groups on a test of receptive grammar, there was a difference between the two young adult groups of participants, and oral language differences could have explained the different performance between these groups. However, the fact that our results were not affected by the inclusion of receptive grammar scores as a co-variate in the analysis, ensures that the obtained results were not due to a structural language problem (e.g., in receptive grammar), but to the experimental manipulation. These analyses also revealed that two types of expressions were particularly difficult for children, cultural idioms and instructive expressions (proverbs), while novel metaphors were processed much faster. These results suggest that the impact of expression transparency tends to diminish with age and as figurative skills stabilize in young adulthood. The analyses of accuracy revealed a similar pattern: more errors were found in the two child groups and in the two groups with autism. Here a main effect of type of expression was also observed, whereby instructive expressions emerged as the most difficult category to process in comparison to both novel metaphors and biological idioms. The analyses of accuracy also revealed that participants with autism were significantly more likely than controls to select the literal interpretation of the figurative expression, as represented in a visual image. In this study, we hypothesized that the participants with autism would display a relative weakness in performance on idiomatic language, despite intact structural language skills. Although individuals with autism obtain high accuracy scores and can obviously solve the task, the current findings confirm that the adult individuals with autism are not reaching an equivalent level of performance to typically developing matched controls, both with regard to accuracy and reaction time. These results are consistent with results from the few previous studies of idiom comprehension in autism, where the participants with HFA performed more poorly and more slowly than their typically developing matched peers, suggesting a different pattern of processing figurative language in general, and idiomatic competences, in particular. These results are also consistent with the finding that the contrast in ability in autism in comparison to controls is more evident on the higher end of the spectrum \[[@pone.0168571.ref060]\]. The current results, however, contradict findings of comparable semantic and language processing skills in studies where the controls and participants with autism are matched on structural language skills \[[@pone.0168571.ref027],[@pone.0168571.ref040]\]. The participants in the current study were highly-verbal individuals with autism carefully matched to the control groups on IQ and verbal skills, including semantic, and oral and reading comprehension skills. The current results thus challenge the idea that the problems in figurative language in autism arise exclusively from verbal status and language skills, also consistent with results in Eigsti, de Marchena, Schuh & Kelley \[[@pone.0168571.ref061]\] and Olofson et al. \[[@pone.0168571.ref062]\] and the review in Vulchanova et al. \[[@pone.0168571.ref010]\]. More importantly, both the reaction latencies and accuracy analyses reveal a different processing pattern in the interaction between modality (auditory/visual) and type of expression, especially evident in the young adult groups. This pattern of responses is suggestive of different processing or perhaps alternative strategies for the comprehension of figurative language in autism, despite overall high accuracy. We consider that this different pattern of processing is due, in part, to a delayed trajectory in developing the necessary skills to process figurative language. This is clearly seen in the similar response pattern of the young adults with autism and the control children on both accuracy and reaction speed. This may suggest that rather than an inability to process figurative language or idioms altogether, in this case, there is a delay in the ability to detect such expressions, and to differentiate between the different categories of figurative language. These results support the findings from another study with the same participants on metaphorical language \[[@pone.0168571.ref025]\]. Yet, even though participants with autism seem to improve with age in terms of reaction time, suggesting faster processing, they still do not reach the same levels of performance as control matched peers. The improvement with age does not only apply to the individuals with HFA, but also to the control groups, which is consistent with Nippold & Duthie \[[@pone.0168571.ref063]\], Levorato & Cacciari \[[@pone.0168571.ref064]\] and Vulchanova et al \[[@pone.0168571.ref008]\], and our original hypothesis, but not with Vogindroukas & Zikopoulou \[[@pone.0168571.ref026]\], who, surprisingly, did not find any reliable difference in performance between the younger and older controls. Furthermore, our study explored the impact of the nature and the transparency of the type of figurative expression. In the study we included biological idioms, which have been argued to be more closely grounded in human experience and interaction with the environment (Searle's *deep background*). We expected such expressions to be more readily accessible and more decomposable, despite their idiomatic nature, compared to cultural idioms, also for the participants with autism. In line with Chahboun et al \[[@pone.0168571.ref025]\], we also expected that novel metaphors would be easier to process. These expectations were confirmed. Even though the participants with autism were significantly delayed in comparison to controls, they displayed similar patterns of difficulties and ease across the test categories as the control groups. Biological idioms and novel metaphors were the easiest to process. We reason that these two categories are the most transparent, and their meaning can be easily inferred from the context, while cultural idioms are less transparent and more "frozen" \[[@pone.0168571.ref065]\]. These findings are consistent with Vulchanova et al. \[[@pone.0168571.ref008]\] and Vega-Moreno \[[@pone.0168571.ref017]\] who claim that inferential skills are not only involved, but mandatory in idiom comprehension. In our study, the figurative expressions were elicited based on a pilot, where we selected the most familiar and frequent expressions for the idioms category, and the least frequent (most novel) for the metaphors category. The current results confirm that novel metaphors were easier to process compared to less transparent expressions, such as idioms. Our findings are partly consistent with Giora, Gazal, Goldstein, Fein, & Stringaris \[[@pone.0168571.ref066]\], who found that familiarity/novelty of the expression were reliable predictors of performance in both typically developing individuals and individuals with HFA. The current results are also consistent with the findings in Norbury \[[@pone.0168571.ref029]\], where participants with autism were facilitated by the presence of context in the comprehension of novel idioms and performed at levels similar to controls. The current findings, however, contradict the results obtained by Mashal & Kasirer \[[@pone.0168571.ref024]\]. These authors found that the most novel figurative expressions were the most difficult to process in both adults and children with HFA. However, this difference might be due to experimental design and choice of experimental stimuli, as is often the case in research on figurative language. Contrary to our expectations, instructive expressions (proverbs) were the most difficult category to process, along with cultural idioms. We included this category in the design because of the expectation that they may not be stored in the mental lexicon, but rather rely on the immediate activation of richer "memory organization packages" \[[@pone.0168571.ref021],[@pone.0168571.ref067]\], thus allowing more readily for interpretations on the fly when encountered in appropriate context. This expectation was not confirmed, however. Moreover, both children and participants with autism were significantly more likely to select the literal interpretations of both instructive expressions and cultural idioms. These results support the findings in Honeck, Welge & Temple \[[@pone.0168571.ref068]\] concerning the processing of proverbs, where participants were faster to make appropriateness judgements when the proverb was related literally to the context compared to when it was related figuratively. These authors interpret the findings as supporting a multi-stage model of proverb comprehension, with literal interpretations becoming available faster than figurative ones. The current results suggest that instructive expressions (proverbs) indeed require some ability to think abstractly \[[@pone.0168571.ref020]\] and cannot be interpreted so readily based on their constituent words. We also find evidence that this skill takes time to acquire, as evident in the child participants. We also speculate that the more culturally-based an expression is, the more reliant it is on learning it in social interaction, thus consistent with the limitations documented in this domain in autism. Regarding the effect of modality, we hypothesized that the control group would not display dissociations between the visual and auditory presentation modalities, while for the participants with autism we expected certain problems when the expressions were presented auditorily and had to be matched to a visual stimulus (a picture) in the cross-modal condition. This was confirmed, but only for the most transparent expressions (biological idioms and novel metaphors). The control groups performed better when biological idioms and novel metaphors were presented in the auditory modality. Thus, it appears that the control group have an advantage for oral language, and can easily integrate oral stimuli with visual contextual support (the images). This was not the case for the participants with autism, and it could suggest a a difficulty even in highly-verbal individuals with autism in multimodal information integration, even when the idiomatic expressions are decomposable (cf. also Ozonoff & Miller, \[[@pone.0168571.ref038]\]). These differential responses to modality in the autism group could be indicating the involvement of different processes in the resolution of the idioms, or different approaches when attempting to respond to this task. These findings point to the possibility of a problem of simultaneous processing in both modalities in this population. An additional possibility may be that there are residual oral language problems also in individuals with intact structural language skills. This is consistent with Eigsti \[[@pone.0168571.ref069]\], where the problems in ASD are attributed to low-level impairments in generic cognitive and processing mechanisms. Alternatively, it can be assumed that "seeing" text is an advantage, especially in the groups with autism. Indeed, there is evidence that orthographic decoding is a strength in autism and this may, in turn, lead to better comprehension outcomes, if a simple view of reading is adopted \[[@pone.0168571.ref070]\]. Another interesting result was the difference in degree of literalness observed in response accuracy. As expected, the younger participants and participants with autism interpreted the stimuli more often literally than the older and the control groups. Figurative language competence takes longer to develop and requires a number of other skills (inferencing, discourse skills; \[[@pone.0168571.ref071]\]), and the older we are, the more resources for processing we have. In the case of participants with autism, both children and young adults were very literal and did not appear to improve in this respect with age (the two age groups in the study performed similarly). This was not the case for the control groups. For children presentation modality did not affect response literalness, while for the older participants there was an advantage in the auditory modality, suggesting that idiom comprehension most probably resides in mature oral language skills. The results of the participants with autism are also consistent with other research suggesting a tendency for literal (compositional) interpretation in autism \[[@pone.0168571.ref026],[@pone.0168571.ref072]\]. The overall results of this study support our hypothesis that idiomatic expressions require more competences than just good structural language. These clearly provide a necessary condition, but may not be sufficient. This is evidenced by the findings supporting the presence of a delay in the two groups with HFA. Additionally, the current study contributes to research studying the processing of idioms specifically, and figurative language, more broadly, in providing empirical evidence on how different categories of figurative expressions are processed in both highly verbal individuals with autism and typically developing individuals. We find support that expressions, which are more firmly grounded in human bodily experience (biological idioms), and are also more transparent (novel metaphors) are processed more readily, despite their idiomatic nature, regardless of participant group. Finally, we find evidence that young adult controls have a clear advantage in processing auditory language stimuli, most likely as a result of subtle differences in oral language skills. The problems observed in the participants with autism are suggestive of either multi-sensory information integration difficulty or simply a residual problem in oral language and need to be addressed in future research potentially leading also to better target intervention for that group. Supporting Information {#sec017} ====================== ###### Examples of expressions in context. (DOCX) ###### Click here for additional data file. ###### Overall model (Reaction times). (RTF) ###### Click here for additional data file. ###### Overall model (Accuracy analysis: Target vs non-target). (RTF) ###### Click here for additional data file. ###### Overall model (Accuracy analysis: Target vs literal responses). (RTF) ###### Click here for additional data file. ###### Overall model (Accuracy analysis: Figurative target vs Figurative non-target responses). (RTF) ###### Click here for additional data file. ###### General linear model analyses with receptive grammar as covariate. (DOCX) ###### Click here for additional data file. ###### Observed powers. (DOCX) ###### Click here for additional data file. We would like to thank the participants and the families for their effort and time in participating in this study. We are also grateful to Martina Micai for help in the data collection. Last but not least, we would like to thank schools and associations for putting us in contact with both the HFA groups and the TD groups. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: **Conceptualization:** MV VV DS SC.**Data curation:** HE SC.**Formal analysis:** MV VV HE SC.**Funding acquisition:** MV VV DS.**Investigation:** SC.**Methodology:** MV VV DS SC.**Project administration:** MV VV DS.**Resources:** MV VV DS HE SC.**Software:** HE SC.**Supervision:** MV VV DS.**Validation:** MV VV HE SC.**Visualization:** MV VV DS SC.**Writing -- original draft:** SC.**Writing -- review & editing:** MV VV DS HE SC. [^3]: Current address: Department of Language and Literature, Norwegian University of Science and Technology, Trondheim, Norway [^4]: Current address: Departamento de Psicología Evolutiva y de la Educación. Universidad de Sevilla, Avda. Camilo José Cela s/n, Seville, Spain

{ "pile_set_name": "PubMed Central" }

INTRODUCTION {#sec1-1} ============ Pre-eclampsia and eclampsia are hypertensive disorders of pregnancy that cause significant morbidity and mortality in the fetus and mother. Globally, pre-eclampsia and eclampsia account for 10-15% of maternal deaths.\[[@ref1]\] Pre-eclampsia is a multi-organ group of related disease processes that occur in up to 5-8% of pregnancies after 20 weeks of gestation. The presentation is variable, but generally includes the combination of maternal hypertension and proteinuria.\[[@ref2]\] Severe pre-eclampsia is defined as any of the following: (i) severe hypertension (systolic blood pressure \>160 mmHg or diastolic blood pressure \>110 mmHg); (ii) proteinuria 5 g per 24 h; (iii) oliguria \<400 ml urine per 24 h; (iv) cerebral irritability or visual disturbances; (v) epigastric or right upper quadrant pain (liver capsule distension); or pulmonary edema.\[[@ref3]\] Eclampsia is the new onset of grandmal seizures occurring during and after pregnancy that do not have another identifiable cause.\[[@ref2]\] The development of eclampsia is associated with increased risk of adverse outcome for both mother and fetus.\[[@ref1][@ref4]\] In Nigeria, eclampsia is the 3^rd^ commonest cause of maternal mortality.\[[@ref5]\] Eclampsia could be prevented in majority of the cases if there was early booking for antenatal care services, permitting early identification of pre-eclampsia and institution of appropriate therapy.\[[@ref6]\] The high maternal morbidity and mortality due to eclampsia in the developing countries has been ascribed to late referral, delay in hospitalization, late transportation, unbooked status of patients, and multiple seizures prior to admission.\[[@ref5]\] Progression of pre-eclampsia to eclampsia may occur during pregnancy, labor or in the postpartum period.\[[@ref4]\] Reviewing the published literature, the authors of this report were unable to find any record of eclamptic seizures occurring during cesarean section under spinal anesthesia. We, therefore, report the case of a primigravida who presented with severe pre-eclampsia, which progressed to eclampsia during cesarean section under spinal anesthesia and was successfully managed. CASE REPORT {#sec1-2} =========== A 26-year-old primipara booked for antenatal care at 36 weeks gestation at the University of Port Harcourt Teaching Hospital. On presentation, she was found to have 2++ of protein in urine and blood pressure was 190/120 mmHg. A diagnosis of severe pre-eclampsia was made, and the patient was subsequently admitted to be managed on diazepam, hydrallazine, and magnesium sulfate. The patient decided to go home to inform her relatives and get her personal belongings. She came back at night and refused medications. She was subsequently booked for emergency cesarean section on the next day 'due to fetal bradycardia and unfavorable cervix.' She was counseled for surgery and spinal anesthesia, which she then accepted. 'She still refused the administration of magnesium sulfate.' 'The platelet count was 130 × 10^9^/liter.' Spinal anesthesia was established at L4, L5 interspace with 1.8 ml of 'preservative-free' 0.5% plain bupivacaine after preloading her with 1 liter of normal saline and oxygen was administered at a flow rate of 6 liters per minute by face mask. The level of block was T8 and surgery commenced. The blood pressure fell from a pre-induction value of 190/110 mmHg to 182/105 mmHg, and the patient had generalized tonic-clonic seizures twice while under spinal anesthesia. 'The first seizure occurred just after delivery of the baby, but before administration of 5 i.u. of bolus intravenous oxytocin.' 'This was about 9 minutes after intrathecal injection of bupivacaine and 5 minutes after commencement of surgery.' 5 mg of intravenous diazepam was administered, and it stopped after 10 seconds. A second fit occurred 5 minutes later, and 10 mg intravenous diazepam was administered. It stopped after about 50 seconds and did not recur. 'Oxygen administration by face mask with the patient breathing spontaneously, which was commenced immediately after establishing the subarachnoid anesthesia, was continued throughout the surgical procedure. A size 3 Guedel oropharyngeal airway was inserted and retained from the onset of the first seizure. Arterial oxygen saturation remained essentially normal, and there was no vomiting or obvious regurgitation.' A live female baby was delivered with Apgar scores of 5 and 6 at 1 and 5 minutes, respectively. 'The baby was resuscitated by the neonatology senior registrar in the operating room. He suctioned the oral cavity, initially ventilated with 100% oxygen by a Rendel-Baker-Soucek pediatric facemask and later placed a size 2.5 mm orotracheal tube through which he continued artificial ventilation of the baby. The baby was wrapped in warm clothing and transferred to the special care baby unit.' She regained consciousness 15 minutes after the second seizure, but was drowsy. She was managed in the first 48 hours after surgery in the intensive care unit. 'Magnesium sulfate was not administered to the patient all through,' and she did not have any other seizure. She was discharged home on the 7^th^ post-operative day with an advice to complete her course of antibiotics, analgesics, and hematinics. DISCUSSION {#sec1-3} ========== The diagnosis of eclampsia was made in this case based on the occurrence of 'generalized' seizures against a background of pre-eclampsia.\[[@ref2]\] 'Spinal myoclonus associated with subarachnoid anesthesia with bupivacaine has been reported,\[[@ref7]\] both in Caucasians\[[@ref8]\] and patients of African origin.\[[@ref9]\] Whereas, our patient had generalized tonic-clonic seizures, spinal myoclonus associated with subarachnoid anesthesia with bupivacaine affected the lower limbs only.' When this patient with severe pre-eclampsia consented to undergo an emergency cesarean section, she was counseled for spinal anesthesia, which she also accepted. Neuroaxial block techniques such as spinal anesthesia are considered a safe method of providing anesthesia for the patient with pre-eclampsia and severe pre-eclampsia. This is due to avoidance of the risks associated with general anesthesia such as exacerbated hypertension, failed intubation, and aspiration.\[[@ref10]\] Recent studies have shown that spinal and combined spinal-epidural anesthesia can be safely administered without significantly increasing the risk to the mother and fetus, even in severe pre-eclampsia.\[[@ref11][@ref12]\] Visalyaputra *et al*.,\[[@ref11]\] showed that while there was a brief period of increased hypotension in severely pre-eclamptic patients receiving spinal versus epidural anesthesia, there were no clinical differences in fetal or maternal outcomes. Judicious volume expansion as was done with normal saline in this case may be considered prior to regional anesthesia, but caution should be used to avoid fluid overload given the increased risk of pulmonary edema.\[[@ref2]\] Although eclampsia may occur in the absence of hypertension,\[[@ref2]\] the rarity of progression from pre-eclampsia to eclampsia during cesarean sections under spinal anesthesia could be due to the tendency for the blood pressure to be lower when spinal anesthesia is used. The intra-operative eclamptic fits were successfully controlled with intravenously administered diazepam. However, magnesium sulfate is the anti-convulsant of choice for treating eclampsia. It is more effective than diazepam, phenytoin, or lytic cocktail.\[[@ref1]\] Magnesium sulfate should have been used even before the cesarean section if not for the patient\'s refusal because it has been shown to reduce the incidence both of eclampsia complicating severe pre-eclampsia and further fits in eclamptic patients.\[[@ref13]\] Eclampsia could be prevented in majority of cases if there was early booking, early identification of pre-eclampsia, and institution of appropriate therapy.\[[@ref6]\] The patient in this case, was residing in Port Harcourt, an urban city, which had two government-owned tertiary health care facilities, and several privately-owned clinics, which offer antenatal care services. She failed to utilize the services available in any of these healthcare facilities. When she presented with severe pre-eclampsia, she initially refused medications, thereby further delaying interventions that could have prevented eclampsia. Eclampsia remains a continuing problem in developing countries despite improvements in antenatal care facilities as seen in this case. Poor utilization of available antenatal care services has been identified in a study as contributing to a high incidence of eclampsia.\[[@ref6]\] It has been suggested that more awareness and enabling factors be created for more women to access antenatal care facilities.\[[@ref5]\] Also, factors preventing pregnant mothers from utilizing available antenatal care services need to be studied so as to maximize the effectiveness of these services in reducing maternal and perinatal morbidity and mortality. CONCLUSION {#sec1-4} ========== 'Progression of pre-eclampsia to eclampsia during caesarean section under spinal anaesthesia is rare,' but it can occur. Although it could be successfully treated with anticonvulsant drugs, it is better to prevent its occurrence' Eclampsia could be prevented in majority of cases by early booking for antenatal care, early identification of pre-eclampsia and prompt institution of appropriate therapy. Factors preventing utilization of available quality antenatal care services by pregnant women should be studied. **Source of Support:** Nil **Conflict of Interest:** None declared

{ "pile_set_name": "PubMed Central" }

Introduction ============ *Mycoplasma pneumoniae* (MP) pneumonia (MPP) is well known for its diverse radiological findings. In a previous report, we developed a system for scoring the affected area on thoracic computed tomography (CT) in adult patents with MPP, and reported that the affected area was associated with the presence of hypoxemia (1). However, to the best of our knowledge, no one has elucidated the correlation between the areas of lung involvement on chest X-ray and the clinical findings in both children and adults with MPP. In the present study, we retrospectively analyzed the data of MPP, to investigate the association between the values obtained by a chest X-ray scoring system and the clinical findings. Materials and Methods ===================== We retrospectively examined the data of MPP patients who visited our hospital during from April 2006 to July 2014. This study was approved by the ethics committee of Kyorin University on July 2014 (approval number: H26-032). For the purpose of the present study, "child" was defined as \<15 years of age, while "adult" was defined as ≥15 years of age. The diagnostic criteria for MPP were as follows: 1) the presence of new abnormal lung shadows on a chest X-ray, and one of the following: 2) a ≥4-fold titer rise \[complement fixation (CF) or particle agglutination (PA) tests\] during the convalescent phase in comparison to the acute phase; or a single PA titer of ＞1:160; or 3) the isolation of MP in sputum culture. The chest X-ray films were divided into three levels: 1) the bronchial bifurcation, 2) the upper level of the diaphragm, and 3) halfway between 1) and 2) ([Fig. 1](#g001){ref-type="fig"}), and total visual scores were defined as follows: grade 0, no opacity; grade 1, 5% opacity; grade 2, 5-24% opacity; grade 3, 25-49% opacity; grade 4, 50-74% opacity; and grade 5, ≥75%. The scoring system that was applied in the present study consisted of a total of six areas, the scores for each area were summed to determine the total score (range, 0-30). Two pulmonologists and one radiologist with ＞10 years of experience independently reviewed each chest X-ray film; the final decisions were made after the examiners reached a consensus. The correlations between the total score and the clinical and/or laboratory data were also evaluated. ![The chest X-ray was divided into three levels: 1) the bronchial bifurcation, 2) the upper level of the diaphragm, and 3) halfway between levels 1) and 2). Each area was calculated as follows: Grade 0, no opacity; Grade 1, \<5% opacity; Grade 2, 5-24% opacity; Grade 3, 25-49% opacity; Grade 4, 50-74% opacity; and Grade 5, \>75% opacity. The total score was calculated as the sum of the scores for the six areas.](1349-7235-56-2845-g001){#g001} Statistical analysis -------------------- The distribution and variance of numeric data were evaluated using the Kolmogorov-Smirnov test and Levene\'s median test, respectively. Categorical data are presented as percentages of the total or numerically, as appropriate. Statistical comparisons of nonparametric data were performed using the Mann-Whitney test. Categorical data were compared using the chi-squared test. All tests were two-sided. p values of ＜0.05 were considered to indicate statistical significance. All statistical analyses were performed using the SPSS software program (version 19.0 for Windows). Results ======= The clinical characteristics of the children and adults with MPP ---------------------------------------------------------------- We identified a total of 71 children and 54 adults with MPP. The characteristics of the two groups were comparable with regard to the proportions of sex, underlying respiratory diseases, the incidence of hypoxemia, and incidence of antecedent treatment with macrolide therapy ([Table 1](#t001){ref-type="table"}). Furthermore, the duration from the onset of symptoms until the referral of the patient to our hospital (with/without treatments at the local hospital) was similar in the two groups (mean±SD: children, 7.1±2.4 days versus adults, 6.9±4.6 days; not significant) ([Table 1](#t001){ref-type="table"}). However, the child group showed significantly higher body temperatures (children, 38.4 ± 0.8 vs. adults, 37.7 ± 1.1°C, p = 0.03) and higher serum levels of lactate dehydrogenase (LDH; 369 ± 142 vs. 237 ± 68.2 IU/L; p ＜0.001) and aspartate aminotransferase (36.4 ± 14.8 vs. 28.0 ± 16.8 IU/L, p ＜0.001) in comparison to the adult group. In contrast, the serum C-reactive protein levels in the adult group were significantly higher in comparison to the child group (adult group, 9.2 ± 9.0 mg/dL vs. child group, 3.0 ± 3.6 mg/dL; p ＜0.001). ###### Comparisons of Clinical Characteristics between the Child and Adult MPP Groups. Child (n=71) Adult (n=54) p value ------------------------------------------------------- -------------- -------------- ---------- Age (years) 7.9 ± 3.6 37.6 ± 18.1 p\<0.001 M:F 34:37 18:36 NS Underlying diseases (%) 11.1 (7/63) 23.4 (11/47) NS Asthma 7 9 NS Emphysema 0 1 NS Lung caner 0 1 NS Initial onset to first visit to our hospital (days)\* 7.1±2.4 6.9±4.6 NS BT (°C) 38.4 ± 0.8 37.7 ± 1.1 p=0.03 Hypoxemia (%) 18.1 (4/22) 12.2 (5/41) NS Antecedent macrolide treatment (%) 46 (23/50) 38.2 (18/47) NS WBC (×10^3^/µL) 8.5 ± 5.5 8.5 ± 4.4 NS CRP (mg/dL) 3.0 ± 3.6 9.2 ± 9.0 p\<0.001 LDH (IU/L) 369 ± 142 237 ± 68.2 p\<0.001 AST (IU/L) 36.4 ± 14.8 28.0 ± 16.8 p\<0.001 ALT(IU/L) 23.9 ± 22.3 22.4 ± 18.2 NS Diagnostic method Single titer (PA≥1:320 or CF≥1:64) 66 (93.0%) 40 (74.1%) p=0.005 Pair (×4) 5 (7.0%) 14 (25.9%) p=0.005 Culture 0 (0) 1 (1.9%) NS AST: aspartate aminotransferase, ALT; alanine aminotransferase, BT: body temperature, CRP: C-reactive protein, F: female, LDH: lactase dehydrogenase, M:male, WBC: white blood cell counts \* Data are presented as mean±standard deviation The radiological findings between children and adults with MPP -------------------------------------------------------------- Regarding the chest X-ray findings, consolidation was a major radiological finding in both groups ([Table 2](#t002){ref-type="table"}). Air bronchogram, bronchial wall thickening, and atelectasis were observed significantly more frequently in the child group than in the adult group. Other findings, such as reticular shadowing, tiny nodules, and pleural effusion, did not differ to a statistically significant extent. ###### Comparisons of Radiological Findings between the Child and Adult MPP Groups. CHILD (%) ADULT (%) p value --------------------------- ----------- ----------- --------- Total number of patients 71 54 Consolidation 62 (87.3) 45 (83.3) NS Air bronchogram 43 (60.6) 18 (33.3) p=0.004 Reticular shadowing 20 (28.2) 11 (20.4) NS Tiny nodules 17 (24.0) 12 (22.2) NS Bronchial wall thickening 24 (33.8) 8 (14.8) p=0.022 Pleural effusion 1 (1.4) 4 (7.41) NS Atelectasis 7 (9.9) 0 (0) p=0.019 The clinical significance of the chest X-ray score and the zonal predominance of MPP ------------------------------------------------------------------------------------ The correlation between hypoxemia and the total affected lung areas (calculated according to the total score) was not statistically significant in either of the groups ([Fig. 2A, D](#g002){ref-type="fig"}). Importantly, no correlation was found between the time from the onset of symptoms to referral to our hospital and the total score in either group (data not shown). Regarding zonal predominance, MPP predominantly affected the middle-to-lower lung fields in both groups ([Fig. 2B, E](#g002){ref-type="fig"}). Interestingly, a moderate positive correlation was found between the total score and the maximum PA titer value (r = 0.409, p ＜0.001), but only in the child group ([Fig. 2C, F](#g002){ref-type="fig"}). The total score was not significantly correlated with the serum inflammatory markers \[white blood cell count (WBC), LDH, and C-reactive protein (CRP)\] in either group (data not shown). ![No significant correlation between hypoxemia and the total score was found in the child group or the adult group (A, C). MPP was predominantly located in the middle-to-lower lung fields in both groups (B, D). A moderate positive correlation was found between the total score and maximum PA titer value (r=0.409, p\<0.001) only in the child group (C, F). \*p value \<0.05. \*\*\*p value \<0.001](1349-7235-56-2845-g002){#g002} The correlations between the serum LDH levels and the proportion of hypoxemia in children and adults with MPP ------------------------------------------------------------------------------------------------------------- No correlation was found between the presence of hypoxemia and the serum LDH levels in either the child group or the adult group; however, a non-significant positive correlation was observed in the latter group (p = 0.074) ([Fig. 3](#g003){ref-type="fig"}). ![No correlation was found between the presence of hypoxemia and the serum LDH level in either group.](1349-7235-56-2845-g003){#g003} Discussion ========== This study provides the first evidence of the relevance of the total affected lung area (calculated as a total score) on chest X-ray films and the clinical and laboratory findings in both children and adults with MPP. Firstly, no correlation was found between the total score and the presence of hypoxemia in either children or adults with MPP. This was contrary to our previous report, in which a scoring method was applied to assess the thoracic CT scans of adults with MPP ([@B1]). This may be because chest X-rays show lower sensitivity in the detection of faint shadowing in comparison to thoracic CT or due to absence of severe MPP patients who required admission to an ICU and/or mechanical ventilation. Secondly, this study clearly demonstrated that the middle-to-lower lung fields were predominantly affected by MPP. Previous studies in which chest X-ray films were analyzed implied that the lower lung field was predominantly affected in both children and adults with MPP ([@B2], [@B3]), while peribronchial and perivascular interstitial infiltration and/or air space consolidation were the common radiological patterns. These trends were quantitatively confirmed in the present study, with the use of the chest X-ray scoring system and by the assessment of the radiological patterns. Although the precise reason why the incidence of atelectasis in children with MPP was significantly higher in comparison to adults with MPP is unknown, it was possibly caused by the obstruction of the airway by a mucus plug due to the increased expression of mucins by the bronchial epithelial cells after MP infection ([@B4]). Thirdly, a positive correlation between the maximum PA titer and the total score was only observed in children with MPP . This discrepancy suggested the possibility that an excessive immune reaction to MP antigens ([@B5]-[@B7]), which was predominantly attributed to the inflammatory process, occurred in children rather than adults. Indeed, the findings of Tanaka et al. ([@B8], [@B9]) and the data from our previous studies ([@B5]-[@B7]) demonstrated that the enhanced cellular-mediated and/or humoral immune response to MP can exaggerate lung inflammation in mouse models. Notably, the total lung score was not correlated with the presence of hypoxemia. This finding was anticipated by general physicians who treat MPP pneumonia. The serum LDH level has been considered a marker for refractory MPP ([@B10]); however, no one has elucidated the significance of LDH levels in predicting hypoxemia. In this regard, the present study demonstrated a lack of correlation between these factors. This study is associated with some limitations. Firstly, the extent of lung abnormalities might have been underestimated due to the low sensitivity of chest X-rays in the detection of tiny nodules and reticular shadowing, which might have affected the relationship between the total score and the laboratory/clinical findings. Furthermore, this study was conducted at a regional referral center; thus, the adult MPP patients seemed to have a greater incidence of underlying respiratory disease in comparison to the typical MPP patients described in previous reports ([@B11]) who were usually treated in local hospitals. This might have affected the results of our study. However, physicians who treat MPP patients have usually recognized that no correlation exists between the extent of lung involvement on chest X-ray and the presence of hypoxemia. The results of the present study are compatible with their position. Conclusion ========== This study showed the first evidence of a correlation between the quantitatively calculated area of lung abnormalities on chest X-ray films and the clinical findings, including the presence of hypoxemia, in both children and adults with MPP. **The authors state that they have no Conflict of Interest (COI).** Takeshi Saraya and Takayasu Watanabe contributed equally to this work. [^1]: Correspondence to Dr.　Takeshi Saraya, <sara@yd5.so-net.ne.jp>

{ "pile_set_name": "PubMed Central" }

1. Introduction {#sec1} =============== Dental pulp is highly vulnerable to various physicochemical and microbiological stimuli, such as acute injury, bacterial activity and metabolites, dental bleaching, and nonpolymerized resin monomers \[[@B1]--[@B3]\]. These stimuli can cause dental pulp cell (DPC) death, eventually leading to irreversible pulp inflammation and necrosis \[[@B1]\]. One main pathogenic mechanism of these stimuli is oxidative stress- (OS-) mediated damage in DPCs \[[@B4], [@B5]\]. OS is characterized by the excessive generation of reactive oxygen species (ROS), e.g., H~2~O~2~, O~2~ ^−^, and ^·^OH \[[@B6]\]. OS stimulated by dental composites and lipopolysaccharides has been reported to induce cell cycle alteration and death of human DPCs (HDPCs) \[[@B7]\]. During dental bleaching, exogenous hydrogen peroxide (H~2~O~2~) released from the bleaching solution can result in the cell death of DPCs \[[@B5]\]. Mitochondria are the main resource and also the major attacking target of ROS \[[@B8]\]. In response to the abovementioned pathological stimuli, the excessively generated ROS will cause rapid depletion of antioxidants and then induce oxidative damage to mitochondria, which subsequently lead to mitochondrial dysfunction in HDPCs \[[@B9]\]. Such a cell death can be attenuated by the application of mitochondria-specific antioxidants \[[@B10]\]. These results indicate that mitochondrial dysfunction is highly involved in OS-induced HDPC death. However, the underlying molecular mechanisms remain largely unknown. The mitochondrial permeability transition pore (mPTP), assembled between the inner and outer mitochondrial membranes, opens with the relatively severe disturbance of intracellular redox and/or Ca^2+^ homeostasis \[[@B11], [@B12]\]. The mPTP opening can lead to a solute exchange between mitochondrial matrix contents and the surrounding cytoplasm, which is commonly linked to mitochondrial dysfunction \[[@B13]\]. Cyclophilin D (CypD) is a critical protein for mPTP opening \[[@B14]\]. CypD-dependent mPTP opening has been shown to play a key role in ROS-induced mitochondrial dysfunction \[[@B15]\] and cell death \[[@B14]\]. The pharmacological inhibition or genetic ablation of CypD can rescue mitochondrial dysfunction and cell damage induced by OS \[[@B16]\]. However, it remains unknown whether the CypD-dependent mitochondrial pathway is involved in the OS-mediated death of HDPCs. In this study, we aimed to identify the potential role of CypD in the regulation of mPTP and mitochondrial dysfunction in the OS-induced HDPC death. 2. Materials and Methods {#sec2} ======================== 2.1. Informed Consent and Ethical Approval {#sec2.1} ------------------------------------------ The human research commission of the School and Hospital of Stomatology, Wenzhou Medical University, Wenzhou, China (2018001), gave approval to the study. For the human tooth collection, a written informed consent was obtained from all the subjects. 2.2. Reagents {#sec2.2} ------------- Cell culture medium and additional supplements were bought from Life Technologies (Grand Island, NY, USA). The antibodies were obtained from Cell Signaling Technology (Beverly, MA, USA). The chamber slides and 4′,6-diamidino-2-phenylindole (DAPI) were from Invitrogen (Carlsbad, CA, USA). MitoSOX Red, TMRM, and MitoTracker Green (MTGreen, Molecular Probes) were from Life Technologies (Grand Island, NY, USA). A terminal deoxynucleotidyl transferase deoxyuridine triphosphate nick-end labeling (TUNEL) kit was from Roche (Mannheim, BW, Germany). 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), an annexin V-fluorescein isothiocyanate (FITC) detection kit, H~2~O~2~, and N-acetylcysteine (NAC) were from Sigma-Aldrich (St. Louis, MO, USA). Cyclosporine A (CsA) was from Cell Signaling Technology (Beverly, MA, USA). An adenosine triphosphate (ATP) assay kit was from Beyotime Institute of Biotechnology (Shanghai, China). 2.3. Cell Culture, Characterization, and Treatment {#sec2.3} -------------------------------------------------- HDPCs were isolated from the dental pulp tissues of noncarious third molars and grown in Dulbecco\'s modified Eagle\'s medium (DMEM) containing 10% fetal bovine serum (FBS), 100 U/mL penicillin, and 100 mg/mL streptomycin (Gibco, Grand Island, NY, USA). Cultures were maintained in a humidified atmosphere containing 5% CO~2~ at 37°C. For HDPC lineage characterization, the morphological analysis was performed, and the cells were immunocytochemically stained for vimentin and keratin. The working concentrations of the compounds were as follows: NAC (2.5 mM) and CsA (2 *μ*M). The final concentration of dimethyl sulfoxide (DMSO) in the culture was less than 0.5% in all the experiments. The cells with or without H~2~O~2~ were treated with the indicated test compounds according to the experimental protocol. 2.4. Cell Viability {#sec2.4} ------------------- HDPCs were seeded in 96-well plates and treated with the indicated reagents. Then, HDPCs were incubated with 10 *μ*L MTT solution (5 mg/mL). After 4 h, the formazan crystals were dissolved by DMSO. The plates were then measured in a microplate reader. The optical density (OD) of the control group in MTT was taken as 100% viable. 2.5. siRNA-Mediated CypD Knockdown in HDPCs {#sec2.5} ------------------------------------------- CypD siRNA targeting human peptidylprolyl isomerase F (PPIF) and the negative control (NC) were purchased from RiboBio (Guangdong, China) and transfected in HDPCs with Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer\'s protocol. 2.6. Measurement of Cell Death by Flow Cytometry and TUNEL Assays {#sec2.6} ----------------------------------------------------------------- The cell death was detected by annexin V labeled with FITC. Propidium iodide (PI) was used to determine cell necrosis. After exposure to various experimental conditions, the cells were trypsinized and labeled with fluorochromes, and then, cytofluorometric analysis was performed with a FACScan (Becton Dickinson, NY, USA). For the TUNEL assays, different groups of cells were grown on a coverslip, incubated with a TUNEL reaction mixture, counterstained with DAPI, and observed under a fluorescence microscope. 2.7. Western Blot Analyses {#sec2.7} -------------------------- HDPCs were collected and lysed in the cell lysis buffer. Proteins were separated by electrophoresis and transferred onto a polyvinylidene difluoride (PVDF) membrane. Proteins bound by primary antibodies were visualized with an appropriate secondary antibody. The protein bands were detected using the Bio-Rad imaging system (Bio-Rad, Hercules, CA, USA) and quantified using NIH ImageJ software. 2.8. Intercellular ATP Level Determination {#sec2.8} ------------------------------------------ The ATP levels were detected by an ATP detection kit according to the manufacturer\'s instructions. The data was measured via a luminescence plate reader. 2.9. Functional Imaging Assays {#sec2.9} ------------------------------ After treatment with the indicated reagents, the cells were incubated with MitoSOX (2.5 *μ*M) or TMRM (100 nM). Images were captured under the fluorescence microscope. The NIH ImageJ software was used to measure and quantify the fluorescence signals. More than 100 clearly identifiable mitochondria in 10-15 randomly selected cells per experiment were measured in three independent experiments \[[@B17], [@B18]\]. 2.10. Detection of Ca^2+^ Level {#sec2.10} ------------------------------- The cells were treated with the indicated reagents. Then, the cells were incubated with 10 *μ*M Fluo-4-AM (Beyotime, Shanghai, China), a Ca^2+^-sensitive fluorescent probe. The Ca^2+^ level was detected under a fluorescence microscope. 2.11. Data Analysis {#sec2.11} ------------------- Data were presented as the mean of three independent replicates ± standard deviation (SD) and considered significant at *p* \< 0.05. Statistical analysis was performed using StatView software (version 5.0.1, SAS Institute, USA). For comparisons between the multiple groups, one-way ANOVA was used followed by individual post hoc Fisher tests when applicable. 3. Results {#sec3} ========== 3.1. H~2~O~2~ Induced Cell Death and Mitochondrial Dysfunction in HDPCs {#sec3.1} ----------------------------------------------------------------------- The cells around the pulp tissue, which showed a spindle shape and fibroblast-like morphology after being cultured for 5 days (Supplementary [Figure 1a](#supplementary-material-1){ref-type="supplementary-material"}), were characterized as primary HDPCs. In addition, the cells exerted strong character of proliferation (Supplementary [Figure 1b](#supplementary-material-1){ref-type="supplementary-material"}) and showed mesenchymal origin by the positive expression of vimentin in the cytoplasm but were negative for keratin (Supplementary Figures [1c](#supplementary-material-1){ref-type="supplementary-material"} and [1d](#supplementary-material-1){ref-type="supplementary-material"}). As shown in [Figure 1(a)](#fig1){ref-type="fig"}, H~2~O~2~ reduced HDPC viability in a time- and dose-dependent manner. The annexin/PI staining assay revealed a significant enhancement in the incidence of cell death. 250 *μ*M H~2~O~2~ markedly enhanced the rate of cell death compared with the control group (Figures [1(b)](#fig1){ref-type="fig"} and [1(d)](#fig1){ref-type="fig"}). The H~2~O~2~-induced DNA damage was further confirmed by the TUNEL staining (Figures [1(d)](#fig1){ref-type="fig"} and [1(e)](#fig1){ref-type="fig"}). Compared with the control group, H~2~O~2~ remarkably increased the protein level of proapoptotic Bax but did not influence antiapoptotic Bcl-2 (Figures [1(f)](#fig1){ref-type="fig"} and [1(h)](#fig1){ref-type="fig"}). For the parameters of mitochondrial dysfunctions, H~2~O~2~ significantly increased the mitochondrial ROS (mtROS) level (MitoSOX staining) (Figures [1(i)](#fig1){ref-type="fig"} and [1(j)](#fig1){ref-type="fig"}) and decreased the mitochondrial membrane potential (MMP) (TMRM staining) (Figures [1(k)](#fig1){ref-type="fig"} and [1(l)](#fig1){ref-type="fig"}) in HDPCs. Furthermore, H~2~O~2~ markedly increased the intracellular Ca^2+^ level (Figures [1(m)](#fig1){ref-type="fig"} and [1(n)](#fig1){ref-type="fig"}) and reduced the ATP level ([Figure 1(o)](#fig1){ref-type="fig"}). In addition, H~2~O~2~ also significantly upregulated the expression of CypD (Figures [1(p)](#fig1){ref-type="fig"} and [1(q)](#fig1){ref-type="fig"}). 3.2. NAC Attenuated H~2~O~2~-Induced Cell Death and Mitochondrial Dysfunction in HDPCs {#sec3.2} -------------------------------------------------------------------------------------- NAC, a nonspecific antioxidant, observably preserved the cell viability in HDPCs treated with H~2~O~2~ ([Figure 2(a)](#fig2){ref-type="fig"}) and significantly mitigated cell death (TUNEL staining) (Figures [2(b)](#fig2){ref-type="fig"} and [2(c)](#fig2){ref-type="fig"}) and Bax expression (Figures [2(d)](#fig2){ref-type="fig"} and [2(e)](#fig2){ref-type="fig"}). NAC attenuated the effects of H~2~O~2~ by significantly suppressing mtROS (Figures [2(f)](#fig2){ref-type="fig"} and [2(g)](#fig2){ref-type="fig"}), increasing MMP (Figures [2(h)](#fig2){ref-type="fig"} and [2(i)](#fig2){ref-type="fig"}), decreasing the intracellular Ca^2+^ level (Figures [2(j)](#fig2){ref-type="fig"} and [2(k)](#fig2){ref-type="fig"}), and enhancing the ATP level ([Figure 2(l)](#fig2){ref-type="fig"}). NAC also decreased the CypD expression (Figures [2(m)](#fig2){ref-type="fig"} and [2(n)](#fig2){ref-type="fig"}). 3.3. Inhibition of CypD by CsA Reversed H~2~O~2~-Induced Cell Death and Mitochondrial Dysfunction in HDPCs {#sec3.3} ---------------------------------------------------------------------------------------------------------- CsA, a pharmaceutical inhibitor of CypD, significantly inhibited the H~2~O~2~-induced cytotoxicity and death, as was evident by the MTT assay ([Figure 3(a)](#fig3){ref-type="fig"}) and TUNEL staining (Figures [3(b)](#fig3){ref-type="fig"} and [3(c)](#fig3){ref-type="fig"}), respectively. Furthermore, CsA significantly downregulated the expression level of Bax and CypD (Figures [3(d)](#fig3){ref-type="fig"} and [3(f)](#fig3){ref-type="fig"}) and abrogated mtROS (Figures [3(g)](#fig3){ref-type="fig"} and [3(h)](#fig3){ref-type="fig"}) and the intracellular Ca^2+^ level (Figures [3(k)](#fig3){ref-type="fig"} and [3(l)](#fig3){ref-type="fig"}). CsA also effectively increased the MMP (Figures [3(i)](#fig3){ref-type="fig"} and [3(j)](#fig3){ref-type="fig"}) and ATP level ([Figure 3(m)](#fig3){ref-type="fig"}). 3.4. Blockade of CypD by siRNA Mitigated H~2~O~2~-Induced Cell Death and Mitochondrial Dysfunction in HDPCs {#sec3.4} ----------------------------------------------------------------------------------------------------------- The genetic knockdown of CypD with siRNA-PPIF was further conducted to validate the effect of CypD on HDPC death and mitochondrial dysfunction. Western blotting proved that siRNA-PPIF significantly downregulated the CypD expression (Figures [4(a)](#fig4){ref-type="fig"} and [4(b)](#fig4){ref-type="fig"}). We found that siRNA-PPIF significantly attenuated the H~2~O~2~-induced cytotoxicity ([Figure 4(c)](#fig4){ref-type="fig"}) and death (Figures [4(d)](#fig4){ref-type="fig"} and [4(e)](#fig4){ref-type="fig"}). siRNA-PPIF significantly inhibited the H~2~O~2~-induced expression of Bax and CypD (Figures [4(f)](#fig4){ref-type="fig"} and [4(h)](#fig4){ref-type="fig"}) and alleviated the mtROS and intracellular Ca^2+^ level (Figures [4(i)](#fig4){ref-type="fig"}, [4(j)](#fig4){ref-type="fig"}, [4(n)](#fig4){ref-type="fig"}, and [4(o)](#fig4){ref-type="fig"}). siRNA-PPIF also protected the mitochondrial function from the adverse effect of H~2~O~2~ (Figures [4(k)](#fig4){ref-type="fig"} and [4(m)](#fig4){ref-type="fig"}). 4. Discussion {#sec4} ============= Substantial evidence has shown that many pathogenic stimuli can cause significant oxidative damage, leading to inflammation and necrosis in dental pulp tissues. However, hitherto, the underlying molecular mechanisms accounting for the OS-induced death of HDPCs remain elusive. In this study, we, for the first time, reported the key role of CypD in the OS-induced death of HDPCs. We found that pharmacological blockage and genetic reduction of CypD could significantly attenuate the oxidative damage to HDPCs. Our findings delineated novel insights into the crucial role of CypD-dependent mPTP opening and mitochondrial dysfunction in the OS-induced HDPC death. Excessive ROS production can be induced in dental pulp by pathogenic stimuli, such as bacterial metabolite, dental bleaching, and unpolymerized resin monomers \[[@B2], [@B3], [@B19]\]. H~2~O~2~ is the most stable and commonly existing form of ROS and is widely used to establish cell OS injury models \[[@B20], [@B21]\]. In this study, we found that H~2~O~2~ significantly decreased the viability of HDPCs. Furthermore, H~2~O~2~ dramatically enhanced the death of HDPCs in comparison with the control group, as indicated by the results of flow cytometry and TUNEL assays. Mitochondrial outer membrane permeation (MOMP) is a key event in apoptotic cell death, which is controlled by the Bcl-2 family proteins \[[@B22]\]. Bax and Bcl-2, Bcl-2 family proteins, are important in regulating apoptotic cell death. Bax is upregulated in response to DNA damage \[[@B23]\] and regulates mitochondria-dependent cell death \[[@B24]\]. In response to apoptotic stimuli, Bax oligomerizes and forms pores perforating the outer mitochondrial membrane \[[@B25], [@B26]\]. MOMP occurs during pore formation, and the size of the pore can vary according to the number of recruited Bax dimers \[[@B27]\]. Bcl-2 plays an important part in promoting cellular survival and bating the actions of proapoptotic proteins \[[@B28]\]. Bcl-2 antagonizes MOMP by blocking Bax oligomerization and pore-forming activity \[[@B29], [@B30]\]. Our immunoblot analyses showed that H~2~O~2~ significantly increased the expression of proapoptotic protein Bax in HDPCs, which was consistent with the previous findings from osteoblastic cells \[[@B18]\] and human fetal lung fibroblast cells \[[@B31]\]. On the other hand, different from these studies, we found that Bcl-2 was not significantly affected by H~2~O~2~, which might be due to the different cell types. Increased ROS triggers cell death by inducing mitochondrial dysfunction \[[@B17], [@B32], [@B33]\]. NAC primarily abrogates mtROS and therefore mitigates myocardial cell death \[[@B20]\]. In the current study, H~2~O~2~ resulted in significant mitochondrial dysfunction, which was indicated by the increased mtROS, enhanced intracellular Ca^2+^ level, and decreased ATP level and MMP in HDPCs. NAC, in this study, not only blunted OS but also attenuated mitochondrial dysfunction in HDPCs. These results further confirmed that ROS-induced mitochondrial dysfunction played a key role in HDPC death. A mitochondria-specific antioxidant, mitoquinone, has been proved to be effective in ameliorating OS-related diseases \[[@B18], [@B34]\]. It would be highly interesting to apply these agents in this HDPC injury model to confirm the role of mitochondrial OS and dysfunction in OS-induced HDPC death in the future study. The mPTP, initiated by severe OS and cytosolic Ca^2+^ overload, is a key regulator in cell death \[[@B35], [@B36]\]. The opening of mPTP leads to rapid loss of MMP (which abolishes all MMP-dependent mitochondrial activities, including ATP synthesis), osmotic breakdown of mitochondrial membranes, and structural breakdown of the organelle, eventually causing the regulated cell death \[[@B13], [@B35], [@B36]\]. In this study, OS was shown to lead to mPTP opening in HDPCs, as reflected by MMP dissipation and Ca^2+^ disorder, which would result in the increase in mtROS and energy failure (decrease in ATP). A similar pattern was also found in H~2~O~2~-treated RAW264.7 macrophages \[[@B37]\]. NAC successfully inhibited mPTP opening and rescued mitochondrial dysfunction, thereby attenuating HDPC death. Taken together, these results showed that (i) mPTP opening was implicated in HDPC death and (ii) OS was the major inducer of mPTP opening-induced death in the H~2~O~2~-treated HDPCs. In our study, we mainly focused on the mPTP opening-induced cell death. Other than CypD, several proteins including the voltage-dependent anion channel (VDAC), adenine nucleotide translocator (ANT), and inorganic phosphate carrier (PHC) are supposed to be the putative modulators of mPTP \[[@B35]\]. However, previous studies using robust genetic tools confirm that CypD is the only vital protein in vivo required for mPTP opening \[[@B14], [@B38]\]. The knockout of VDAC, ANT, or PHC fails to protect cells from the mPTP opening \[[@B39]--[@B42]\]. In addition, mitochondrial F~1~F~O~ ATPase is also demonstrated to be involved in mPTP formation \[[@B43]\]. However, this conclusion is challenged by further reports by calculating their ion conductance and selectivity and especially the persistence of the mitochondrial permeability transition in the absence of the c-subunit of human F~1~F~O~ ATPase \[[@B44]--[@B46]\]. A very recent study largely gives structural explanation for the abovementioned inconsistent phenomena \[[@B47]\]. This study shows that there is no conventional mPTP formation but a CsA-sensitive channel when c-subunit is knocked out, still contributing to the depolarization of the inner mitochondrial membrane \[[@B47]\]. Further evidence will help to confirm the role of F~1~F~O~ ATPase in mPTP. In this study, we mainly focused on CypD and hypothesized that CypD was a key molecule in the H~2~O~2~-induced HDPC death. As expected, CsA and CypD siRNA inhibited the H~2~O~2~-induced mPTP opening, attenuated mitochondrial dysfunction, and prevented the death of HDPCs. Moreover, our results found that the CypD expression level significantly increased in the H~2~O~2~-treated HDPCs. This was consistent with a previous study showing that the increased expression of CypD contributed to the oxidative injuries in endothelial cells induced by a thyroid-stimulating hormone \[[@B16]\]. An *in vivo* study also confirmed that OS-induced cell death was mediated by CypD overexpression in the fibroblasts of patients with X-linked adrenoleukodystrophy \[[@B48]\]. Nonetheless, the overexpression of CypD was not always indispensable in its mediated cell death \[[@B49]\]. CypD may mediate the mPTP opening through three main mechanisms: (1) physical relationships with pore components \[[@B50]\], (2) ANT-independent pore formation \[[@B39]\], and (3) mitochondrial crista remodeling on the complete discharge of cytochrome C \[[@B51]\]. Consequently, CypD-mediated mPTP opening, especially in response to OS and Ca^2+^ overload, can be attenuated by either CypD genetic deletion \[[@B14]\], CypD peptidyl-prolyl *cis*- or *trans*-isomerase activity inhibition, or CsA molecular conformation \[[@B39]\]. All findings underscore a vital role of CypD-driven mPTP opening in OS-induced mitochondrial dysfunction and cell death of HDPCs. However, we only analyzed the role of CypD in H~2~O~2~-induced toxicity in dental pulp in vitro, and in vivo evidence should be achieved to corroborate the role of CypD. The upstream molecules regulating CypD remain largely unknown in the H~2~O~2~-induced death of HDPCs. A previous study indicates that gallic acid potentiates the extracellular signal-regulated kinase (ERK) phosphorylation, leading to a decrease in CypD expression, which contributes to the neuroprotective effect on cerebral ischemia/reperfusion injury \[[@B52]\]. Some studies also show that an increased thyroid-stimulating hormone can activate CypD by inhibiting sirtuin-3 (SIRT3) in endothelial cells \[[@B16]\]. Therefore, the search for vital upstream regulators of CypD in HDPCs, such as ERK, SIRT3, or p53, needs to be further explored in the future study. 5. Conclusions {#sec5} ============== Our data offered new insights into the role of CypD-mediated mitochondrial dysfunction in the H~2~O~2~-induced HDPC death (Supplementary [Figure 2](#supplementary-material-1){ref-type="supplementary-material"}) and the possible usage of the CypD inhibitor in the clinical treatment of dental pulp injury in the future. This work was supported by the Natural Science Foundation of China (81870757, 81500817, and 81870777), Zhejiang Provincial Natural Science Foundation of China (LY16H140005), Health Science and Technology Project of Zhejiang Province (2016KYB184), and Wenzhou Public Technical Research Medical Program (2015Y0507). Data Availability ================= The data used to support the findings of this study are available from the corresponding authors upon request. Conflicts of Interest ===================== The authors declare that there is no conflict of interest regarding the publication of this paper. Authors\' Contributions ======================= Shengbin Huang, Bingbing Zheng, and Xing Jin contributed equally to this work. Supplementary Materials {#supplementary-material-1} ======================= ###### Figure S1: characterization of HDPCs. Figure S2: working hypothesis. ###### Click here for additional data file. ![H~2~O~2~-induced cell death and mitochondrial dysfunction in the HDPCs. (a) Cell viability was determined by MTT reduction in the HDPCs with or without the presence of H~2~O~2~. (b, c) Flow cytometric quantification of cell death. HDPCs were treated for 24 h with culture medium or H~2~O~2~ (150 *μ*M, 250 *μ*M). (d, e) TUNEL staining and assay. (f) Representative immunoreactive bands with different densities for Bcl-2 and Bax in the HDPCs in the presence of H~2~O~2~. Quantification of immunoreactive bands for Bax (g) and Bcl-2 (h) relative to *β*-actin. Representative images showing MitoSOX staining (i) and quantification (j) in the indicated groups. Representative images with TMRM staining (k) and quantification (l) in the indicated groups. Representative images showing Fluo-4-AM staining (m) and quantification (n) in the indicated groups. ATP (o) in the indicated groups. (p) Densitometry of immunoreactive bands for CypD in the HDPCs in the presence of H~2~O~2~. (q) Quantification of immunoreactive bands for CypD relative to *β*-actin. HDPCs were treated for 24 h with H~2~O~2~ (250 *μ*M) (+) or culture medium (-). Data represent the mean values ± SD of three independent experiments.](OMCL2019-1729013.001){#fig1} ![NAC attenuated H~2~O~2~-induced cell death and mitochondrial dysfunction in the HDPCs. (a) Cell viability determined by MTT reduction in the HDPCs in the presence of H~2~O~2~ with or without NAC. (b) TUNEL staining and (c) assay after NAC treatment. (d) Representative immunoreactive bands for Bax in the HDPCs with (+) or without (-) NAC treatment in the presence of H~2~O~2~ (+) or culture medium (-). Quantification of immunoreactive bands for Bax (e) relative to *β*-actin. Representative images showing MitoSOX staining (f) and quantification (g) in the indicated groups. Representative images showing TMRM staining (h) and quantification (i) in the indicated groups. Representative images showing Fluo-4-AM staining (j) and quantification (k) in the indicated groups. ATP (l) in the indicated groups. (m) Densitometry immunoreactive bands for CypD in the HDPCs with (+) or without (-) NAC treatment in the presence of H~2~O~2~ (+) or culture medium (-). (n) Quantification of immunoreactive bands for CypD relative to *β*-actin. HDPCs were treated for 24 h with (+) or without (−) NAC (2.5 mM) in the presence of H~2~O~2~ (250 *μ*M) (+) or culture medium (−). Data represent the mean values ± SD of three independent experiments.](OMCL2019-1729013.002){#fig2} ![CsA attenuated H~2~O~2~-induced cell death and mitochondrial dysfunction in the HDPCs. (a) Cell viability determined by MTT reduction in the HDPCs in the presence of H~2~O~2~ with or without CsA. (b) TUNEL staining and (c) assay after CsA treatment. (d) Representative immunoreactive bands for Bax and CypD in the HDPCs with (+) or without (-) CsA treatment in the presence of H~2~O~2~ (+) or culture medium (-). Quantification of immunoreactive bands for Bax (e) and CypD (f) relative to *β*-actin. Representative images showing MitoSOX staining (g) and quantification (h) in the indicated groups. Representative images showing TMRM staining (i) and quantification (j) in the indicated groups. ATP (m) in the indicated groups. Representative images showing Fluo-4-AM staining (k) and quantification (l) in the indicated groups. HDPCs were treated for 24 h with (+) or without (−) CsA (2 *μ*M) in the presence of H~2~O~2~ (250 *μ*M) (+) or culture medium (−). Data represent the mean values ± SD of three independent experiments.](OMCL2019-1729013.003){#fig3} ![CypD siRNA attenuated H~2~O~2~-induced cell death and mitochondrial dysfunction in the HDPCs. (a) Representative immunoreactive bands for CypD in the HDPCs with CypD siRNA (+) or NC (-) treatment at a final concentration of 50 nM for 48 h. Quantification of immunoreactive bands for CypD (b) relative to *β*-actin. (c) Cell viability determined by MTT reduction in the HDPCs in the presence of H~2~O~2~ with or without siRNA-PPIF. (d) TUNEL staining and (e) assay after siRNA-PPIF treatment. (f) Representative immunoreactive bands for Bax, Bcl-2, and CypD in the HDPCs with (+) or without (-) siRNA-PPIF treatment in the presence of H~2~O~2~ (+) or culture medium (-). Quantification of immunoreactive bands for Bax (g), Bcl-2 (h), and CypD (i) relative to *β*-actin. Representative images showing MitoSOX staining (j) and quantification (k) in the indicated groups. Representative images showing TMRM staining (l) and quantification (m) in the indicated groups. ATP (n) in the indicated groups. Representative images showing Fluo-4-AM staining (o) and quantification (p) in the indicated groups. HDPCs were treated for 24 h with H~2~O~2~ (250 *μ*M) (+) or culture medium (−) in the CypD siRNA group (+) or the NC group (−). Data represent the mean values ± SD of three independent experiments.](OMCL2019-1729013.004){#fig4} [^1]: Academic Editor: Manuela Corte-Real

{ "pile_set_name": "PubMed Central" }

![](glasgowmedj75363-0001){#sp1 .161} ![](glasgowmedj75363-0002){#sp2 .162} ![](glasgowmedj75363-0003){#sp3 .163} ![](glasgowmedj75363-0004){#sp4 .164} ![](glasgowmedj75363-0005){#sp5 .165} ![](glasgowmedj75363-0006){#sp6 .166} ![](glasgowmedj75363-0007){#sp7 .167} ![](glasgowmedj75363-0008){#sp8 .168} ![](glasgowmedj75363-0009){#sp9 .169} ![](glasgowmedj75363-0010){#sp10 .170} ![](glasgowmedj75363-0011){#sp11 .171} ![](glasgowmedj75363-0012){#sp12 .172} ![](glasgowmedj75363-0013){#sp13 .173} ![](glasgowmedj75363-0014){#sp14 .174} ![](glasgowmedj75363-0015){#sp15 .175} ![](glasgowmedj75363-0016){#sp16 .176} ![](glasgowmedj75363-0017){#sp17 .177} [^1]: Read before a meeting of the Monklands Medical Society, on 24th November, 1914.

{ "pile_set_name": "PubMed Central" }

Introduction {#s1} ============ Glutaredoxins (Grxs) are oxidoreductases sharing a conserved 3D structure with members of the thioredoxin (Trx) superfamily. They are present in most living organisms except in some bacterial and archaeal phyla (Rouhier et al., [@B35]; Alves et al., [@B1]). The primary function of Grxs was long thought to be the reduction of disulfide bonds and more particularly those formed between reduced glutathione (GSH) and a protein cysteinyl residue, a process known as glutathionylation, but as explained below, specific Grx members could rather serve as iron-sulfur (Fe-S) cluster transfer proteins (Rouhier, [@B31]). Glutathionylation is a post-translational modification which potentially fulfills several functions. It can constitute an intermediate of some reaction mechanisms, but it can also modulate protein function and serve as a signaling mechanism or protect cysteine residues from irreversible oxidation in plants (Zaffagnini et al., [@B47]) and in animals (Dalle-Donne et al., [@B14]). Although a few additional proteins, as some specific thioredoxins, can catalyze deglutathionylation reactions, Grxs are likely the major deglutathionylation system, at least in plants (Bedhomme et al., [@B5]; Chibani et al., [@B7]). They are usually regenerated via a NADPH/glutathione reductase (GR)/GSH system, but a few Grxs can be reduced by ferredoxin- or NADPH-dependent thioredoxin reductases (Johansson et al., [@B23]; Zaffagnini et al., [@B48]). Depending on the Grx isoforms and target proteins or substrates considered, Grxs can use different catalytic mechanisms, called monothiol and dithiol mechanisms (Rouhier et al., [@B35]). The monothiol mechanism is used for the reduction of glutathionylated proteins and requires *a priori* only the catalytic cysteine (the first or more N-terminal of the two active site cysteines, referred to as Cys~A~). It performs a nucleophilic attack on the protein-glutathione adduct, the Grx becoming glutathionylated. This oxidized Grx is regenerated by reduction via a GSH molecule. The dithiol mechanism requires the catalytic cysteine but also a recycling cysteine which could be either the second active site cysteine (Cys~B~) or an additional extra active site cysteine (Cys~*C*~). If the target protein is glutathionylated, the first step is similar to the monothiol mechanism, but the glutathione-adduct formed on Grx is solved by one of these recycling cysteines instead of GSH. The resulting disulfide is then either reduced by two molecules of GSH or by a thioredoxin reductase. If the target protein has an intra or inter-molecular disulfide, the mechanism is comparable to the one used by Trxs. The catalytic cysteine forms a transient mixed disulfide with one of the cysteines of the target protein, which is then resolved by the recycling cysteine. As above, final regeneration occurs via two molecules of GSH or via a thioredoxin reductase. Based on the active site sequence, Grxs were initially categorized into two classes, a dithiol (CP\[Y/F\]C motif) and a monothiol (CGFS motif) class (Rodriguez-Manzaneque et al., [@B29]) that were subsequently renamed classes I and II, respectively (Couturier et al., [@B10]). Analyses of Grx distribution and evolution in archaea, bacteria, and eukaryotes suggested that Grx domains of classes I and II have evolved through duplication and divergence from one initial gene present in the last common ancestor of all organisms (Alves et al., [@B1]). Two additional classes are found in land plants (Couturier et al., [@B10]). The Grx isoforms belonging to class III are specific to land plants and those from class IV are also present in few algae and animals. There is no biochemical or functional information on the latter Grxs yet. From genetic analyses, it appears that class III Grx isoforms also named CC-type Grxs or ROXY play a role in plant development and in pathogen defense mechanisms, reviewed in (Meyer et al., [@B25]). These functions are related to their ability to interact with TGA transcription factors, likely regulating their redox states. However, the biochemical properties of these Grxs have not been studied in detail yet, because of the difficulty to produce soluble recombinant proteins (Couturier et al., [@B9]). Only one study demonstrated that *Arabidopsis thaliana* GrxS10 is able to regenerate *in vitro* the mitochondrial type IIF peroxiredoxin (Finkemeier et al., [@B16]). The Grxs from class II are particular considering their involvement in the maturation of Fe-S clusters and in the regulation of iron homeostasis (Rouhier et al., [@B32]), which is likely related to their capacity to bind labile Fe-S clusters and to transfer them to target proteins (Bandyopadhyay et al., [@B4]). On the other hand, though many studies performed with class II Grxs from various sources indicated that they have no or very poor efficiency for the reduction of the traditionally tested glutathionylated substrates using a GR/GSH regeneration system, at least one study indicated that they have a reductase activity. Indeed, *Chlamydomonas reinhardtii* Grx3 is able to reduce *in vitro* a glutathionylated A~4~-glyceraldehyde-3-phosphate dehydrogenase (Zaffagnini et al., [@B48]). Interestingly, it can be regenerated by the ferredoxin-thioredoxin reductase but not by the GSH/GR system. The existence of alternative regeneration systems may explain why the characterization of such activities for class II Grxs has been retarded. Overall, the demonstrated capacity of several class I Grxs to provide electrons to a battery of enzymes, as peroxiredoxins and methionine sulfoxide reductases, enzymes that go through a catalytic cycle which involves glutathionylation of their catalytic cysteine (Rouhier et al., [@B34]; Vieira Dos Santos et al., [@B43]; Tarrago et al., [@B42]), suggests that they are the central and favorite deglutathionylating agents in cells. In plants, class I regroups six different glutaredoxin members called GrxC1, C2, C3, C4, S12, and C5, the latter being restricted to Brassicaceae (Couturier et al., [@B13]). The cytoplasmic GrxC1 and the plastidial GrxC5 are able to bind an \[Fe~2~S~2~\] cluster, at least *in vitro* into recombinant proteins (Rouhier et al., [@B36]; Couturier et al., [@B13]). The biochemical and structural characterization of the two plastidial isoforms, GrxS12 and GrxC5, confirmed that the apoforms of these Grxs reduce glutathionylated proteins with catalytic properties quite similar to other plant and non-plant class I Grxs (Couturier et al., [@B11], [@B13]). Moreover, poplar GrxS12 was shown to be sensitive and temporarily inactivated after treatments with oxidizing compounds such as hydrogen peroxide (H~2~O~2~), nitrosoglutathione (GSNO), and oxidized glutathione (GSSG) (Zaffagnini et al., [@B46]). Considering the very electronegative redox potential of the glutathione adduct formed on the catalytic Cys~A~ (−315 mV at pH 7.0), the reduction of GrxS12 necessitates a quite low redox potential of the GSH/GSSG couple in the chloroplast. Thus, GrxS12 may act as a stress-related redox regulator allowing glutathione to play a signaling role in some oxidizing conditions by maintaining the glutathionylation of its target proteins over the duration of the stress period. Several class I Grxs, *C. reinhardtii* Grx1 and Grx2 (Gao et al., [@B19]), *A. thaliana* GrxC1, C2 (Riondet et al., [@B28]) and GrxC5 (Couturier et al., [@B13]) and poplar GrxC1, GrxC4, and GrxS12 (Rouhier et al., [@B33], [@B36]; Couturier et al., [@B11]; Zaffagnini et al., [@B46]) have been partially or extensively characterized at the biochemical level, but some members have not yet been studied and there is no thorough and comparative study allowing to understand if the existence of several members is only related to their sub-cellular localization or to their expression pattern or whether some biochemical properties can partially explain their diversity. From the biochemical analyses performed here with the four dithiol class I Grxs existing in poplar, GrxC1 to GrxC4 (the fifth member, GrxS12, has a monothiol active site and was characterized previously), two sub-groups with distinct properties emerged. In addition, this work provides new insights on the existence of different oxidation forms involving the active site cysteines (Cys~A~ and Cys~B~) or an additional C-terminal cysteine (Cys~*C*~) and on their potential role for the reaction mechanism. Materials and methods {#s2} ===================== Site-directed mutagenesis ------------------------- The construction of pET3d expression plasmids containing poplar GrxC1, C2, C3, and C4 was described previously (Rouhier et al., [@B33], [@B36]). All cysteine residues of GrxC2, C3 were individually substituted into serines using two complementary mutagenic primers (listed in Table [S1](#SM7){ref-type="supplementary-material"}). The corresponding mutated proteins are called GrxC2 C23S, C26S, or C80S, GrxC3 C37S or C40S. The recombinant proteins GrxC1 C31S, C34S, C88S, and GrxC4 C27S and C30S were prepared previously (Rouhier et al., [@B33], [@B36]). Two additional mutants, in which only the catalytic cysteine is remaining, GrxC1 C34/88S and GrxC2 C26/80S, have also been cloned. Heterologous expression and purification of recombinant proteins in *E. coli* ----------------------------------------------------------------------------- For protein expression, cultures of the *E. coli* BL21(DE3) strain, containing the pSBET plasmid and co-transformed with the different recombinant pET3d plasmids, were successively amplified up to 2.4 L in LB medium at 37°C supplemented with 50 μg/mL of ampicillin and kanamycin. Induction of protein expression was performed at exponential phase by adding 100 μM isopropyl β-D-thiogalactopyranoside for 4 h at 37°C. After centrifugation (20 min at 4400 g), the cell pellets were resuspended in about 20 mL of TE NaCl buffer (30 mM Tris-HCl pH 8.0, 1 mM EDTA, 200 mM NaCl) and eventually conserved at −20°C. Cell lysis was performed by sonication and the soluble and insoluble fractions were separated by centrifugation for 45 min at 27,000 g. Purification of proteins was carried out in three steps. The soluble fraction was first precipitated by ammonium sulfate from 0 to 40% and then to 80% of the saturation. The 40--80% ammonium sulfate-precipitated fraction was subjected to gel filtration chromatography (ACA44 gel) equilibrated with TE NaCl buffer. After dialysis against TE (30 mM Tris-HCl pH 8.0, 1 mM EDTA) buffer and concentration, the interesting fractions were loaded to a DEAE (diethylaminoethyl) sepharose column equilibrated in TE buffer. GrxC1 WT, GrxC2 WT and their corresponding cysteinic mutants passed through the column, whereas GrxC3 WT, GrxC4 WT and their respective cysteinic mutants were retained and eluted using a linear 0--0.4 M NaCl gradient. The purest fractions, as judged by SDS-PAGE analysis, were pooled and dialyzed against TE buffer by ultrafiltration in Amicon cells equipped with a YM10 membrane. Finally, the fractions were concentrated and stored at −20°C until further use. Protein purity was checked by SDS-PAGE, and protein concentrations were determined spectrophotometrically using the corresponding molar extinction coefficients at 280 nm of 10,095 M^−1^ cm^−1^ for GrxC1 WT and its monocysteinic mutants, 9970 M^−1^ cm^−1^ for GrxC1 C34/88S, 8605 M^−1^ cm^−1^ for GrxC2 WT and its monocysteinic mutants, 8480 M^−1^ cm^−1^ for GrxC2 C26/80S, 7575 M^−1^ cm^−1^ for GrxC3 WT, 7450 M^−1^ cm^−1^ for GrxC3 C37S and C40S, 4595 M^−1^ cm^−1^ for GrxC4 WT and 4470 M^−1^ cm^−1^ for GrxC4 C27S and C30S. Enzymatic activities -------------------- The thioltransferase activities \[2-hydroxyethyl-disulfide (HED) and dehydroascorbate (DHA) assays\] were measured as described previously (Couturier et al., [@B13]) using as purified non-reduced proteins. Briefly, measurements were performed at 25°C in steady-state conditions by following NADPH oxidation at 340 nm in the presence of a Grx reducing system which is composed of NADPH, GR, and GSH. The reaction was started by adding Grx after a 3-min pre-incubation time and Grx activity was corrected by subtracting the spontaneous reduction rate observed in the absence of Grx. Because of the spontaneous reaction of GSH with the substrates, we can generally not work at saturating concentrations of all substrates using these assays. Enzyme and substrate concentrations used are indicated in the legend of figures and tables. The activity was expressed as nmol of NADPH oxidized/nmol of Grx/s using a molar extinction coefficient of 6220 M^−1^ cm^−1^ at 340 nm for NADPH. Three independent experiments were performed at each substrate concentration, and *k*~*cat*~ and apparent *K*~*m*~ values were calculated by non-linear regression using the Michaelis--Menten equation using the program GraphPad Prism 4. Preparation and analysis of oxidized Grxs ----------------------------------------- Around 10 mg of protein was reduced using 40 mM DTT in 500 μl of 30 mM Tris-HCl pH 8.0 buffer for 2 h at 25°C followed by desalting on G25 column pre-equilibrated with 30 mM Tris-HCl pH 8.0 buffer. Oxidized proteins were prepared by incubating 500 μM reduced poplar Grxs in *ca* 200 μL with a 10 fold excess GSSG or H~2~O~2~ or a 5 fold excess GSNO for 2 h at 25°C before desalting on G25 columns. The protein oxidation state was then analyzed by electrospray ionization mass spectrometry analysis as described previously (Couturier et al., [@B13]). For alkylation assays, reduced Grxs treated or not with oxidants were diluted to 10 μM into 50 μ l of 30 mM Tris-HCl pH 8.0 buffer and precipitated on ice for 30 min with one volume of 20% trichloroacetic acid (TCA). After centrifugation (10 min at 13,000 *g*) and washing with 100% acetone, the pellet was resuspended into 10 μl of 100 mM Tris-HCl pH 8.0, 1% SDS containing 2 mM of methoxyl-PEG maleimide of 2 kDa (mPEG maleimide) which alkylates free thiol groups. The protein mixture was then separated on non-reducing 15% SDS-PAGE. For midpoint redox potential titrations, Cys~A~ glutathionylated forms were obtained by the above described GSNO treatment of the double cysteinic mutants (GrxC1 C34/88S, GrxC2 C26/80S) and of the monocysteinic mutants (GrxC3 C40S and GrxC4 C30S). The intramolecular disulfide bond (Cys~A~-Cys~B~) containing proteins were obtained by the above described H~2~O~2~ treatment but using GrxC1 C88S, GrxC2 C80S and GrxC3 and GrxC4 WT isoforms. P*K*~*a*~ determination of the catalytic cysteine ------------------------------------------------- The p*K*~*a*~ measurements of N-terminal active site cysteines (Cys~A~) have been performed with WT Grx isoforms following a procedure described in (Gallogly et al., [@B17]). Briefly, 3 μM reduced Grx was incubated with or without 300 μM iodoacetamide (IAM) in 100 mM sodium citrate or phosphate buffers ranging from pH 2.0 to 7.0. Following this pre-incubation step, Grx activity was determined by adding an aliquot of the pre-incubation mixture to the HED assay described above. The percentages of remaining activity at each pH were determined by comparing the activity of the enzyme incubated with and without IAM and an adaptation of the Henderson--Hasselbach equation (Gallogly et al., [@B17]) was used for p*K*~*a*~ value determination using the program GraphPad Prism 4. Midpoint redox potential (*E*~*m*~) determination ------------------------------------------------- Oxidation-reduction titrations using the fluorescence of the adduct formed between protein free thiols and monobromobimane (mBBr) were carried out at ambient temperature in 200 μl of 100 mM HEPES pH 7.0 buffer containing 10 μM of oxidized proteins, either glutathionylated or with an intramolecular disulfide-bond, and defined mixtures of oxidized and reduced DTT or glutathione for more positive values to set the ambient potential (*E*~*h*~). Total concentration of DTT or glutathione was 2 mM. After 2 h incubation, mBBr was added at a final concentration of 2 mM and the reaction was carried out in the dark for 20 min. The reaction mixture was then precipitated on ice for 30 min with one volume of 20% TCA. After centrifugation (10 min at 13,000 g) and washing with 100% acetone, the pellet was resuspended into 400 μl of 100 mM Tris-HCl pH 8.0, 1% SDS. Fluorescence emission of the resulting solution was measured at 472 nm after excitation at 380 nm using a variant Cary Eclipse (Agilent). Values were transformed into percentages of reduced protein and fitted to the Nernst equation using non-linear regression for *E*~*m*~ value determination. Results {#s3} ======= Kinetic comparison of poplar class I Grxs ----------------------------------------- An amino acid sequence alignment of the predicted mature forms of all poplar class I Grxs reveals several conserved residues, particularly those around the active sites and those involved in glutathione binding (Figure [1](#F1){ref-type="fig"}). However, GrxC1 and GrxC2 present CGYC and CPFC active site motifs, respectively, different from the CPYC active site found in GrxC3 and GrxC4. Interestingly, contrary to GrxC3 and GrxC4, GrxC1 and GrxC2 isoforms possess an additional cysteine (Cys~*C*~) located in their C-terminal part in a IGGCD motif, as in human Grx1, *E. coli* Grx3 and the two other class I Grxs found in plants, GrxC5 and GrxS12 (Figure [1](#F1){ref-type="fig"}). With the aim of comparing the redox and kinetic properties of all dithiol Grxs (GrxC5 is not found in poplar) and understanding why these duplicated genes have been retained during evolution, we have performed a thorough biochemical analysis of the corresponding recombinant proteins. ![**Amino acid alignment of class I glutaredoxins**. The alignment was done using ClustalW from sequences of the recombinant proteins that are deprived of putative targeting sequences. The strictly conserved amino acids are depicted in white on black. Other conservative amino acid changes are indicated in black on gray. Residues involved in GSH binding based on the 3D structures of poplar GrxC4 and GrxS12 and Arabidospis GrxC5 are marked by an asterisk.](fpls-04-00518-g0001){#F1} The catalytic properties of each isoform were first determined by measuring their reductase activity using the classical HED and DHA assays, which measure the reduction of glutathionylated-β-mercaptoethanol or of dehydroascorbate, respectively. All proteins display a better catalytic efficiency (*k*~*cat*~/*K*~*m*~) in the HED test with a difference of two orders of magnitude (Table [1](#T1){ref-type="table"}). In both assays, the kinetic parameters determined for each protein follow the same tendency. GrxC3 and C4 have quite comparable turnover numbers, with *k*~*cat*~ values about 10 fold higher than those of GrxC1 and C2. In terms of catalytic efficiencies, the dichotomy between GrxC1/C2 and GrxC3/C4 is clear but it is often below this factor of 10 because GrxC3 and C4 have higher apparent *K*~*m*~ values for both substrates. Overall, GrxC1 always exhibited the lowest catalytic efficiency and GrxC4 the highest. Concerning the apparent affinity for GSH, although there are differences in the *K*~*m*~ values for GSH between the four isoforms, ranging from \~2.0 mM for GrxC4 to \~3.9 mM for GrxC2, the GrxC1/C2 and GrxC3/C4 dichotomy is not visible using this parameter. ###### **Kinetic parameters of poplar dithiol class I glutaredoxins**. **PtGrxC1** **PtGrxC2** **PtGrxC3** **PtGrxC4** **PtGrxS12[^a^](#TN1){ref-type="table-fn"}** **AtGrxC5[^a^](#TN1){ref-type="table-fn"}** --------- ------------------------------------------ ------------- ------------- ------------- ---------------------------------------------- --------------------------------------------- ------------- β-ME-SG *K*~*m*~ (mM) 0.10 ± 0.01 0.06 ± 0.01 0.50 ± 0.06 0.32 ± 0.04 0.31 ± 0.04 0.20 ± 0.02 *k*~*cat*~(s^−1^) 7.9 ± 0.2 11.7 ± 0.1 77.0 ± 4.4 90.3 ± 4.9 23.1 ± 0.9 1.21 ± 0.03 *k*~*cat*~/*K*~*m*~(x 10^5^ M^−1^.s^−1^) 0.79 1.95 1.54 2.82 0.745 0.06 DHA *K*~*m*~ (mM) 0.55 ± 0.06 0.37 ± 0.03 1.14 ± 0.13 0.60 ± 0.09 0.38 ± 0.19 0.21 ± 0.03 *k*~*cat*~(s^−1^) 0.6 ± 0.1 1.3 ± 0.1 6.5 ± 0.4 8.2 ± 0.6 1.7 ± 0.2 0.23 ± 0.01 *k*~*cat*~/*K*~*m*~(x 10^3^ M^−1^.s^−1^) 1.09 3.51 5.70 13.67 4.6 1.1 GSH *K*~*m*~ (mM) 2.76 ± 0.26 3.87 ± 0.53 3.60 ± 0.42 1.96 ± 0.28 4.0 ± 0.5 nd *k*~*cat*~(s^−1^) 13.8 ± 0.6 30.1 ± 2.3 131.3 ± 8.3 102.9 ± 6.5 92.8 ± 5.3 nd *k*~*cat*~/*K*~*m*~(x 10^3^ M^−1^.s^−1^) 5.00 7.78 36.47 52.50 23.20 nd The apparent K~m~ values for β-ME-SG (glutathionylated β-mercaptoethanol) were determined using a HED concentration range of 0.1--1 mM in presence of 2 mM GSH and 100 nM enzyme for GrxC1, 50 nM for GrxC2, 10 nM for GrxC3 and GrxC4. The apparent K~m~ values for DHA were determined using a DHA concentration range of 0.1--1 mM in presence of 2 mM GSH and 2 μM enzyme for GrxC1, 1 μM for GrxC2, 200 nM for GrxC3, and 150 nM for GrxC4. The apparent K~m~ values for GSH were determined using a GSH concentration range of 0.25--5 mM in the presence of 0.7 mM HED and 100 nM enzyme for GrxC1, 50 nM for GrxC2, 10 nM for GrxC3 and GrxC4. The k~cat~ and K~m~ values were calculated by non-linear regression using the Michaelis--Menten equation. Values are the means ± SD of three replicates. Data extracted from the following references Couturier et al. ([@B11], [@B13]), Zaffagnini et al. ([@B46]). nd, not determined. Then, the involvement for the catalytic mechanism of the different cysteine residues present in each isoform was investigated by comparing the activity of WT proteins to their corresponding cysteinic mutants using similar assays at fixed substrate concentrations (Figure [2](#F2){ref-type="fig"}). As expected, all monocysteinic mutants for the catalytic cysteine Cys~A~ (GrxC1 C31S, GrxC2 C23S, GrxC3 C37S, and GrxC4 C27S) were inactive, confirming that it is indispensable for the reductase activity. On the contrary, mutating Cys~*C*~, in positions 88 and 80 in GrxC1 and GrxC2, respectively, did not affect catalytic activity. Finally, the mutation of Cys~B~ has a different impact depending on the isoforms. The protein variants for this cysteine in GrxC1 and C2 (GrxC1 C34S and GrxC2 C26S) were 5 fold more efficient and 2--3 fold more efficient than the corresponding WT proteins in the HED and DHA assays, respectively. On the other hand, the mutation of this cysteine in GrxC3 and GrxC4 has almost no impact, since, only a slight decrease was observed in the HED assay for the GrxC3 C40S variant. ![**Reductase activity of GrxC1, C2, C3, and C4 and of cysteine-to-serine variants. (A)** HED reduction was measured in the presence of 2 mM GSH and 0.7 mM HED using either 100 nM GrxC1, 20 nM for GrxC1 C34S, 50 nM GrxC2, 10 nM for GrxC2 C26S, 10 nM GrxC3, or 10 nM GrxC4. **(B)** DHA reductase activity was measured in the presence of 2 mM GSH and 1 mM DHA using either 2 μM GrxC1, 1 μM for GrxC1 C34S, 1 μM GrxC2, 250 nM GrxC2 C26S, 200 nM GrxC3, or 150 nM GrxC4. The data are expressed as nmol NADPH oxidized per nmol enzymes per s and represent means ± SD of three separate experiments.](fpls-04-00518-g0002){#F2} Oxidative modifications of cysteine residues -------------------------------------------- Several observations prompted us to analyze the sensitivity of all dithiol class I Grxs to oxidizing treatments and the resulting changes in their redox state. First, the mutation of Cys~B~ increased the catalytic efficiency of GrxC1 and C2 but not of GrxC3 and C4 suggesting possible difference in the reaction mechanisms. Second, although Cys~*C*~ does not influence their catalytic activity, the migration of the purified proteins in non-reducing SDS PAGE gels indicated the formation of covalent linkages for GrxC1 and C2 that are not visible for GrxC3 and C4 (data not shown). For this purpose, reduced proteins have been treated for 2 h with a 10 fold excess of GSSG or H~2~O~2~, or a 5 fold excess of GSNO. The resulting redox state was analyzed by mass spectrometry and by protein migration on non-reducing SDS PAGE with or without thiol alkylation by 2 kDa mPEG maleimide, which allows distinguishing reduced and oxidized forms of Grxs. Since reduced mPEG-alkylated proteins and covalent dimers of GrxC1 and GrxC2 migrate roughly at the same position, controls without alkylation have been included (Figure [3](#F3){ref-type="fig"}). With non-alkylated proteins, we confirmed that an H~2~O~2~ treatment, but not the GSSG and GSNO treatments, promoted the formation of covalent dimers of GrxC1 and GrxC2 (Figures [3A,B](#F3){ref-type="fig"}) but not of GrxC3 and GrxC4 (Figures [3C,D](#F3){ref-type="fig"}). Only a small part of GrxC2 dimer is visible after these 2 h GSSG and GSNO treatments. From the procedure that included an alkylation step, we can further conclude that all treatments led essentially to a complete oxidation of all thiol groups present on all Grxs since almost no shift of migration was observed. A minor form with one free thiol group is visible for GrxC2 after a GSSG treatment. It is even more discrete upon GSNO treatment. For GrxC3 and C4, these oxidized monomeric forms are interpreted as containing intramolecular disulfides between active site cysteines (Cys~A~-Cys~B~) (Figures [3C,D](#F3){ref-type="fig"}). For GrxC1 and C2, these two cysteines should also form this intramolecular disulfide, and Cys~*C*~ is either modified by a glutathione or a nitroso (in the case of GSNO) adduct as confirmed by mass spectrometry analyses (see below) or it is involved in an intermolecular disulfide upon H~2~O~2~ treatment. From similar experiments performed with the GrxC1 C88S and GrxC2 C80S mutants (Figures [3E,F](#F3){ref-type="fig"}), the absence of covalent dimers indicated that Cys~*C*~ is indeed involved in dimerization. Moreover, in the context of these variants, after GSSG and GSNO treatments, the proteins mainly contain an active site disulfide bond whereas a small part of the proteins has one remaining free thiol, likely corresponding to glutathionylated intermediates. ![**Sensitivity of poplar Grxs to oxidative treatments**. Pre-reduced Grxs were subjected to 2 h oxidative treatments with H~2~O~2~, GSSG, and GSNO as described in the "materials and methods" section. Prior to separation on SDS-PAGE gels in non-reducing conditions, GrxC1 WT **(A)**, GrxC2 WT **(B)**, GrxC3 WT **(C)**, GrxC4 WT **(D)**, GrxC1 C88S **(E)**, and GrxC2 C80S **(F)** were alkylated or not with 2 kDa mPEG maleimide. The shift observed following the alkylation of the free thiol groups is larger than expected (ca. 4 kDa instead of 2 kDa), but this has been already observed previously (Chibani et al., [@B8]). The numbers on the right correspond to the number of alkylated thiols that remained free after reducing or oxidizing treatment of Grxs. The stars in panels **(A)** and **(B)** indicate the position of covalent dimer.](fpls-04-00518-g0003){#F3} In order to identify the type of protein modification formed by these oxidative treatments, all proteins have been analyzed by mass spectrometry (Table [2](#T2){ref-type="table"} and Figures [S1](#SM1){ref-type="supplementary-material"}--[S6](#SM6){ref-type="supplementary-material"}). First, all reduced proteins gave a single species with a mass decrease of *ca* 131 Da compared to the calculated theoretical molecular masses. It does undoubtedly correspond to the cleavage of the first methionine as expected from the presence of an alanine in second position. The results obtained with GrxC3 and GrxC4 are more easily interpreted since these proteins contain only two cysteines. In the presence of H~2~O~2~, and in accordance with the absence of detected free thiols in the previous experiments (Figures [3C,D](#F3){ref-type="fig"}), GrxC3 and GrxC4 are exclusively detected as a monomeric oxidized form with a decrease of \~2 Da confirming the formation of a Cys~A~-Cys~B~ intramolecular disulfide bond. The same forms were primarily obtained upon GSSG and GSNO treatments but a minor monomeric form with a mass increment of 305 Da corresponding to one glutathione adduct was also detected (Table [2](#T2){ref-type="table"}). It was not detected on gels presumably because of the presence of minute amounts. Nevertheless, this suggests that the glutathionylated form represents an intermediate for the formation of the intramolecular disulfide. ###### **Electrospray ionization mass spectrometry analysis of the redox states of poplar class I Grxs**. **Protein** **Theoretical size (Da)** **Theoretical size without Met (Da)** **DTT~red~ treatment** **GSSG treatment** **Relative percentage values (%)** **GSNO treatment** **Relative percentage values (%)** **H~2~O~2~ treatment** **Relative percentage values (%)** **Deduced redox state** -------------- --------------------------- --------------------------------------- ------------------------ ----------------------- ------------------------------------ ----------------------- ------------------------------------ ------------------------ ------------------------------------ ------------------------- ---- ------------------ PtGrxC1 WT 12,514.3 12,383.1 12,382.7 **12,686.3 (+303.6)** **76.4** 12,380.0 (−2.7) 5.7 **12,412.2 (+29.5)** **72.1** SH SH SH 12,993.8 (+611.1) 23.6 **12,686.3 (+303.6)** **81.8** 24,756.5 (×2) 27.9 12,993.4 (+610.7) 12.5 S-S SH S-S SO~2~H S-S SSG SSG SH SSG S-S S-S (interchain) PtGrxC2 WT 12,178.9 12,047.7 12,047.3 12,045.0 (−2.3) 11.1 12,044.9 (−2.4) 23.7 12,077.2 (+32.2) 50 SH SH SH **12,351.5 (+304.2)** **68.4** **12,350.7 (+303.4)** **76.3** 24,088.9 (×2) 50 12,657.9 (+610.6) 20.5 S-S SH S-S SO~2~H S-S SSG SSG SH SSG S-S S-S (interchain) PtGrxC3 WT 12,516.2 12,385.0 12,384.7 **12,382.7 (−2.0)** **66.3** **12,382.9 (−1.8)** **75.9** 12,382.7 (−2.0) 100 SH SH 12,690.0 (+305.3) 33.7 12,690.0 (+305.3) 24.1 S-S SSG SH PtGrxC4 WT 12,526.1 12,394.9 12,394.4 **12,392.0 (−2.4)** **62.5** **12,392.1 (−2.5)** **90.8** 12,392.3 (−2.7) 100 SH SH 12,699.6 (+305.2) 37.5 12,699.2 (+304.8) 9.2 S-S SSG SH PtGrxC1 C88S 12,498.2 12,367.0 12,366.6 **12,364.7 (−1.9)** **64.0** **12,364.5 (−2.1)** **79.6** 12,364.6 (−2.0) 100 SH SH 12,671.8 (+305.2) 36.0 12,672.0 (+305.4) 20.4 S-S SSG SH PtGrxC2 C80S 12,162.8 12,031.7 12,030.6 **12,029.3 (−1.3)** **81.5** 12,028.8 (−1.8) 100 12,028.7 (−1.9) 100 SH SH 12,336.5 (+305.9) 18.5 S-S SSG SH Reduced proteins and proteins treated with GSNO, GSSG, or H~2~O~2~ were analyzed by mass spectrometry. The mass differences are indicated between parentheses. The mass accuracy is generally ±0.5 Da. Comparing mass differences and migration profiles of corresponding alkylated proteins under non-reducing conditions enabled to deduce the redox state of each cysteine residue. When several oxidized forms are detected and with clear different abundance, the major form is indicated in bold. The relative percentage of each form has been deduced from the spectra assuming equal ionization of all GRX forms. The results obtained for GrxC1 and GrxC2 are slightly more complex. In response to an H~2~O~2~ treatment, two different species are formed for both proteins. In accordance with the results from alkylation experiments, a major form corresponds to a disulfide-bridged dimer having also an intramolecular Cys~A~-Cys~B~ disulfide bond. The detected monomeric forms, which have no free thiols or a very small fraction as observed for GrxC2, exhibited a mass increment of 29.5 Da for GrxC1 and 32.2 Da for GrxC2, suggesting the presence of two oxygen atoms. Considering that this overoxidation is not observed for GrxC1 C88S and GrxC2 C80S variants which only form monomeric proteins with an intramolecular disulfide (as indicated by the mass decrease of 2 Da), the results obtained for the WT proteins are best interpreted as the formation of an intramolecular disulfide between the active site cysteines (Cys~A~-Cys~B~) and of a sulfinic acid on Cys~*C*~. In the presence of GSSG or GSNO, GrxC1 and GrxC2 exist only under monomeric forms but with variable oxidation states. Except for the GSSG-treated GrxC1, there is a minor form with a mass decrease of about 2 Da consistent with the formation of an intramolecular disulfide. Another minor form, found in all samples except the GSNO-treated GrxC2, exhibited a mass increment of 610.5--611 Da, which corresponded to the presence of two glutathione adducts. However, the major species formed by these two treatments is a monomer with a mass increment of about 303.5 Da which is consistent with the presence of one glutathione adduct on Cys~*C*~ in addition to the Cys~A~-Cys~B~ disulfide bond. Accordingly, the major species detected in protein variants mutated for Cys~*C*~ (Grx C1 C88S and GrxC2 C80S) is the one with an intramolecular disulfide (decrease of 2 Da upon GSSG or GSNO treatments). The glutathionylated forms detected by mass spectrometry were visible on gels (see Figures [3E,F](#F3){ref-type="fig"}). The *in vitro* observation of intermolecular disulfide bonds in GrxC1 and C2, involving Cys~*C*~, raised the question of their possible reduction. Thus, we tested the ability of DTT but also of a physiological GSH reducing system composed of NADPH, GR, and GSH to reduce this intermolecular disulfide, the formation of which was initiated by a 2h H~2~O~2~ treatment. As shown in Figure [4](#F4){ref-type="fig"}, both DTT and GSH are able to reduce GrxC1 and GrxC2 dimers to a similar extent. Whether minute amounts of reduced monomeric Grx initially present or formed can catalytically increase the reduction of dimeric Grxs in the presence of GSH cannot be excluded. Overall, it indicates that the formation of these disulfide-bridged forms of GrxC1 and GrxC2 isoforms is reversible, notably by GSH, a relevant cellular physiological reductant. ![**GSH-dependent reduction of the intermolecular disulfide in poplar GrxC1 and GrxC2**. The ability of GSH to reduce poplar disulfide-bridged dimer GrxC1 and disulfide-bridged dimer GrxC2 generated by a 2 h H~2~O~2~ treatment was evaluated in the presence of 1 mM DTT and 1 mM GSH coupled to NADPH and GR. The lanes Ox represent the initial oxidized proteins. Proteins were separated on SDS-PAGE gels under non-reducing conditions.](fpls-04-00518-g0004){#F4} Redox properties of class I glutaredoxins ----------------------------------------- As in other thiol-dependent oxidoreductases, the catalytic efficiency of class I Grxs should be at least partially governed by thermodynamic parameters as the redox midpoint potentials of dithiol-disulfide couples and by the reactivity of the catalytic cysteine which is dependent on its p*K*~*a*~. Hence, we determined the Cys~A~ p*K*~*a*~ of the four Grx isoforms using a method relying on iodoacetamide which is an alkylating reagent reacting with thiolates but not thiols. The pH-dependent inactivation of WT Grx by iodoacetamide was followed by comparing the activity of WT proteins in the HED assay, after pre-incubation of the protein in different buffers ranging from pH 2.0 to 7.0 in the presence or not of iodoacetamide. From these titration curves, we obtained p*K*~*a*~ values of 5.3 ± 0.1 for GrxC1, 5.0 ± 0.1 for GrxC2, 4.6 ± 0.1 for GrxC3, and 4.6 ± 0.1 for GrxC4 (Figure [5](#F5){ref-type="fig"}). These results indicate first that at physiological pH, the thiolate ion will be the dominant species for all Grxs. Although the differences are not huge, it also confirmed the existence of two subgroups, GrxC1 and C2 having higher p*K*~*a*~compared to GrxC3 and C4. ![**p*K*~*a*~ determination of the catalytic cysteine of poplar GrxC1, C2, C3, and C4**. Reduced GrxC1 **(A)**, GrxC2 **(B)**, GrxC3 **(C)**, and GrxC4 **(D)** WT proteins were incubated in different buffers ranging from pH 2.0 to 7.0 in presence or not of iodoacetamide (IAM) prior to measurement of their activity by using the HED assay. The percentages of remaining activity at each pH were determined by comparing the activity of the enzyme incubated with and without IAM and an adaptation of the Henderson--Hasselbach equation was used for p*K*~*a*~ value determination. The data are represented as mean ± SD of three separate experiments.](fpls-04-00518-g0005){#F5} Two disulfide bond forms may be catalytically relevant for all these Grxs, the one with a glutathionylated Cys~A~ and the one with a Cys~A~-Cys~B~ intramolecular disulfide bond. For measuring the midpoint redox potentials (*E*~*m*~) of each of these disulfide/dithiol couples, we used different protein variants and two distinct oxidative treatments in order to discriminate between these two oxidized forms. To obtain proteins containing an intramolecular disulfide bond, reduced Grxs possessing only the two active site cysteines (GrxC1 C88S, GrxC2 C80S, GrxC3 WT, and GrxC4 WT) have been treated with H~2~O~2~ before desalting on G25 columns. On the other hand, glutathionylated proteins have been obtained by treating reduced proteins in which only Cys~A~ was remaining (GrxC1 C34/88S, GrxC2 C26/80S, GrxC3 C40S, and GrxC4 C30S) with GSNO before desalting. The *E*~*m*~ values determined at pH 7.0 for the intramolecular disulfide range from −233 mV for GrxC3 to −263 mV to GrxC1 (Figure [6](#F6){ref-type="fig"}). On the other hand, only little variations were visible when measuring *E*~*m*~ values for the glutathione adduct formed on the catalytic cysteine. They varied from −242 mV for GrxC3 to −254 mV for GrxC1. Overall, the intramolecular active site disulfide bond and glutathionylated forms of GrxC2 and GrxC4 present similar *E*~*m*~ values. A 10 mV difference was measured between both forms for GrxC1 and C3, the more electronegative redox potential being the active site disulfide in GrxC1 and the glutathione adduct in GrxC3. However, the differences are small and should not explain the observed differences in catalytic efficiencies. In reducing cellular environments, all these oxidation forms should be compatible with glutathione reduction. Considering their possible targeting to secretory pathways, this may not be the case for Grx3 and C4 if they are in the endoplasmic reticulum or in the apoplasm. ![**Determination of redox potential of oxidized poplar GrxC1, C2, C3, and C4**. Representative curve of oxidation-reduction titration of Cys~A~-Cys~B~ intramolecular disulfide forms (black circles) and glutathionylated Cys~A~ (white circles) forms of poplar GrxC1 **(A)**, GrxC2 **(B)**, GrxC3 **(C)**, and GrxC4 **(D)** at pH 7.0. The titration was carried out using a total DTT or GSH concentration of 2 mM for 2 h and resulting free thiol groups were labeled by mBBr. Values in the text are the means ± SD of three replicates.](fpls-04-00518-g0006){#F6} Discussion {#s4} ========== Duplication and neo-functionalization into class I glutaredoxins ---------------------------------------------------------------- Among all living organisms, higher plants have the highest number of genes coding for Grxs. In addition to specific isoforms such as those belonging to the class III, land plants also possess an expanded class I, usually with 5--6 isoforms (Couturier et al., [@B10]). From previous phylogenetic analyses, it was quite clear that, there are three Grx couples, GrxC1/C2, GrxC3/C4, and GrxC5/S12 that probably evolved by more or less recent duplication (Couturier et al., [@B10]). Interestingly, GrxC1 and C2 proteins are cytoplasmic proteins (Riondet et al., [@B28]), GrxC3 and C4 have N-terminal extensions predicted to target the proteins to secretory pathways but the final destination is not known, and GrxC5 and S12 are chloroplastic (Couturier et al., [@B11], [@B13]). The case of GrxC1 and C2 is interesting. Whereas GrxC2 prototypes are found in all land plants (mosses, lycophytes, monocots, dicots), GrxC1 is specifically found in dicots. A redundancy between both isoforms is suggested in *A. thaliana* by the observation that single knock-out mutants in *grxc1* or *grxc2* are aphenotypic, whereas a double mutant is lethal at an early stage after pollinization (Riondet et al., [@B28]). Considering that GrxC1, but not GrxC2, is able to bind an Fe-S cluster (at least *in vitro*), this would rather suggest that the essential nature of these two paralogs is linked to their reductase activity. These results point to the need of investigating at the genetic level a possible redundancy among the GrxC3/C4 and GrxC5/12 couples. This study primarily focused on the biochemical properties of dithiol class I Grxs since a detailed structure-function analysis was previously achieved on the two other prototypes of class I Grxs, the poplar monothiol GrxS12 and the Arabidopsis dithiol GrxC5 (Couturier et al., [@B11], [@B13]). GrxC5 isoforms are only found in species of the Brassicaceae family. We have demonstrated in particular that, although being very similar in sequence due to their recent duplication, a single substitution in the active site sequence (from WCSYS in GrxS12 to WCSYC in GrxC5) explains the capacity of GrxC5, but not GrxS12, to bind an \[Fe~2~-S~2~\] cluster into a dimer. Among other class I Grxs, only GrxC1 can also exist under a holoform binding an Fe-S cluster (Rouhier et al., [@B36]). In this case, the mutational analysis showed that the presence of the glycine in the CGYC active site instead of a proline in other Grxs (CPFC in GrxC2 or CPYC in GrxC3 and C4) was sufficient to explain this property. Overall, this indicates that subtle sequence differences affecting key residues may provide different properties to otherwise quite close proteins. This prompted us to further analyze the various glutaredoxins of this class as GrxC2 and GrxC3 have been only poorly studied so far. Two subgroups in dithiol Grxs exhibit distinct kinetic properties ----------------------------------------------------------------- Considering the divergence observed in the amino acid sequences, the first aim of this work was to perform a thorough comparative analysis of the biochemical properties of GrxC1, GrxC2, GrxC3, and GrxC4 from poplar. All these proteins exhibit deglutathionylation or reductase activities using classical HED and DHA assays with catalytic efficiencies (10^5^ M^−1^ s^−1^ for HED and 10^3^--10^4^ M^−1^ s^−1^ for DHA) in the same range as those previously reported for plant and non-plant class I Grxs (Table [3](#T3){ref-type="table"}). Two subgroups can however be distinguished, GrxC1 and GrxC2 appeared to be less efficient catalysts than GrxC3 and GrxC4. The difference is particularly clear when considering the turnover numbers (k~cat~) and the k~cat~/*K*~DHA~ and k~cat~/*K*~GSH~. This is not visible with k~cat~/*K*~β\ −*ME*−*SG*~ because the differences in the turnover number are compensated in this case by differences in apparent affinity in favor of GrxC1 and C2. An observation that may be interesting to understand these differences is that GrxC1/C2 and GrxC3/C4 should have quite different charge distribution at the protein surface that could account for substrate recognition and thus for the observed differences in catalytic efficiency. Indeed, their behavior on ion exchange chromatography indicates that GrxC1 and GrxC2 should be more basic having a pI close or higher than 8, whereas GrxC3 and GrxC4 should be more acidic having a pI value below 8 (theoretical pI estimations are indeed comprised between 5 and 6). ###### **Kinetic and biochemical parameters of selected plant and non-plant class I Grxs**. **Name** **Active site sequence** **HED** **DHA** **GSH** **p*K*~*a*~** ***E*~*m*~ (pH7.0) (mV)** **References** ---------- -------------------------- ------------- -------------- ------------- --------------- --------------------------- ---------------- ----------- -------------------- -------------- ------------------ ------ ----------------------------------------------------------------------- AtGrxC1 YCGYC 0.74 ± 0.01 38.91 ± 0.43 5.3 × 10^4^ 0.08 ± 0.01 3.03 ± 0.82 3.8 × 10^4^ nd nd nd nd nd Riondet et al., [@B28] AtGrxC2 YCPYC 0.34 ± 0.02 18.44 ± 0.57 5.4 × 10^4^ 0.08 ± 0.01 3.26 ± 0.03 4.1 × 10^4^ nd nd nd nd nd Riondet et al., [@B28] ScGrx1 YCPYC 0.14 3.5 2.5 × 10^4^ nd Nd nd 6.2 17.1 (HED) 2.75 × 10^3^ 3.2 to 4.0 ± 0.2 nd Discola et al., [@B15] ScGrx2 YCPYC 0.6 85 1.4 × 10^5^ nd nd nd 0.9 129 (HED) 1.43 × 10^5^ 3.1 to 3.5 ± 0.2 nd Discola et al., [@B15] CrGrx1 HCPYC 0.34 ± 0.04 30.40 ± 4.31 8.9 × 10^4^ 0.39 ± 0.03 1.67 ± 0.09 4.3 × 10^3^ 2.6 ± 0.5 161.6 ± 15.0 (HED) 6.1 × 10^4^ 3.9 ± 0.1 nd Zaffagnini et al., [@B48] CrGrx2 YCPYC 0.20 ± 0.04 7.10 ± 0.27 3.5 × 10^4^ 0.17 ± 0.05 0.84 ± 0.06 4.9 × 10^3^ 3.7 ± 0.7 26.5 ± 4.1 (HED) 7.1 × 10^3^ 4.8 ± 0.1 nd Gao et al., [@B19] HsGrx1 TCPYC 1.07 8.16 7.6 × 10^3^ nd Nd nd 2.2 293 (HED) 1.4 × 10^5^ 3.6 −232 Johansson et al., [@B23]; Jao et al., [@B22]; Sagemark et al., [@B38] HsGrx2 SCSYC 0.11 1.3 1.2 × 10^4^ nd Nd nd 5.9 71.3 (HED) 1.2 × 10^4^ 4.6 −221 Jao et al., [@B22]; Sagemark et al., [@B38]; Gallogly et al., [@B17] nd, not determined. In order to further explain the differences between poplar class I Grxs, we determined the p*K*~*a*~ value of their catalytic cysteine, as it was proposed that reactivity of human and Chlamydomonas class I Grxs is partially related to the p*K*~*a*~ of N-terminal active site cysteine (Gallogly et al., [@B17]; Gao et al., [@B19]). The values obtained for poplar Grxs are globally close to those reported for other plant and non-plant class I Grxs (Table [3](#T3){ref-type="table"}), including the value of 3.9 determined for poplar GrxS12 using the same procedure (Zaffagnini et al., [@B46]; Roos et al., [@B30]). Interestingly, a difference of ca 0.5 unit is visible between GrxC1/C2 (p*K*~*a*~ around 5) and GrxC3/C4 (p*K*~*a*~ around 4.6). Such differences between paralogs have been described already since *C. reinhardtii* Grx1, *S. cerevisiae* Grx2, and human Grx1 exhibit higher turnover numbers often translated into higher catalytic efficiencies compared to *C. reinhardtii* Grx2, *S. cerevisiae* Grx1, and human Grx2, respectively (Table [3](#T3){ref-type="table"}). For the human and *C. reinhardtii* enzymes, it is also associated to a difference in p*K*~*a*~ values of *ca* 1 unit, the enzymes with the lowest p*K*~*a*~ being the most efficient. According to the Brønsted theory, in a thiol-disulfide exchange reaction, each unit decrease in the p*K*~*a*~ value of the leaving group is responsible for a 4 fold increase in the second-order rate constant (Szajewski and Whitesides, [@B40]). This could fit with the conclusion that differences in p*K*~*a*~ values between class I poplar Grx isoforms are at least partially responsible for the differences in catalytic efficiencies observed. On the other hand, only small differences were observed for the oxidation-reduction midpoint potentials of both the glutathione adduct or the intramolecular disulfide and they are not discriminating GrxC1/C2 vs. GrxC3/C4. The measured values (from −230 to −260 mV) are in the same range than those reported for *E. coli* Grx1 (−233 mV) and human Grx1 and 2 (−232 and −221 mV) (Table [3](#T3){ref-type="table"}) (Aslund et al., [@B2]; Sagemark et al., [@B38]). The role of the recycling Cys~B~: monothiol or dithiol reaction mechanism? -------------------------------------------------------------------------- Interestingly, the differences in protein activities between GrxC1 or C2 and GrxC3 or C4 are attenuated when Cys~B~ of GrxC1 or C2 is removed (see Figure [2](#F2){ref-type="fig"}). This has been reported also for Arabidopsis GrxC5, yeast Grx1, human Grx1 and Grx2, and pig Grx (Yang and Wells, [@B45]; Yang et al., [@B44]; Johansson et al., [@B23]; Gallogly et al., [@B17]; Discola et al., [@B15]; Couturier et al., [@B13]). In clear contrast, mutating this cysteine in yeast Grx2, *E. coli* Grx1 and Grx3 led to a decrease of Grx activity (Bushweller et al., [@B6]; Nordstrand et al., [@B26]; Peltoniemi et al., [@B27]; Discola et al., [@B15]; Saaranen et al., [@B37]). For *E. coli* Grx1 and by extension for Grxs for which mutation of this cysteine decreased their activity, this decrease was attributed to a change in glutathione recognition, binding, or affinity, because Cys~B~ would determine the glutathione specificity of the glutathionylated Grx reduction step (Saaranen et al., [@B37]). For GrxC1 and GrxC2 and other eukaryote Grxs mentioned above in which the mutation of this cysteine increased their activity, the data indicates that the presence of Cys~B~ slows down the reaction for some reason. It may eventually modify the p*K*~*a*~ of Cys~A~. This has been experimentally confirmed for a pig Grx, for which the p*K*~*a*~ of the catalytic Cys22 is 3.8 in a WT protein but 4.9 and 5.9 in C25S and C25A variants (Yang and Wells, [@B45]). Another possibility is that a Cys~A~-Cys~B~ intramolecular disulfide is formed during activity measurements either because it constitutes a regular step of the reaction cycle in some situations, especially for dithiol substrates or because it is a side reaction reflecting the competition between Cys~B~ and GSH for the reduction of the glutathionylated Cys~A~ (Gallogly et al., [@B18]). Then, the reduction of this Cys~A~-Cys~B~ intramolecular disulfide intermediate would require two molecules of GSH vs. one molecule of GSH for the glutathionylated form and add two steps in the reaction mechanism. Thus, it seems unfavorable in terms of efficiency and energetic cost. We have previously analyzed possible variations in the redox state of the dithiol AtGrxC5 (WCSYC active site) in response to 15 min treatments with GSSG or oxidized DTT by measuring the tryptophan intrinsic fluorescence of the protein, since the oxidation of the catalytic cysteine quenched the fluorescence of the adjacent tryptophan (Couturier et al., [@B13]). It was concluded that the dithiol AtGrxC5 cannot form a Cys~A~-Cys~B~ intramolecular disulfide first because only the GSSG treatment led to the formation of a glutathionylated protein and to fluorescence quenching and second because the WT protein was crystallized with a glutathionylated adduct. However, at that time, we did use neither H~2~O~2~ and GSNO treatments, nor prolonged incubation, nor alkylation assays. In fact, reconsidering the results of mass spectrometry showing, in addition to the peaks with one or two glutathione adducts, a peak with a decrease of 2 Da between the reduced protein and the GSSG-treated protein, it is very likely that a small fraction of the protein contained an intramolecular disulfide formed between active site cysteines (Couturier et al., [@B13]). Here, we have clearly observed that a prolonged oxidative treatment of reduced class I Grxs with H~2~O~2~, GSSG, or GSNO led preferentially to the formation of an intramolecular disulfide bond between Cys~A~ and Cys~B~. For the H~2~O~2~ treatment, it likely goes first through the formation of a sulfenic acid on Cys~A~ before Cys~B~ performs a nucleophilic attack on it. For the GSSG or GSNO treatments, it is very likely that a nucleophilic attack of Cys~A~ onto these oxidized glutathione species leads to the formation of a glutathione adduct on Cys~A~ which is then slowly reduced by a Cys~B~ nucleophilic attack leading to the formation of the intramolecular disulfide and to the concomitant release of a GSH molecule. It is worth mentioning here that these *in vitro* experiments have been done in particular with the aim of exploring all possible post-translational modifications but the conditions are not exactly physiological first because the concentrations of the oxidants may not be realistic and more importantly because this is done in the absence of GSH, which is normally found at high amounts and is highly reduced in most cell compartments. The presence of GSH would certainly prevent to a large extent the formation of these intramolecular disulfides because it may preferentially reduce either the sulfenic acids or the glutathione adducts over the Cys~B~ recycling cysteine. This is certainly the case for GrxC3 and GrxC4 since mutating CysB has no effect on the activity. The formation of an intramolecular disulfide in these proteins could thus constitute uniquely a side reaction, protecting the cysteine or inactivating the proteins only under specific conditions. On the contrary, if we consider that the increase of protein activity observed for CysB mutated GrxC1 and GrxC2 in HED and DHA assays in the presence of a fully reduced GSH pool reflects the formation of this intramolecular disulfide, it may be physiologically relevant for GrxC1 and GrxC2. This is interesting in several respects. First, while the current dogma is that the preferential substrates of Grxs are glutathionylated proteins, recent studies performed with mammalian Grx2 indicated that they can reduce dithiol substrates, e.g., proteins with intra- or intermolecular disulfide (Hanschmann et al., [@B20]; Schutte et al., [@B39]). In this case, the Grxs should employ a reaction mechanism similar to the one used by Trxs involving both Cys~A~ and Cys~B~. Second, based on the observation that some Grxs can be reduced by thioredoxin reductases (Johansson et al., [@B23]; Zaffagnini et al., [@B48]; Couturier et al., [@B12]), another possibility could be that in specific physiological situations or sub-cellular compartments where glutathione is depleted, absent or oxidized, the formation of an intramolecular disulfide may be determinant either to favor the use of an alternative reducing system as the one constituted by NADPH and thioredoxin reductase or to simply constitute a protective mechanism, preventing the irreversible overoxidation of this residue and thus protein inactivation. In conclusion, the role of Cys~B~ in dithiol Grxs is yet uncertain and may depend on the Grx isoform as this residue influences differentially the protein activity. The fact that the formation of an intramolecular disulfide bond between Cys~A~ and Cys~B~ is possible in all cases lets open the possibility to use both monothiol and dithiol catalytic mechanisms, contrary to Grxs possessing only one cysteine residue. The role of the semi-conserved Cys~*C*~ in GrxC1 and GrxC2 isoforms ------------------------------------------------------------------- The Cys~*C*~ cysteine residue, present in the IGGCD motif (at position 88 and 80 in GrxC1 and C2, respectively), is found in several other class I Grxs, including plant GrxC5 and S12 as well as in human Grx1 and *E. coli* Grx3. Moreover, it is also present in many class II Grxs. It was shown for instance that a Cys~A~-Cys~*C*~ intramolecular disulfide bond can be formed in *S. cerevisiae* Grx5 and *C. reinhardtii* Grx3 (Tamarit et al., [@B41]; Zaffagnini et al., [@B48]), suggesting that Cys~*C*~ can serve as a recycling cysteine in the absence of Cys~B~. Reduced glutathione does not seem to be an efficient reductant for this intramolecular disulfide bond and in the case of CrGrx3, it is reduced by a ferredoxin-thioredoxin reductase system (Zaffagnini et al., [@B48]). Although GrxS12 has a monothiol WCSYS active site sequence, a Cys~A~-Cys~*C*~ intramolecular disulfide has not been observed whatever the oxidizing conditions tested. Moreover, in the crystal structure of a glutathionylated GrxS12, the sulfur atom of Cys87 is 9.6Å away from the Cys27-SG adduct (Couturier et al., [@B11]). Hence, this cysteine may not serve as a recycling cysteine in class I Grxs and does not *a priori* participate to the reaction mechanism. Consistently, its mutation in GrxC1 and C2 did not influence protein activity, as previously observed for poplar GrxS12, Arabidopsis GrxC5, or *E. coli* Grx3 (Nordstrand et al., [@B26]; Couturier et al., [@B11], [@B13]). However, considering its position very close to the active site residues and to residues involved in GSH binding an effect on activity could have been observed. Indeed, it is positioned between the so-called GG kink which is important in determining the backbone geometry of the following amino acids, and a Thr or an Asp which forms hydrogen bonds with GSH (see Figure [1](#F1){ref-type="fig"}) (Couturier et al., [@B11]; Li et al., [@B24]). In the GrxS12 structure, the backbone amino group of Cys~*C*~ also directly forms a hydrogen bond with the glutamyl group of GSH (Couturier et al., [@B11]), but the substitution of the cysteine into serine would not disrupt this interaction, preventing to definitely rule out a GSH stabilizing effect of this cysteine. On the other hand, it is probable that modifying the redox state of this cysteine might have more pronounced effects influencing for example GSH binding. It has been shown for example that, under non-reducing conditions, *E. coli* Grx3 exists under an expected monomeric form but also forms disulfide-bridged dimer or multimers via Cys~*C*~ (Aslund et al., [@B3], [@B2]). In addition, previously reported oxidative treatments of human Grx1, which contains three extra active site cysteines including Cys~*C*~, led to the identification of several possible post-translational modifications, i.e., intramolecular disulfide, disulfide-bonded dimers and oligomers, glutathione adducts, nitrosylation, some of them inhibiting its activity (Hashemy et al., [@B21]). Depending on the conditions, Cys~*C*~ was found to be either nitrosylated or glutathionylated or involved into a disulfide. Similarly, we have observed redox changes of GrxC1 and C2 upon treatment with oxidants that could further provide clues about the possible function of this cysteine, if any. In the presence of H~2~O~2~, this cysteine was found to be either overoxidized or involved in disulfide-bridged homodimers. Since this dimerization only occurs in presence of H~2~O~2~, the formation of an intermediate sulfenic acid is likely required before intermolecular disulfide bond formation which is consistent with the presence of a small part of overoxidized proteins. Interestingly, GSH can efficiently reduce these disulfide-bridged dimers, making this reaction physiologically reversible. In the presence of GSSG or GSNO, Cys~*C*~ is also prone to oxidative modification, but in the form of a glutathione adduct. Although we cannot test the impact of these modifications on Grx activity (we cannot selectively modify Cys~*C*~ without also oxidizing Cys~A~and Cys~B~), we can speculate that the different reversible oxidation forms of Cys~*C*~ observed here might represent regulatory or signaling intermediates transiently modifying the functions of these Grxs by affecting for example the nature or the time-course reduction of the target proteins. Conflict of interest statement ------------------------------ The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. This work was supported by the French National Research Agency through the Laboratory of Excellence ARBRE (ANR-12-LABXARBRE-01). The authors would like to thank Bruno Dastillung and Arnaud Hecker for their help with some experiments. Supplementary material {#s5} ====================== The Supplementary Material for this article can be found online at: <http://www.frontiersin.org/journal/10.3389/fpls.2013.00518/abstract> ###### **Electrospray ionization mass spectrometry analysis of PtGrxC1 WT**. Spectra of whole protein were determined for reduced protein before **(A)** and after treatment with GSSG **(B)**, GSNO **(C)**, or H~2~O~2~ **(D)** as described in the Methods section. ###### Click here for additional data file. ###### **Electrospray ionization mass spectrometry analysis of PtGrxC2 WT**. Spectra of whole protein were determined for reduced protein before **(A)** and after treatment with GSSG **(B)**, GSNO **(C)**, or H~2~O~2~ **(D)** as described in the Methods section. ###### Click here for additional data file. ###### **Electrospray ionization mass spectrometry analysis of PtGrxC3 WT**. Spectra of whole protein were determined for reduced protein before **(A)** and after treatment with GSSG **(B)**, GSNO **(C)**, or H~2~O~2~ **(D)** as described in the Methods section. ###### Click here for additional data file. ###### **Electrospray ionization mass spectrometry analysis of PtGrxC4 WT**. Spectra of whole protein were determined for reduced protein before **(A)** and after treatment with GSSG **(B)**, GSNO **(C)**, or H~2~O~2~ **(D)** as described in the Methods section. ###### Click here for additional data file. ###### **Electrospray ionization mass spectrometry analysis of PtGrxC1 C88S**. Spectra of whole protein were determined for reduced protein before **(A)** and after treatment with GSSG **(B)**, GSNO **(C)**, or H~2~O~2~ **(D)** as described in the Methods section. ###### Click here for additional data file. ###### **Electrospray ionization mass spectrometry analysis of PtGrxC2 C80S**. Spectra of whole protein were determined for reduced protein before **(A)** and after treatment with GSSG **(B)**, GSNO **(C)**, or H~2~O~2~ **(D)** as described in the Methods section. ###### Click here for additional data file. ###### **Primers used for site-directed mutagenesis**. ###### Click here for additional data file. [^1]: Edited by: Andreas Meyer, University of Bonn, Germany [^2]: Reviewed by: Mirko Zaffagnini, University of Bologna, Italy; Christophe Riondet, UMR 5096, France [^3]: This article was submitted to Plant Physiology, a section of the journal Frontiers in Plant Science.

{ "pile_set_name": "PubMed Central" }

Introduction {#sec1-1} ============ Diffusion tensor imaging (DTI) is a MRI technique that enables the measurement of the diffusion of water in tissue in order to produce neural tract images (**[Fig. 1](#F1){ref-type="fig"}**). The idea of using diffusion data to produce images of neural tracts was first proposed by Aaron Filler & colleagues in March of 1991. Several months later (1992) the first DTI image showing neural tracts curving through the brain was produced. Recent development of magnetic resonance imaging (MRI) has made it possible to delineate detailed anatomical and functional information of both normal and pathological brain tissues in a noninvasive manner. Among them, diffusion tensor MR imaging (DTI), has attracted great interest in that it can provide neural ﬁber information by use of the ﬁber--tracking technique and can also detect the presence of tumor cell inﬁltration into the normal white matter around the tumor boarder. The target in the treatment of malignant brain tumors is to maximize tumor resection volume, information on neural ﬁber preservation and tumor cell inﬁltration is critical for preoperative neurosurgical planning \[**[@R1]**\]. DTI data can be used to visualize the major white matter (WM) tracts of the brain \[**[@R2]--[@R4]**\]. DTI is an MR technique that can indirectly evaluate the integrity of WM by measuring water diffusion and its directionality in three dimensions \[**[@R5]**\]. DTI has been applied to differentiate edema from tumor, in patients with brain tumor for tumor characterization and to assess structural properties of the adjacent tracts \[**[@R6]--[@R11]**\]. Knowledge of the anatomical relationship between tumor and WM tracts could improve preoperative risk analysis and decrease the risk of WM tract injury during surgery. Multiple studies investigated the relationships between fractional anisotropy (FA) values, mean diffusivity (MD) values and the histological tumor cell inﬁltration within the region of interest where the FA and MD were measured \[**[@R12]--[@R14]**\]. These studies have demonstrated that FA values tend to decrease in locations where the neural ﬁbers have been inﬁltrated by tumor cells, suggesting that relatively low FA values indicate tumor cell invasion into white matter, where neural ﬁbers reside. The relationships between tumor cell density and FA or MD are still controversial, as both positive and negative correlation between these parameters has been reported \[**[@R13],[@R15],[@R16]**\]. Thus, the development of preoperative histological evaluation of tumors by noninvasive imaging techniques would greatly assist the preoperative diagnosis of the lesion and also preoperative surgical planning of the operation. For example, when a tumor biopsy of a glioma is required, it is extremely important that the target for biopsy is the location with the highest cell density, to allow accurate histological diagnosis. ![1 The first images of tractography from our research project. Fibers in red color are left--right. Fibers in green color are anterior--posterior. Fibers in blue color are superior--inferior](JMedLife-04-372-g001){#F1} Methods {#sec1-2} ======= The principal application of DTI is in the imaging of white matter where the location, orientation, and anisotropy of the tracts can be measured. DTI allows clinicians to look at anisotropic diffusion in white-matter tracts, but it is limited in demonstrating spatial and directional anisotropy. Advanced methods such as color coding and tractography (fiber tracking) have been used to investigate the directionality. If diffusion gradients (i.e. magnetic field variations in the MRI magnet) are applied in at least 3 directions, 6 directions improve the accuracy that describes the 3-dimensional shape of diffusion. The fiber direction is indicated by the tensor's main eigenvector which can be color--coded, yielding a cartography of the tracts\' position and direction: red for left--right, blue for superior--inferior, green for anterior--posterior. Tractography potentially solves a problem for a neurosurgeon in terms of minimizing functional damage and determining the extent of diffuse infiltration of pathologic tissue to minimize residual tumor volume. In this way, tractography facilitates preoperative planning. Tractographic images may help to clarify whether a tumor is compressing, abutting, or infiltrating the contiguous white--matter tracts. DTI identify different tumor components, and to differentiate tumor invasion from normal brain tissue or edema. DTI has demonstrated a potential in distinguishing gliomas and solitary metastasis in the brain parenchyma. Significantly higher MD, compared with levels in normal--appearing white matter, has been demonstrated in the peritumoral regions of both gliomas and metastases. Peritumoral MD of metastases and meningioma is significantly higher than that of gliomas. The FA and MD measurement in region of interest can be selected anatomically by referring to the MR images of preoperative surgical planning and postoperative results in order to match the location from which the histological specimen can be obtained. FA and MD values were can be calculated using a computer with the image analyzing software like MedINRIA. After surgical resection, specimens can be ﬁxed in formalin and embedded with parafﬁn; hematoxylin and eosin-stained specimens were checked to determine the histological tumor type. Cell counting can be performed under a light microscope at standard magnification like ×200. A standard area for the tumor cell density can be established and can be used a program like ImageJ for counting. Also, Ki--67 labeled cells can be counted and the percentage of Ki--67 labeled cells can be calculated within the observed ﬁeld. \[**[@R17]**\] The localization of [tumors](http://en.wikipedia.org/wiki/Tumor) in relation to the white matter tracts (infiltration, deflection), has been one the most important initial applications. In surgical planning for some types of brain tumors, surgery is aided by knowing the proximity and relative position of tumor. Discussions and Results {#sec1-3} ======================= Tractography combined with functional MRI may potentially help in preoperative planning of brain tumors by mapping areas of active infiltration. The recent development of DTI allows for direct examination of the brain microstructure. In the surgery of patients with brain tumors, preservation of vital cerebral function is as important as maximizing tumor resection. The associated morbidity of aggressive resections can be significantly reduced by carefully preservation of vital cerebral function, and the quality of life of these patients will be largely improved. In general, cerebral tumor may alter the adjacent WM in three different ways: by (1) displacing the WM tracts but with relative preservation of the fibers, (2) infiltrating the WM tracts, and (3) disrupting of the WM tracts. Intracranial tumors may involve both functional cortical gray and white matter tracts. Preoperative characterization of tumor tissue using noninvasive imaging modalities is one of many necessary steps for establishing a successful treatment strategy against malignant brain tumors. It is widely accepted that tumor cell density correlates with tumor malignancy. In cases where a tumor biopsy is required, the target is preferentially chosen at the lesion with the highest tumor cell density, to achieve accurate diagnosis and grading of the disease. In cases of contrast agent--enhancing tumors, usually the target is set at the location with distinct enhancement on MRI, based on the known fact that the blood--brain barrier is disrupted within the lesion with higher malignancy, enabling contrast agents to enhance the area \[**[@R18]**\]. Perfusion weighted MR imaging (PWI), chemical shift imaging (CSI) and positron emission tomography (PET) have been proposed for retrieving information on biological characteristics of the tumor in a noninvasive fashion \[**[@R19],[@R20]**\]. The FA and MD can be calculated and the results indicated an FA signiﬁcantly lower in the low-grade gliomas than the malignant lymphomas. \[**[@R17]**\] This observation is similar to the conclusion of Toh et al. (2008) \[**[@R21]**\]. The average indicated MD signiﬁcantly higher in low--grade gliomas and also in high-grade gliomas compared to malignant lymphomas. These ﬁndings suggest that MD is more sensitive than FA in distinguishing gliomas from malignant lymphomas. Tumor density of the tumor core shows positive correlation with FA but negative correlation with MD. These ﬁndings are compatible with those previously reported by Beppu et al. (2003, 2005) \[**[@R15],[@R16]**\] but contradict those of Stadlbauer et al. (2006) \[**[@R13]**\]. Although a direct comparison of FA and MD was not performed in the study by Stadlbauer et al., they showed that tumor cell density has negative correlation with FA and positive correlation with MD \[**[@R13]**\]. Thus it seems fair to speculate that the data in the study of Stadlbauer et al. also shows an inverse relationship between FA and MD values \[**[@R13]**\]. In addition, an inverse relationship between FA and MD has also been reported for multiple sclerosis \[**[@R22]**\]. One possible explanation for FA and MD show inverse correlation is the presence of tumor cells within the tissue prevents water diffusing in a single direction, which would show both high FA and MD values. Previous studies have focused the use of diffusion tensor MR imaging (DTI) for ﬁber--tracking \[**[@R23]--[@R30]**\] and evaluating tumor inﬁltration near the tumor boarder \[**[@R12]--[@R14]**\]. These assessments are crucial for achieving maximum tumor resection of malignant brain tumors. Ki--67 labeling index is positively correlated with FA and negatively correlated with MD. Tumor cell density at the core of the tumor and Ki--67 labeling index show good positive correlation, suggesting that tumor cell density and Ki--67 labeling index have diagnostic value in malignant brain tumors. \[**[@R17]**\]. As FA and MD (in combination) seem to represent the histology of these tumors, in particular of tumor cell density. It is expected that locations with higher FA and lower MD will have higher tumor cell density, as well as have a relatively high Ki--67 labeling index, indicating higher malignant potential. It should be noted that higher FA values could also implicate the contamination of high FA resulting from intact neural ﬁbers. \[**[@R17]**\] Conclusions {#sec1-4} =========== As a conclusion, data from multiple studies of DTI suggest to represent tumor tissue characteristics in a noninvasive fashion. When the enhancement by contrast agent is vague, other indicators are necessary to assess tumor cell density within the tumor. The use of diffusion tensor MR imaging (DTI), however, for this purpose still requires investigation. **Acknowledgments** This investigation was supported by a Grant for Scientiﬁc Research from the Ministry of Education and Science of Romania (subject numbers: 42--149/2008).

{ "pile_set_name": "PubMed Central" }

Background {#Sec1} ========== Enterohaemorrhagic *Escherichia coli* O157:H7 (EHEC O157) is an important intestinal pathogenic bacterium that can causes diarrhea, hemorrhagic colitis, and in 10% of cases of systemic hemolytic uremic syndrome. EHEC O157 is the most extensively studied EHEC and is responsible for regular outbreaks of foodborne illness worldwide \[[@CR1]\]. EHEC O157 colonization involves the formation of attaching and effacing (A/E) lesions on the intestinal epithelium, which are characterized by loss of microvilli and allow intimate attachment of the bacterium to the host cell membrane \[[@CR2]\]. A/E lesion formation genes are localized on a pathogenicity island, known as the locus for enterocyte effacement (LEE), which encodes a bacterial type III secretion system (T3SS) and is capable of injecting bacterial effector proteins into the host cell cytoplasm \[[@CR3]\]. The LEE contains 41 genes that are organized in five major operons (LEE1, LEE2, LEE3, LEE5, and LEE4) \[[@CR3]\]. LEE1, LEE2, and LEE3 encode the major structural components of the T3SS, LEE4 encodes several secreted proteins \[[@CR4]--[@CR6]\], and LEE5 encodes Tir and Intimin \[[@CR7]\]. ORF1 on the LEE1 operon encodes the master regulator Ler (LEE encoded regulator). Ler is capable of activating LEE2 to LEE5 \[[@CR8], [@CR9]\]. Transcriptional regulation of the LEE is extremely complex. The regulatory system of the LEE involves at least three kinds of regulators: LEE-encoded regulators (including Ler \[[@CR9]\], GrlA and GrlR \[[@CR10]\]), global regulators (such as H-NS, IHF and Fis \[[@CR11]\]), and horizontally transferred regulators (such as EivF, EtrA, and GrvA). Although the complexity of LEE regulation in EHEC O157 has been acknowledged, the mechanism by which LEE regulation occurs in not fully understood. In EHEC O157, biofilm formation is regulated by a complex network of regulatory cascades. The biofilm master regulator CsgD (curli specific gene D) is a key transcriptional response regulator controlling the formation of curli fimbriae and cellulose production \[[@CR12]\]. Biofilm formation can also be regulated by cyclic diguanylate (c-di-GMP) concentration \[[@CR13]\]. c-di-GMP is a common second messenger in bacteria, which is synthesized by diguanylate cyclases (DGCs) and degraded by c-di-GMP-specific phosphodiesterases (PDEs) \[[@CR14]\]. Among these genes, *yeaI*, which is a DGC-encoding gene, increases the c-di-GMP concentration in *E. coli* BW25113 \[[@CR15]\] and promotes biofilm formation in uropathogenic *E.coli* CFT073 \[[@CR16]\]. Colonization of the mammalian gastrointestinal tract brings bacteria into contact with a strong acid barrier in the stomach and organic acids in the intestine \[[@CR17]\]. To reach their site of colonization, EHEC must traverse the acidic environment of the stomach. Although the environment within the large intestine is less acidic, EHEC must survive volatile organic acids produced via anaerobic fermentation by the local microbiota \[[@CR18]\]. Several distinct acid resistance (AR) pathways have been identified in *E. coli*, and are present in EHEC \[[@CR19]\]. Acid resistance and/or induction of acid tolerance may better enable pathogens to survive gastrointestinal acidity and ultimately cause disease, and may thus enhance virulence \[[@CR20]--[@CR22]\]. The gene *asr* is important for adaptation to the acidic stomach, as *asr* mutants are unable to establish colonies in the stomach \[[@CR23]\]. The periplasmic chaperones HdeA and HdeB are also important for cell survival at low pH \[[@CR24]\] by protecting periplasmic proteins from aggregation at low pH, which is crucial considering the high permeability of the outer membrane \[[@CR25]\]. Transcription of the *hdeAB* operon is activated by RpoS and GadE, and repressed by H-NS and MarA \[[@CR26]\]. GadE, GadW, and GadX also play a critical role in the transcriptional regulation of the glutamate-dependent acid resistance (GDAR) system in *E. coli* K-12 MG1655 \[[@CR27]\]. Two-component signal transduction systems (TCSs) enable bacteria to sense environmental stimuli and transfer this information across the cytoplasmic membrane to the cytoplasm \[[@CR28]\]. A typical TCS consists of a sensor histidine kinase (HK) and its cognate DNA-binding response regulator (RR). The membrane HK typically has extracellular and cytoplasmic domains linked via a transmembrane domain. Upon ligand binding to the extracellular domain and subsequent conformational change, auto-phosphorylation of the conserved histidine residue in the cytoplasmic domain takes place. The phosphate is then transferred to the aspartic residue on the RR. Phosphorylation of the RR activates an output domain that can modulate gene expression \[[@CR29]\]. Most RRs are transcriptional factors, and once phosphorylated they bind to target promoters, activating or repressing transcription \[[@CR28]\]. Recently, a functional TCS, RstAB, has been implicated in the regulation of bacterial virulence in *Vibrio alginolyticus*, *Salmonella typhimurium*, *Photobacterium damselae*, *Clostridioides difficile*, and avian pathogenic *E. coli* \[[@CR30]--[@CR35]\]. The regulatory function of the TCS protein RstA on bacterial virulence in EHEC O157 remains unclear. Therefore, in the present study, we investigated the global effects of RstA on gene expression in EHEC O157. Genes whose expression was affected by RstA were identified and grouped into different clusters of orthologous group (COG) categories. We aimed to contribute to the understanding of the regulatory function of RstA in EHEC O157, especially with regard to virulence, which may impact future disease control and treatment effort against this important pathogen. Results {#Sec2} ======= Transcriptional data analysis {#Sec3} ----------------------------- To gain an understanding of RstA regulation at the global level, we systematically catalogued the transcriptomes of the EHEC O157:H7 strain EDL933 wild type strain (WT) and the Δ*rstA* mutant using high-throughput Illumina RNA-seq analysis. After filtering low quality reads, a total of 17,129,356 to 23,535,490 reads were obtained for the EHEC O157 WT and the Δ*rstA* mutant, respectively. Approximately 99.5% of the total reads for the EHEC O157 WT and 99.3% of those for the Δ*rstA* mutant were uniquely mapped to the reference genome (Additional file [1](#MOESM1){ref-type="media"}: Table S1). A total of 1237 genes were differentially expressed in the Δ*rstA* mutant compared to that in the EHEC O157 WT of these, 892 and 345 genes were categorized as up- and down-regulated, respectively (Additional file [2](#MOESM2){ref-type="media"}: Excel files S1 and S2). These results indicate that RstA acts as both an activator and repressor in EHEC O157. We selected 10 of the differentially regulated genes at random for validation by qRT-PCR using the same culture conditions. The qRT-PCR results correlated well with the RNA-seq data, indicating that the RNA-seq data were robust and valid (Fig. [1](#Fig1){ref-type="fig"}).Fig. 1Confirmation of RNA-seq results by qRT-PCR. RNA-seq results were validated by comparing EHEC O157 WT and Δ*rstA* mutant strains using qRT-PCR to measure the relative expression of 10 randomly selected genes that were differentially expressed according to RNA-seq. The trend of expression was similar according to both RNA-seq and qRT-PCR for these 10 genes, validating the RNA-seq data. Data are presented as mean ± SD; n = 3 Genes with differential regulation in the wild type and mutant strains were classified using the NCBI COG functional categories annotation system. The COG categories that were significantly enriched in the group of up-regulated genes were primarily involved in the cell wall, membrane, envelope biogenesis, translation, ribosomal structure and biogenesis, carbohydrate transport and metabolism, nucleotide transport and metabolism, energy production and conversion, lipid transport and metabolism, and amino acid transport and metabolism. The COG categories that were significantly enriched in the list of down-regulated genes included posttranslational modification, protein turnover, and chaperones (Fig. [2](#Fig2){ref-type="fig"}, Additional file [1](#MOESM1){ref-type="media"}: Fig. S1).Fig. 2Clusters of orthologous group (COG) analysis of RstA-regulated genes in EHEC O157. Bars represent the number of up-regulated (blue) or down-regulated (orange) genes in the Δ*rstA* mutant compared to those in the EHEC O157 WT strain. The significant enrichment of a given COG in the sets of up- or downregulated genes was determined using one-tailed Fisher's exact test with Benjamini--Hochberg false discovery rate correction RstA regulates the LEE pathogenicity island {#Sec4} ------------------------------------------- Based on RNA-seq results, the expression of majority of LEE genes (from the LEE1 to LEE5 operon) were significantly downregulated in the Δ*rstA* mutant relative to the EHEC O157 WT strain (Fig. [3](#Fig3){ref-type="fig"}a, Additional file [2](#MOESM2){ref-type="media"}: Excel file S3). To determine whether RstA is involved in the virulence of EHEC O157, qRT-PCR was performed to measure the expression of seven representative LEE genes, including *ler* (the master regulator of LEE genes), *escT* (LEE1), *escC* (LEE2), *escN* (LEE3), *eae* (intimin, LEE5), *tir* (intimin receptor, LEE 5), and *espB* (LEE 4) in EHEC O157 WT and the Δ*rstA* mutant strains. The transcript levels of these representative LEE genes were down-regulated in the Δ*rstA* mutant compared to those in the EHEC O157 WT (Fig. [3](#Fig3){ref-type="fig"}b). We then evaluated the adherence of the Δ*rstA* mutant to HeLa cells, and found that deletion of *rstA* significantly reduced bacterial adherence to HeLa cells compared with that of the EHEC O157 WT (Fig. [3](#Fig3){ref-type="fig"}c). Both EHEC O157 WT and the Δ*rstA* mutant exhibited a similar growth rate, indicating the difference in adherence capacity between these two strains was not due to different growth rates (Fig. [3](#Fig3){ref-type="fig"}d). We found consistent results using fluorescent actin staining (FAS), which suggested that the Δ*rstA* mutant formed fewer pedestals on HeLa cells than the EHEC O157 WT (Fig. [3](#Fig3){ref-type="fig"}f, g). Mouse colonization experiments were used to determine the adherence capacity of these bacterial strains in vivo. The amount of the Δ*rstA* mutant recovered from the colon of infected mice was significantly lower than that of the EHEC O157 WT strain at 6 h post-infection (Fig. [3](#Fig3){ref-type="fig"}e). These differences could be restored to wild-type levels when a complementary plasmid pTRC99a-RstA was introduced into the Δ*rstA* mutant. Collectively, these results suggest that RstA is a positive regulator of bacterial virulence in EHEC O157.Fig. 3Effect of *rstA* on EHEC O157 adherence and LEE genes expression. **a** A heat map representing differential regulation of virulence genes in the EHEC O157 WT and mutant strains. The z-score indicates whether the genes were upregulated (red) or downregulated (green). **b** qRT-PCR analysis of changes in LEE genes expression in EHEC O157 WT, Δ*rstA* mutant, and *rstA* complementary strain. **c** Adherence of EHEC O157 WT, the Δ*rstA* mutant, *rstA* complementary strain to HeLa cells. **d** Growth of EHEC O157 WT, the Δ*rstA* mutant, *rstA* complementary strain in LB medium. **e** Adherence capacity of EHEC O157 WT, Δ*rstA* mutant and *rstA* complementary strain in the distal colon of mice at 6 h. Statistical significance was assessed using the Mann--Whitney rank-sum test. **f** Detection of AE lesion formation by EHEC O157 WT, Δ*rstA* mutant, and *rstA* complementary strain by FAS in HeLa cells at 3 h. The HeLa cell actin cytoskeleton (green) and nuclei of bacterial and HeLa cells (red) are shown. AE lesions are indicated by arrowheads. **g** The number of pedestals/infected HeLa cell of EHEC O157 WT, Δ*rstA* mutant, and *rstA* complementary strain (n = 50 cells). In **b**--**d**, data represent mean ± SD (n = 3). \*P ≤ 0.05, \*\*P ≤ 0.01, \*\*\*P ≤ 0.001 (Student's t-test) RstA is involved in the regulation of EHEC O157 acid tolerance {#Sec5} -------------------------------------------------------------- The immediate challenge facing EHEC O157 in an infected human host is survival in the extreme acidic environment of the stomach. In the present study, we used RNA-seq to determine that genes encoding acid resistance proteins (HdeAB, Asr and GadEWX) and we found that these genes were down-regulated in the Δ*rstA* mutant compared to those in the EHEC O157 WT strain (Fig. [4](#Fig4){ref-type="fig"}a, Additional file [2](#MOESM2){ref-type="media"}: Excel file S4). We therefore compared the survival of EHEC O157 WT and the Δ*rstA* mutant when exposed to acidified LB broth (pH 3.0).The viable cells were recovered on LB agar plates after incubation in the acidified broth at 37 °C for 0--6 h, and the number of CFUs were determined as a ratio to the initial inoculum (CFU/ml at 0 h). The survival assay showed that the survival rate of EHEC O157 WT were much higher than the Δ*rstA* mutant after 2 h in acidified LB broth (Fig. [4](#Fig4){ref-type="fig"}b), indicating that RstA plays a significant role for EHEC O157 survival in low pH environment.Fig. 4Effect of *rstA* on EHEC O157 acid tolerance. **a** A heat map representing differential regulation of acid tolerance related genes in the WT and mutant strains. The z-score indicates whether the genes were upregulated (red) or downregulated (green). **b** Survival assay of EHEC O157 WT and the Δ*rstA* mutant strain in acid challenge after incubation in acidified broth (pH 3.0) for 0--6 h. Data represent mean ± SD (n = 3). \*P ≤ 0.05, \*\*P ≤ 0.01 (Student's t-test) RstA impacts biofilm formation {#Sec6} ------------------------------ In the present study, we used RNA-seq to compare the transcriptomes of the EHEC O157 WT and the Δ*rstA* mutant strains, and identified several upregulated diguanylate cyclase genes in the Δ*rstA* mutant strain. These genes participate in the formation of the ubiquitous second messenger, cyclic-di-GMP (c-di-GMP) (Fig. [5](#Fig5){ref-type="fig"}a, Additional file [2](#MOESM2){ref-type="media"}: Excel file S5), which promotes biofilm formation in many bacteria \[[@CR36]\].Fig. 5Effect of *rstA* on EHEC O157 biofilm formation. **a** A heat map representing differential regulation of diguanylate cyclase genes in the EHEC O157 WT and the Δ*rstA* mutant strain. The z-score indicates whether the genes were upregulated (red) or downregulated (green). **b** Biofilm formation in EHEC O157 WT, Δ*rstA* mutant, and *rstA* complementary by crystal violet staining. **c** Intracellular c-di-GMP concentration of EHEC O157 WT, the Δ*rstA* mutant and *rstA* complementary strain. In **b**, **c**, Data represent mean ± SD (n = 3). \*P ≤ 0.05, \*\*P ≤ 0.01 (Student's t-test) To evaluate the effect of RstA in biofilm formation in EHEC O157, a crystal violet staining assay was performed to quantify biofilm formation by EHEC O157 WT, the Δ*rstA* mutant, and the complementary strain. Biofilm formation was significantly increased in the Δ*rstA* mutant compared to that in the EHEC O157 WT and the complementary strain (Fig. [5](#Fig5){ref-type="fig"}b), indicating that RstA is a negative regulator of biofilm formation in EHEC O157. We then measured the concentration of intracellular c-di-GMP using HPLC, and found significantly increased concentrations of intracellular c-di-GMP in the Δ*rstA* mutant (Fig. [5](#Fig5){ref-type="fig"}c). Taken together, these findings support the conclusion that RstA inhibits biofilm formation in EHEC O157 by controlling the biosynthesis of c-di-GMP. RstA box analysis in EHEC O157 {#Sec7} ------------------------------ RstA has been shown to bind to the conserved motif TACATNTNGTTACA, which is termed the RstA box and is present in the promoter region of many RstA-activated or repressed genes in *E. coli* \[[@CR37]\]. Furthermore, the consensus TACA repeat sequence is necessary for RstA binding \[[@CR38]\]. Then, we searched for this RstA box-like sequence (TACANNNNNNTACA, N = 5--6) along the entire EHEC O157 EDL933 genome, and found 19 possible targets in the intergenic region. Among these targets, 14 are located in the promoter region and 8 were identified here for the first time (Additional file [1](#MOESM1){ref-type="media"}: Table S4). No RstA box was found in the promoter region of any of the LEE operons (P~LEE1~, P~LEE2/3~, P~LEE4~, P~LEE5~). Electrophoretic mobility shift assay (EMSA) results confirmed that RstA cannot directly bind to the promoter region of LEE operons (Fig. [6](#Fig6){ref-type="fig"}a--d). This suggests that RstA may activates LEE genes expression indirectly via an unknown regulator(s). With increasing concentrations of RstA protein, we observed slowly migrating bands for the promoter region of *asr*, *hdeA* and *yeaI*, but not *rpoS* (negative control) under the same conditions. This indicates that RstA enhances acid tolerance by directly regulating the expression of *hdeA* and *asr*, and represses biofilm formation by regulating the concentration of c-di-GMP via *yeaI* (Fig. [6](#Fig6){ref-type="fig"}e--h).Fig. 6EMSA of the binding of RstA. EMSA of the binding of RstA to P~LEE1~ (a), P~LEE2/3~ (**b**), P~LEE4~ (**c**), and P~LEE5~ (**d**), *rpoS* (**e**, negative control), P~*asr*~ (**f**), P~*hdeA*~ (**g**) and P~*yeaI*~ (**h**). PCR products were added to the reaction mixtures at 40 ng each. RstA protein was added to the reaction buffer in each assay in lanes 2--5 at 0.25, 0.5, 1, and 2 μM, respectively. No protein was added in lane 1. No binding was observed in **a**--**e**, while binding was observed in **f**--**h** Discussion {#Sec8} ========== RstA is a well-known TCS regulator that decreases bacterial adhesion and virulence in different bacterial species, including avian pathogenic *E. coli*, *Edwardsiella ictaluri*, *Photobacterium damselae* and *Clostridioides difficile*. In the present study, we investigated the effects of RstA on the global gene expression of EHEC O157 using RNA-seq. We validated the RNA-seq results using qRT-PCR to evaluate the changes in expression of 10 randomly selected genes in the WT and mutant strains. For all examined genes, the fold change detected by RNA-seq had the same trend as that observed by real-time PCR. These results are consistent with the results of previous experiments, validating our use of RNA-seq and verifying the results obtained here. In EHEC O157, 33 response regulators and 30 sensor kinases have been assumed to exist on the basis of genome sequence analysis results. Several response regulators have been reported to regulate the expression of virulence genes in EHEC O157. In the present study, we observed that inactivation of RstA results in significant downregulation of LEE genes expression in EHEC O157. Several other genes related to virulence were also regulated by RstA, including *stx1*, *stx2*, *nleA*, *nleB*, *nleB2*, *nleC*, and *nleL*, which implies that RstA is a global virulence regulator in EHEC O157 (Additional file [2](#MOESM2){ref-type="media"}: Excel file S6). According to the results of the adherence, FAS, and colonization assays, RstA contributes to EHEC O157 adherence in vitro and colonization in vivo. These results taken together indicate that RstA is a transcriptional activator of virulence in EHEC O157. Therefore, it was somewhat surprising that no RstA box was found in the promoter region of these genes. The EMSA results also confirm that RstA does not directly bind to the promoters of LEE1, LEE2/3, LEE4, and LEE5. This suggests that the observed positive regulation of LEE genes by RstA occurs via an indirect mechanism, possibly with an unknown intermediate regulator. To be able to establish colonization in a host, EHEC O157 must survive the acidic conditions in the stomach before it reaches the intestine \[[@CR39]\]. In the present study, RNA-seq comparison of mutant and WT strains revealed that several important acid tolerance genes (*hdeAB* operon, *asr* and *gadEWX*) were down-regulated in the Δ*rstA* mutant. We performed a survival assay and found that the Δ*rstA* mutant was more acid sensitive than the EHEC O157 WT. According to our EMSA results, *rstA* can directly bind to the promoter region of both *asr* and *hdeA*. These results suggest that *rstA* plays important roles in acid tolerance during host colonization. Previous studies have shown that changes in temperature, pH, and starvation, dramatically affect *rstA* expression in *V. alginolyticus* \[[@CR30]\]. This suggests that RstAB may sense environmental pH changes to regulate acid tolerance genes. However, whether *rstA* responds directly to low pH as a signal to activate acid tolerance pathways is unknown, and requires further investigation. Biofilm formation is mediated by bacterial surface structures that are regulated by environmental conditions. RstAB inhibits biofilm formation in *Salmonella enterica* and promotes biofilm formation in *Vibrio alginolyticus* \[[@CR30], [@CR40]\]. In the present study, we demonstrated that disruption of *rstA* in EHEC O157 also results in significant up-regulation of several c-di-GMP synthesis genes, suggesting that biofilm formation is influenced by c-di-GMP concentration. As expected, the Δ*rstA* mutant exhibited increased biofilm formation ability and intracellular c-di-GMP concentration. Among these up-regulated c-di-GMP synthesis genes, *yeaI* was directly regulated by *rstA*, and thus *rstA* can increase biofilm formation by increasing the concentration of c-di-GMP. Our RNA-seq results show that 66 regulator genes were differentially expressed (40 up-regulated and 26 down-regulated) in the Δ*rstA* mutant compared with WT strain (Additional file [2](#MOESM2){ref-type="media"}: Excel file S7). COG analysis of these genes indicated that these regulators participate in multiple biological processes (including intracellular trafficking, secretion, vesicular transport, amino acid transport metabolism signal transduction mechanisms, and transcription). This suggests that *rstA* can be both activator and repressor, and is capable of regulating more complex pathways than expected. Among the up-regulated regulators, an RstA box was found in the promoter region of *narP* (*z3450*), which regulates nitrate/nitrite respiration \[[@CR41]\]. The regulatory effect of *rstA* on other regulators needs to be confirmed experimentally. Conclusions {#Sec9} =========== The present study has contributed to our understanding of the EHEC O157 RstAB regulon, and identified a number of novel genes and functions that are affected by *rstA*. We found that RstA positively regulates virulence and acid tolerance, but negatively regulates biofilm formation in EHEC O157. In summary, the RstAB TCS in EHEC O157 plays a major role in the regulation of virulence, acid tolerance, and biofilm formation. Further research is required to reveal the mechanisms by which RstA regulates LEE genes. This may identify novel gene targets to control infections caused by this pathogen, which is particularly important given the emergence of drug resistance. Methods {#Sec10} ======= Bacterial strains and media {#Sec11} --------------------------- Bacterial strains, plasmids and primers used in this study are listed in Additional file [1](#MOESM1){ref-type="media"}: Tables S2 and S3. The Δ*rstA* mutant was constructed using the λ-Red recombination system and confirmed by PCR amplification and sequencing. A complementary strain was constructed by cloning RstA into the plasmid pTRC99a, and the resulting constructs were electroporated into EHEC O157 Δ*rstA* mutant. Antibiotics were added at the following final concentrations as required: 100 μg/ml ampicillin, 25 μg/ml chloramphenicol, 50 μg/ml nalidixic acid. RNA isolation, purification and sequencing {#Sec12} ------------------------------------------ Overnight cultures of EHEC O157 wild type (EHEC O157 WT) and the Δ*rstA* mutant were 1:100 subcultured in 20 ml of fresh Dulbecco's modified Eagle medium (DMEM, virulence-inducing medium for EHEC O157, Hyclone; \#SH30022.01) without antibiotics at 37 °C with shaking at 180 rpm, until the exponential growth phase was reached (OD~600~ = 0.6--0.8). Total RNA was extracted using TRIzol Reagent (Invitrogen; \# 15596026) and purified using the RNeasy Mini Kit (Qiagen; \#74104). The RNA was quantified and qualified using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA), a NanoDrop (Thermo Fisher Scientific Inc.), and 1% agarose gel electrophoresis. One microgram total RNA with a RIN value \> 6.5 was used for library preparation. rRNA (including 16S and 23S rRNA) was depleted from total RNA using The Ribo-off rRNA Depletion Kit (Bacteria) (Vazyme; \#N407). Libraries were constructed by VAHTSTM Total RNA-seq (H/M/R) Library Prep Kit for Illumina^®^ (Vazyme; \#NR603) according to manufacturer's instructions. Libraries with different indices were multiplexed and loaded on an Illumina HiSeq instrument according to manufacturer's instructions (Illumina, San Diego, CA, USA). The sequences were processed and data were analyzed by GENEWIZ, Inc (Suzhou, China). All sequence data have been deposited in the NCBI SRA database under the accession codes SRR9678084, SRR9678085, SRR9678086, and SRR9678087. Quantitative RT-PCR (qRT-PCR) {#Sec13} ----------------------------- Total RNA was extracted as previously described. First-strand cDNA was synthesized using the PrimeScript 1st Strand cDNA Synthesis Kit (Takara; \#D6110 A), according to the manufacturer's instructions. Primers for qPCR are listed in Additional file [1](#MOESM1){ref-type="media"}: Table S3. The 16S rRNA gene (*rrsH*) was used as a reference to standardize expression across the samples \[[@CR42]\]. Samples were amplified by PCR and amplicons were detected using SYBR green dye and an Applied Biosystems ABI 7500 sequence detection system (Applied Biosystems, CA, USA). The relative difference in gene expression was calculated using the cycle threshold method (2^−ΔΔct^) \[[@CR43]\]. Data were collected from at least three biological replicates. Bacterial adherence assay {#Sec14} ------------------------- Overnight cultures were subcultured in DMEM at 37 °C until they reached an OD~600~ of 0.6--0.8 for adaptation. Before infection, HeLa cells were washed three times with phosphate-buffered saline (PBS). The cell culture medium was replaced with fresh DMEM without antibiotics or fetal bovine serum. Cells were then infected with bacteria in DMEM at a multiplicity of infection (MOI) of 100:1. After incubation with HeLa cells for 3 h, unattached bacteria were removed by washing with PBS six times. The HeLa cells were then lysed with 0.1% SDS in H~2~O. Lysates were plated onto LB agar plates to count the number of viable adhered bacteria. Each experiment was carried out at least three times. Gut colonization assay {#Sec15} ---------------------- Six-week-old female BALB/c mice were provided with food and water ad libitum before infection. In each group, female BALB/c mice (n = 10) were orally infected with 10^9^ CFU of bacteria in 100 μl PBS. The infected mice were anaesthetized and euthanized via cervical dislocation at 6 h after infection. The distal colons were excised and the luminal contents were removed. Each distal colon of the intestine was washed with PBS three times to remove unattached bacteria, and then weighed and homogenized in 0.5 ml of PBS. The homogenates were diluted, and samples of the O157 WT strain, the Δ*rstA* mutant, and complementary strain were plated on LB agar containing nalidixic acid (50 μg/ml), chloramphenicol (25 μg/ml), or ampicillin (100 μg/ml), respectively, to determine the number of CFU per gram of organ tissue. Fluorescent actin staining {#Sec16} -------------------------- Fluorescent actin staining (FAS) assays were performed as described previously \[[@CR44]\]. Overnight cultures were subcultured in DMEM at 37 °C until they reached an OD~600~ of 0.6--0.8 for adaptation. HeLa cells were grown on coverslips for 24 h at 37 °C with 5% CO~2~. HeLa cells on coverslips were then infected with bacteria in DMEM at a multiplicity of infection (MOI) of 100:1. After incubation for 3 h at 37 °C and 5% CO~2~, the coverslips were washed with PBS and the bacteria were fixed with formaldehyde, and the cells were permeabilized with 0.2% Triton-X and stained with fluorescein isothiocyanate-labeled phalloidin to visualize actin filaments. Bacteria and HeLa cell nuclei were stained with propidium iodide. AE lesions formed by each strain were calculated for at least 50 HeLa cells. Quantitative biofilm assay {#Sec17} -------------------------- Biofilm formation was quantified by crystal violet staining, as previously described \[[@CR22]\]. Overnight cultures were diluted in fresh medium (1:100) and incubated in 96-well polystyrene microtiter plates at 37 °C for 24 h. The loosely associated bacteria were removed by washing with PBS three-times, and the remaining bacteria were stained with 0.5% crystal violet for 5 min. The biofilm was then destained by adding 200 μl of 95% ethanol to each well, and quantified using an enzyme-linked immunosorbent assay plate reader at 590 nm. Each experiment was carried out at least three times. High performance liquid chromatography {#Sec18} -------------------------------------- c-di-GMP was quantified using HPLC as described previously \[[@CR45]\]. Overnight cultures were subcultured in LB medium at 37 °C until they reached an OD~600~ of 0.6 for adaptation. Approximately 100 mg of cells were harvested in a pellet by centrifugation. The pellet was washed with PBS and resuspended in H~2~O. The suspension was heated at 95 °C for 15 min, followed by sonication. Ethanol was added to the sample to a final concentration of 70%. After centrifugation, the supernatant was pooled, frozen, and subsequently lyophilized overnight. The lyophilized flakes were resuspended in 1 ml of H~2~O and filtered through a 0.2 µm pore size filter. HPLC was performed using a 5 μm, 4.6 × 250 mm reverse phase column (Agela Venusil XBP-C18, VX952505-0) at room temperature with detection at 253 nm, on a Surveyor Plus HPLC System (Thermo Finnigan). Each experiment was carried out at least three times. Acid tolerance assay {#Sec19} -------------------- Overnight cultures were washed with PBS three times then diluted to a concentration of 10^6^ CFU/ml in LB acidified to pH 3.0 with HCl. Then cultures were incubated at 37 °C for 0 to 6 h with shaking at 180 rpm. A 100 μl aliquot was removed from the flask and suitable dilutions were plated on LB agar once every hour. Experiments were performed independently three times. EMSA {#Sec20} ---- The 6 × His-tagged RstA protein was expressed and purified in *E. coli* BL21 (DE3). DNA target fragments were amplified by PCR and purified using a SPARKeasy Gel DNA Extraction Kit (Sparkjade; \#AE0101-C). Purified PCR fragments (40 ng) were incubated at 25 °C for 30 min with 6× His-tagged RstA protein at concentrations ranging from 0 to 2 µM in 20 μl reactions containing binding buffer (1 mM Tris--HCl \[pH 7.5\], 0.2 mM dithiothreitol, 5 mM MgCl~2~, 10 mM KCl, and 10% glycerol, 30 mM acetyl phosphate). The protein-DNA fragments were electrophoretically separated on a native polyacrylamide gel at 4 °C and 80 V/cm. The gel was stained for 10 min in a solution of 0.1% GelRed (Biotium; \#41000), and protein bands were visualized by ultraviolet transillumination. Statistical analysis {#Sec21} -------------------- Statistical analysis was conducted using MedCalc (v12.3.0.0). The mean ± SD from three independent experiments was calculated. Differences between two mean values were evaluated by two-tailed Student's *t* test. Statistical significance was assessed with the Mann--Whitney rank-sum test in mouse colonization experiments. The significant enrichment of a given COG in the sets of up- or downregulated genes was determined using one-tailed Fisher's exact test with Benjamini--Hochberg false discovery rate correction \[[@CR46]\]. A P value \< 0.05 was considered to indicate statistical significance. Supplementary information ========================= {#Sec22} **Additional file 1: Figure S1.** Distribution of differentially expressed genes in the EHEC O157 EDL933 genome, **Table S1.** Transcriptional reads analysis of EHEC O157 WT and the Δ*rstA* mutant, **Table S2.** Strains and plasmids used in this study, **Table S3.** Primers used in this study (5′--3′), **Table S4.** RstA box analysis in EHEC O157. **Additional file 2: Excel file S1.** Down-regulated genes in the Δ*rstA* mutant compared with EHEC O157:H7 wild type strain, **Excel file S2.** Up-regulated genes in the Δ*rstA* mutant compared with EHEC O157:H7 wild type strain. **Excel file S3.** Differently expressed LEE genes in the Δ*rstA* mutant compared with EHEC O157:H7 wild type strain, **Excel file S4.** Differently expressed acid tolerance genes in the Δ*rstA* mutant compared with EHEC O157:H7 wild type strain, **Excel file S5.** Differently expressed diguanylate cyclases and phosphodiesterases genes in the Δ*rstA* mutant compared with EHEC O157:H7 wild type strain, **Excel file S6.** Differently expressed Stxs and Nle genes in the Δ*rstA* mutant compared with EHEC O157:H7 wild type strain, **Excel file S7.** Differently expressed regulatory genes in the Δ*rstA* mutant compared with EHEC O157:H7 wild type strain. EHEC O157 : enterohemorrhagic *Escherichia coli* O157:H7 TCS : two-component system HUS : hemolytic uremic syndrome A/E lesions : attaching and effacing lesions LEE : locus for enterocyte effacement T3SS : type III secretion system c-di-GMP : cyclic diguanylate DGCs : diguanylate cyclases PDEs : phosphodiesterases GDAR : glutamate-dependent acid resistance COG : clusters of orthologous group HK : histidine kinase RR : response regulator **Publisher\'s Note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Yutao Liu and Shujie Li contributed equally to this work Supplementary information ========================= **Supplementary information** accompanies this paper at 10.1186/s13099-019-0335-4. We wish to thank Wenchao Lin from Tianjin Biochip Corporation for his technical assistance in RNA-seq data analysis. We thank Xi Hou for her help with statistical analysis. BY designed the research; BY, YL, SL, WL, PW, PD, LL, JW, PY, QW, TX and YX performed the experiments; YL, WL, PW, PD, LL, JW, and PY collected the data; SL, QW, TX and YX analyzed the data; BY, YL and SL wrote the manuscript. All authors read and approved the final manuscript. This research was funded by the National Science Foundation of China (NSFC) Program (Grant Nos. 31530083 and 31800125) and the Natural Science Foundation of Tianjin (Grant No. 17JCQNJC09300). The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Assays performed in vivo were carried out according to the standards set forth in the Guide for the Care and Use of Laboratory Animals. The experimental protocols were approved by the Institutional Animal Care Committee at Nankai University. The research protocol was revised and approved by that committee. Not applicable. The authors declare that they have no competing interests.

{ "pile_set_name": "PubMed Central" }

INTRODUCTION {#sec1-1} ============ Oral lichen planus (OLP) is a T-cell mediated inflammatory disease of the oral mucosa. The etiology of OLP is unknown and it occurs in 0.5% -1.9% of the population.\[[@ref1][@ref2]\] Interleukin-8(IL-8) is an important mediator of host response to injury, trauma and inflammation\[[@ref3]\] and plays a role in activation of neutrophils, neutrophil chemotactic factor, T-cells and basophils.\[[@ref4]\] Interlukin-8 is produced by different cells such as monocytes/macrophages, T-cells, neutrophils, endothelial cells, fibroblasts and keratinocytes in inflammatory and pathological processes.\[[@ref3][@ref5][@ref6]\] Healthy tissues have insignificant amounts of IL-8 but its concentration quickly reaches 10--100 times its baseline level in response to pro-inflammatory cytokines such as tumor necrosis factor-alpha (TNF-α) or IL-1β, bacterial or viral products, and cellular stress.\[[@ref7]\] Keratinocytes in OLP patients can synthesize IL-1 and TNF-α.\[[@ref8][@ref9]\] Moreover, mononuclear cells infiltrated in the tissues in mucosal OLP patients, as well as mononuclear cells in the peripheral blood of these patients, can produce TNF-α.\[[@ref8][@ref9][@ref10]\] Furthermore, following an increase in releasing both locally and systemically of IL-1 and TNF-α, keratinocytes, macrophages, T-cells, endothelial cells and fibroblasts in OLP lesions release significant amounts of IL-8. Interlukin-8 results in greater infiltration of T-cells such as cytotoxic T-cells at the site of OLP. Thus, IL-8 may be involved in the pathogenesis of OLP.\[[@ref11]\] Studies have shown that serum levels of IL-6 and IL-8 are higher in OLP patients compared to healthy individuals.\[[@ref11][@ref12]\] Moreover, serum level of IL-8 is a more specific indicator of OLP than IL-6.\[[@ref11]\] Diabetes mellitus (DM) is a metabolic disease caused by impaired insulin production or resistance to it, and is characterized by the abnormal metabolism of glucose, proteins and lipids.\[[@ref13]\] In fact, type II DM is an inflammatory disease and inflammatory cytokines are involved in its pathogenesis.\[[@ref14]\] Serum level of IL-8 in type II DM is significantly higher than that in healthy individuals.\[[@ref15][@ref16]\] At present, a correlation between DM and OLP seems probable.\[[@ref17]\] The Prevalence of DM in OLP patients has been reported as1.6% -- 85%.\[[@ref18][@ref19][@ref20]\] Studies in this regard have been variable in the methodology and results. Jolly showed that 85% of patients with OLP in his study had abnormal glucose tolerance test results.\[[@ref21]\] Seyhan *et al*., in their study stated that half of the patients with OLP had glucose metabolism impairment and one-fourth of these patients had DM.\[[@ref22]\] However, other studies failed to show a higher incidence of abnormal glucose tolerance test in OLP patients compared to the healthy population.\[[@ref23][@ref24]\] Since IL8 is an important inflammatory mediator involved in both conditions, the first objective of this study was to measure and compare the serum level of IL8 in four groups of OLP, DM, OLP+DM and healthy control. Considering the existing controversies regarding the possible association of DM and OLP, the second objective of this study was to assess the fasting blood sugar (FBS) and 2-h postprandial blood sugar in OLP patients. MATERIALS AND METHODS {#sec1-2} ===================== This analytical cross sectional study was conducted on three groups of OLP, type II DM and control. Considering the 80% power ofthe study, d = 50 and α = 0.05, thesample size was calculated to be 35 in OLP, 20 in type II DM and 20 in control groups. Samples were selected through convenience sampling among those presenting to Isfahan School of Dentistry and clinics in the city of Isfahan. The OLP group was divided into two subgroups of OLP alone and OLP + type II DM. Thus, 30 OLP,5OLP and type II DM,20 type II DM and 20healthy control individuals were evaluated. The healthy controls were selected among those presenting for check up and had no history of systemic disease or drug intake. An internal medicine specialist confirmed the health of control subjects. The DM patients were selected among subjects suspected for DM, who referredto a diabetes treatment center for a definite diagnosis. These subjects had FBS ≥110 mg/dL or 2-h postprandial blood sugar ≥200 mg/dL and reported polyphagia, polydipsia and polyuria.\[[@ref25]\] These subjects had no other systemic disease and reported no drug intake. The diagnosis of type II DM was made by an internal medicine specialist. The OLP patients were selected among those presenting to Isfahan School of Dentistry and clinics in the city of Isfahan. Following observation of white striations and red streaks and inclusion of OLP in the list of differential diagnoses (according to the clinical criteria) by an oral medicine specialist, a biopsy sample was taken from the lesion. Upon confirmation of OLP by pathologist according to the World Health Organization criteria,\[[@ref26]\] the patient was included in the study. Patients with lichenoid reactions due to medications or contact and those with graft versus host disease (frequently Bone marrow graft) or cutaneous lesions were excluded. Fasting and 2-hpostprandial blood sugar tests were requested for subjects with definite diagnosis of OLP. Based on the test results, they were assigned to OLP or OLP+DM group until 35 samples were recruited. All participants underwent a thorough physical examination to ensurethe absence of inflammation in other parts of the body. Tobacco users, alcohol consumers and patients with periodontitis were excluded. In general, patients with a history of systemic conditions other than DM and OLP and those on any type of medication were excluded from the study. After 8 h of fasting, 3cc of blood was drawn from each subject between 7 and 9 a.m.(to prevent errors due to the circadian rhythm). Within 5--10 minutes, subjects had a usual breakfast (carbohydrate-rich) and underwent 2-h postprandial blood sugar test two hours after the breakfast. Within this 2-h period, subjects were requested to refrain from severe physical exercise. Immediately after clotting, blood samples were centrifuged at 3000 rpm for 15 minutes to isolate the serum. Immediately afterwards, the FBS and 2-h postprandial blood sugar were measured. The remaining serum isolated from the fasting blood sample was stored at 20°C for measurement of IL-8 using the ELISA Kit (Quantikine ELISA, R and D Systems, Inc., USA). The obtained data were analyzed using SPSS (IBM Stasistics 20), one-way ANOVA and post -- hocleast significant difference (LSD) test. Normality of the data was checked using Kolmogorov- Smirnov test. RESULTS {#sec1-3} ======= A total of 20 DM, 20 healthy control and 35 OLP individuals were included; 14.28% of OLP patients had type II DM. Furthermore, 20% of OLP patients had impaired fasting glucose (FBS of 100- 125 mg/dL). As seen in [Table 1](#T1){ref-type="table"}, the highest mean concentration of serum IL-8 was seen in OLP+DM group followed by OLP, DM (6.1 ± 1.91 pg/mL) and control (5.3 ± 1.23 pg/mL) groups. ###### The mean (±standard deviation) and range of changes in serum interleukin-8 concentration in the four groups and result of *post hoc* least significant difference test ![](DRJ-13-413-g001) One-way ANOVA showed significant differences among groups in terms of serum level of IL8 (*P* = 0.011). Post-hoc LSD test was then applied and showed a significant difference in serum level of IL-8 between the control and OLP groups (*P* = 0.002). The difference between the OLP+DM and control subjects was also significant in terms if serum level of IL-8 (*P* = 0.022). However, although the mean serum level of IL-8 in the DM group was higher than that in the control group, this difference was not significant (*P* = 0.087). The serum level of IL8 in OLP+DM patients was higher than that in DM group but this difference was not significant either (*P* = 0.219). The serum level of IL8 was higher in OLP+DM group compared to the OLP group but not significantly (*P* = 0.615). The mean levels of FBS and 2-h postprandial blood sugar in OLP+DM patients were the highest followed by DM, OLP and control groups, respectively \[[Table 2](#T2){ref-type="table"}\]. ###### The mean levels of fasting blood sugar and 2-h postprandial blood sugar in the four groups ![](DRJ-13-413-g002) DISCUSSION {#sec1-4} ========== To the best of the authors\' knowledge, no previous study has compared FBS, 2-h postprandial blood sugar and IL-8 levels in the serum of OLP, type II DM, OLP+DM and control subjects simultaneously; thus, the current study is probably the first one. The first objective of this study was comparing the serum level of IL-8 in OLP subjects and healthy individuals. The results showed that the mean serum level of IL-8 in OLP patients was significantly higher than that in controls (*P* \< 0.05). Sun *et al*.,\[[@ref11]\] in 2005 concluded that the serum level of IL-8 was significantly higher in OLP patients than controls and thus, IL-8 is a more sensitive marker than IL-6 for diagnosis of OLP. Rhodus *et al*. in 2005\[[@ref27]\] and Zhang *et al*. in 2008\[[@ref28]\] stated that the serum and salivary levels of IL8 were significantly higher than those in control group. Our findings are in agreement with the results of the above-mentioned studies. The following explanation can better elucidate the reason behind higher serum level of IL-8 in OLP patients compared to healthy controls. In general, there is a strong theory that T-cell-mediated autoimmune reactions are involved in the pathogenesis of OLP. The localand systemic release of different cytokines from the oral mucosa and blood is responsible for initiation and progression of this disease.\[[@ref29]\] Histological analysis has shown band-like infiltrates of T lymphocytes in the upper lamina propriaof OLP lesions.\[[@ref30]\] Inflammatory cytokines such as TNF-αand IL-1 stimulate T -- cells, monocytes, macrophages, keratinocytes, neutrophils, and endothelial cells and result in secretion of IL-8 in the lesion and in the peripheral blood.\[[@ref31][@ref32]\] Cell culture studies have shown that OLP keratinocytes produce higher levels ofTNF-α than gingivitis keratinocytes.\[[@ref9][@ref10]\] Thus, greater infiltration of inflammatory cells especially T -- cells as well as increased secretion of TNF-α and IL-1 in these patients can lead to higher secretion of IL-8; which explains the high serum levels of IL-8 in OLP patients compared to healthy individuals. Based on previous studies, OLP plus DM may occur concomitantly.\[[@ref18][@ref19][@ref20]\] Thus, the second objective of this study was to assess the serum level of IL-8 as a possible relevant factor in DM patients compared to healthy individuals. In the current study, the mean serum level of IL-8 in DM patients (6.1 ± 1.91 pg/mL) was higher than that in controls (5.3 ± 1.23pg/mL); however, this difference was not significant (*P* \> 0.05). According to a study by Abou-Shousha *et al*.,\[[@ref14]\] serum level of IL-8 in DM patients was significantly higher than that in healthy controls (*P* = 0.032). They suggested that serum level of IL-8 could be used as a predictor of DM-related micro- and macro-vascular diseases especially in high-risk individuals. Other studies also showed significantly higher serum level of IL-8 in type II DM compared to healthy individuals.\[[@ref15][@ref16]\] Although in the above-mentioned studies the serum level of IL-8 in diabetic patients was significantly higher than that in controls, this difference in our study was not significant. Such difference in results might be due to the variable sample sizes, differences in the nutritional regimens of healthy individuals, genetic variations and ethnic characteristics of subjects in different geographical locations. In general, it seems that high glucose level in diabetic patients results in attachment of monocytes to endothelial cells and subsequent production of IL-8.\[[@ref15][@ref33]\] The third objective of the current study was to measure the serum level of IL-8 in patients with OLP+DM in comparison with OLP plu DM patients as well as healthy controls. In our study, a significantly higher level of serum IL-8 was found in OLP+DM patients compared to healthy controls (despite the small sample size of the former group)-(*P* \< 0.05). No previous study has been conducted in this regard to compare with. Since the occurrence of DM and glucose impairments has been reported in OLP patients,\[[@ref18][@ref19][@ref20]\] the fourth objective of this study was to measure the levels of FBS and 2-h postprandial blood sugar in OLP patients in comparison with other groups. The mean level of FBS and 2-h postprandial blood sugar was the highest in OLP+DM patients followed by DM, OLP and control groups, respectively. In our study, 14.28% of OLP patients had DM and 20% had impaired fasting glucose (FBS of 100--125 mg/dL). Despite the higher mean level of FBS in OLP patients compare tocontrols, this difference did not reach statistical significance. Seyhan *et al*.\[[@ref22]\] found a significant difference in FBS between OLP and control subjects; whereas, some other studies failed to show a higher frequency of abnormal glucose tolerance test in OLP patients compared to the general population.\[[@ref23][@ref24]\] Ansar *et al*.\[[@ref34]\] reported that the mean FBS in OLP patients was 102.5 ± 32 mg/dL. This value was close to the value reported in our study (99.85 ± 14.4 mg/dL). Such controversy in the results of the afore-mentioned studies may be due to different methodology and study designs. Higher FBS levels in OLP patients compared to healthy controls points to the possible role of glucose intolerance in the pathogenesis of OLP.\[[@ref35]\] In addition, the ascending trend of the serum level of IL8 in control, DM, OLP, and OLP+DM patients may indicate the role of this factor in the pathogenesis of OLP and DB; moreover, it may have a synergistic effect on both conditions. In this study, it was not possible to detect internal inflammation (if any); thus, the IL-8 value was not exclusively indicative of OLP or DM. Small sample size in the OLP+DM group was another limitation of our study which was due to the small number of OLP patients and difficulty in encouraging them to participate in the study. Future studies with larger sample sizes are required on OLP patients to compare the serum level of IL-8 in different types of OLP and also to assess its role in the clinical course of the disease. CONCLUSION {#sec1-5} ========== The results showed higher serum levels of IL-8 in OLP+DM and OLP patients compared to healthy controls. Thus, IL-8 should be considered as an effective inflammatory factor in the pathogenesis of this disease. Administration of medications balancing the level of IL8 and decreasing its serum level may be a step forward to alleviate the symptoms of OLP and OLP+DM patients. Financial support and sponsorship {#sec2-1} --------------------------------- Grant Number: 394661. Conflicts of interest {#sec2-2} --------------------- The authors of this manuscript declare that they have no conflicts of interest, real or perceived, financial or non-financial in this article. This study financially supportedby the dental Research commitee of Isfahan University of MedicalSciences.

{ "pile_set_name": "PubMed Central" }

*GLAUCOMIDES BROMELICOLA* Foissner, [@b12] has been recently described in the context of a comprehensive project on the diversity and ecology of ciliates from tank bromeliads (Foissner [@b12]). Bromeliads are species-rich epiphytic and terrestrial plants that are wide-spread in Central and South America. Many are able to store water in a reservoir formed by the tightly overlapping bases of their rosette leaves. Such phytotelmata (Varga [@b34]), also known as tanks or cisterns, may contain up to 30 litres of water; they represent a highly specialized aquatic habitat (Armbruster et al. [@b1]; Foissner et al. [@b16]; Kitching [@b21]). In contrast to larger aquatic habitats, tank bromeliads are disconnected from each other. Thus, phytotelmata represent natural model systems for studying major functional characteristics of food webs and for testing community theories (Armbruster et al. [@b1]; Kitching [@b22]). As yet, ca. 40 new ciliate taxa have been discovered in tank bromeliads and several have already been described (Dunthorn et al. [@b4]; Foissner [@b6],b[@b7], [@b10], [@b12]; Foissner and Stoeck [@b13]; Foissner and Wolf [@b14]; Foissner et al. [@b16], [@b15]; Omar and Foissner [@b28], [@b29]). *Glaucomides bromelicola* belongs to the Bromeliophryidae and is distantly related to *Glaucoma* (Foissner [@b12]; Foissner and Stoeck [@b13]). While the genus *Glaucoma* radiated in more common freshwater habitats, *G. bromelicola* is widespread in tank bromeliads, commonly occurring in several bromeliad species from southern Chile to Mexico (Foissner [@b12]). The original description focused on morphological and molecular characteristics while the ecology received little attention. Details of the ecology of ciliates from tank bromeliads are hitherto known only from one species, *Bromeliothrix metopoides* (Weisse et al. [@b42]). Accordingly, the goal of this study was to investigate the functional ecology of *G*. *bromelicola* in comparison to that of *B. metopoides*. In contrast to *B. metopoides* and many other colpodean ciliates, *G*. *bromelicola* does not form cysts (Foissner [@b12]), which renders this species vulnerable to extinction following desiccation in the tanks. Water volume (typically ca. 0.01--3 liters), light, pH (usually ranging from 4.0 to 7.0), nutrients, and allochtonous organic input from leaf litter are the most important abiotic factors characterizing phytotelmata of tank bromeliads (Brouard et al. [@b2]; Janetzky [@b20]; Laessle [@b23]; Lopez et al. [@b24]; Marino et al. [@b25]). We hypothesized that *G*. *bromelicola* should have evolved adaptations different from those of *B. metopoides* to compensate for the lack of cyst formation and promote its survival in astatic aquatic environments. We focused on food (quantity and quality), competition, and pH as major environmental variables that are amenable to experimental manipulation in the laboratory. As *G*. *bromelicola* often co-occurs with *B. metopoides*, the ciliate species that is second in common in tank bromeliads, the traits that enable coexistence are of special interest. To this end, we performed competition experiments in the laboratory. We did not consider temperature in this study because typical temperature variation in the tanks of bromeliads is in the range of (only) 10 °C (summarized by Weisse et al. [@b42]), i.e. presumably of minor importance relative to the other physico-chemical variables. Materials and Methods ===================== Origin and maintenance of stock cultures ---------------------------------------- *Glaucomides bromelicola* Foissner, [@b12] represents a new tetrahymenid genus and species that is frequent in tank bromeliads of Central and South America (Foissner [@b12]). Similar to *B. metopoides* Foissner, [@b10], *G. bromelicola* forms macrostomes when bacterial food is depleted. The common in vivo cell size of cultivated specimen is ca. 57 × 30 μm (microstomes), respectively, 80 × 50 μm (macrostomes). The organisms used in this study were collected from tanks of small tree bromeliads either in a subtropical forest of Brazil (*B. metopoides*, *Polytomella* sp., heterotrophic bacteria) or in the botanical garden of Puerto Plata, Dominican Republic (*G. bromelicola*). Details of the origin of our study organisms have been reported by Foissner ([@b10], [@b12]). Cultures were established in Eau de Volvic (French table water) enriched with some sterilized, crushed wheat grains to promote growth of indigenous bacteria and bacterivorous flagellates. Monospecific protist cultures were obtained by repeated dilution with Volvic and pipetting of individual target cells, respectively, by removing unwanted predators or competitors. Ciliate stock cultures were maintained in "filter caps" culture flasks (Biomedica) with 50 ml of Eau de Volvic enriched with one to two wheat grains. The stock cultures were kept in an incubator at 22.5 °C under a 14:10 h light--dark cycle; pH was ca. 7.5. New cultures were inoculated once per week by transferring 25 ml of the aged culture to a new flask containing 25 ml Eau de Volvic and one new wheat grain. We kept *G. bromelicola* in nonaxenic stock cultures with and without the as-yet-undescribed flagellate *Polytomella* sp.; this flagellate, which was isolated together with *B. metopoides*, is variable in cell size (10--40 μm), has four flagella at the basis of a distinct papilla, four contractile vacuoles, lacks an eyespot, and has the nucleus in the anterior body half (Foissner [@b10]). The genus *Polytomella* comprises several colorless nutritionally versatile species that thrive on acetate and other organic acids, peptone, and yeast extract (de la Cruz and Gittleson [@b3]; Pringsheim [@b30]). We added two wheat grains to 50 ml of Eau de Volvic to provide an organic-rich medium for our *Polytomella* sp. cultures. Ciliate cell numbers in stock cultures usually ranged from 0.5 to 2.0 × 10^4^ cells/ml. Growth experiments with different food organisms ------------------------------------------------ We measured specific growth rates (μ) of *G. bromelicola* in response to different food quality (heterotrophic bacteria and the flagellates *Polytomella* sp. and *Cryptomonas* sp.) and quantity. *Polytomella* sp. coexists with the ciliate in tank bromeliads, *Cryptomonas* sp. is a cryptophyte common in many freshwater bodies and the preferred food of many planktonic ciliates (Skogstad et al. [@b31]; Weisse and Müller [@b38]). If not specified, cell numbers of *G. bromelicola* include microstomes and macrostomes in the following. In asexually reproducing ciliates, μ is a direct proxy of their fitness (Weisse [@b35]). The ciliate was inoculated together with the respective food organism(s) into 50-ml culture flasks. Treatments with food organisms but without ciliates served as controls. Target food levels lower than satiating (\< 2 mg C/L; experiments reported in Fig. [3](#fig03){ref-type="fig"}) were obtained by diluting strongly growing cultures with Eau de Volvic. Flagellate prey concentration ranged from a few hundred cells/ml to \> 80,000 cells/ml (equivalent to \< 0.05--\> 2 mg C/L), depending on the experimental conditions. Bacterial levels ranged from 10^6^ to 10^8^ cells/ml. Except for the experiments with *Cryptomonas* sp., the bacterial flora used in this study was enriched from the original habitat. The bacteria present in the *Cryptomonas* cultures remained unidentified but were, most likely, qualitatively different from those used in the other experiments. The initial experimental ciliate abundance was 20--110 cells/ml. All growth experiments were performed in the dark to prevent photoautotrophic food (the flagellate *Cryptomonas* sp.) from growing. Similarly, we removed remnants from wheat grains from the experimental containers to limit bacterial growth. The experiments lasted from several days to several weeks. ![Numerical response of the ciliate *Glaucomides bromelicola*. (A) With the auxotrophic flagellate *Polytomella* sp. as food. (B) With the phototrophic flagellate *Cryptomonas* sp. as food. The solid lines represent the fit to Eq. 4 (see text).](jeu0060-0578-f1){#fig01} ![Population dynamics of the ciliate *Glaucomides bromelicola* fed bacteria and the flagellate *Cryptomonas* sp. at four different food levels (A--D). Symbols represent means of triplicates; error bars denote 1 SD.](jeu0060-0578-f2){#fig02} ![Population dynamics of the ciliates *Bromeliothrix metopoides* and *Glaucomides bromelicola* fed bacteria and the flagellate *Polytomella* sp. (A) *B. metopoides* in single growth experiments. (B) *G. bromelicola* in single growth experiments. (C) Both ciliates in pairwise growth (= competition) experiments. Symbols represent means of triplicates; error bars denote 1 SD.](jeu0060-0578-f3){#fig03} Samples (5 ml) were taken from the experimental containers at 24 h intervals and fixed with acid Lugol\'s iodine (final concentration 2% v/v). Ciliate cell numbers were determined microscopically either in counting chambers of 3 ml volume or in Sedgewick Rafter chambers of 1 ml volume. Flagellate cell numbers were also counted microscopically, either together with the ciliates or separately in a 1-ml Sedgewick chamber. At higher abundance (\> 10^4^ cells/ml), flagellates were also counted and sized electronically by means of an automatic particle counter (CASY 1-model TTC; Schärfe System, Reutlingen, Germany; Weisse and Kirchhoff [@b37]). Bacterial levels were measured in Formalin-fixed (2% v/v) samples (2 ml) taken together with the ciliate samples. Bacterial cell numbers were assessed by flow cytometry after staining with the green fluorescent nucleic acid stain SYTO-13 (Molecular Probes, Invitrogen^™^, Carlsbad, CA, USA), using a FACSCalibur flow cytometer (Becton Dickinson BD Biosciences, San Jose, CA) equipped with an argon ion laser emitting light at 488 nm. Cell size of ciliates and flagellates was measured with Lugol\'s fixed material using an inverted microscope and a semi-automatic image analysis system (LUCIA version 4.51, Laboratory Imaging, Prague, Czech Republic). Flagellates were also measured in unfixed material. The automatic particle counter yielded an independent estimate of flagellate cell size, which was used mainly to determine the cell volume of living cells. Both methods yielded similar results. The cell volume of the prey flagellates used in this study (*Polytomella* sp. and *Cryptomonas* sp.) was converted to carbon units assuming the allometric equation provided by Menden-Deuer and Lessard ([@b26]), i.e. pg C/cell = 0.216 × cell volume^0.939^. To calculate carbon biomass of bacteria, we assumed a conservative estimate of 26 fg C/cell for our cultivated bacteria (Troussellier et al. [@b33]). Cell volume (*V*) of the ciliates was determined from length (*l*) and width (*w*) measurements, assuming a prolate spheroid shape: where *b* is cell breadth (in μm, as *l* and *w*). Measurements were made on 50 ciliates each at the end of several experiments. As we could not measure the third dimension, we assumed that *b* is equal to 0.6 × *w* (Foissner [@b12]). Competition experiments ----------------------- The experiments described above assumed that *G. bromelicola* is the only ciliate feeding on bacterial and flagellate food. However, in the tank bromeliads, this ciliate is often encountered together with *B. metopoides*, a small colpodean ciliate (Dunthorn et al. [@b4]; Foissner [@b10]). Accordingly, we compared the growth response of both ciliates when kept alone to that of pairwise experiments, i.e. with the potential competitor present. In each of four different competition experiments, we offered a combination of bacteria and *Polytomella* sp. as food; initial food levels were saturating for both ciliate species (\> 2 mg C/L, Weisse et al. [@b42]; this study) but declined in the course of the experiments (Fig. [3](#fig03){ref-type="fig"}, 5). To investigate the effect of competition on growth and survival of *B. metopoides,* we added *G. bromelicola* as competitor to the experimental containers (50-ml culture flasks) with *B. metopoides* and vice versa. To ensure that ciliate growth rates with and without competitor were directly comparable, we inoculated, in the first competition experiment, from a vigorously growing *B. metopoides* culture, one treatment without competitor (Fig. [3](#fig03){ref-type="fig"}A) and another one with its competitor, *G. bromelicola* (Fig. [3](#fig03){ref-type="fig"}C); similarly, we inoculated simultaneously from a *G. bromelicola* culture two different treatments without (Fig. [3](#fig03){ref-type="fig"}B) and with its competitor, *B. metopoides* (Fig. [3](#fig03){ref-type="fig"}C). Each treatment was then split into three replicates. The flagellate *Polytomella* sp. (initial abundance ca. 50,000 cells/ml, equivalent to ca. 2.7 mg C/L) served as food for the macrostomes; bacteria (initial abundance ca. 10^8^ cells/ml, equivalent to ca. 2.6 mg C/L) were present and served as food for the microstomes, but were unimportant as food for the macrostomes. This first competition experiment, therefore, consisted of three different predator--prey treatments: (A) *B. metopoides* plus *Polytomella* sp. and bacteria; (B) *G. bromelicola* plus *Polytomella* sp. and bacteria; (C) *B. metopoides* and *G. bromelicola* plus *Polytomella* sp. and bacteria. The duration of this experiment (1 wk) turned out to be too short to reveal whether the ciliates may outcompete each other. Further, we did not monitor bacterial levels. To this end, we repeated this competition experiment with several modifications and extended the duration up to 4 wk; we also measured bacterial concentration together with flagellate and ciliate abundance. In the second competition experiment (Fig. [5](#fig05){ref-type="fig"}A, B), we provided *Polytomella* sp. and bacteria at similar initial satiating amounts as in the previous experiment (ca. 50,000 *Polytomella*/ml, ca. 10^8^ bacteria/ml) but did not add additional food in the course of the experiment. Due to ciliate grazing, food limitation was likely after several days. To reduce food limitation, we added flagellate food 5 d after the beginning of the third competition experiment ([Fig.5](#fig05){ref-type="fig"}C, D). Finally, in a fourth competition experiment ([Fig.5](#fig05){ref-type="fig"}E, F), we added food upon each sampling occasion, beginning on experimental day 5. Initial ciliate abundance ranged from 30 to 40 cells/ml in all competition experiments. The general experimental conditions, the analyses of protist and bacterial cell numbers, and the calculation of protist growth rates were identical to the growth experiments reported above. Response to pH -------------- We investigated the growth and survival of *G. bromelicola* over pH ranging from 4.0 to 9.0; pH was measured using a microprocessor pH-mV meter (model pH 526; WTW, Weilheim, Germany) to the nearest 0.01 unit. The pH sensor was 2-point calibrated with standard buffer solutions of pH = 6.87 and pH = 9.18 before each series of measurement. Ciliates and their prey (provided at satiating amounts) were acclimated to the experimental conditions in steps of 0.5 pH unit change per day for 2--5 d. We measured and adjusted the pH in each experimental container twice per day; when pH differed by more than 0.2 from the target pH, it was adjusted by addition of small amounts (15--35 μl) of 0.1 mol/l NaOH or HCl (Weisse and Stadler [@b39]; Weisse et al. [@b41], [@b44]). The general experimental design followed that of the growth experiments. We used an incubator with a constant temperature of 22.5 °C without illumination for all experiments. Except for the competition experiments 2--4 (results reported in [Fig.5](#fig05){ref-type="fig"}), which were performed in duplicate, all experiments were run in triplicate. Results reported are mean values ± 1 standard deviation (SD). Data analysis and statistics ---------------------------- Ciliate growth rate (μ) was calculated from changes in cell numbers, assuming exponential growth over the experimental period according to where *N*~0~ and *N*~*t*~ are ciliate numbers at the beginning and at the end of the experimental period, respectively. Please note that 'experimental period' does not denote the total duration of an experiment, but specific periods of 1--4 d each during which the ciliate population increased exponentially. Details are reported in the Results section. Ciliate growth rates were related to the geometric mean prey concentration (*P*) during the experimental period (Frost [@b17]; Heinbokel [@b18]) according to Eq. 3: where *P*~0~ and *P*~t~ are the initial and final prey concentrations (cells/ml) during incubations. Ciliate growth rates were fit to Eq. 4, which includes a positive *x*-axis intercept, using the Marquardt-Levenberg algorithm (SigmaPlot; SPSS Inc., Chicago, IL). where μ is the growth rate (/d), μ~max~ is the maximum growth rate (/d), *P* is the geometric mean prey concentration (Eq. 3), *k* is a constant (cells/ml), and *x*′ is the *x*-axis intercept (i.e. threshold concentration, cells/ml, where μ = 0). This equation is similar to the Michaelis--Menten model and Holling\'s type II functional response (Holling [@b19]), but assumes a positive *x*-axis intercept where population growth equals mortality (Weisse et al. [@b43]). Accordingly, the constant *k* of Eq. 4 is similar, but not identical to the half saturation constant known from Michaelis--Menten kinetics. The curve resulting from Eq. 4 that describes the change in predator density as a function of change in prey density is known as numerical response (Solomon [@b32]). One-way ANOVA and Student\'s *t*-test were used to test for significant differences between several pairwise treatments. Results were considered statistically significant if *p* \< 0.05. Results ======= Response to food supply ----------------------- With *Polytomella* sp. and bacteria as food, *G. bromelicola* showed a typical numerical response curve (Fig. [1](#fig01){ref-type="fig"}A). The nonlinear curve fit (Eq. 4) yielded a maximum growth rate (μ~max~) of 3.55/d, a constant *k* = 7,035 *Polytomella* cells/ml, and an *x*-axis intercept (*x*′ in Eq. 4, where μ = 0) of 868 *Polytomella* cells/ml. All parameter estimates and the curve fit were statistically significant (*p* \< 0.001, *R*^2^ = 0.882). When converted to carbon units, food levels at near-to-maximum growth rates (μ = 3.0) were 1.92 mg C/L, *k* was reached at 0.38 mg C/L, and *x*′ amounted to 0.05 mg C/L. The percentage of flagellate-feeding macrostomes ranged from 1.0% to 8.5% of the total *G. bromelicola* cell number. The calculations above ignore the presence of bacteria that were, most likely, primarily used as food by the microstomes. Bacterial levels ranged from 1.1 to 10.5 × 10^6^ cells/ml in the numerical response experiment, equivalent to 0.03--0.28 mg C/L. As bacterial abundance was positively correlated with *Polytomella* abundance, the threshold food concentration of *G. bromelicola* was likely close to 0.1 mg C/L, and μ~max~ was recorded when total food concentration (i.e. *Polytomella* sp. plus bacteria) reached ca. 2.2 mg C/L. We did not assess the numerical response of *G. bromelicola* to bacteria as sole food, because we had observed in our routine cultures that ciliate growth rates are lower in the absence of the flagellate (Weisse, unpubl. data, but see below, Response to pH). When we replaced *Polytomella* sp. by *Cryptomonas* sp., *G. bromelicola* needed approximately 3 d to adapt to the new food (Fig. [2](#fig02){ref-type="fig"}). Irrespective of the food level, ciliate growth rates were low to moderate during the first 2--3 d of the experiment and then increased. Differences between *Cryptomonas* abundance in the experimental containers with ciliates and controls without ciliates were small and mostly statistically not significant during this period, suggesting that *G. bromelicola* did not at all or only sparsely feed upon the flagellates. However, at day 7, *Cryptomonas* levels were significantly lower (pairwise *t*-tests, *p *\< 0.001 in each case) in each experimental container than in the respective controls, and ciliate numbers had increased, relative to the first 4 d of the experiment (Fig. [2](#fig02){ref-type="fig"}A--D). Only at the lowest flagellate abundance (ca. 15,000 cells/ml), the ciliates did not grow during the period of observation, although bacteria were present in moderate abundance at the beginning of the experiment and continuously increased up to 9 × 10^6^ cells/ml to the end of the observation period (Fig. [2](#fig02){ref-type="fig"}A). The decline in bacterial levels recorded in Fig. [2](#fig02){ref-type="fig"}C, D demonstrates that *G. bromelicola* microstomes fed on bacteria, because bacterial levels did not decline in the controls without ciliates. Similarly, the initial bacterial growth in treatments A and B was lower than in the respective controls (data not shown). The difference between *Cryptomonas* sp. abundance in the experimental containers and controls indicates that the flagellate was ingested by *G. bromelicola* at the lowest flagellate abundance tested. However, the cryptophyte biomass of ca. 0. 4 mg C/L did not support positive ciliate growth. Including bacteria, the total concentration of potential food was ca. 0. 6 mg C/L at the end of this experiment. When we plotted the ciliate growth rate measured between experimental days 4--7 vs. the mean *Cryptomonas* abundance during this period, we obtained a numerical response curve with a statistically significant (*R*^2^ = 0.896) curve fit (Fig. [1](#fig01){ref-type="fig"}B). However, comparing Fig. [1](#fig01){ref-type="fig"}B to A demonstrates that μ~max~ of *G. bromelicola* was significantly lower (0.60/d) with *Cryptomonas* sp. as food and the cellular threshold prey concentration was approximately tenfold higher for *Cryptomonas* sp. than for *Polytomella* sp. In terms of biomass, the difference was lower, ca. 0. 30 mg C/L (*Cryptomonas* sp.), respectively, 0.05 mg C/L (*Polytomella* sp.). Including bacteria, we estimated a total threshold concentration of ca. 0.5 mg C/L for *G. bromelicola* fed *Cryptomonas* sp. and bacteria. Tables [1](#tbl1){ref-type="table"}, [2](#tbl2){ref-type="table"} summarize the results obtained in both numerical response experiments. ###### Parameter estimates of the numerical response curves (see text, Eq. 4) of *Glaucomides bromelicola* fed the flagellates *Polytomella* sp., respectively, *Cryptomonas* sp. Variable Coefficient SE Coefficient SE ---------- ------------- -------------------------------------------- ------------- ------------------------------------------- μ~max~ 3.55 0.23[\*\*\*](#tf1-3){ref-type="table-fn"} 0.60 0.14[\*\*](#tf1-2){ref-type="table-fn"} *x*′ 868 170[\*\*\*](#tf1-3){ref-type="table-fn"} 10,678 670[\*\*\*](#tf1-3){ref-type="table-fn"} *k* 7,035 1,470[\*\*\*](#tf1-3){ref-type="table-fn"} 12,364 6,782[\*](#tf1-1){ref-type="table-fn"} *R*^2^ 0.88 0.44[\*\*\*](#tf1-3){ref-type="table-fn"} 0.90 0.07[\*\*\*](#tf1-3){ref-type="table-fn"} \**p* \< 0.102. \*\**p* \< 0.01. \*\*\**p* \< 0.0001. μ~max~, maximum growth rate (per d), *x*′, threshold food concentration (cells/ml), *k*, a constant (cells/ml), *R*, coefficient of determination, SE, standard error. ###### Key variables characterizing the life strategies of the ciliates *Bromeliothrix metopoides* and *Glaucomides bromelicola* from tank bromeliads Variable *B. metopoides* *G. bromelicola* Source -------------------------------------- -------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------- Taxonomic affiliation (class, order) Colpodea, Colpodida Oligohymenophorea, Tetrahymenida Foissner ([@b10], [@b12]) Cell length 20--55 μm 25--80 μm Foissner ([@b10], [@b12]), Weisse et al. ([@b42]) Average cell volume in cultures 8,570 μm^3^ 8,250 μm^3^ Weisse et al. ([@b42]), This study Macrostomes Yes Yes Foissner ([@b10], [@b12]) Cysts Yes No Foissner ([@b10], [@b12]) Division chains Yes No Foissner ([@b10], [@b12]) Food organisms Bacteria; *Polytomella* sp. Bacteria; *Polytomella* sp.; other flagellates Weisse et al. ([@b42]), This study Sensitivity to competition Moderate Low This study Growth rate threshold 1.4[a](#tf2-1){ref-type="table-fn"}--2.3[b](#tf2-2){ref-type="table-fn"}mg C/L 0.05[](#tf2-3){ref-type="table-fn"}--0.1[a](#tf2-1){ref-type="table-fn"}mg C/L ∼0.5[d](#tf2-4){ref-type="table-fn"}mg C/L Weisse et al. ([@b42]), This study μ~max~ (/d) 3.55 4.71 Weisse et al. ([@b42]), This study pH tolerance \< 4--\> 9 \< 4--\> 9 Weisse et al. ([@b42]), This study pH optimum ∼8 7--8 Weisse et al. ([@b42]), This study With bacteria and *Polytomella* sp. as food, i.e. microstome and macrostome feeding. With bacteria as sole food, i.e. only microstome feeding. With *Polytomella* sp. as food, ignoring the bacterial background. With bacteria and *Cryptomonas* sp. as food, i.e. microstome and macrostome feeding. Competition for food -------------------- When kept alone with their prey, both ciliates started growing exponentially after an initial lag phase (Fig. [3](#fig03){ref-type="fig"}A, B). In contrast to *B. metopoides* (Fig. [3](#fig03){ref-type="fig"}A), exponential growth of *G. bromelicola* continued to the final day of observation (Fig. [3](#fig03){ref-type="fig"}B); μ calculated from linear regression of ln cell numbers vs. time was 1.19 ± 0.12/d (days 1--6, *n* = 6, *R*^2^ = 0.960) for *B. metopoides*, respectively, 1.37 ± 0.07/d (days 2--7, *n* = 6, *R*^2^ = 0.989) for *G. bromelicola*. Highest growth rates from day to day were reached for both ciliates during days 3--4, with μ~max~ = 2.45 ± 0.16/d (*B. metopoides*) and μ~max~ = 2.01 ± 0.20/d (*G. bromelicola*). In the latter case, average ciliate cell number increased from 84.3 to 632.0 cells/ml during days 3--4; due to the large scale of the *y*-axis, this is difficult to deduce from Fig. [3](#fig03){ref-type="fig"}B. In the pairwise experiment (Fig. [3](#fig03){ref-type="fig"}C), *B. metopoides* reached the same continued growth rate, 1.20 ± 0.09/d, as in the single growth experiment, but for a shorter period (days 1--4, *n* = 4, *R*^2^ = 0.989); during days 4--6, μ of *B. metopoides* was ≤ 1.0/d. Similarly, highest day to day growth rate was reached during days 3--4; however, μ~max~ = 1.51 ± 0.02/d in the competition experiment was significantly reduced (*p* \< 0.001) compared to the single growth experiment. Maximum cell number in the former was 4,700 cells/ml, compared to 8,060 cells/ml in the single growth experiment. Final abundance of *B. metopoides* was 6,041 ± 374 cells/ml (single growth), respectively, 777 ± 166 cells/ml (pairwise growth). These differences in the duration of exponential growth and final cell numbers were affected by a switch to mass encystment observed in the competition experiment during days 6--7 (Fig. [4](#fig04){ref-type="fig"}). On day 7, cyst abundance in the competition experiment was significantly higher (*p* \< 0.001) than in the single growth experiment. The opposite was true for days 2 and 4; we recorded no difference in cyst abundance between the two treatments during days 0, 1, 5, and 6. ![Cyst formation of the ciliate *Bromeliothrix metopoides* in the experiments shown in Fig. [3](#fig03){ref-type="fig"}A and C.](jeu0060-0578-f4){#fig04} Continued exponential growth rate of *G. bromelicola* in the pairwise experiment (Fig. [3](#fig03){ref-type="fig"}C) was identical to that measured in the single growth experiment (1.36 ± 0.08/d; days 3--7, *n* = 5, *R*^2^ = 0.988); similarly, μ~max~ = 1.88 ± 0.10/d reached during days 3--4 and was not different (*p* = 0.369) from the single growth experiment. Considering that *G. bromelicola* grew exponentially, differences in its final abundance reported in Fig. [3](#fig03){ref-type="fig"}B, C were minor. Both ciliate species formed macrostomes in the competition experiments, with typical percentages ranging from 2% to 7% of the total cell number. In summary, the first competition experiment demonstrated some effect of *G. bromelicola* on *B. metopoides*, but virtually no effect of the latter on the former. The experimental duration was too short to reveal if *B. metopoides* may be outcompeted by *G. bromelicola*. In the following three competition experiments, we extended the duration up to 4 wk; we also measured bacterial concentration together with flagellate and ciliate abundance. Results shown in Fig. [5](#fig05){ref-type="fig"} demonstrate that both ciliates may coexist over several weeks. In the second competition experiment, without feeding (Fig. [5](#fig05){ref-type="fig"}A, B), both ciliates coexisted with small, but relatively stable populations (*B. metopoides*: 242 ± 150 cells/ml, *G. bromelicola* 313 ± 285 cells/ml) after 2 wk, when the abundance of the food flagellate *Polytomella* sp. remained \< 1,000 cells/ml, and bacterial levels were close to 0.2 × 10^8^ cells/ml (Fig. [5](#fig05){ref-type="fig"}A). Starting 1 wk after the beginning of the experiment, the abundance of encysted *B. metopoides* exceeded that of motile cells and continuously increased until the end of the experiment (Fig. [5](#fig05){ref-type="fig"}B). Maximum abundance of *G. bromelicola* (25,715 ± 3,692 cells/ml) was more than 10-fold higher than that of *B. metopoides* (1,753 ± 85 cells/ml). Both ciliates were, most likely, food limited in this experiment after the initial week. ![Competition experiment with the ciliates *Bromeliothrix metopoides* and *Glaucomides bromelicola* (bottom panels) fed bacteria and the flagellate *Polytomella* sp. (top panels). (A, B) No additional food was added in the course of the experiment. (C, D) Additional food was added on day 5 of the experiment (indicated by dashed lines and arrows in Fig. [5](#fig05){ref-type="fig"}C). (E, F) Additional food was added on each sampling occasion, beginning on day 5 of the experiment. Bacterial abundance is related to the left *y*-axis in Fig. [5](#fig05){ref-type="fig"}A, C, E; flagellate abundance is related to the right *y*-axis in Fig. [5](#fig05){ref-type="fig"}A, C, E. Symbols represent means of duplicates, error bars denote 1 SD.](jeu0060-0578-f5){#fig05} In the third competition experiment, we added flagellate food 5 d after the beginning of the experiment (Fig. [5](#fig05){ref-type="fig"}C). The ciliates coexisted in this experiment for the following 2 wk with higher cell numbers than in the previous experiment (Fig. [5](#fig05){ref-type="fig"}D). The average abundance of *G. bromelicola* (12,608 ± 4,553 cells/ml) was ca. 14× higher than that of *B. metopoides* (884 ± 337 cells/ml) during the second and third week of the experiment. Different from the second competition experiment without additional feeding, motile cells of *B. metopoides* prevailed over encysted cells throughout this experiment. In the fourth competition experiment, with food added upon each sampling occasion from experimental day 5 onwards, food levels of bacteria and *Polytomella* sp. were higher than in the previous experiments (Fig. [5](#fig05){ref-type="fig"}E). Ciliate growth rates cannot be calculated directly from the results shown in Fig. [5](#fig05){ref-type="fig"}F because the experimental containers were diluted 1:3 upon each addition of food. However, the nearly constant ciliate cell numbers recorded for both species during experimental days 5--19 suggest that the ciliate populations increased with a relatively constant rate close to 0.55/d. Similar to the previous experiment, (i) the average abundance of *G. bromelicola* (4,534 ± 1,539 cells/ml) was ca. 10-fold higher than that of *B. metopoides* (456 ± 58 cells/ml), and (ii) motile cells of *B. metopoides* prevailed over encysted cells. Individual and average ciliate cell size was variable in the different growth and competition experiments, mainly depending on food and the percentage of macrostomes in the population (data not shown). Assuming typical average dimensions of *l* = 42 μm and *w* = 25 μm and accounting for 10% shrinkage due to fixation, we calculated an average cell volume of *G. bromelicola* of 8,250 μm^3^ under food replete conditions. The average cell size of *G. bromelicola* measured in the present investigation was thus smaller than the in vivo cell size reported by Foissner ([@b12]) for well-fed laboratory specimens. Response to pH -------------- We investigated the pH response of *G. bromelicola* over pH ranging from 4 to 9 with combined bacterial and flagellate (*Polytomella* sp.) food, respectively bacteria without any other food organism (Fig. [6](#fig06){ref-type="fig"}). To avoid food limitation, prey organisms were provided in satiating amounts (\> 2 mg C/L). With the combined food, *G. bromelicola* was relatively insensitive to pH, reaching high growth rates (\> 2.4/d) over the entire pH range tested; growth rates peaked at pH 7--8, where μ was not different (*t*-test, *p* = 0.073). Highest growth rates, μ~max~ = 3.05/d, obtained in the pH experiment with *Polytomella* sp. plus bacterial food at pH 7--8 were close to μ~max~ measured in the numerical response experiment with the same food organisms (cf. Fig. [1](#fig01){ref-type="fig"}A). If bacteria were the sole food, the ciliates (i) reached lower maximum growth rates, (ii) the pH optimum was shifted to acidic conditions, and (iii) the highest pH barely supported positive growth rates (Fig. [6](#fig06){ref-type="fig"}). ![pH response of the ciliate *Glaucomides bromelicola* fed bacteria only, respectively bacteria plus the flagellate *Polytomella* sp. Bars represent means of triplicates, error bars denote 1 SD.](jeu0060-0578-f6){#fig06} Discussion ========== Different life strategies of *Glaucomides bromelicola* and *Bromeliothrix metopoides* enable their stable coexistence in tank bromeliads ---------------------------------------------------------------------------------------------------------------------------------------- Results of this study and a related recent study (Foissner [@b10], [@b12]; Weisse et al. [@b42]) demonstrate contrasting life strategies of the two common ciliates from tank bromeliads (summarized in Table [2](#tbl2){ref-type="table"}). As *G. bromelicola* cannot form cysts, this species is threatened by desiccation that may be common in many small bromeliads (Kitching [@b22]; Marino et al. [@b25]). Accordingly, a *G. bromelicola* population can escape extinction only via dispersal into new suitable habitats. The preferred flagellate prey, *Polytomella* sp., is common in tank bromeliads but it was not found in every reservoir (Foissner, unpubl. data). The ability to adapt to different prey such as *Cryptomonas* sp. may enable survival of the ciliate in less suitable habitats. The adaptation period of ca. 3 d that we observed (Fig. [2](#fig02){ref-type="fig"}) appears to be short enough to prevent extinction of the population. Although μ~max~ was significantly reduced with *Cryptomonas* sp. as food (μ~max~ = 0.60/d), it is still in a range typical for many aquatic ciliates known to feed on *Cryptomonas* sp. (Skogstad et al. [@b31]; Weisse [@b35]; Weisse and Müller [@b38]). Secondly, the food threshold of *G. bromelicola*, i.e. the prey level that is needed to sustain the population, is relatively low, compared to most planktonic ciliates (Weisse [@b35]), and orders of magnitudes lower than that of *B. metopoides* (Weisse et al. [@b42]). This nutritional versatility may compensate for the lack of cyst formation in *G. bromelicola*. *Bromeliothrix metopoides* is unable to use *Cryptomonas* sp. and other flagellates and small ciliates as food (Weisse et al. [@b42]). With its preferred food, *Polytomella* sp., the threshold concentration is close to 1.4 mg C/L (Table [2](#tbl2){ref-type="table"}), i.e. higher than that of any other aquatic ciliate investigated thus far under comparable experimental conditions (Weisse et al. [@b42]). The microstomes of *B. metopoides*, which feed on bacteria, require an even higher food threshold (2.3 mg C/L) to proliferate. In contrast, microstomes of *G. bromelicola* can ingest bacteria at concentrations \< 10^7^/ml, equivalent to ca. 0.2 mg C/L (Fig. [2](#fig02){ref-type="fig"}C, D). The lower affinity to its preferred bacterial and flagellate food is likely the cause why *B. metopoides* is the inferior competitor when paired with *G. bromelicola*. However, encystment and excystment enable *B. metopoides* to coexist with *G. bromelicola* over weeks (Fig. [5](#fig05){ref-type="fig"}). During short periods of time, the former may even reach higher growth rates than the latter (Table [2](#tbl2){ref-type="table"} and Fig. [3](#fig03){ref-type="fig"}). We repeatedly observed greater than five cell divisions per day of *B. metopoides* in our laboratory cultures (Foissner [@b10]; Weisse et al. [@b42]), similar to the shortest generation times known from small terrestrial colpodeans (reviewed by Weisse et al. [@b42]). If the ciliates reach such high growth rates in situ remains at present unknown. Why are both ciliates restricted to tank bromeliads? ---------------------------------------------------- A central issue of this research was to explain why the most common ciliates from tank bromeliads appear to be restricted to this peculiar habitat. Of the four environmental variables studied (pH, food quantity and quality, competition), pH was seemingly of minor importance, as both ciliate species were widely tolerant to changing pH (Weisse et al. [@b42]; this study). However, recent experimental evidence with freshwater ciliates, flagellates, and microeukaryotes revealed that the realized pH niche may be confined to a small range, relative to the fundamental pH assessed in the laboratory under optimized conditions. This is because there is a significant interaction of pH with food and temperature (Moser and Weisse [@b27]; Weisse [@b35]; Weisse et al. [@b43], [@b40], b[@b42]). Interaction of pH with food quality was obvious for both ciliate species of this study; the pH niche was narrower and, in the case of *G. bromelicola*, restricted to the pH range typical of bromeliad reservoirs if bacteria were the only food (Fig. [6](#fig06){ref-type="fig"}; Weisse et al. [@b42]). It appears that (metabolism of) the flagellate *Polytomella* sp. conditioned the medium in a favorable way for *G. bromelicola*, promoting high growth rates even under alkaline conditions. The unusually high food threshold and its demand for specific food items (*Polytomella* sp. and some bacteria) are the main reasons why *B. metopoides* cannot thrive in other environments (Weisse et al. [@b42]), in spite of its ability to form cysts and endure desiccation of its natural habitat. Summarizing the sparse literature on the abundance of bacteria and flagellates in tank bromeliads, Weisse et al. ([@b42]) concluded that, while it is likely that the high food levels needed by *B. metopoides* are occasionally met in situ, it remains at present unknown if such favorable conditions are persistent in tank bromeliads. The results of this study further suggest that motile cells of this species are inferior competitors, relative to other ciliates with higher affinity to prey common in bromeliads and elsewhere. Weisse et al. ([@b42]) concluded that *B. metopoides* has reached a narrow peak along its fitness landscape and any deviation from the optimum conditions will reduce its fitness. This conclusion does not apply to *G. bromelicola*; based upon its feeding and growth characteristics reported in this study, this species should be able to live in a wider array of aquatic habitats than *B. metopoides*. Indeed, *G. bromelicola* was found in samples from 11 different bromeliad species, while *B. metopoides* occurred in only three bromeliad species (Dunthorn et al. [@b4]). The lack of cyst formation severely limits dispersal of *G. bromelicola*. Long range dispersal of ciliates and other microbes is only possible in the form of resting stages resistant to desiccation (Foissner [@b8], [@b9], [@b11]; Weisse [@b36]). We hypothesize that dispersal of *G. bromelicola* over narrow ranges is primarily mediated via insects specifically attracted to bromeliads, explaining why this species is endemic in tank bromeliads. A corollary of this conclusion is that it supports the earlier notion that the high degree of endemicity observed in the bromeliad tanks points to speciation within this habitat (Dunthorn et al. [@b4]; Foissner [@b12]; Foissner et al. [@b16]). Both species used in the present investigation possess a micronucleus (Foissner [@b10], [@b12]). However, conjugation was observed only in *G. bromelicola* (Foissner, unpubl. data), while most colpodids very likely lack sex at all (Foissner [@b5]). Thus far, no attempts have been made at mating in the laboratory. The genetic population structure of tank bromeliads is an area that awaits future research. We appreciate the constructive comments by two anonymous reviewers on the manuscript. This study was financially supported by the Austrian Science Fund (FWF, project P20360-B17).

{ "pile_set_name": "PubMed Central" }

![](brjcancer00051-0226.tif "scanned-page"){.1122}

{ "pile_set_name": "PubMed Central" }

Introduction {#Sec1} ============ The *All-Pairs Shortest Paths* (APSP) problem asks to find distances between all pairs of vertices in a graph. For a directed graphs with weights in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {R}$\end{document}$, there is a classical *O*(*n*^3^) time algorithm \[[@CR11], [@CR29]\]. Currently best upper bound for this problem is due to \[[@CR30]\] who showed an $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$O\left (\frac {n^{3}}{2^{{\Omega }(\log n)^{0.5}}}\right )$\end{document}$ algorithm. It is asymptotically faster than *O*(*n*^3^/log*cn*) for any *c* \> 0 (see survey \[[@CR6]\] for earlier algorithms). Showing any algorithm that would work in *O*(*n*^3−*𝜖*^) time for some *𝜖* \> 0 is a major open problem \[[@CR30]\]. If we consider unweighted, directed graphs there are subcubic algorithms that exploit fast matrix multiplication. For the undirected graph \[[@CR24]\] presented the optimal $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time algorithm, where *ω* \< 2.373 is the matrix multiplication exponent \[[@CR16]\]. For the directed case \[[@CR35]\] presented an *O*(*n*^2.575^) time algorithm that is based on the fast rectangular matrix multiplication. Moreover, if we are interested in small integer weights from the set {−*M*,...,*M*} we have *O*(*M*^0.68^*n*^2.575^) algorithm \[[@CR35]\]. Because APSP in undirected graphs can be solved in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$, diameter, radius, shortest cycle, etc. can be determined in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time as well. It is surprising that for a directed case, where merely *O*(*n*^2.575^) APSP is known there are also $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ algorithms for determining these properties. After a long line of improvements \[[@CR9]\] showed an $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(M n^{\omega })$\end{document}$ time algorithms for finding minimum weight perfect matching, shortest cycle, diameter and radius (some of these results were already known \[[@CR21]\]). Also, \[[@CR9]\] showed an application of their techniques that improves upon \[[@CR31]\] $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(M n^{\omega } t)$\end{document}$ time algorithm for the following problem: *determine the set of vertices that lie on some cycle of length at mostt*. Cygan et al. \[[@CR9]\] managed to solve this problem in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(M n^{\omega })$\end{document}$ time using Baur-Strassen's theorem. All of these algorithm are effective merely in the case of a dense graphs. For graphs with the small number of edges there are more efficient algorithms (e.g., APSP in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(|V||E|)$\end{document}$ time \[[@CR27]\]). But these algorithms are Θ(*n*^3^) when \|*E*\| = Θ(*n*^2^). Related Work {#Sec2} ------------ ### Distance Queries {#Sec3} Yuster and Zwick \[[@CR33]\] considered the weighted, directed graphs with weights in {−*M*,...,*M*}. They showed an algorithm that needs $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(M n^{\omega })$\end{document}$ preprocessing time. After preprocessing each distance *δ*(*u*, *v*) in the graph can be computed exactly in *O*(*n*) query time. In the special case *M* = 1 they showed $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ algorithm that solves *Single Source Shortest Paths* (SSSP). This is the best known algorithm for a dense, weighted graph. We will match their bounds (up to the polylogarithmic factors) using Frobenius normal form. Next we will extend that approach so it will return more information about a graph in the same query/preprocessing time. ### Counting Cycles {#Sec4} For a given graph *G* and *k* determining whether *G* contains a simple cycle of length exactly *k* is NP-hard (in particular determining whether a graph contains a Hamiltonian cycle is NP-complete). However, if we fix *k* to be a constant this problem can be solved in polynomial time. Alon et al. \[[@CR4]\] introduced a color coding technique. For a fixed *k* if a graph *G*(*V*, *E*) contains a simple cycle of size exactly *k* then such a cycle can be found in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(|V|^{\omega })$\end{document}$ time. Unfortunately, their algorithm depends exponentially 2^*O*(*k*)^ on the length of the cycle and in consequence is inapplicable for large *k*. In the next years, \[[@CR5]\] showed (using a different technique) that for *k* ≤ 7 one can count the number of cycles of length exactly *k* in a graph in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(|V|^{\omega })$\end{document}$ time. In \[[@CR32]\] it is shown that for any even *k*, cycles of length *k* can be found in *O*(\|*V* \|^2^) time in undirected graphs (if they contain such a cycle). Alon et al. \[[@CR5]\] showed more methods that depend solely on a number of edges in a graph. For example for odd *k* they showed $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$O\left (E^{2-\frac {2}{k + 1}}\right )$\end{document}$ algorithm for finding cycles of length *k*. However, for dense graphs these results are worse than \[[@CR4]\]. From the other hand, to detect whether a non-simple cycle of length exactly *k* exists one can use the folklore algorithm. It starts by taking the adjacency matrix *A* of a graph *G*. Subsequently, in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time compute *A*^*k*^ by repeated squaring. If Tr \[*A*^*k*^\] \> 0 then there exists a non-simple cycle of length *k*.[1](#Fn1){ref-type="fn"} Yuster \[[@CR31]\] considered the following problem: *for every vertex in a graph find a shortest cycle that contains it*. He called this problem *All-Nodes Shortest Cycle* (ANSC). He showed a randomized algorithm that solves ANSC for undirected graphs with weights {1,...,*M*} in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(\sqrt {M} n^{(\omega + 3)/2})$\end{document}$ time. He noted that for simple digraphs (directed graphs with no anti-parallel edges) it reduces to All-Pairs Shortest Path problem. The fastest known APSP algorithm for unweighted, directed graphs runs in *O*(*n*^2.575^) due to \[[@CR35]\]. Here, we will show how to solve ANSC in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ for general, unweighted, directed graphs. Unfortunately, our techniques will allow us merely to find the length of such a cycle. But we can return the set of points, that lie on some cycle of a given length. Independently to our work \[[@CR3]\] proved that ANSC can be solved in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ for unweighted, undirected graphs using a completely different technique. Yuster \[[@CR31]\] also considered following problem: *given a graph and an integert*. *LetS*(*k*) *denote the set of all vertices lying in a cycle of length*≤ *k*. *DetermineS*(*t*). He considered directed graphs with weights in {−*M*,...,*M*} and showed $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(M n^{\omega } t)$\end{document}$ algorithm. Recently, \[[@CR9]\] improved his algorithm. They showed that for a fixed *t* ∈ \[0,*nM*\] the set *S*(*t*) can be computed in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(M n^{\omega })$\end{document}$ randomized time. We show, that for unweighted (*M* = 1) directed graphs we can compute sets *S*(1),*S*(2),...,*S*(*D*) in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time with high probability. Preliminaries {#Sec5} ============= Let *T*(*n*) be the minimal number of algebraic operations needed to compute the product of *n* × *n* matrix by an *n* × *n* matrix. We say that *ω* is the exponent of square matrix multiplication. For now the best known upper bound on *ω* is due to \[[@CR16]\]: **Theorem 1** {#FPar1} ------------- \[[@CR16]\] *For every𝜖* \> 0,*T*(*n*) \< *O*(*n*^*ω*\ +\ *𝜖*^) *whereω* \< 2.37287*.* In this paper, we will omit *𝜖* in definition and will assume that *O*(*n*^*ω*^) operations are needed to multiply two matrices. The best lower bound for the exponent of matrix multiplication is *ω* ≥ 2. For convenience in this paper we will assume that *ω* \> 2. The $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}$\end{document}$ notation hides polylogarithmic factors in complexity. We will use it to emphasize that all our algorithms need polylogarithmic number of calls to the fast matrix multiplication algorithm. In this paper we will consider the Word RAM model of computation with a word size *O*(log *n*). The $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {F}$\end{document}$ denotes a small finite field and often we will assume this field to be $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {Z}_{p}$\end{document}$ for a prime number *p* with *O*(log *n*) bits. Note, that in Word RAM model arithmetic operations in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {Z}_{p}$\end{document}$ can be done in constant time. In some situations we will need to stress that we want to work on larger integers. We will use *W* to denote the upper bound on the largest number in the field. In this case the arithmetic operations in Word RAM model take *O*(log *W*) time. For matrices $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A \in \mathbb {F}^{n\times k}$\end{document}$ and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$B \in \mathbb {F}^{n \times l}$\end{document}$ the $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A \oplus B \in \mathbb {F}^{n \times (k+l)}$\end{document}$ is the concatenation of their columns. $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$C_{a,b} \in \mathbb {F}^{n\times (b-a)}$\end{document}$ denotes a matrix constructed by concatenating columns *c*~*a*~,*c*~*a*+ 1~,...,*c*~*b*~ of matrix $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$C \in \mathbb {F}^{n \times m}$\end{document}$. In this paper we will consider the unweighted, directed graph *G* unless stated otherwise. We will denote by *V* (*G*) and *E*(*G*) the sets of vertices and edges of graph *G*. The adjacency matrix *A*(*G*) ∈{0,1}^*n*×*n*^ of a directed graph *G* with *n* vertices is defined as: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A(G)_{i,j} = \left\{\begin{array}{lll} 1 & \text{if} i \rightarrow j \in E(G) \\ 0 & \text{if} i \rightarrow j \notin E(G). \end{array}\right. $$\end{document}$$ Introduction to Cyclic Subspaces and Connection to Frobenius Matrices {#Sec6} ===================================================================== We will start with a motivational example of application of cyclic subspaces. We have a constant-recursive sequence $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$a_n = c_0 a_{n-1} + c_1 a_{n-2} + {\ldots} + c_{r-1} a_{n-r} $$\end{document}$$ of order *r*, where all *c*~*i*~ are constants and initial conditions *a*~0~,...,*a*~*r*− 1~ are given. Perhaps, the most familiar example is the Fibonacci sequence *F*~*n*~ = *F*~*n*− 1~ + *F*~*n*− 2~. We can define the *companion matrix* of our general sequence as: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C = \left[\begin{array}{llll} 0&\ldots&0 & -c_0\\ 1&\ddots& & -c_1\\ &\ddots& 0&{\vdots} \\ 0& & 1&-c_{r-1} \end{array}\right]. $$\end{document}$$ The crucial property of the matrix *C* is that we can generate the next element of sequence with multiplication by *C*: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C^T \left[\begin{array}{lll} a_{n-r} \\ {\vdots} \\ a_{n-2} \\ a_{n-1} \end{array}\right] = \left[\begin{array}{lll} a_{n-r + 1} \\ {\vdots} \\ a_{n-1} \\ a_{n} \end{array}\right]. $$\end{document}$$ For example, for Fibonacci sequence we have: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left[\begin{array}{lll} 0 & 1 \\ 1 & 1 \end{array}\right] \left[\begin{array}{lll} F_k \\ F_{k + 1} \end{array}\right] = \left[\begin{array}{lll} F_{k + 1} \\ F_{k + 2} \end{array}\right]. $$\end{document}$$ Now, if we want the *a*~*n*+ 1~ element of the series we can square the companion matrix and get: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C^T C^T \left[\begin{array}{lll} a_{n-r} \\ {\vdots} \\ a_{n-2} \\ a_{n-1} \end{array}\right] = C^T \left[\begin{array}{lll} a_{n-r + 1} \\ {\vdots} \\ a_{n-1} \\ a_{n} \end{array}\right] = \left[\begin{array}{lll} a_{n-r + 2} \\ {\vdots} \\ a_{n} \\ a_{n + 1} \end{array}\right]. $$\end{document}$$ And analogously to get the *a*~*n*\ +\ *k*− 1~ element we need to compute the *k*-th power of the matrix *C*. In linear algebra such transformations are known as the *cyclic subspaces* generated by the vector *a*. For our purposes, we will restrict ourself to a conclusion, that some columns of a companion matrix occur in its powers. These properties are well known in the linear algebra theory (see \[[@CR10], [@CR13]\] for more cyclic properties). Consequences of Frobenius Normal Form {#Sec7} ------------------------------------- Let $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {F}$\end{document}$ be a commutative field. For any matrix $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A \in \mathbb {F}^{n\times n}$\end{document}$ there exists an invertible *U* over $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {F}$\end{document}$ such that: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$U^{-1} A U = F = \left[\begin{array}{lllll} C_1 & & & & 0 \\ & C_2 & & \\ & & C_3 & & \\ & & & {\ddots} & \\ 0 & & & & C_k \end{array}\right]. $$\end{document}$$and *F* is the Frobenius-canonical-form[2](#Fn2){ref-type="fn"} of *A*. The diagonal block *C*~*i*~ is called the companion matrix: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C_i = \left[\begin{array}{lllllll} 0& &{\ldots} & &0 & -c_0\\ 1&0& & &0 & -c_1\\ &1&{\ddots} & &{\vdots} & -c_2\\ & &{\ddots} &0& & {\vdots} \\ & & &1&0 & -c_{r-2}\\ 0& & & &1 & -c_{r-1} \end{array}\right] \in \mathbb{F}^{r\times r} . $$\end{document}$$ Each companion matrix corresponds to the monic polynomial $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$C_{i}(x) = x^{r} + c_{r-1}x^{r-1} + {\ldots } + c_{0} \in \mathbb {F}\left [x\right ]$\end{document}$ (similarly to the sequence example) and this polynomial is called the *minimal polynomial* of *A*. Each minimal polynomial has a property that *C*~*i*~(*A*) = 0. To guarantee that matrix has only one decomposition into Frobenius normal form we require that every polynomial must divide the next one, i.e., *C*~*i*~(*x*)\|*C*~*i*+ 1~(*x*). The final list of polynomials is called the *invariant factors* of matrix *A* \[[@CR25]\]. Storjohann \[[@CR25]\] proposed the deterministic algorithm to compute the Frobenius canonical-form efficiently. ### **Theorem 2** {#FPar2} \[[@CR25]\] *The Frobenius canonical-form of a matrix can be computeddeterministically using*$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$*fieldoperations.* Moreover, there are also probabilistic algorithms that compute this form in expected $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time over small fields \[[@CR10]\]. In this paper, all algorithms are deterministic if we the upper bound on the number of distinct walks is *W*. Then, due to the time of a single field operation we need additional *O*(log *W*) factor in the complexity. However, since we are mainly interested in determining if a cycle/walk of a given length exists in a graph, we can set a sufficiently small field $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {Z}_{p}$\end{document}$ (*p* has *O*(log *n*) bits). This way when algorithm returns nonzero we are sure that there exists some walk. If algorithm returns zero, then with high probability there will be no such walk. Cyclic Subspaces {#Sec8} ---------------- Frobenius decomposition can be used to get the desired power of a matrix (analogously to the diagonal decomposition): $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A^k = (U F U^{-1})^k = U F (U^{-1} U) F {\cdots} F (U^{-1} U) F U^{-1} = U F^k U^{-1}. $$\end{document}$$ Moreover, we will use the property that the power of block diagonal matrix *F* is block diagonal: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F^k = \left[\begin{array}{lllll} C_1^k & & & & 0 \\ & C_2^k & & \\ & & C_3^k & & \\ & & & {\ddots} & \\ 0 & & & & C_l^k \end{array}\right] .$$\end{document}$$ Now, we need a property of companion matrices that will enable us to power them efficiently (Fig. [1](#Fig1){ref-type="fig"}). Fig. 1Visualisation of the cyclic property (Definition 1) ### **Definition 1** (Cyclic Property) {#FPar3} Let *v*~1~,...,*v*~*n*~ be the columns of a matrix $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$C \in \mathbb {F}^{n\times n}$\end{document}$. Let *v*~*n*+ 1~,...,*v*~2*n*~ be the columns of matrix *C*^*n*+ 1^. If, for every 1 ≤ *k* ≤ *n* the columns of matrix *C*^*k*^ are *v*~*k*~,*v*~*k*+ 1~,...,*v*~*k*\ +\ *n*~ then the C has a **cyclic property**. It turns out, that companion matrices have a cyclic property. ### **Theorem 3** (Folklore [@CR13], see \[[@CR17]\] for generalization) {#FPar4} *Every companion matrix has a cyclic property.* We will illustrate this property with an example: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C = \left[\begin{array}{lllll} 0 & 0 & 0 & 0 & 1 \\ 1 & 0 & 0 & 0 & 2 \\ 0 & 1 & 0 & 0 & 3 \\ 0 & 0 & 1 & 0 & 4 \\ 0 & 0 & 0 & 1 & 5 \end{array}\right], C^{2} = \left[\begin{array}{lllll} 0 & 0 & 0 & 1 & 5 \\ 0 & 0 & 0 & 2 & 11 \\ 1 & 0 & 0 & 3 & 17 \\ 0 & 1 & 0 & 4 & 23 \\ 0 & 0 & 1 & 5 & 29 \end{array}\right], C^{3} = \left[\begin{array}{lllll} 0 & 0 & 1 & 5 & 29 \\ 0 & 0 & 2 & 11 & 63 \\ 0 & 0 & 3 & 17 & 98 \\ 1 & 0 & 4 & 23 & 133 \\ 0 & 1 & 5 & 29 & 168 \end{array}\right], $$\end{document}$$$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C^{4} = \left[\begin{array}{lllll} 0 & 1 & 5 & 29 & 168 \\ 0 & 2 & 11& 63& 365 \\ 0 & 3 & 17& 98& 567 \\ 0 & 4 & 23& 13& 770 \\ 1 & 5 & 29& 16& 973 \end{array}\right], C^{5} = \left[\begin{array}{lllll} 1 & 5 & 29 & 168 & 973 \\ 2 & 11 & 63 & 365 & 2114 \\ 3 & 17 & 98 & 567 & 3284 \\ 4 & 23 & 133 & 770 & 4459 \\ 5 & 29 & 168 & 973 & 5635 \end{array}\right] C^{6} = \left[\begin{array}{lllll} 5 & 29 & 168 & 973 & 5636 \\ 11 & 63 & 365 & 2114 & 12243 \\ 17 & 98 & 567 & 3284 & 19019 \\ 23 & 133 & 770 & 4459 & 25824 \\ 29 & 168 & 973 & 5635 & 32634 \end{array}\right] $$\end{document}$$ The matrix *C*^*i*^ has 4 columns identical to matrix *C*^*i*+ 1^. *C* has coefficients of order equal to dimension (dimension is 5 and maximum coefficient is 5). After powering to the 5th power, the coefficients can be of order 5^5^. Over a finite field $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {Z}_{p}$\end{document}$, all those coefficients will have *O*(log *p*) bits. Matching Distance Queries on Directed Unweighted Graphs {#Sec9} ======================================================= In this section, we will present an algorithm that matches the best known upper bounds of \[[@CR33]\] for distance queries in directed unweighted graphs and uses Frobenius matrices. We take the adjacency matrix *A* of a graph *G* (i.e., *n* × *n* matrix with *a*~*u*,\ *v*~ = 1 when (*u*, *v*) ∈ *G* and 0 otherwise). The *k*-th power of the adjacency matrix of the graph *G* holds the number of walks, i.e., an *a*~*u*,\ *v*~ element of *A*^*k*^ is *the count of distinct walks fromutovof lengthkin the graph*. **Observation 1** (Folklore, \[[@CR8]\]) {#FPar5} ---------------------------------------- Let *A* ∈{0,1}^*n*×*n*^ be the adjacency matrix of a directed graph G. The (*A*^*k*^)~*u*,\ *v*~ is the number of distinct walks from u to v of length exactly k in the graph G. Hence, the shortest path between vertices *u*, *v* is the smallest *k* such that *A*^*k*^ has nonzero element *a*~*u*,\ *v*~. This will allow us to forget about graph theory interpretation for a brief moment and focus only on finding such *k* with algebraic tools. In this section we will proof the following Lemma. **Lemma 1** {#FPar6} ----------- *Given a matrix*$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$A \in \mathbb {F}^{n \times n}$\end{document}$*.There exists an algorithm that after some preprocessing, can answer queries for any givenpair of indicesi*, *j* ∈{1,...,*n*} *and integerk* ∈{1,...,*n*}, *such that:query returns an element* (*A*^*k*^)~*i*,\ *j*~,*preprocessing takes*$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$*fieldoperations and query takesO*(*n*) *field operations.* *The algorithm is deterministic.* To proof this Lemma we decompose matrix *A* into the Frobenius normal form. Storjohann \[[@CR25]\] showed an algorithm that returns *U* and *F* deterministically in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ field operations (note that matrix inverse can also be computed in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ field operations). To better explain the idea, in the next section we will consider a simple case when a number of invariant factors of *A* is exactly 1. Then in Section [4.2](#Sec11){ref-type="sec"} we will show how to generalize it to multiple invariant factors. Single Invariant Factor {#Sec10} ----------------------- In that situation, the matrix *F* is a companion matrix $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$C \in \mathbb {F}^{n\times n}$\end{document}$. First, we compute the (*n* + 1)-th power of the companion matrix *F*^*n*+ 1^. This can be done by using $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ field operations by repeated squaring (compute *F*, *F*^2^,*F*^4^,...,*F*^*n*+ 1^ with *O*(log *n*) matrix multiplications). Let *v*~1~,...,*v*~*n*~ be the columns of matrix *UF* and *v*~*n*+ 1~,...,*v*~2*n*~ be the columns of matrix *UF*^*n*+ 1^. Note, that because matrix *F* has a cyclic property, the columns *v*~*k*~,...,*v*~*k*\ +\ *n*− 1~ construct *UF*^*k*^ (see Fig. [2](#Fig2){ref-type="fig"}). Fig. 2Construction of *UF*^*k*^ from matrices *UF* and *UF*^*n*+ 1^ This step took just two matrix multiplications, because we need to multiply *U* times *F* and *F*^*n*+ 1^. The preprocessing phase takes only $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ field operations. Now, if a query asks for the number of distinct walks from vertices *u* to *v* of length exactly *k* we: select *u*-th row of matrix *UF*^*k*^ (*n* numbers),select *v*-th column of matrix *U*^− 1^,multiply them in by using *O*(*n*) multiplications (dot product of two *n*-dimensional vectors). This will give us the *u*, *v* element of matrix *UF*^*k*^*U*^− 1^ = *A*^*k*^. To get the length of the shortest path (i.e., the minimal *k* such that (*A*^*k*^)~*u*,\ *v*~ \> 0), we will modify our matrices slightly to get the number of walks of length ≤ *k*. At the end, we will use in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n)$\end{document}$ query tim (by using binary search) and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ preprocessing time. Basically, for a given *k* we need to get the *u*, *v* element of matrix *A* + *A*^2^ + ⋯ + *A*^*k*^. It suffices to add consecutive columns of matrix *UF* ⊕ *UF*^*n*+ 1^ = *v*~1~ ⊕ *v*~2~ ⊕... ⊕ *v*~2*n*~ in the following manner:[3](#Fn3){ref-type="fn"}$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$M^{\prime} = \left[\begin{array}{ccccccc} v_1 & v_1 + v_2 & v_1 + v_2 + v_3 & {\ldots} & {\sum}^k_{i = 1} v_i & {\ldots} & {\sum}^{2n}_{i = 1} v_i \end{array}\right] \in \mathbb{F}^{n \times 2n} .$$\end{document}$$ Now, to get *A* + *A*^2^ + ⋯ + *A*^*k*^ one would need to multiply $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$M^{\prime }_{k,k+n-1} U^{-1}$\end{document}$ and subtract $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$M^{\prime }_{1,n} U^{-1}$\end{document}$ for a balance.[4](#Fn4){ref-type="fn"} The naive algorithm can transform matrices *U* and *F* to matrix *M*^′^ in *O*(*n*^2^) field operations during preprocessing. During query, we will need to compute two dot products (*u*-th row of $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$M^{\prime }_{k,k+n-1}$\end{document}$ times *v*-th column of *U*^− 1^ and *u*-th row of $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$M^{\prime }_{1,n}$\end{document}$ times *v*-th column of *U*^− 1^) and subtract them. We have an algorithm that for a given vertices *u*, *v* ∈ *G* and integer *k* ∈{1,...,*n*} can answer: *how many walks fromutovof length less or equalkare in the graphG* in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n)$\end{document}$ query time and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ preprocessing time. Because the result of the query is increasing in *k* we can use binary search. We can determine the first *k* for which the query will answer nonzero value in *O*(log *n*) tries. Hence, in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n)$\end{document}$ we can find the length of the shortest path. This generalized query can also return the number of walks of length exactly *k*, i.e., *q*(*u*, *v*, *k*) − *q*(*u*, *v*, *k* − 1). We matched the result of \[[@CR33]\] for unweighted graphs with a single invariant factor. In the next section, we will show how to generalize this technique for graphs with any number of invariant factors. Multiple Invariant Factors {#Sec11} -------------------------- Now, we will consider a case when *k* ≥ 1, i.e., matrix *F* has multiple invariant factors. First of all, we need to note that this generalization is not perfect and will allow to compute the number of walks of length up to *D* (the longest distance in a graph, i.e., diameter). In a real world applications of our framework (detecting cycles, determining distance between nodes, etc.) one does not need to consider walks longer than the longest possible distance in a graph. It is natural that the diameter is considered to be a bound of an output in graph problems \[[@CR1], [@CR2], [@CR7], [@CR9]\]. ### Relation of the Graph Diameter and Frobenius Normal Form {#Sec12} We begin with relating the graph diameter to the Frobenius normal form. It turns out that the graph diameter is bounded by the degree of a smallest invariant factor. #### **Lemma 2** {#FPar7} \[[@CR8]\] *Given a directed, unweighted graph* G *with a diameter* D*. Letμdenote the degree of the smallest invariant factor (i.e.,the dimension of the smallest block in the Frobeniusmatrix* F*) of an adjacency matrix of the graph* G*. ThenD* ≤ *μ.* This theorem is well known in literature \[[@CR8]\]. We include the proof of this theorem for completeness. #### Proof {#FPar8} For a contradiction assume that *D* \> *μ* and let *u*, *v* ∈ *G* be vertices such that *δ*(*u*, *v*) = *D*. Let *A* be the adjacency matrix of *G*. We know, that there is the minimal polynomial of degree *μ*: Term $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$a^{k}_{i,j}$\end{document}$ denotes the *i*, *j* element of the matrix *A*^*k*^. Now, consider the elements *u*, *v* of each matrix. The diameter *D* \> *μ* and *δ*(*u*, *v*) = *D*, so for every *k* ≤ *μ* the elements $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$a^{k}_{u,v} = 0$\end{document}$ (because there is no walk of length less than *D* from *u* to *v*). Now, if we multiply the minimal polynomial by the matrix *A* we get: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A^{\mu+ 1} = a_0 A + a_1 A^2 + a_2 A^3 + {\ldots} + a_{\mu-1} A^{\mu} . $$\end{document}$$ Hence $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$a^{\mu + 1}_{u,v} = 0$\end{document}$, because every element in the sequence $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$a^{k}_{u,v} = 0$\end{document}$ for *k* ≤ *μ*. By repeating this reasoning, we get that for every *k* \> 0 the element $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$a^{k}_{u,v} = 0$\end{document}$. So, for every achievable pair of vertices, there must be some *k* ≤ *μ*, such that $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$a^{k}_{u,v} \ne 0$\end{document}$ and diameter is bounded by *μ*. □ The bounds of this inequality are tight. There are graphs with diameter *D* = *μ* and graphs with *μ* = *n* and arbitrary small diameter \[[@CR8]\]. Our algorithms are able to return walks up to the length *μ*. We use the bound on *D* solely because it is easier to interpret diameter than the *smallest degree of the invariant factor*. ### Generalization to Multiple Invariant Factors {#Sec13} Let *k* denote the number of blocks in the Frobenius matrix *F* and *μ* be the number of columns of the smallest block. To multiply the matrix *U* by *F* we can start by multiplying strips of matrix *U* by appropriate blocks of *F* and concatenate them later (see Fig. [3](#Fig3){ref-type="fig"}). Fig. 3Multiplication of the *UFU*^− 1^. Example for 3 invariant factors. We know that matrix *F* is block and consists of *F*~1~,*F*~2~,*F*~3~. We divide matrix *U* into strips *U*~1~,*U*~2~,*U*~3~ that corespond to blocks of *F*. The observation is that we can compute *U*~1~ ⋅ *F*~1~, *U*~2~ ⋅ *F*~2~ and *U*~3~ ⋅ *F*~3~ independently and concatenate them into matrix *UF* We start by splitting the matrix *U* into *k* strips with rows corresponding to the appropriate blocks of *F* (strip *U*~*i*~ has as many columns as block *F*~*i*~). Then we multiply *UF* and have *k* strips: *U*~1~*F*~1~,*U*~2~*F*~2~,...*U*~*k*~*F*~*k*~ (each with at least *μ* columns). Next, we multiply *UF*^*μ*^ and also keep *k* strips: $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$U_{1} F_{1}^{\mu }, U_{2} F_{2}^{\mu },\ldots ,U_{k} F_{k}^{\mu }$\end{document}$. Our goal is to get a data structure such that if we need *UF*^*k*^, we can quickly choose appropriate columns and append them. The matrix *U*~*i*~*F*~*i*~ has *l*~*i*~ columns: $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$v_{1},\ldots , v_{l_{i}}$\end{document}$. Because *F*~*i*~ is a companion matrix, the $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$U_{i} F_{i}^{\mu }$\end{document}$ has the cyclic property (Definition 1). And the matrix $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$U_{i} F_{i}^{\mu }$\end{document}$ has columns: $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$v_{\mu },\ldots , v_{\mu +l_{i}}$\end{document}$. Note, that there are some duplicate columns in *U*~*i*~*F*~*i*~ and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$U_{i} F_{i}^{\mu }$\end{document}$, when *μ* \< *l*~*i*~. Hence, we only need to keep columns $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$v_{1},\ldots ,v_{\mu +l_{i}}$\end{document}$ for this strip. We do this for all strips *U*~1~*F*~1~,...*U*~*k*~*F*~*k*~ (see Fig. [4](#Fig4){ref-type="fig"}). Fig. 4Combining strips into a single matrix. The height of the matrix in the schema is scalled down. We computed *U*~1~*F*~1~ and $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$U_{1} F_{1}^{\mu }$\end{document}$. Now we noted that companion matrices have a cyclic property so some of the rows in the strips are repeated. So in the single strip we can store only subsequent columns We are left with a matrix that has at most 2*n* columns (because $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$l_{1} + \mu + l_{2} + \mu + {\ldots } l_{k} + \mu = k \mu + {\sum }^{n}_{i = 1} l_{i} = n + k\mu \le 2n$\end{document}$). To generate it we need to power *F* to *μ* and do multiplications *U* ⋅ *F* and *U* ⋅ *F*^*μ*^. This can be computed in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ field operations via fast matrix multiplication and repeated squaring. ### Queries with Multiple Invariant Factors {#Sec14} When a query for the number of walks of length *k* from node *u* to *v* comes, we do: For each strip *i* take the *u*-th row of $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$U_{i} F_{i} \oplus U_{i} F_{i}^{\mu }$\end{document}$ and concatenate them (see Fig. [5](#Fig5){ref-type="fig"}) into vector $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\bar {u}$\end{document}$,Take *v*-th column of *U*^− 1^ matrix and denote it $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\bar {v}$\end{document}$,Return the dot product $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\bar {u} \cdot \bar {v}$\end{document}$.Fig. 5Schema of obtaining vector $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\bar {u}$\end{document}$ (marked red) from 3 strips. We are given the row number and the power *k* and the lenghts *l*~*i*~ of each strip. At the end we concatenate them Because *l*~1~ + *l*~2~ + ... + *l*~*k*~ = *n* the vector $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\bar {u} \in \mathbb {F}^{n}$\end{document}$. Query needs *O*(*n*) field operations. Finally, this dot product is *a*~*u*,\ *v*~ element of the matrix *UF*^*k*^*U*^− 1^, for a fixed *k* ≤ *μ* because $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\bar {u}$\end{document}$ is the concatenation of original vector *u*. Analogously to Section [4.1](#Sec10){ref-type="sec"} one can extend this result to return the number of walks of length less or equal *k*. This matches (up to the polylogarithmic factor) the result of \[[@CR33]\]. We will omit the details of this observation because in the next section we will extend this framework even further. Almost Optimal Query {#Sec15} ==================== In the previous section, we showed how to preprocess a matrix *A* with $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ field operations in such a way that in query that uses *O*(*n*) field operations we can return a number (*A*^*k*^)~*i*,\ *j*~. However, in linear time *O*(*n*) we return only a single number. The goal of this section is to get far richer information in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n)$\end{document}$ query time (with extra factors from field operations). **Theorem 4** {#FPar9} ------------- *LetA* ∈{0,1}^*n*×*n*^*be a matrix such that the degree of smallest invariant factor isμ. Thereexists a deterministic algorithm that after some preprocessing can answer queries for anygiveni*, *j* ∈{1,...,*n*}*:Query returns* {*a*~*k*~\|1 ≤ *k* ≤ *μ*},*wherea*~*k*~ = (*A*^*k*^)~*i*,\ *j*~ ,*Preprocessing takes* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega }\log {W})$\end{document}$ *and* *query takes* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n\log {W})$\end{document}$ *time,* *where* W *is an upper bound ona*~*k*~*for allk* ∈{1,...,*μ*}*.* Note, that this theorem has some immediate application in graph algorithms (see Section [6](#Sec18){ref-type="sec"}). Hankel Matrix {#Sec16} ------------- Now, we will focus on the proof of Theorem 4. First, we need to introduce the Hankel matrix and its properties. $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$H = \left( \begin{array}{cccccc} c_1 & c_2 & {\ldots} & c_n \\ c_2 & c_3 & {\ldots} & c_{n + 1} \\ {\vdots} & {\vdots} & & {\vdots} \\ c_n & c_{n + 1} & {\ldots} & c_{2n-1} \end{array}\right) $$\end{document}$$ Hankel matrix is defined by its first row and last column (2*n* − 1 numbers define *n* × *n* Hankel matrix). The numbers from the previous row are left-shifted by one and the new number is added at the end. Hankel matrices have some similarities to Topelitz and Circulant matrices. The basic property we need is that the product of Hankel matrix and vector can be computed in *O*(*n* log *n*) time (see \[[@CR15], [@CR26]\]) even though explicitly writing the Hankel matrix as *n* × *n* matrix takes *O*(*n*^2^) time. The algorithm takes 2*n* − 1 parameters that define the Hankel matrix and *n* parameters that define the vector. The technique is based on the Fast Fourier Transformation \[[@CR15], [@CR26]\]. Using Hankel Matrices to Get Richer Query {#Sec17} ----------------------------------------- To proof the Theorem 4 we will modify only the last step in Section [4.2.3](#Sec14){ref-type="sec"}. The algorithm from Section [4.2.3](#Sec14){ref-type="sec"} concatenates the strips *U*~*i*~*F*~*i*~ and builds a single vector. Subsequently, that vector is multiplied by a column of matrix *U*^− 1^. But we can also do it in a different order: first we multiply the strip by a section of matrix *U*^− 1^ and sum the results at the end. Thus, we perform *k* (number of strips) dot products of smaller vectors (see Fig. [6](#Fig6){ref-type="fig"}). Fig. 6Multiplication of strips by *U*^− 1^ matrix. As you can see, matrix *U*^− 1^ can be splited into sections, that multiply only *U*~*i*~*F*~*i*~ strips Consider the query for a number of walks of length exactly *k*. The strips in the matrix *U*^− 1^ do not depend on *k* (vector (*u*~0~,...,*u*~*l*~)). However, the vector taken from *U*~*i*~*F*~*i*~ (vectors (*x*~*i*~,...,*x*~*i*\ +\ *l*~)) will be left shifted if we want to compute the next one. $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{array}{ccccccccr} (x_0 & x_1 & x_2 & {\ldots} & x_l )\\ (x_1 & x_2 & {\ldots} & x_l & x_{l + 1} )\\ (x_2 & {\ldots} & x_l & x_{l + 1} & x_{l + 2})\\ {\vdots} & & & & {\vdots} \\ (x_{\mu} & &{\ldots} & & x_{\mu+l}) \end{array} \times \left( \begin{array}{cc}{u_0}\\{\vdots}\\{u_l} \end{array}\right) $$\end{document}$$ As you can see, the subsequent rows can be written as the Hankel matrix (we need to add zeros to get a square matrix, but it will not influence asymptotic complexity since there will be at most *O*(*n*) of them). By using the *fast Hankel matrix-vector multiplication* we can compute *μ* values for every strip *i* in time *O*(*l*~*i*~ log *l*~*i*~) (*l*~*i*~ was defined as the length of *i*-th strip). At the end, we need to sum all results into a single array. Therefore, the number of operations is $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$O\left ({\sum }_{i = 1}^{k} l_{i} \log l_{i} \right )$\end{document}$. Because $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${\sum }^{k}_{i = 1} l_{i} = n$\end{document}$ the algorithm needs *O*(*n* log *n*) field operations. This proves Theorem 4. Here, we silently assumed that the number of walks is bounded by *W*. Note, that for large *W*, the algorithm needs to output *O*(*n*^2^ log *W*) bits and the complexity of every arithmetic operation needs to be multiplied by log *W*. If one is only interested in the deciding if an entry of some power of adjacency matrix is nonzero, we can use a standard randomization technique to eliminate log *W* factors from the running time. ### **Corollary 1** {#FPar10} *LetA* ∈{0,1}^*n*×*n*^*be a matrix such that the degree of smallest invariant factor isμ. Thereexists an algorithm that after some preprocessing can answer queries for any giveni*, *j* ∈{1,...,*n*}*:Query returns* {*a*~*k*~\|1 ≤ *k* ≤ *μ*},*wherea*~*k*~ = 1 *if* (*A*^*k*^)~*i*,\ *j*~*is nonzero,Preprocessing takes* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ *and* *query takes* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n)$\end{document}$ *time.* *The algorithm is randomized with one-sided bounded error.* ### Proof {#FPar11} At the beginning we will randomly select a prime number with *O*(log *n*) bits. We can write the matrix A as a polynomial: $$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tilde{A}_{i,j} = \left\{\begin{array}{lll} x_{i,j} & \text{if} A_{i,j} = 1 \\ 0 & \text{otherwise}, \end{array}\right. $$\end{document}$$where *x*~*i*,\ *j*~ are unique variables. Now we can apply Schwartz-Zippel Lemma \[[@CR23], [@CR34]\]. The $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\tilde {A}^{k}_{i,j}$\end{document}$ is the polynomial of degree at most *n* (because *k* ≤ *μ* ≤ *n*) and if we compute it over $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\mathbb {Z}_{p}$\end{document}$ for *p* = *O*(*n*^2^) we can correctly determine if $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\tilde {A}^{k}_{i,j}$\end{document}$ is a nonzero polynomial with a constant probability. We can repeat the above procedure *O*(log *n*) times to get a correct result in all entries with high probability. □ Applications {#Sec18} ============ In this section we will show how to use Theorem 4 to improve known algorithms on graphs. First we will develop a data structure that returns a number of distinct walks efficiently. **Lemma 3** {#FPar12} ----------- *LetG* = (*V*, *E*) *be a directed, unweighted graph with* n *vertices and a diameter* D*. There exists andeterministic algorithm that after some preprocessing can answer queries for any givenu*, *v* ∈ *V:Query returns* {*w*~*i*~\|1 ≤ *i* ≤ *D*},*wherew*~*i*~*is the number of distinct walks from* u *to* v *of length exactly* i,*Preprocessing takes* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega }\log {W})$\end{document}$ *and* *query takes* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n\log {W})$\end{document}$ *field* *operations.whereWis the upper bound onw*~*i*~*for alli* ∈{1,...,*D*}. Proof {#FPar13} ----- We encode the graph *G* as an adjacency matrix *A*(*G*). We use the Theorem 4 to construct the data structure that given a query (*u*, *v*) outputs (*A*^*k*^)~*u*,\ *v*~ for all 1 ≤ *k* ≤ *μ*. Finally, we use Observation 1 to note, that (*A*^*k*^)~*u*,\ *v*~ is equal to the number of distinct walks from u to v of length exactly k. Moreover we use Lemma 2 to bound the number *D* ≤ *μ*, so we will always output more numbers (but we can truncate them in *O*(*n*) time). Finally we note, that the preprocessing and query of Theorem 4 matches the statement and construction of adjacency matrix is *O*(*n*^2^). □ This algorithm is a significant improvement over \[[@CR33]\]: One can use Lemma 3 to find the distance between *u*, *v* by linearly scanning the array and returning the first *k* such that *w*~*k*~ \> 0,Lemma 3 can count cycles. In contrast the \[[@CR33]\] cannot, because the distance from *u* to *u* is always 0 (see Section [6.1](#Sec19){ref-type="sec"}),Lemma 3 is almost optimal, i.e., when *D* = *O*(*n*) then query will need to output *O*(*n* log *W*) bits. From the other hand, Lemma 3 is merely a functional improvement and it does not break the $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ of the *Single Source Shortest Paths* (SSSP) for dense, directed graphs. Now we will show the application of Lemma 3. We begin with almost optimal algorithm to compute the number of all walks between all pairs of vertices. We are not aware of any other research concerning this problem. **Definition 2** (All-Pairs All Walk problem) {#FPar14} --------------------------------------------- Given a directed, unweighted graph *G* with a diameter *D*. The task is to return an array *A*, such that for every pair of vertices *u*, *v* ∈ *G* and every *k* ∈{1,...,*D*} an element *A*\[*u*, *v*, *k*\] is the number of distinct walks from *u* to *v* of length *k*. The folklore solution to this needs *O*(*Dn*^*ω*^ log *W*) time (where *W* is an upper bound on number of walks) and works as follows: *take the adjacency matrixAof graphGand save it inA*\[*u*, *v*, 1\]. *Then, square it to getA*^2^*and save it inA*\[*u*, *v*, 2\]. *Continue until you fill out complete table*. In the worst case this algorithm needs *D* = *O*(*n*) matrix multiplications, thus it needs *O*(*Dn*^*ω*^) field operations. At the first glance it is surprising that we can improve it to $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{3})$\end{document}$ field operations. **Theorem 5** {#FPar15} ------------- *All-Pairs All Walk problem admits an*$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{3} \log {W})$\end{document}$*algorithm(where* W *is upper bound on number of walks between every pair of vertices).* Proof {#FPar16} ----- We will apply the Lemma 3 algorithm. The preprocessing takes $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time. Then, for every pair of vertices *u*, *v* ask a query. A single query takes $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n \log {W})$\end{document}$ time. Next we will save it in the table *A*\[*u*, *v*\] (query gives D numbers *w*~1~,...,*w*~*D*~, such that *w*~*i*~ is the number of walks of length i and save it *A*\[*u*, *v*, *i*\] := *w*~*i*~). Because there are *O*(*n*^2^) pairs and for each pair we need $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n\log {W})$\end{document}$ time, the complexity of our solution is $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{3}\log {W})$\end{document}$. The algorithm is almost optimal because the output in the worst case may be *O*(*n*^3^ log *W*) (we may need *O*(log *W*) bits to encode a single entry in the table). □ Counting and Determining the Lengths of Cycles {#Sec19} ---------------------------------------------- We will use Theorem 4 to solve All-Nodes Shortest Cycle (ANSC) problem efficiently. ### **Definition 3** (All-Nodes Shortest Cycles [@CR31]) {#FPar17} Given a directed, unweighted graph *G*. The problem *All-Nodes ShortestCycle* asks to output for every vertex *v* the length of the shortest cycle that contains *v*. ### **Lemma 4** {#FPar18} *There exists a deterministic algorithm that for a given unweighted, directed* *G* *with a diameter* *D:* *For every vertexureturnsDnumbers:*$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${c_{u}^{1}}, {c_{u}^{2}}, \ldots {c_{u}^{D}}$\end{document}$,*where*$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${c_{u}^{k}}$\end{document}$*isthe number of non-simple cycles of length exactlyk, that contain vertex u*,*Algorithm works in* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega } \log {W})$\end{document}$ *time* *(where* *W* *is an upper bound on* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${c^{k}_{u}}$\end{document}$ *).* ### Proof {#FPar19} We will use Theorem 4. We start by preprocessing the graph *G* in time $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega }\log {W})$\end{document}$. Theorem 4 allows us to ask for a number of walks from *u* to *v* and receive *D* numbers: $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$w_{u,v}^{k}$\end{document}$. So, we ask for the number of walks from vertex *u* to the same vertex *u*. This is exactly the number of non-simple cycles of given length that contain vertex *u*. Because we need to ask only *n* queries (it is the number of vertices in a graph) and each query takes $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n\log {W})$\end{document}$ time we have $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega }\log {W} + n^{2}\log {W})$\end{document}$ = $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega }\log {W})$\end{document}$ algorithm. □ If we are only interested in deciding if the numbers $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${c_{u}^{i}}$\end{document}$ are nonzero, instead of Theorem 4 we can use Corollary 1. It introduces the one-sided randomization but allows us to shave log *W* factors in the running time. ### **Corollary 2** {#FPar20} *There exists a randomized algorithm that for a given unweighted, directed* *G* *with* *a diameter* *D* *:* *For every vertexureturnsDnumbers:*$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${c_{u}^{1}}, {c_{u}^{2}}, \ldots {c_{u}^{D}}$\end{document}$, *where*$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${c_{u}^{k}}$\end{document}$*is1 if there exists a non-simple cycle of length exactlyk, that contain vertex uor 0 otherwise,Algorithm works in* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ *time* *with one sided bounded error.* Now we will show how to improve upon \[[@CR31]\] $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{(\omega + 3)/2})$\end{document}$ algorithm with Corollary 2. ### **Theorem 6** {#FPar21} *All-Nodes Shortest Cycles admits an* $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ *randomized* *time algorithm.* ### Proof {#FPar22} We use Lemma 2 to compute the table *S*\[*v*\]. For every vertex we search for the first nonzero element linearly. This with high probability is the length of the shortest cycle that contains it. Because the output contains *O*(*n*^2^) numbers the complexity is equal to the preprocessing time $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$. □ Also the Corollary 2 improves upon \[[@CR9], Theorem 45\] for unweighted graphs. ### **Corollary 3** {#FPar23} *Given a directed, unweighted graph* G *with a diameter* D*. LetS*(*c*) *denote the set of vertices that lie in the cycle of length exactly* c*.In*$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$*timewe can return the setsS*(1),...,*S*(*D*) *with constant probability of success.* ### Proof {#FPar24} Similarly to the proof of Theorem 6, we can scan the output to compute the set *S*(*c*) that contains all vertices that lie on some cycle of length ≤ *c*. Then, by linear scan we can return the sets *S*(1),...,*S*(*D*). □ Conclusion and Open Problems {#Sec20} ============================ We introduced the framework based on Frobenius normal form and used it to solve some problems on directed, unweighted graphs in matrix multiplication time. The main open question is to use this framework to prove that APSP on such graphs can be solved in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ or at least *O*(*n*^2.5^). The promising way is to use the algorithms that determine operators of matrices of polynomials (e.g., determinant, solving linear system \[[@CR14], [@CR18]\]). Additionally, algorithms for a black-box polynomial degree determination seem to be a promising path. Another interesting problem is to use this framework to obtain additive approximation for APSP. Currently, the best additive approximation of APSP is due to \[[@CR20]\]. However, no additive approximation of APSP is known that would work in $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(n^{\omega })$\end{document}$ time. Application in dynamic algorithm also seems to be a promising approach. Frandsen and Sankowski \[[@CR12]\] showed an algorithm, that dynamically preserves Frobenius normal form in *O*(*kn*^2^) time. Our algorithms use fast Hankel matrix-vector multiplication that is based on FFT. Reif and Tate \[[@CR19]\] presented an $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$O(\sqrt {n})$\end{document}$ time per request algorithm for FFT. Can we use these approaches to obtain a faster dynamic algorithm? Finally, it remains open how to apply the Frobenius normal form in the weighted directed graphs with small, integer weights {−*M*,...,*M*}. Cygan et al. \[[@CR9]\] took degree *M* polynomials and used \[[@CR18]\] algorithms to get $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\widetilde {O}(M n^{\omega })$\end{document}$ time radius and diameter detection. We suspect that similar technique can be applied to Frobenius normal form framework. Tr \[*A*\] denotes the trace of a matrix *A*, i.e., the sum of elements on the main diagonal. Sometimes this form is called rational-canonical form. Operation ⊕ denotes concatenation. *X*~*a*,\ *b*~ denotes a matrix constructed by concatenating columns *x*~*a*~,*x*~*a*+ 1~,...,*x*~*b*~ of a matrix *X*. Extended Abstract of this paper was presented at STACS 2017 \[[@CR22]\] This article is part of the Topical Collection on *Special Issue on Theoretical Aspects of Computer Science (STACS 2017)* **Publisher's Note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This work is part of a project TOTAL that has received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (grant agreement No 677651). Piotr Sankowski is supported by the Polish National Science Center, grant no 2014/13/B/ST6/00770. Karol Węgrzycki is supported by the grants 2016/21/N/ST6/01468 and 2018/28/T/ST6/00084 of the Polish National Science Center. Part of this work was done while authors were visiting the Simons Institute for the Theory of Computing.

{ "pile_set_name": "PubMed Central" }

![](hosplond72205-0024){#sp1 .52} ![](hosplond72205-0025){#sp2 .53} [^1]: The two previous articles appeared on March 8 and 29.

{ "pile_set_name": "PubMed Central" }

Cancer of the thyroid is the most common endocrine malignancy. [@b1-asm-5-428] Most patients with well-differentiated thyroid carcinoma present with an asymptomatic thyroid nodule with or without regional lymphadenopathy or rarely with distant metastasis at diagnosis. Western series report a prevalence rate of 4% for metastasis outside the neck,[@b1-asm-5-428] but in an endemic iodine deficiency area like ours the overall prevalence of distant metastasis was 20%.[@b2-asm-5-428] In 17% of cases, metastases were detected synchronously and at multiple sites. Metastatic disease as the presenting manifestation is exceedingly rare.[@b3-asm-5-428] Patients with metastatic disease are usually asymptomatic and rarely present with organ-specific symptoms. We describe 4 patients with peculiar presentations. In 3 patients the metastatic disease presented with unusual manifestations for the first time and one had a follicular thyroid carcinoma in a struma ovarii after total thyroidectomy. Patients and Methods ==================== Fifty-five consecutive patients, 15 male (27%) and 40 female (73%) with differentiated thyroid carcinoma, who had more than two follow-up visits in the Thyroid Clinic at Nehru Hospital, Postgraduate Institute of Medical Education and Research, Chandigarh, over the last 10 years were included in this study. Their mean (±SD) duration of follow-up was 4.0±3.6 years ranging from 0.25 to 10 years. Their age ranged between 13--70 years with mean (±D) age of 39.5±13.3 years. Twenty-nine (53%) patients had multinodular goitre, 24 (44%) had a solitary thyroid nodule and one had a diffuse goitre; another had no obvious thyroid swelling clinically. All were euthyroid at diagnosis except for 5 patients (9%), 4 of whom were thyrotoxic and one hypothyroid. Total thyroidectomy was done in 46 patients (84%), hemithyroidectomy in 8 (15%) and in 1 (2%) surgery was not done due to refusal. Analysis of histopathological data revealed that 35 (64%) patients had papillary, 5 (9%) had follicular variant of papillary thyroid carcinoma and 15 (27%) had follicular carcinoma. All patients except one, after total thyroidectomy, received an ablative dose of ^131^I and were on suppressive doses of L-thyroxine. Four patients (7%) had post-operative permanent hypoparathyroidism. We describe 4 peculiar cases out of 55 who had unusual presenting manifestations of the thyroid cancer ([Table 1](#t1-asm-5-428){ref-type="table"}). Case 1 ====== A 30-year-old-man from an iodine sufficient area was admitted with sudden onset of flaccid quadriparesis. He had no previous history or family history of similar episodes. However, he had a history of loss of 10 kg weight in 4 months with increased appetite, hyperdefecation, palpitation and tremulousness. On examination, his body mass index (BMI) was 20.3 kg/m^2^ and had pulse rate of 112/min, and BP of 110/60 mm Hg. He had fine tremors, Plummer's nail and a soft, diffuse grade II goiterwith bruit and generalized, firm lymphadenopathy. His muscle bulk was decreased with a power of ⅖ in the lower limbs and ⅗ in the upper limbs, and deep tendon reflexes were diminished. Sensory examination was normal and there was no cranial nerve deficit. He had no thyroid associated ophthalmopathy or dermopathy. Other systemic examinations were normal. Laboratory studies at admission showed serum Na^+^ 140 mEq/L, K+ 2.4 mEq/L, Cl- 109 mEq/L, HCO3- 26 mEq/L, Mg^++^ 2.5 mEq/L, phosphate 3.8 mg/dL, T3 3.19 ng/mL (normal, 0.8--2.1), T4 253 ng/mL (normal, 40--120) and TSH was 0.05 μU/mL (normal, 0.3--6.0). Single random urine electrolytes were Na^+^ 78 mEq/L, K^+^ 28 mEq/L, Cl 95 mEq/L. Radioactive iodine uptake at 24 hours was 35% (normal, 15--35%) and whole body ^131^I scan revealed increased tracer uptake in the thyroid, lymph nodes and L5 vertebrae. Thyroid microsomal antibody (TMA) titres were insignificant. He was managed with parenteral potassium supplementation following which the patient's muscle power improved to normal within 8 hours. He was put on neomercazole and propranolol and after achieving euthyroidism. Total thyroidectomy with lymph node resection was performed. Histopathology confirmed follicular thyroid carcinoma with lymph node metastasis. He received an ablative dose of ^131^I and is on L-thyroxine suppressive therapy. There was no recurrence of such episode on follow-up in the last 2 years. Case 2 ====== A 55-year-old woman, resident of an iodine sufficient area, had diffuse goiter for the past 25 years. She presented with rapidly increasing swelling in the right frontal region along with an increase in size of the goiter for the last 4 years. On examination, her BMI was 20 kg/m^2^, she had a pulse rate of 80/min, and no tremors. She had diffuse, firm and globular goiter (grade IV) with multiple tattoo marks and did not have cervical lymphadenopathy. The swelling in the right frontal region (6×8×6 cm) was globular, firm and pulsatile ([Figure 1a](#f1a-asm-5-428){ref-type="fig"}) and she had another firm, pulsatile swelling in the left shoulder region. On investigation, thyroid function tests and serum biochemistry were unremarkable. An x-ray skull lateral view showed destruction of the frontal bone with adjacent soft tissue mass. CT scan head showed a solitary calvarial metastasis destroying the frontal bone with intracranial as well as extracranial extension into the adjacent soft tissue. ^131^I whole body scan showed significant increased tracer uptake in the thyroid region and skull ([Figure 1b](#f1b-asm-5-428){ref-type="fig"}). 99mTcMDP bone scan showed photopenic areas in the skull, left shoulder, multiple vertebrae and the seventh rib. She underwent total thyroidectomy uneventfully and histopathology confirmed follicular carcinoma and FNAC from osseous lesions showed metastases. She was subjected to ^131^I ablation for skeletal metastases followed by L-thyroxine suppressive therapy. Case 3 ====== A 70-year-old man from an iodine sufficient area presented with progressive loss of vision in the left eye of one-year duration and pain on movement of the globe. He was not a known diabetic, nor had any history of thyroid disease or ocular trauma. On examination, he was sick looking with a BMI of 27 kg/m^2^. His vitals were stable and he was clinically euthyroid. Ocular examination revealed loss of light reflex, dilated pupil and choroidal mass in the left eye and a normal right eye. There was no visible or palpable neck swelling or lymph nodes anywhere else. For the painful blind eye due to intractable glaucoma he underwent enucleation. The histopathology of the excised tissue revealed metastatic papillary thyroid carcinoma involving the posterior choroid and superficial retina ([Figure 2a](#f2a-asm-5-428){ref-type="fig"}, [2b](#f2b-asm-5-428){ref-type="fig"}). Anterior segment and optic nerves were uninvolved, while the retina was detached posteriorly. Immunostaining for thyroglobulin was positive in the tumor tissue. Color Doppler ultrasound of the neck done subsequently revealed neovascularization and 15×15 mm nodule in the right lower pole of the thyroid and a 2×2 cm lymph node near the common carotid artery. Fine needle aspiration cytology from the thyroid swelling and lymph node revealed papillary thyroid carcinoma ([Figure 2c](#f2c-asm-5-428){ref-type="fig"}). He underwent total thyroidectomy and radical node resection followed by an ablative dose of ^131^I. Whole body ^131^I scan done subsequently found no residual disease. He was on a suppressive dose of L-thyroxine and doing well for the last 3 years. Case 4 ====== A 47-year-old woman from an iodine sufficient area presented with an euthyroid solitary nodule. It had poor ^131^I-uptake on scan and had features of follicular neoplasm on FNAC. Her thyroid profile was normal. She was planned for hemithyroidectomy but underwent total thyroidectomy for hemorrhage into the nodule on exploration. The detailed histology of the nodule excluded malignancy. She received 150 μg L-thyroxine daily for 3 years. However, even without thyroid hormone replacement she later remained euthyroid and had no palpable goiter and had normal serum T3 and T4 levels. Six years later, she complained of irregular cycles and experienced hot flashes and after screening was put on hormone replacement therapy (HRT) with conjugated equine estrogen and medroxyprogesterone continuously. With this therapy, hot flashes decreased but after 6 months she had inter-menstrual bleeding. A tender right fornix and 5×6 cm firm mobile mass was found on per vaginal examination and it was confirmed by pelvic ultrasound. Bilateral salpingo-oophorectomy and hysterectomy were performed. Histopathology revealed follicular carcinoma in a struma ovarii ([Figure 3](#f3-asm-5-428){ref-type="fig"}). No other teratomatous elements were seen. The left ovary was normal. Estrogen and progesterone receptor status by peroxidase anti-peroxidase immunochemistry on ovarian tissue after microwave retrieval were found to be negative. An ^131^I whole body scan after 6 weeks of laparotomy revealed uniform low uptake in residual thyroid tissue (2.4% at 24 h) and no tracer uptake elsewhere. Her unsuppressed serum thyroglobulin level was 4.5 ng/mL (normal \<10 ng/mL) and she had low serum T3 and T4 (0.55 and 40 ng/mL, respectively) and an elevated thyroid stimulating hormone (TSH) of 40.8 μU/mL, indicating hypothyroidism. She received 4500 rads external beam radiotherapy to the pelvic region in view of the capsular invasion and ascites, together with L-thyroxine therapy. She had no evidence of metastasis during her 10-year follow-up. Discussion ========== Most patients with thyroid cancer are euthyroid and thyroid cancer manifesting with thyrotoxicosis is rare.[@b4-asm-5-428],[@b5-asm-5-428] Leiter et al described the first patient with adenocarcinoma of the thyroid with functioning metastasis and postoperative thyrotoxicosis in 1946.[@b6-asm-5-428] Since then a handful of cases (less than 50) are described. Four out of 55 patients in our study were toxic at presentation, three had pre-existing MNG and one had a functioning metastasis and presented with periodic paralysis. The different mechanisms by which thyroid malignancy causes thyrotoxicosis include excessive production of thyroid hormones by functioning thyroid and/or metastatic tissue, excessive release of thyroid hormones by infiltration of the thyroid gland by the tumor tissue, underlying Graves' disease, and iodine-induced toxicosis as a part of diagnostic or therapeutic interventions.[@b6-asm-5-428]--[@b9-asm-5-428] The association between thyrotoxicosis and periodic paralysis was initially described by Rosenfeld in 1902.[@b10-asm-5-428] The prevalence being 8% to 34% in oriental males and 0.2% in females, while in non-Asian population the reported prevalence is 0.1% to 0.2%.[@b11-asm-5-428]--[@b13-asm-5-428] Thyrotoxicosis usually precedes or occurs concurrently with periodic paralysis in 80% of cases, like ours.[@b14-asm-5-428] The disorder is clinically indistinguishable from familial hypokalemic periodic paralysis. The pathogenesis of periodic paralysis in thyrotoxicosis includes an alteration in either Na+/K+ ATP-ase pump activity due to thyrotoxicosis or increased (2 adrenergic receptor sensitivity. Increased insulin sensitivity and altered voltage gated calcium channels are also postulated.[@b15-asm-5-428] Follicular thyroid carcinoma with thyrotoxic periodic paralysis as a presenting manifestation has not been reported previously. Differentiated thyroid cancer commonly metastasizes to lymph nodes, bones, lung, brain and other organs. Osseous metastasis occurs in 3% to 12% of patients with differentiated follicular and papillary thyroid cancer.[@b14-asm-5-428] However, ^131^I scan or 99mTcMDP scan can reveal osseous metastasis in the majority (74%) of these patients.[@b14-asm-5-428] Of the two, a relatively higher prevalence of bone metastasis occurs in follicular (15.2%) than in papillary carcinoma (0.6%).[@b16-asm-5-428] Frequent sites of osseous metastases are sternum, vertebrae, pelvis and ribs. Case 2 presented with a pulsating metastatic deposit in the frontal bone, which though classical, is very rare as the presenting manifestation. The overlying skin was also involved as follicular carcinoma has a greater preponderance than papillary carcinoma for cutaneous metastases and the majority of skin metastases from thyroid cancer are localized to the head and neck area with underlying bone involvement.[@b17-asm-5-428] However, one such case has been described in which the sacrum was the site of pulsating metastases.[@b18-asm-5-428],[@b19-asm-5-428] Although secondaries are the commonest malignancy of the choroid, their occurrence is quite rare. In a study of 8712 patients with carcinoma, only 6 had metastasis to the choroid.[@b20-asm-5-428] The reasons for rarity of choroid metastasis are early diagnosis of cancer and right angle branching of the ophthalmic artery, and therefore malignant emboli lodge into the brain and meninges by means of a straight route.[@b21-asm-5-428] The other explanation in patients with disseminated malignancy could be the clinical picture dominated by a poor general condition, slow growth of choroid metastasis and death of the patient before attention is attracted to the ocular condition. Routine ophthalmic or pathological examinations of the eye are usually omitted in patients with terminal carcinomatosis.[@b20-asm-5-428] The majority of emboli travel to the 20-odd short posterior ciliary arteries rather than the long anterior branches. Therefore the posterior region of the choroid near the macula is the site of predilection. The commonest malignancies metastasizing to the choroid are malignancy of the breast, lungs, colorectal region and urinary bladder.[@b20-asm-5-428] Choroid metastasis of thyroid carcinoma is rare. Of 227 cases of choroid metastasis only one was from thyroid.[@b22-asm-5-428] A review of the literature shows 14 reported thyroid cancers metastasizing to the eye; papillary thyroid carcinoma has been reported thrice before.[@b22-asm-5-428]--[@b24-asm-5-428] Patients with choroid metastasis may present with progressive loss of vision, limitation of ocular movement and retro-orbital/orbital pain, though the majority are asymptomatic. Ocular examination usually reveals neovascularization of the iris, raised intraocular pressure, and exudative retinal detachment with an orange-colored nodule. B-mode ultrasound, MRI and biopsy in a appropriate setting clinch the diagnosis. The treatment modalities are ^131^I ablation, photocoagulation, cryopexy and rarely enucleation.[@b24-asm-5-428] Our case is unique because the choroid metastasis antedated the diagnosis of thyroid carcinoma and he had no clinically obvious thyroid swelling or regional lymphadenopathy. Struma ovarii is a slow growing ovarian neoplasm within thyroid tissue because it is the only or the predominant (\>50%) constituent. The ovarian thyroid is histologically and functionally identical to the cervical thyroid. The tumour is a highly specialized subclass of benign cystic teratoma; 95% remain benign while the remainder undergo malignant transformation, with peak frequency during the fifth decade of life.[@b25-asm-5-428],[@b26-asm-5-428] The left ovary is more frequently involved than the right and in 6% of instances struma ovarii are bilateral. Struma ovarii often present as an abdominal mass, with lower abdominal pain, ascites and, uncommonly, hyperthyroidism (5%). Past or concomitant thyroid enlargement (18%) has been described with struma ovarii, which may cause difficulty in its diagnosis. Retrospectively, the fact that this patient remained euthyroid without L-thyroxine replacement for 4 years, could have given a clue to the presence of a functioning struma ovarii. This was further substantiated when the patient became hypothyroid (TSH 40.8 μU/mL) promptly after the removal of the ovarian mass. The majority of malignant struma present as follicular carcinoma, while papillary, anasplastic, and Hurthle cell carcinoma has also been described.[@b27-asm-5-428] It is noteworthy that the lack of demonstration of estrogen and progesterone receptors in struma in our case can be explained by the complete transformation of ovarian tissue into struma, as normal thyroid follicular cells do not express estrogen receptors.[@b27-asm-5-428] In conclusion, differentiated thyroid cancer can present with atypical manifestations. A routine preoperative physical examination of the thyroid gland should be done in all cases of metastatic lesions of unknown origin. Clinical suspicion and aggressive work-up and treatment are usually rewarding. The high incidence of metastatic disease in an iodine deficiency area like ours is postulated due to the advanced stage at presentation or inherent aggressive biological behavior. ![Pulsating metastasis on the right side of the frontal area with prominent blood vessels.](asm-5-428f1a){#f1a-asm-5-428} ![^131^I whole body scan showing significantly increased tracer uptake in the thyroid region and the skull.](asm-5-428f1b){#f1b-asm-5-428} ![Low-power photograph showing papillary configuration of the tumor in relation to the choroid (H&E×160).](asm-5-428f2a){#f2a-asm-5-428} ![Photomicrograph showing retinal detachment (H&E×220)](asm-5-428f2b){#f2b-asm-5-428} ![Optically clear nuclei of tumor cells (H&E×440).](asm-5-428f2c){#f2c-asm-5-428} ![High power phtotomicrograph showing follicular thyroid carcinoma in struma ovarii and infiltration by malignant cells in capsular lymphatics (H&E×440).](asm-5-428f3){#f3-asm-5-428} ###### The clinical profile of 4 patients who had unusual features as a presenting manifestation. Patient Age Sex Cause of referral Functional status Histopathology --------- ----- ----- --------------------------------------- ------------------------------------- ---------------------------------- 1 30 M Hypokalemic paralysis Toxic Follicular 2 55 F Pulsatile scalp swelling Euthyroid Follicular 3 70 M Metastasis to globe Euthyroid Papillary 4 47 F Solitary thyroid nodule, ovarian mass Euthyroid after total thyroidectomy Struma ovarii with follicular Ca

{ "pile_set_name": "PubMed Central" }

INTRODUCTION ============ As a clean and carbon-neutral energy source, solar energy has drawn increasing attention in recent years. To date, solar energy utilization has largely revolved around photovoltaic power generation ([@R1]--[@R7]). Although significant progress has been made in photovoltaics, electricity alone only accounts for less than 20% of the global delivered energy consumptions, and the remaining energy is provided by solid, liquid, or gaseous fuels ([@R8]). Therefore, novel approaches that convert solar energy into fuels offer significant opportunities to reduce fossil fuel consumption in industrial and transportation sectors, thereby providing potentially effective means to mitigate anthropogenic CO~2~ emissions. Solar fuels can be produced from water and CO~2~ by photocatalytic splitting under mild conditions ([@R9]--[@R14]). However, the efficiency and yields of existing photocatalytic processes are far from satisfactory due to the high thermodynamic and activation barriers of these reactions and challenges in band structure optimization for photocatalysts, electron-hole recombination, and photocatalyst stability ([@R13]). In addition, product separation for these processes can be problematic ([@R14]). As an alternative to photocatalytic processes, solar-thermal water/CO~2~ splitting has shown excellent potential ([@R15]--[@R21]). For instance, solar-to-hydrogen efficiency can be as high as 18% for solar-thermal water splitting, whereas state-of-the-art photocatalytic processes have an energy conversion efficiency of less than 0.2% ([@R22]). Typical solar-thermal water/CO~2~-splitting processes convert concentrated solar energy in a two-step, cyclic redox scheme: A redox material (usually a metal oxide) is first decomposed (or reduced) under an inert environment to (partially) release its lattice oxygen (MO~*x*~ → MO~*x*−δ~ + δ/2O~2~) at elevated temperatures (\>1200°C), and then the reduced metal oxide is reoxidized (or regenerated) by putting it in contact with water or CO~2~ to replenish its lattice oxygen, producing hydrogen or carbon monoxide (MO~*x*−δ~ + δH~2~O/CO~2~ → MO~*x*~ + δH~2~/CO). The resulting hydrogen (H~2~) and/or carbon monoxide (CO) can be used as gaseous fuels or converted into liquid fuels or chemicals ([@R23], [@R24]). Such a two-step redox scheme eliminates the need for gas separation and has demonstrated notably higher CO/H~2~ production rates (1.35 mmol~CO~ kg~oxide~^−1^ s^−1^ and 0.71 mmol~H2~ kg~oxide~^−1^ s^−1^) when compared to state-of-the-art photocatalytic approaches \[\~0.07 mmol~CO~ kg~cat~^−1^ s^−1^ ([@R21], [@R25])\]. The challenge for solar-thermal processes resides in the high reaction temperatures for metal oxide decomposition. Reasonable decomposition rates are only achieved at temperatures above 1500°C for nonvolatile monometallic metal oxides ([@R15], [@R21], [@R26], [@R27]). Although recent studies on mixed metal oxides have led to notable decreases in the decomposition temperatures, most solar-thermal redox processes require an operating temperature above 1200°C ([@R17], [@R26]--[@R30]). In addition to the high reaction temperatures, water or CO~2~ conversions in these schemes are far from satisfactory (typical below 10%), posing significant challenges for process heat integration and downstream product separations ([@R31]). Introducing reducing agents can lower the decomposition (or reduction) temperature ([@R32]). In such an open-loop thermochemical approach, reducing agents such as carbonaceous fuels are used as oxygen "sinks" to facilitate lattice oxygen extraction from redox materials ([@R15]). Although significantly lower reduction temperatures (\<800°C) have been demonstrated, steam or CO~2~ conversion in the subsequent water- or CO~2~-splitting step is still very low ([@R33]--[@R35]). To address this challenge, we proposed a hybrid solar-redox scheme for water splitting, wherein concentrated solar energy is used to drive the reduction reaction in the presence of methane (CH~4~) that undergoes partial oxidation (POx) to produce Fischer-Tropsch (F-T) ready syngas. Using perovskite-supported Fe~3~O~4~ as the redox material, up to 77% steam-to-hydrogen conversion was achieved ([@R16], [@R36]). Although such a methane-assisted solar-thermal water-splitting scheme is applicable for CO~2~ splitting, further increases in water/CO~2~-splitting conversion is desirable from process efficiency and product separation standpoints. Another challenge is that iron oxide--based materials tend to overoxidize syngas products in the methane POx step, limiting the syngas selectivity to less than 70%. Herein, we report high-performance redox materials for CO~2~ splitting using the aforementioned hybrid solar-redox concept. Perovskite nanocomposites (NCs) exhibit exceptional efficacy for both CO~2~ splitting and methane POx: Over 98% CO~2~-to-CO conversion was achieved in CO~2~ splitting, and syngas selectivity reached 96% in methane POx. As a stand-alone process, methane-assisted CO~2~ splitting offers an effective approach for both CO~2~ utilization and F-T/methanol ready syngas production. As an example, the resulted syngas and CO products can be readily used for green acetic acid production ([Fig. 1](#F1){ref-type="fig"} ). Compared to conventional acetic acid synthesis, the proposed scheme has the potential to reduce fossil fuel consumption by 67% and produce 84% less CO~2~ when compared to coal-based process. ![Hybrid solar-redox process and its application for green acetic acid production from methane and CO~2~.\ This process has three sections: hybrid solar-redox for syngas and CO coproduction, methanol synthesis from the syngas, and acetic acid production.](1701184-F1){#F1} RESULTS ======= Rationale for redox material selection -------------------------------------- Performance of the redox material is critical to the hybrid solar-redox scheme. Desirable properties for the redox materials include high syngas yield and productivity, excellent CO~2~ conversion and CO productivity, and good redox stability. Ideal redox materials should also be affordable, environmentally benign, and attrition-resistant. Among the various metal oxides investigated to date, iron oxide--containing materials are the most promising ones because of their low cost, abundance, and low toxicity ([@R16], [@R17], [@R30]--[@R36]). However, pure iron oxides are not suitable for the proposed redox reactions from a thermodynamics viewpoint. A methane-assisted thermochemical CO~2~-splitting scheme can be described by the reactions listed in [Table 1](#T1){ref-type="table"}. From a thermodynamic standpoint, the redox material can be considered as an oxygen "source" in the methane POx step and an oxygen sink in the CO~2~-splitting step. Hence, a redox pair (MeO~*x*~/MeO~*x*−1~) with a high equilibrium oxygen pressure (P~O2~) can lead to overoxidation of syngas (Eqs. 3 and 4) and low CO~2~ conversion (Eq. 7). In contrast, redox pairs with a low P~O2~ will result in low methane conversion (Eq. 2). CO~2~ splitting, on the other hand, is monotonically favored at lower P~O2~. The relationship among equilibrium P~O2~, syngas yield, and CO~2~ conversion, derived from minimizing Gibbs free energy of the reaction set (Eqs. 2 to 6), is illustrated in [Fig. 2](#F2){ref-type="fig"}. As can be seen, equilibrium P~O2~ of Fe/FeO redox pair, which is a commonly used redox material, is away from the ideal (high performance) region for the hybrid solar-redox scheme. This is confirmed by a number of studies that reported low syngas selectivity and limited CO~2~ conversions ([@R16], [@R34]--[@R38]). Thermodynamic analysis suggests that no first-row transition metal oxide has redox properties located in this region ([@R39]). Unlike monometallic oxides whose redox properties are fixed, mixed metal oxides offer potentially tunable thermodynamic properties. Using iron oxide (FeO) as an example, it can form mixed oxides with strontium oxide (SrO) at various Sr/Fe atomic ratios ([@R40]). Although SrO is not reducible, formation of Sr-Fe mixed oxides changes the coordination environments and electron density of oxygen anions (O^2−^). Thus, equilibrium P~O2~ of Sr-Fe mixed oxides can be tailored by adjusting Sr/Fe ratios ([Fig. 2](#F2){ref-type="fig"}). For example, spinel-type SrFe~2~O~4~ exhibits lower P~O2~ than FeO. Further increases in Sr/Fe ratios leads to the formation of perovskite oxides such as Brownmillerite-type Sr~2~Fe~2~O~5~ and Ruddlesden-Popper (R-P)--phase Sr~3~Fe~2~O~6~ (Sr~*n*+1~Fe~*n*~O~3*n*+1−δ~ with *n* = 2), both having notably lower P~O2~ than FeO ([Fig. 2](#F2){ref-type="fig"}). Among the three Sr-Fe mixed oxides investigated, SrFe~2~O~4~, with a marginally improved redox performance over FeO, is not an excellent redox material because of low equilibrium syngas yields (\<90%) and CO~2~ conversions (\<95%) at the temperature range of interest. To compare, both Sr~2~Fe~2~O~5~ and Sr~3~Fe~2~O~6~ can be effective for methane POx and CO~2~ splitting, with Sr~3~Fe~2~O~6~ offering improved redox properties for the hybrid solar-redox scheme. ###### Main reactions that occur in methane POx and CO~2~ splitting in the redox scheme. **Methane POx** **CO~2~ splitting** ------------------------------------------ ----------------------------------------- MeO~*x*~ ⇆ MeO~*x*−1~ + ^1^/~2~O~2~ (1) CO~2~ ⇆ CO + ^1^/~2~O~2~ (7) CH~4~ + ^1^/~2~O~2~ ⇆ CO + 2H~2~ (2) MeO~*x*−1~ + ^1^/~2~O~2~ ⇆ MeO~*x*~ (8) CO + ^1^/~2~O~2~ ⇆ CO~2~ (3) 2H~2~ + O~2~ ⇆ H~2~O (4) 2CO ⇆ C + CO~2~ (5) CO + H~2~O ⇆ H~2~ + CO~2~ (6) ![Thermodynamic analysis of methane POx/CO~2~-splitting reactions and selected redox materials.\ Standard Gibbs free energy change (Δ~r~*G*°) of the oxidation reactions (①2Fe + O~2~ = 2FeO; ②^2^/~3~SrO + ^4^/~3~Fe + O~2~ = ^2^/~3~SrFe~2~O~4~; ③^4^/~3~SrO + ^4^/~3~Fe + O~2~ = ^2^/~3~Sr~2~Fe~2~O~5~; and ④2SrO + ^4^/~3~Fe + O~2~ = ^2^/~3~Sr~3~Fe~2~O~6~) and corresponding equilibrium oxygen partial pressures (P~O2~), for ≥95% CO~2~ conversions (blue region) and ≥90% methane-to-syngas yield (green region) at 1 atm. The overlapping (cyan) zone is the ideal (high performance) region for both methane POx and CO~2~-splitting reactions.](1701184-F2){#F2} Redox performance of Sr~3~Fe~2~O~7**−**δ~ ----------------------------------------- As-prepared Sr~3~Fe~2~O~7−δ~ (SF~7~) exhibits excellent redox activity (figs. S1 and S2), but it deactivates continuously (fig. S3) at 900°C under a weight hourly space velocity (WHSV) of 60,000 cm^3^ g~SF7~^−1^ hour^−1^. Specifically, syngas productivity (eq. S1) dropped by 82% in the methane POx step, and CO productivity (eq. S2) decreased by 72% in the CO~2~-splitting step over 15 redox cycles. It is noted that even with significant deactivation, CO productivity in the 16th cycle (2.6 mol kg~SF7~^−1^) is still an order of magnitude higher than that in state-of-the-art solar-thermal CO~2~-splitting schemes \[\<0.3 mol kg~oxide~^−1^ for ceria and perovskites ([@R21], [@R29], [@R41], [@R42])\]. Sr~3~Fe~2~O~7−δ~ is also superior to CeO~2~-modified Fe~2~O~3~ in a H~2~-assisted thermochemical CO~2~-splitting scheme, where CO productivity is about 1.1 mol kg~oxide~^−1^ ([@R35]). To shed light on the cause of deactivation, we acquired energy-dispersive x-ray spectroscopy (EDS) mapping and x-ray diffraction (XRD) patterns of as-prepared and spent samples (fig. S4). XRD measurements and thermogravimetric analysis suggest that the as-prepared sample has an R-P--phase Sr~3~Fe~2~O~6.75~ (tetragonal *I*4/*mmm*), whose oxygen stoichiometry decreases to about 6 in an inert atmosphere at 900°C. Metallic iron, iron carbides, Fe~3~O~4~, and SrO were detected after the first reduction. It was found that CO~2~ can fully reoxidize metallic iron back to the R-P phase, but a significant portion of Sr~3~Fe~2~O~6~ becomes unreducible in the 16th reduction (fig. S4). Elemental mappings indicate that significant phase segregation and sintering occurred in redox reactions (fig. S4); this may have caused the redox activity to decrease. To retard sintering and deactivation, the redox active-phase (SF~7~) was dispersed in an inert and earth-abundant calcium manganese oxide (Ca~0.5~Mn~0.5~O) phase. Sr~3~Fe~2~O~7−δ~-Ca~0.5~Mn~0.5~O NCs ------------------------------------ Formation of Sr~3~Fe~2~O~7−δ~-Ca~0.5~Mn~0.5~O NCs (SF~7~-CM NCs) is revealed by electron microscopy (figs. S5 and S6), and reducibility of the sample is characterized by hydrogen temperature-programmed reduction (H~2~-TPR) and in situ XRD (figs. S7 and S8). The redox stability of SF~7~-CM NCs is shown in [Fig. 3](#F3){ref-type="fig"}. Both syngas productivity in the methane POx step and CO productivity in the CO~2~-splitting step decreased slightly in the first 21 cycles and then stabilized in the subsequent 9 cycles at 900°C at a WHSV of 24,000 cm^3^ g~SF7-CM~^−1^ hour^−1^. Specifically, syngas productivity decreased from 4.42 to 4.28 mol kg~SF7-CM~^−1^, whereas CO productivity varied from 1.74 to 1.53 mol kg~SF7-CM~^−1^ over the first 21 cycles. In the CO~2~-splitting step, CO productivity was significantly higher for SF~7~-CM NCs than for CeO~2~-Fe~2~O~3~ \[1.1 mol kg~oxide~^−1^ ([@R35])\]. In the last 10 cycles, syngas selectivity was higher than 95% in the methane POx step. Moreover, the molar ratio of H~2~/CO remained around 1.96, and coke formation was negligible between 20 and 30 cycles. The excellent stability of the NCs is attributed to the confinement effect that involves dispersing an active phase in an inert medium. XRD measurements demonstrated that Ca~0.5~Mn~0.5~O acted as a dispersing medium ([Fig. 3](#F3){ref-type="fig"}), which itself had no activity toward the redox reactions. In addition, elemental mappings demonstrate that all elements are uniformly distributed in as-prepared and spent samples (figs. S9 and S10). ![Redox stability of SF~7~-CM NCs.\ (**A**) Productivity of H~2~, CO, and CO~2~ in methane POx and CO in CO~2~ splitting. (**B**) XRD patterns of spent samples. Reaction conditions: *m*~SF7-CM~ = 0.25 g; *T* = 900°C, *F* = 74.4 μmol s^−1^, *P* = 1 atm, *y*~CH4~ = 0.1, and *y*~CO2~ = 0.1. a.u., arbitrary units.](1701184-F3){#F3} [Figure 4A](#F4){ref-type="fig"} shows typical gaseous product elution profiles in the methane POx step of the 30th cycle. After feeding CH~4~ to the NC bed, CO~2~ coeluted with CO and H~2~, but its elution profile was different from those of CO and H~2~. The peak production rates for CO~2~, CO, and H~2~ were 0.51, 4.14, and 8.03 mmol kg~SF7-CM~^−1^ s^−1^ at 16, 250, and 250 s, respectively. Such a pattern is typical for the redox-based methane POx ([@R43]). The gaseous compound elution profiles in the CO~2~-splitting step of the 30th cycle are presented in [Fig. 4B](#F4){ref-type="fig"}. The breakthrough time of CO~2~, which is the time when CO~2~ flow rate at the outlet of a reactor reaches 2% of that at the inlet, was 35 s, and it coincided with the time at which the CO production rate maximized (22 mmol~CO~ kg~SF7-CM~^−1^ s^−1^). The peak CO production rate in this open-loop thermochemical scheme was one order of magnitude higher than that in conventional solar-thermal CO~2~ splitting, which was carried out at temperatures 200°C higher than that in the former ([@R21]). It should be noted that the breakthrough curve of CO~2~ depends on factors like the flow rate of feed, amount of redox materials, and inlet concentration of CO~2~. As shown in [Fig. 4B](#F4){ref-type="fig"}, CO~2~ flow rate leveled off after 180 s. In the first 3 min, the average CO production rate was 8.23 mmol kg~SF7-CM~^−1^ s^−1^, significantly higher than that (\<3 mmol kg~oxide~^−1^ s^−1^) in a recently reported methane-assisted thermochemical scheme ([@R33]). ![Instantaneous production rates of gaseous products from the redox reactions using SF~7~-CM NCs.\ (**A**) Elution profiles of H~2~, CO, and CO~2~ in methane POx at the 30th cycle. (**B**) Elution profiles of CO and CO~2~ in CO~2~ splitting at the 30th cycle. Inset in (B) is the cumulative CO molar fraction in the product and moles of CO produced in CO~2~ splitting for a short feeding period (the blue dot denotes the total moles of CO produced for complete regeneration). Reaction conditions: *m*~SF7-CM~ = 0.25 g; *T* = 900°C, *F* = 74.4 μmol s^−1^, *P* = 1 atm, *y*~CH4~ = 0.1 (A), and *y*~CO2~ = 0.1 (main figure) and 0.05 (inset) (B).](1701184-F4){#F4} One key issue with thermochemical CO~2~-splitting studies conducted to date is the incomplete CO~2~ conversion (\<90%). For SF~7~-CM NCs, near-complete CO~2~ conversion was achieved during the early stage of the reaction (fig. S11). However, the average CO~2~ conversion in the first 3 min was only 32.3% in the 30th cycle ([Fig. 4B](#F4){ref-type="fig"}). To improve CO~2~ conversion, we attempted to shorten the duration of the CO~2~-splitting step. As shown in [Fig. 4B](#F4){ref-type="fig"}, CO~2~ conversion was above 98%, and the average CO production rate was 13.8 mmol kg~SF7-CM~^−1^ s^−1^ for a shorter feeding period. Such a short regeneration period was adequate to replenish over 80% of the active lattice oxygen of the reduced NCs. To the best of our knowledge, such a high CO~2~ conversion has yet to be reported for thermochemical CO~2~ splitting. These findings clearly demonstrate that this NC outperforms the other redox materials for thermochemical CO~2~ splitting. SrFeO~3−δ~-CaO NCs ------------------ To further verify the role of the dispersing medium, we prepared another NC with SrFeO~3−δ~ as the active phase and CaO dispersing medium. SrFeO~3~ has the highest oxygen capacity among the R-P--type perovskite family (Sr~*n*+1~Fe~*n*~O~3*n*+1,~ *n* = ∞) while still having desirable redox properties ([Fig. 2](#F2){ref-type="fig"}). XRD and Rietveld refinement do not indicate significant Ca substitution of Sr in SrFeO~3−δ~ phase in the NCs, and excess Sr atoms were incorporated into the crystal structure of CaO (figs. S12 and S13). Redox testing indicated that the CaO phase significantly enhanced the redox kinetics, leading to more than 10-fold faster methane POx and CO~2~ splitting (figs. S14 and S15). Apart from the faster redox kinetics, the redox stability of SrFeO~3−δ~-CaO NCs (SF~3~-C NCs) was also excellent at 900°C. After the first two cycles, syngas (H~2~/CO ≅ 2) productivity and selectivity (eq. S3) maintained at around 5.8 mol kg~SF3-C~^−1^ and 96%, respectively, in the methane POx step (fig. S16). Meanwhile, CO productivity in the CO~2~-splitting step decreased slightly over the first 10 cycles (less than 9%) and then remained unchanged between 10 and 16 cycles (fig. S16). The stability was further confirmed by XRD analysis, which demonstrated that the 1st and 16th reduced samples exhibited identical XRD patterns with iron reduced to its metallic form (fig. S17). As demonstrated earlier, it is possible to achieve high CO~2~ conversion by partially replenishing the active lattice oxygen. Meanwhile, high methane conversion can be reached with partially reoxidized NCs (for example, 90% reoxidation). For these POx/splitting cycles, both methane and CO~2~ conversions were high compared to complete reduction/oxidation cycles but with slightly lowered syngas and CO productivity. The redox stability of SF~3~-C NCs that underwent partial reduction and oxidation at 980°C with a WHSV of 12,000 cm^3^ g~SF3-C~^−1^ hour^−1^ is shown in [Fig. 5](#F5){ref-type="fig"}. In the methane POx step, syngas (H~2~/CO ≅ 2.13) productivity maintained at around 4.63 mol kg~SF3-C~^−1^ after 13 cycles with insignificant CO~2~ production (\<0.08 mol kg~SF3-C~^−1^). After the redox material stabilized, methane conversion reached 59%, and syngas selectivity was around 90% (fig. S18 and table S1). In the CO~2~-splitting step, CO~2~ conversion was 98%, the average CO production rate was 9.13 mmol kg~SF3-C~^−1^ s^−1^, and CO productivity was 2.17 mol kg~SF3-C~^−1^ in the last 17 cycles. A further increase in syngas yield can be realized by tuning the redox window of SF~3~-C NCs. Over partially reoxidized SF~3~-C NCs, methane conversion of 90% was achieved with 89% syngas selectivity (table S2) at 980°C with a WHSV of 2045 cm^3^ g~SF3-C~^−1^ hour^−1^. Moreover, CO~2~ conversion was nearly 100% in the CO~2~-splitting step before methane POx (fig. S19). ![Redox stability of SF~3~-C NCs.\ (**A**) Productivity of syngas and CO~2~ and H~2~/CO molar ratio in the methane POx step. (**B**) Productivity of CO and CO molar fraction in the product during the CO~2~-splitting step; the dashed line denotes the CO mole fraction of 0.95. In all 30 redox cycles, methane conversion is greater than 58%, and CO~2~ conversion is above 94%. Reaction conditions: *m*~SF3-C~ = 0.25 g; *T* = 980°C, *F* = 37.2 μmol s^−1^, *P* = 1 atm, *y*~CH4~ = 0.1, and *y*~CO2~ = 0.1.](1701184-F5){#F5} DISCUSSION ========== The hybrid solar-redox scheme offers various attractive options for CO~2~ utilization and liquid fuel/chemical production. Here, acetic acid production was investigated as an example. Direct conversion of methane and CO~2~ to acetic acid is thermodynamically unfavorable, albeit being atom-economic ([@R44]). State-of-the-art approach is the syngas route: Coal is first gasified to produce syngas, and then the syngas is cleaned and conditioned to remove excess CO and CO~2~. This is followed by methanol synthesis (CO + 2H~2~ ⇆ CH~3~OH) and subsequent carbonylation of methanol (CH~3~CHO + CO ⇆ CH~3~COOH). Methane is not an ideal feedstock for acetic acid synthesis via the conventional syngas route because the CO/H~2~ ratio is too low in methane-derived syngas. The solar-driven redox process, as shown in [Fig. 1](#F1){ref-type="fig"} , produces a pure CO stream and a syngas stream with ideal compositions for methanol synthesis. Thus, the need for costly separation processes is eliminated. The solar-driven redox process has three main sections ([Fig. 1](#F1){ref-type="fig"} and fig. S20): (i) the hybrid solar-redox section where syngas and CO are produced, (ii) the methanol synthesis section that converts syngas into methanol ([@R45]), and (iii) the acetic acid synthesis section in which methanol and CO are converted to acetic acid by the Cativa process ([@R46]). Using ASPEN Plus, the overall energy required for the process was calculated. This process was compared with a coal-based scheme (fig. S21) ([@R47]). For each metric ton of acetic acid produced, 20.4 GJ~thermal~ is required in the solar-driven process, whereas 37.8 GJ~thermal~ is needed in the coal-based process ([Fig. 6](#F6){ref-type="fig"}). Regarding the productivity, the former produces 2.7 times as much acetic acid as the latter per gigajoule of feed fuel ([Fig. 6](#F6){ref-type="fig"}). Simulation results suggest that production of 1 metric ton of acetic acid results in 0.37 metric ton of CO~2~ in the solar-driven process but 2.2 metric tons CO~2~ in the coal-based approach. ![Comparison of the conventional and hybrid solar-redox processes for acetic acid production.\ The hybrid solar-redox process (see [Fig. 1](#F1){ref-type="fig"} ) is compared with the traditional coal-based process to obtain syngas. For the coal-based process, all energy demand is provided by coal. For the hybrid solar-redox process, the total energy demand is 20.4 GJ~thermal~ per metric ton of acetic acid (AcOH), and 39.2% (8 GJ~thermal~ metric ton~AcOH~^−1^) is provided by solar energy.](1701184-F6){#F6} We report iron-containing mixed metal oxides as effective CO~2~-splitting and methane POx agents in a cyclic redox scheme. Unsupported Sr~3~Fe~2~O~7−δ~ demonstrated extraordinary methane POx and CO~2~-splitting activity at 900°C. However, it deactivated over redox cycles. Dispersing Sr~3~Fe~2~O~7−δ~ in a Ca~0.5~Mn~0.5~O matrix significantly enhanced the redox stability of the R-P--structured Sr~3~Fe~2~O~7−δ~. Further investigation of the perovskite-structured SrFeO~3−δ~ compositing with CaO indicated that the inert CaO phase not only enhanced the redox kinetics but also improved its redox stability at 900° to 980°C. At 980°C, the redox activity decreased slightly in the first 13 cycles and then remained unchanged in the last 17 cycles. In the last 17 cycles, the average CO production rate was 9.13 mmol kg~SF-C~^−1^ s^−1^, and CO productivity was 2.17 mol kg~SF3-C~^−1^. The former is 6.8 times higher than state-of-the-art solar-thermal CO~2~-splitting schemes (1.35 mmol kg~oxide~^−1^ s^−1^) carried out at significantly higher temperatures (above 1200°C), and the latter is sevenfold higher than that in solar-thermal schemes (\<0.3 mol kg~oxide~^−1^). For partially reoxidized SF~3~-C NCs, methane conversion reached 90% with 89% syngas selectivity at 980°C with a WHSV of 2045 cm^3^ g~SF3-C~^−1^ hour^−1^. Moreover, CO~2~-to-CO conversion maintained at near 100% in the CO~2~-splitting step. The exceptionally high CO~2~ utilization efficiency and high syngas yield make the redox materials and the hybrid solar-redox scheme potentially attractive. Process simulations indicate that the fossil energy consumption for acetic acid production can be reduced by 67% by the hybrid solar-redox process when compared to the state-of-the-art one while reducing CO~2~ emission by as much as 84%. MATERIALS AND METHODS ===================== Synthesis of the redox materials -------------------------------- For the synthesis of SF~7~-CM NCs \[18 weight % (wt %) SF~7~\], 0.83 g of strontium nitrate \[Sr(NO~3~)~2~\], 3.71 g of calcium nitrate \[Ca(NO~3~)~2~·4H~2~O\], 4.94 g of manganese nitrate \[Mn(NO~3~)~2~·4H~2~O\], and 18.90 g of citric acid (C~6~H~8~O~7~) were dissolved in 80 ml of deionized (DI) water. The solution was then heated up to 40°C under agitation (600 rpm) on a hot plate, followed by maintaining at this temperature for 30 min. After that, 9.18 g of ethylene glycol (C~2~H~6~O~2~) was added to the above solution. The resulted solution was subsequently heated up to 80°C and then kept at this temperature until a polymeric gel (ca. 25 cm^3^) was formed. Iron oxide dispersion was prepared by ultrasonicating 0.50 g of nanoparticles in 51.5 g of 22.3 to 77.7 wt % ethylene glycol/water solution. The dispersion was then transferred to a 50-ml centrifuge tube, followed by centrifuging at 3500 rpm for 5 min. Around 30 ml of the upper part of the nanoparticle suspension was mixed with 20 ml of gel. The mixture was ultrasonicated for 15 min and then heated up to 80°C, followed by holding at this temperature for 2 days. Afterward, the gel was dried at 120°C overnight to evaporate the residual water. The dry precursor was put in a high-purity alumina crucible, which was then placed in a high-purity alumina ceramic tube that was mounted on an MTI furnace (GSL-1500X-50-UL), into which air continuously flew. The tube was heated up from room temperature to 1200°C for 300 min and then held at this temperature for 12 hours, followed by cooling down to 300°C for 300 min. For the synthesis of SF~3~-C NCs (26 wt % SrFeO~3−δ~), 0.70 g of Sr(NO~3~)~2~, 0.90 g of iron nitrate \[Fe(NO~3~)~2~·9H~2~O\], 4.57 g of Ca(NO~3~)~2~·4H~2~O, and 11.97 g of C~6~H~8~O~7~ were dissolved into 80 ml of DI water. The solution was then heated up to 40°C under agitation (600 rpm), followed by maintaining at this temperature for 30 min. After that, 5.80 g of C~2~H~6~O~2~ was added. The resulted solution was further heated up to 80°C and then kept at this temperature until a polymer gel formed. Afterward, the gel was dried at 120°C overnight to evaporate the residual water. Finally, the dry precursor was calcined at 1200°C in air for 12 hours, as described before. For the synthesis of Sr~3~Fe~2~O~7−δ~ and SrFeO~3−δ~, these two redox materials were prepared by the Pechini method as described above. The molar ratio of Sr/Fe in the precursor solutions was 3:2 and 1:1 for Sr~3~Fe~2~O~7−δ~ and SrFeO~3−δ~, respectively. The molar ratio of C~2~H~6~O~2~/C~6~H~8~O~7~/total cations was 3.75:2.5:1 in the gel. The calcination temperature was 1200°C. Material characterization ------------------------- The oxygen stoichiometry of as-prepared redox materials was determined by H~2~-TPR, which was carried out on a TA Instruments thermogravimetric analyzer (SDT Q600). About 50 mg of the sample was loaded into an alumina crucible. Before measurements, the as-prepared sample was treated at 900°C for 1 hour in 20 volume percent (volume %) O~2~/80 volume % Ar (74.4 μmol s^−1^). After treatment, the sample was cooled down and subsequently heated up to 100°C and then kept at this temperature for 30 min in 10 volume % H~2~/90 volume % Ar (74.4 μmol s^−1^). Then, it was heated up to 1150°C at a ramping rate of 10°C min^−1^. To obtain O~2~- or CO~2~-regenerated sample, the sample was first reduced in 10% CH~4~/90% Ar (74.4 μmol s^−1^) and then oxidized in 10% O~2~ (or CO~2~)/90% Ar (74.4 μmol s^−1^) at 900°C, followed by cooling down to room temperature. The oxygen stoichiometry of redox materials under inert conditions was estimated from the amount of lattice oxygen released from the as-prepared sample (after 20% O~2~ pretreatment) at a temperature range of \~25° to 900°C in Ar. Around 580 mg of the sample was loaded to a U-type quartz tube (Ø~in~ = 4 mm), and quartz wool was packed on both ends of the oxide bed to keep the particles in place. The reactor was heated up to 900°C in Ar (22.32 μmol s^−1^). The effluent stream was monitored by an MKS Cirrus 2 quadrupole mass spectrometer (QMS). Rietveld analysis of the sample's XRD patterns at room temperature before and after treating at 900°C in nitrogen was also performed. The XRD patterns were acquired on an Empyrean PANalytical XRD using a Cu Kα radiation (λ = 0.1542 nm) operating at 45 kV and 40 mA, which was equipped with an XRK 900 reactor chamber. A sample pellet was placed on a glass ceramics holder (Ø = 15 mm) in the high-temperature chamber, and the sample was scanned at a 2θ range of 20° to 110° with a step size of 0.012°. The sample was heated up from 25° to 900°C in 30 min and then held at this temperature for 1.5 hours in nitrogen. Finally, the sample was cooled down to room temperature in nitrogen. Powder XRD measurements of as-prepared and spent samples under ambient conditions were performed on a Bragg-Brentano x-ray diffractometer (Rigaku SmartLab) at a 2θ range of 20° to 80° with a step size of 0.05°, using graphite monochromatic Cu Kα radiation (λ = 0.1542 nm) with a nickel filter and operating at 40 kV and 44 mA. To examine the phase properties during the reduction of SF~7~-CM NC, we performed in situ crystallographic analyses of the sample on an Empyrean PANalytical XRD using a Cu Kα radiation operating at 45 kV and 40 mA, which is equipped with a disk-shaped high-temperature chamber. A sample pellet was placed on an alumina sample holder (Ø = 15 mm) in the high-temperature chamber, and the sample was scanned at a 2θ range of 20° to 80° with a step size of 0.1° and a scanning time of 0.1 s at each step. The sample was heated up from 25° to 675°C at an interval of 50°C in 5% H~2~/95% He with a ramping rate of 5°C min^−1^. At each step, the temperature was held for 5 min. High-resolution transmission electron microscope (TEM) images were obtained with a field-emission scanning transmission electron microscope (STEM) (JEOL 2010 F) operating at 200 kV. STEM images and mapping were obtained on an aberration-corrected (S)TEM (FEI Titan 80-300) operating at 200 kV. The sample was prepared by drop-casting a suspension of sample in ethanol (200 proof, anhydrous) on a carbon-coated copper grid and then drying it under ambient conditions. Scanning electron microscope (SEM) images were obtained on a Verios 400 (FEI) field-emission microscope and a JSM-6010LA. To determine the bulk composition of NC, samples were digested with hot concentrated nitric acid. The aliquots were filtered with a 0.45-μm filter (Whatman), and the metal concentrations were analyzed by using an inductively coupled plasma (ICP) optical emission spectrometry (PerkinElmer ICP Optical Emission Spectrometer model 8000). N~2~ physisorption was performed at −196°C on a volumetric adsorption analyzer (Micromeritics ASAP 2020). Before the analysis, the sample was outgassed under vacuum at 300°C for 2 hours. The specific surface area was calculated by using the Brunauer-Emmett-Teller method in the relative pressure range from 3 × 10^−6^ to 0.2. Reactor setup and redox experiments ----------------------------------- The redox performance was evaluated in a microreactor (U-type quartz tube; Ø~in~ = 4 mm) that was vertically placed inside an electric furnace. The furnace was equipped with a K-type thermocouple whose tip touched the outside of the quartz tube at the location of the redox material. The gas flows were controlled by MCQ mass flow controllers (Alicat). The composition of the effluent stream was monitored by QMS. In a typical experiment, 0.25 g of the redox material (250 to 425 μm) was loaded into the reactor, and quartz wool was packed on both ends of the oxide bed to keep the particles in place. The reactor was heated up to the targeted temperature in Ar (66.96 μmol s^−1^), and then these conditions were maintained for around 30 min to remove the residual air in the reaction system. After that, a CH~4~ stream (7.44 μmol s^−1^) was introduced to the Ar stream, and methane POx lasted for 15 min, followed by purging the reactor with Ar (66.96 mmol s^−1^) for 10 min. Next, a CO~2~ stream (7.44 μmol s^−1^) was introduced to the Ar stream, and CO~2~ splitting lasted for 10 min, followed by purging the reactor with Ar (66.96 μmol s^−1^) for 10 min. The above POx-splitting cycle was repeated 30 times. In the redox test with SF~3~-C NCs at 980°C, the amount of coke formed in the POx step was determined by in situ oxidizing the NCs in 10% O~2~/90% Ar (37.2 μmol s^−1^) at 900°C while measuring the CO~*x*~ production with QMS. The first reduced sample was recovered after the first methane POx step, and the last oxidized sample was collected after the redox experiment was completed. The last reduced sample was prepared by reducing part of the last oxidized sample in 10 volume % CH~4~/90 volume % Ar (74.4 μmol s^−1^) at 900°C for 15 min. We thank A. Shafiefarhood and Y. Liu for collecting the TEM images, Y. Gao for acquiring the SEM micrographs, N. Galinsky for collecting the in situ XRD patterns, and T. Y. Li for the Rietveld analysis of XRD patterns at the North Carolina (NC) State University. **Funding:** This work was supported by the NSF (CBET-1254351 and CBET-1510900) and the Kenan Institute at NC State University. **Author contributions:** F.L. conceived and supervised the study. J.Z. and F.L. designed the study and wrote the manuscript. J.Z. carried out the experiments and analyzed the data. V.H. carried out the ASPEN Plus simulation and collected the XRD data. **Competing interests:** The authors declare that they have no competing interests. **Data and materials availability:** All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors. Supplementary material for this article is available at <http://advances.sciencemag.org/cgi/content/full/3/8/e1701184/DC1> Supplementary Text fig. S1. Characterization of Sr~3~Fe~2~O~7−δ~. fig. S2. Redox kinetics on Sr~3~Fe~2~O~7−δ~. fig. S3. Redox stability of Sr~3~Fe~2~O~7−δ~. fig. S4. Characterization of as-prepared and spent Sr~3~Fe~2~O~7−δ~ under ambient conditions. fig. S5. Characterization of as-prepared and spent (oxidized) SF~7~-CM NCs. fig. S6. STEM image and EDS elemental mapping of as-prepared SF~7~-CM NCs. fig. S7. H~2~-TPR of SF~7~-CM NCs. fig. S8. In situ XRD patterns of as-prepared SF~7~-CM NCs during reduction in hydrogen atmosphere at a temperature range of 25° to 675°C. fig. S9. STEM image and EDS elemental mapping of spent (reduced) SF~7~-CM NCs after 30 cycles. fig. S10. STEM image and EDS elemental mapping of spent (oxidized) SF~7~-CM NCs after 30 cycles. fig. S11. CO~2~ splitting on SF~7~-CM NC after methane POx. fig. S12. Characterization of SrFeO~3−δ~. fig. S13. Characterization of SF~3~-C NCs. fig. S14. Redox kinetics on SrFeO~3−δ~. fig. S15. Redox kinetics on SF~3~-C NCs. fig. S16. Redox stability of SF~3~-C NCs. fig. S17. Characterization of as-prepared and spent SF~3~-C NCs under ambient conditions. fig. S18. Redox kinetics over SF~3~-C NCs in a short redox period. fig. S19. Redox kinetics over SF~3~-C NCs. fig. S20. Block diagram for the hybrid solar-redox (HSR)--AcOH. fig. S21. Block diagram for AcOH synthesis using coal-slurry gasifier (CG). fig. S22. Comparison in the overall energy demand and production of AcOH for CG and HSR processes. fig. S23. CO~2~ footprint of the two processes. table S1. Product analysis of the last redox cycle over SF~3~-C NCs. table S2. Product analysis of a short POx over SF~3~-C NCs at 980°C. table S3. Description of the ASPEN Plus model. table S4. ASPEN Plus modules, property methods, and databanks. table S5. Comparison of CG and HSR-AcOH. References ([@R48]--[@R52]) [^1]: Present address: School of Chemical Engineering and Technology, Xi'an Jiaotong University, Xi'an, Shaanxi 710049, China.

{ "pile_set_name": "PubMed Central" }

![](hosplond70322-0011){#sp1 .191} ![](hosplond70322-0012){#sp2 .192}

{ "pile_set_name": "PubMed Central" }

Introduction ============ Retinitis pigmentosa (RP) is the most common form of hereditary retinal degeneration, occurring in around 1 in 3,500 people \[[@r1]\]. It is characterized by progressive degeneration of the peripheral retina (leading to night blindness), loss of the peripheral visual fields, and an abnormal electroretinogram. RP is clinically and genetically heterogeneous, with all three modes of Mendelian inheritance: autosomal dominant RP (adRP), autosomal recessive RP (arRP), X-linked RP (xlRP), as well as mitochondrial inheritance running in families. To date some 47 loci have been identified for nonsyndromic RP and for these 36 of the causative genes have been identified (see [RetNet](http://www.sph.uth.tmc.edu/RetNet/disease.htm) for full list). These include genes encoding components of the phototransduction cascade and the visual cycle, by which the chromophore component of rhodopsin is recycled, as well as retinal transcription factors, structural proteins, and proteins thought to be involved in intracellular transport within photoreceptors \[[@r2]\]. In addition, mutations in four ubiquitously expressed splicing factors, [p]{.ul}re-m[R]{.ul}NA [p]{.ul}rocessing factor 8 (PRPF8 \[[@r3]\]), PRPF31 \[[@r4]\], PRPF3 \[[@r5]\], and PIM1 associated protein (PAP-1) \[[@r6],[@r7]\] have also been described in dominant RP patients \[[@r8]\]. Patients with splicing factor mutations do not appear to exhibit any disease phenotype other than retinal degeneration, and the basis of this specificity remains to be determined. Splicing is a complex process that results in the precise excision of introns from pre-mRNA by a macromolecular structure called the spliceosome \[[@r9],[@r10]\]. The majority of introns are spliced by the major (U2-dependent) spliceosome, which consists of auxiliary protein factors and four small nuclear ribonucleoprotein particles (snRNPs): U1, U2, U5, and U4/U6. However, a small percentage of introns (about 1 in 700 for primates) \[[@r11]\] are spliced by the minor (U12-dependent) spliceosome, which differs from the major spliceosome in four of the snRNPs it contains (U11, U12, U4atac, and U6atac with canonical U5). Signals in intron and exon sequences further define intron and exon recognition and the execution of splicing. In both U2 and U12 introns, canonical (GT/AG) and various noncanonical consensus splice site sequences are used. U12-dependent introns are characterized by a more highly conserved branch site and lack of a polypyrimidine tract \[[@r12],[@r13]\]. It has been suggested that the splicing of U12-dependent introns may be a rate-limiting step in gene expression \[[@r14]\]. PRPF8 is a 220 kDa protein that is highly conserved both in sequence and size, varying between 220 and 280 kDa in different organisms. It is a component of the U5snRNP and U5·U4/U6 tri-snRNP \[[@r15]\]. PRPF8 can also be photochemically cross-linked to the 5′ splice site, the branchpoint and the 3′ splice site in the pre-mRNA \[[@r16]--[@r18]\] and to the U5 and U6 snRNAs \[[@r19],[@r20]\]. PRPF8 is therefore thought to be at the catalytic center of the spliceosome \[[@r21],[@r22]\]. This close association with crucial RNA sequences and proteins in the spliceosome suggests that PRPF8 could directly affect the function of the catalytic core, perhaps acting as a splicing cofactor. Unlike most other known splicing factors, as a component of the U5 snRNPs, PRPF8 participates in both U2 and U12 splicing \[[@r15]\]. Mutations in PRPF8 cause a severe form of dominant RP \[[@r3]\]. To date, 16 different mutations, including missense, premature stop and deletions, have been identified \[[@r3],[@r23]--[@r26]\]. They cluster in a highly conserved region within the last exon, encoding the C-terminus of the protein, making it unlikely that these transcripts will be removed by nonsense-mediated decay \[[@r27]\]. This is supported by real-time PCR analysis that has shown similar expression of both wild-type and mutant alleles in cell lines from RP patients carrying a nonsense mutation \[[@r28]\]. This remarkable clustering of mutations and the relatively high number of missense changes could imply that the mutations alter a specific domain of particular importance to retinal function. Crystal structural analysis of the last 273 amino acids of *Caenorhabditis elegans* PRPF8 shows that the RP mutations are contained in a C-terminal extension that is not thought to affect the overall structure of this domain \[[@r29]\]. However, it is hypothesized that the C-terminal peptide (2310--2335) constitutes a direct binding surface between PRPF8 and other partners, and mutations on this surface could therefore lead to disruption of critical protein--protein interactions necessary for function \[[@r29]\], a finding recently confirmed by Boon et al. \[[@r30]\]. PRPF31 is a 61 kDa splicing factor first identified in yeast \[[@r31]\]. It is a U4/U6 snRNP-associated protein that promotes tri-snRNP association between U4/U6 snRNP and U5 snRNP by direct interactions with a 102 kDa U5-specific protein \[[@r32]\]. At present, there are 19 known RP-causing mutations in PRPF31. Only three of these are missense changes while the remainder encode severely truncated or grossly abnormal transcripts (six deletions ranging from one base pair to the whole gene, five splice-site mutations, two insertion/deletion events, one duplication, and one insertion) \[[@r24],[@r33]--[@r41]\]; these changes may be targeted by nonsense-mediated decay \[[@r27]\]. In support of this, allele-specific measurement of PRPF31 transcripts have revealed a strong reduction in the expression of mutant alleles, with no truncated proteins being detected. Blocking of nonsense-mediated mRNA decay restored the amount of mutant PRPF31 mRNA but did not restore the synthesis of mutant proteins \[[@r42]\]. This suggests that mutations in PRPF31 cause RP due to haploinsufficiency, an observation further supported by the finding that high-expressing alleles of *PRPF31* from the normal parent compensate for a potentially RP-causing mutation on the opposite chromosome \[[@r43]\]. This phenomenon explains the unusually high level of nonpenetrance associated with RP due to mutations in *PRPF31* \[[@r44]\]. Three hypotheses can be put forward to explain how mutations in splicing factors might cause RP in the absence of any other defect. The first is that mutations specifically affect the splicing of one or several retina-specific genes. An obvious target would be rhodopsin (RHO), which makes up 80% of the protein content of the rod outer segment \[[@r45]\]. The constant process of disc shedding and renewal of the rod outer segments means that demand for *RHO* mRNA might be high, which could be met through new synthesis as well as a function of mRNA stability. Consistent with this first hypothesis reduced splicing of an exon 3--4 minigene construct of rhodopsin was associated with cells transfected with a mutated version of PRPF31 \[[@r46]\]. A second hypothesis would be that mutations cause a generalized splicing deficiency but this is only pathological in the fast-metabolizing, end-differentiated retina \[[@r47]\]. This could be consistent with a defect in the translocation of these splicing factor proteins to the nucleus, as proposed by Deery et al. \[[@r48]\], or with a deficit in U12-dependent spliceosome function, as proposed by McKie et al. \[[@r3]\]. A third hypothesis is that PRPF8 and PRPF31 have another function unrelated to mRNA splicing, as proposed for another disease-implicated splicing factor, survival of motor neuron 1 (SMN1) \[[@r49]\]. The finding of mutations in four spliceosomal components leading to adRP would however tend to suggest involvement of the whole tri-snRNP, rather than separate components of it, in this alternative function. Hypotheses one and two, that the RP-causing mutations cause splicing deficiency, may be more consistent with the haploinsufficiency mechanism proposed for PRPF31 mutations, rather than the dominant negative effect implied by the clustered missense mutations found in PRPF8. In this study we compared the efficiency with which U2 and U12 introns (both canonical and noncanonical) were spliced in cell lines derived from RP patients carrying PRPF8 or PRPF31 mutations and controls. The study also aimed to determine whether RP-causing mutations resulted in measurable splicing defects in lymphoblastoid cell lines derived from these patients. Methods ======= Cell culture ------------ Samples of blood from individuals carrying mutations in PRPF8 were collected by D. Mackey (Melbourne, Australia) and R. Ramesar (Cape Town, South Africa; see [Table 1](#t1){ref-type="table"} for details) and deposited with European Collection of Cell Cultures (ECACC; Salisbury, UK), where cell lines were then prepared by EBV transformation. All individuals with PRPF8 mutations had RP. Individuals recruited were otherwise healthy. Peripheral venous blood was obtained from patients and other family members by venepuncture using lithium heparin or sodium citrate vacutainers. Samples were stored and transported at room temperature for a maximum of 5 days before immortalisation with EBV. All blood samples were collected from patients after obtaining informed consent from each participant and receiving local ethics committee approval. The remaining cell lines were obtained directly from ECACC. These included cell lines from patients and asymptomatic carriers with PRPF31 mutations and also unaffected relatives of PRPF31 mutation carriers, who were used as controls. The individuals from whom these lines were prepared were again otherwise healthy. Cells were grown in RPMI1640 medium with L-glutamine supplemented with 10% heat-inactivated FBS (Gibco, Carlsbad, CA) at 37 °C in 5% CO~2~ and maintained according to ECACC protocols. ###### Cell lines used in this study with their corresponding mutations and patient characteristics **Group** **Cell line** **Mutation** **Sex** **Age** ---------------- ---------------------- ----------------------------------- --------- --------- Control AG0318 Control M 22 AG0300 Control F 32 AG0309 Control F 34 AG0267 Control M 46 AG0326 Control F 58 AG0261 Control F 77 AGO296 Control F 56 PRPF31 Carrier AG0353 Exon 11, 11 bp Del, nt 1115--1125 M 22 AG0311 Exon 11, 11 bp Del, nt 1115--1125 F 23 AG0298 Exon 11, 11 bp Del, nt 1115--1125 M 58 PRPF31 Severe AG0306 Exon 11, 11 bp Del, nt 1115--1125 F 47 AG0325 Exon 11, 11 bp Del, nt 1115--1125 M 56 HG0003 IVS 6 +3 A\>G F 62 HG0005 Whole gene deletion F 65 AG0316 Exon 11, 11 bp Del, nt 1115--1125 F 72 PRPF8 HG0007 H2309P F 33 HG0008 H2309P F 67 HG0009 HG0011 HG0012 H2309P F 72 H2309R F 60 H2309R F 30 Samples were collected by D.Mackey (Melbourne, Australia), or R.Ramesar (Cape Town, South Africa), or obtained from the ECACC as stated in Materials and Methods. RNA extraction and cDNA synthesis --------------------------------- Total RNA was extracted from pelleted cells using the Qiagen RNeasy extraction kit according to the manufacturer's protocol. The process included a 15-min treatment with DNaseI (Qiagen) to digest any contaminating genomic DNA. Next, 1 µl of RNA was mixed with 1 µg random primer (Invitrogen, Glasgow, UK) and 10 µl nuclease-free water. This mixture was incubated at 70 °C for 10 min and chilled on ice. To each reaction was added 8 µl of master mix: 4 µl 5X Moloney Murine Leukemia Virus Reverse Transcriptase (M-MLV RT) buffer, 2 µl 0.1 M dithiothreitol (DTT), 2 µl 10 mM deoxynucleosides (dNTPs; all Invitrogen), and 0.25 µl RNAsin (Promega, Hampshire, UK). This then was equilibrated at 37 °C for 2 min. Next, 0.5µl Moloney Murine Leukemia Virus (M-MLV) reverse transcriptase (Invitrogen) was added to each reaction, followed by incubation at 37 °C for 1 h and at 95 °C for 5 min. For each cell line a control cDNA synthesis without reverse transcriptase was performed and the presence of genomic DNA was excluded by failure to PCR with GAPDH primers shown in [Table 2](#t2){ref-type="table"}. ###### Primers used in real time PCR analysis **Primer name** **Sequence (5′-3′)** **Primer name** **Sequence (5′-3′)** ---------------------- -------------------------- ---------------------- ------------------------ GAPDH ref F AACAGCGAGACCCACTCCTC APRT ex 2--3 F GGGCCGCATCGACTACAT GAPDH ref R CATACCAGGAAATGAGCTTGACAA APRT ex 2--3 R AGCCCAGTCCAAGCTCCTG 18S F CGGCTTTGGTGACTCTAGATAACC APRT in 2--ex 3 F AACCAGGTACCCCTTGCCAC 18S R AAGTTGATAGGGCAGACGTTCG APRT in 2--ex 3 R CAAAGAGGAAGCCTCGGGAG β-actin ex 2--3 F CCTGGCACCCAGCACAA PLCδ ex 3--4 F AGATCATCCACCACTCAGGCTC β-actin ex 2--3 FR CCGATCCACACGGAGTACTT PLCδ ex 3--4 R TGTCAGCTTTTCGCAAGCAG β-actin in 2-ex 3 F AGCTGTCACATCCAGGGTCC PLCδ in 3--ex 4 F CTTGGTAGGTTCCAGGGTTCCT β-actin in 2-ex 3 FR CCGATCCACACGGAGTACTT PLCδ in 3--ex 4 R TTGTCAGCTTTTCGCAAGCA PRPF8 ex 25--26 F AGGCAAGAGGCCATTGCTC STK11 ex 3--4 F TGAGGAGGTTACGGCACAAAA PRPF8 ex 25--26 R GATTCGAGGAATGCCACGAT STK11 ex 3--4 R GCCACACACGCAGTACTCCAT PRPF8 ex 25--in 25 F AAGAGCCCATTGCTCAGAACA STK11 in 3--ex 4 F AGCTGTGTGTCCTTAGCGCC PRPF8 ex 25--in 25 R ACTCCACACGGTTCAAAGGC STK11 in 3--ex 4 R GCACACTGGGAAACGCTTCT RPGR ex 10--11 F TCTATCAGCACGTATGCGGC C3G ex 15--16 F TCCCCAGAGGAGCTCATCAA RPGR ex 10--11 R AAACAAGCAGAAAGGCCAAGAG C3G ex 15--16 R TTGCTGACGCGCTTCTTG RPGR in 9--ex 10 F TTCTGTGGATTTATGCTGCAGG C3G in 15--ex 16 F GCCGACCACATGGCTATTTC RPGR in 9--ex 10 R TTTTGCCACACCACGATGAG C3G in 15--ex 16 R GACGCGCTTCTTGAATGTGTC ZNF198 ex 8--9 F GCGCCAAGTGATATTCAGTTGA AP2A1 ex 19--20 F CCTCGGTGCAGTTCCAGAAT ZNF198 ex 8--9 R TGCTGCAGAACTGATGCACTTT AP2A1 ex 19--20 R GCTTGGTCTGCACAGCCAG ZNF198 in 8--ex 9 F TGCTTCTGTAAAAGTGGCGTGT AP2A1 ex 19--in 19 F CTCGGTGCAGTTCCAGAATTTC ZNF198 in 8--ex 9 R TTTGCTGCAGAACTGATGCAC AP2A1 ex 19--in 19 R GACGTGGAGAGGCAGGAGG The table lists the primer name and corresponding amplified exon. Primers were designed using Primer Express 1.5 (ABI) as described in Materials and Methods. Real-time PCR ------------- Ten introns (in ten different genes) were selected for analysis to represent both U2 and U12 introns with canonical and noncanonical donor and acceptor sites. Primer Express 1.5 software (ABI Prism 7000 User's Guide, Applied Biosystems, Foster City, CA) was employed to design primer pairs. Pairs spanned two adjacent exons or an exon and adjacent intron and were used to amplify a product of 80--140 bp, with an annealing temperature between 58 and 60 °C (see [Table 2](#t2){ref-type="table"} for complete primer list). Primer pair concentrations were optimized for real-time PCR analysis according to (ABI, Applied Biosystems) recommendations. GAPDH and 18S were used as endogenous references. cDNAs of highly expressing transcripts were diluted to obtain comparable levels of cycle threshold (C~t~). Real-time PCR was performed using the SYBR Green PCR Core Reagents Kit (ABI, Applied Biosystems) as described by Ponchel et al. \[[@r50]\], and all samples were run in duplicate on an ABI PrismTM 7700 Sequence Detector System. Expression levels of each exon-exon and exon-intron were normalized to GAPDH using a standard ΔC~t~ method. Statistical analyses -------------------- The ratio of spliced to unspliced RNA was calculated for each data point as relative amount of exon-exon product divided by relative amount of exon-intron product. Descriptive statistics on the ratios for each gene and cell line were performed and graphs were produced in Excel. For each cell line the replicates were replaced by their respective means for four groups: normal cells; those from PRPF31 mutation carriers without RP; those from affected patients with PRPF31 mutations; and those with PRPF8-RP. These were compared using one-way ANOVA to detect any overall difference between the four groups, and the LSD post-hoc multiple comparison test was applied to examine individual intergroup differences, using the SPSS statistics base 17.0, SPSS inc, Chicago, IL. In addition, multivariate tests were performed in SPSS to analyze inherent group differences between the control and the mutant groups for U2 and U12 introns, and further, a one-sided bootstrap test was performed. Serum starvation ---------------- Cells were incubated in a serum-free medium for 24 h. The medium was then replaced with serum-containing medium. Samples were taken at 0, 2, 4, 6, 12, 24, and 72 h after another addition of serum. Total RNA was then extracted and cDNA synthesized for real-time PCR analysis as described in the previous section. Transfections ------------- A plasmid minigene containing mouse retinal outer segment membrane protein-1 (Rom1) intron 2 was constructed by Dr J. Wu (Saint Louis University, St. Louis, MO) and was obtained as a gift from Professor Eric Pierce, Scheie Eye Institute, Philadelphia, PA. Two donor site sequence changes were introduced using the Stratagene (La Jolla, CA) site-directed mutagenesis kit (according to manufacturer's instructions as described above), which altered the wild-type canonical GT site to AT and GA. Cells were transfected using the Effectene transfection reagent (Qiagen) following the manufacturer's protocol. Results ======= Pre-mRNA splicing in lymphoblastoid cell lines derived from patients with PRPF8 and PRPF31 adRP ----------------------------------------------------------------------------------------------- To determine whether mutations in splicing factors associated with RP cause detectable splicing defects in non-retinal tissues, we obtained lymphoblastoid cell lines from four groups of individuals: RP patients carrying the H2309P or H2309R mutation in PRPF8; those with the R372_A375delfs mutation in PRPF31 who have severe RP; those carrying the PRPF31 R372_A375delfs PRPF31 mutation but not manifesting RP \[[@r4]\]; and controls (see [Table 1](#t1){ref-type="table"}). Ten exon-intron-exon sequences were chosen for analysis. These included four "house-keeping" genes containing U2 introns with consensus splice sites, β-actin (*ACTB*), glyceraldehyde 3-phosphate dehydrogenase (*GAPDH*), the splicing factor *PRPF8* gene itself, and *RPGR*, mutations in which also lead to RP \[[@r51]--[@r53]\]. Also included were two genes containing U2 introns with non-consensus splice sites, *ZNF198* (a zinc-finger protein encoding gene), and adenine phosphoribosylase (*APRT*) . Finally four genes were selected containing [U12](http://www.sanger.ac.uk/Users/rd/U12) introns, one with a consensus GT/AG splice site from phospholipase C delta (*PLCδ*), and three with non-consensus splice sites, serine-threonine kinase 11 (*STK11*), ras guanine nucleotide releasing factor C3G and adaptor-related protein complex 2 α-1 (*AP2A1*; see [Table 3](#t3){ref-type="table"}). The U12 introns were chosen from a database of confirmed [U12](http://www.sanger.ac.uk/Users/rd/U12) introns \[[@r54]\]. ###### Genes selected for analysis of splicing **Intron type** **Splice site sequence** **Gene** **Exons** ----------------- -------------------------- ---------- -------------------- U2 GT/AG β-actin 2--3 GT/AG GAPDH 7--8 GT/AG PRPF8 25--26 GT/AG RPGR 10--11;in 9--ex 10 GC/AG ZNF198 8--9 GC/AG APRT 2--3 U12 GT/AG PLCδ 3--4 AT/AC STK11 3--4 AT/AC C3G 15--16 AT/AG AP2A1 19--20 This table summarizes gene intron type, splice site sequence and amplified exon for each gene used in analysis. The splicing efficiency of these introns in control and patient cell lines was determined by examining ratios of spliced versus unspliced transcripts using real-time PCR. Relative levels of exon-exon product were divided by relative levels of exon-intron product to produce an estimate of splicing efficiency. Values for the three groups of cell lines were then compared with values obtained in control cell lines, and results analyzed using ANOVA and LSD tests as described in the previous section. Results are shown in [Figure 1](#f1){ref-type="fig"}, [Figure 2](#f2){ref-type="fig"}, and [Table 4](#t4){ref-type="table"}. ![**Splicing ratios of U2.** Splicing ratios of U2 containing introns *RPGR* (**A**) and *GAPDH* (**B**) RNA in cell lines derived from control and splicing-factor mutated cell lines. Group 1 represents controls, group 2 represents PRPF31 carriers, group 3 represents PRPF31 severe, and group 4 represents PRPF8 cell lines. The splicing ratios were obtained as described in the Methods section. Means from four to six repeat experiments are shown ±SEM. For individual cell line characteristics see [Table 1](#t1){ref-type="table"}.](mv-v14-2357-f1){#f1} ![**Splicing ratio of U12.** Splicing ratio of U12 containing intron *STK11* RNA in cell lines were derived from control and splicing-factor mutated cell lines. Group 1 represents controls, group 2 represents PRPF31 carriers, group 3 represents PRPF31 severe, and group 4 represents PRPF8 cell lines. The splicing ratios were obtained as described in the Methods section. Means from four to six repeat experiments are shown ±SEM. For individual cell line characteristics see [Table 1](#t1){ref-type="table"}.](mv-v14-2357-f2){#f2} ###### Ratios of spliced to unspliced RNA in groups of control and splicing factor mutant cell lines for selected pre-mRNA introns **Intron and splice site** **Gene** **Control** **PRPF31 carrier** **PRPF31 severe** **PRPF8** ---------------------------- ---------- ------------- -------------------- ------------------- ------------------ U2 GT/AG ACTIN 328 406 315 402 U2 GT/AG GAPDH 1683 1385 **2993 p\<0.05** 1433 U2 GT/AG PRPF8 198 173 184 140 U2 GT/AG RPGR 78 105 **23 p\<0.05** 95 U2 GC/AG ZNF198 154 166 128 124 ↓ U2 GC/AG APRT 8 10 7 6.5 ↓ U12 GT/AG PLCd 18 19 21 9 ↓ U12 AT/AC SKT11 39 37 32 **28 p\<0.05 ↓** U12 AT/AC C3G 17 24 9 13 ↓ U12 AT/AG AP2A1 36 39 40 29 ↓ Significant differences obtained by the LSD test are highlighted in bold. A significant increase in splicing efficiency was observed for GAPDH in PRPF31 severe cell lines. In contrast a significant decrease in splicing efficiency was observed for RPGR splicing in the same cell line. In PRPF8 cell lines a significant decrease in splicing efficiency was seen for SKT11, as well as an overall decline in splicing for all U12 introns and non-canonical U2 introns. U2 introns ---------- No significant differences in splicing ratios were observed between control and PRPF8 mutant cell lines for any U2 introns. There were also no significant differences in splicing ratios of U2 introns between control cell lines and those deriving from asymptomatic patients with PRPF31 mutations. In contrast, PRPF31 mutant cell lines from patients with a severe phenotype showed significantly lower splicing ratios for *RPGR* with respect to controls ([Figure 1](#f1){ref-type="fig"}) and significantly higher splicing ratios for *GAPDH*. These data are summarized in [Table 4](#t4){ref-type="table"}. U12 introns ----------- None of the splicing ratios assayed for U12 introns in PRPF31 mutant cell lines, either those from RP patients or from asymptomatic carriers, showed significant differences from those obtained in control cell lines. However, PRPF8 mutant cells showed significantly decreased splicing ratios for intron 3 of SKT11 with respect to controls ([Figure 2](#f2){ref-type="fig"}). *AP2A1*, *PLC-delta*, and *C3G* introns did not show statistically significant differences between control and PRPF8 mutant cell lines; however, a general trend could be seen toward a decrease of splicing ratios in PRPF8 mutants compared to controls for all U12 introns. This trend also extended to the two noncanonical U2 introns. These data are summarized in [Table 4](#t4){ref-type="table"}. Multivariate tests ------------------ Multivariate tests were performed to investigate these findings. These tests failed to ascertain the apparent decrease in splicing ratios in PRPF8 mutant cells for U12 introns. This could be due to the sample sizes being too small, or to the parametric test relying on incorrect assumptions. Therefore a bootstrap test was performed, in which the data were resampled, a test statistic was computed, and the results were compared to the data test statistic. The one-sided p-value for the U12 introns comparing controls versus PRPF8 mutants was 0.067, whereas the p-value for controls versus PRPF31 severe was much higher (0.30). This confirmed that the trend toward decreased splicing deficiency seen for PRPF8 mutant cell lines (which just fell short of significance) was not reproduced in PRPF31 mutant cell lines. Analysis of splicing ratios under high metabolic demand ------------------------------------------------------- One possible explanation for the retina-specific manifestations of splicing-factor RP is the high metabolic rate of photoreceptors \[[@r47]\] and the consequent inability of the mutated PRPF8 or PRPF31 to cope with the accompanying splicing demands; this might become especially apparent for rate-limiting U12 intron splicing \[[@r14]\]. To investigate whether splicing is affected during periods of high demand on the splicing apparatus, we starved control and patients' cell lines of serum for 24 h. The splicing efficiency was investigated by real-time PCR following the reintroduction of serum (at 0, 2, 4, 6, 12, 24, and 72 h), which would be expected to result in a substantial increase in the rate of cell division and growth. No significant differences were found (data not shown). Study of the splicing of transfected retina-specific gene, *Rom-1* ------------------------------------------------------------------ In a further attempt to determine if splicing could be affected by conditions of high splicing demand, we transfected cell lines with plasmids encoding ROM-1 (exon 1 to exon 2), both in the wild-type form and with mutations introducing a noncanonical splice donor sites (canonical GT mutated to noncanonical AT and GA). The spliced versus unspliced ratios of all constructs were again measured by real-time RT--PCR. No significant differences in splicing efficiency between the control and splicing factor mutant cell lines for any of these constructs were detected (data not shown). Discussion ========== These experiments aimed to distinguish between two hypotheses: that splicing factor mutations cause RP due to a retina-specific defect, either in splicing or in some other function involving these proteins; or that splicing factor RP results from a defect in splicing that is subpathological elsewhere in the body but which leads to a cumulative defect in the metabolically active, irreplaceable retinal tissue. To distinguish between these hypotheses, we assayed the ratio of spliced to unspliced RNA as a measure of splicing efficiency in lymphoblastoid cell lines derived from splicing factor RP patients and normal control cell lines. This assay was applied to a series of introns representing a range of intron types. Results did reveal significant differences in splicing ratios in these cells, but these differences were not consistent between the different splicing factor genes. In cells from PRPF31-RP patients, we observed a decrease in the splicing efficiency of *RPGR* intron 9 and an increase in the splicing efficiency of *GAPDH* intron 7. These are both U2 introns with canonical donor and acceptor sites. In contrast, cells from patients with PRPF8-RP showed no significant differences in splicing ratios for U2 introns but showed a consistent downward trend in splicing efficiency of all U12 and noncanonical U2 introns assayed (statistical significance was only attained for intron 3 of *STK11*). Cells from asymptomatic carriers of PRPF31 mutations revealed no significant differences in splicing ratio compared with control cells, consistent with the findings of Vithana et al. \[[@r43]\], who showed that high expression of the second allele of PRPF31 compensated for the haploinsufficiency. The lack of consistency between findings in the cell lines from patients with the two different mutant genes for splicing factor may be the result of the small number of cell lines assayed, together with the relatively high level of individual variation in splicing efficiency between cell lines. The finding of a difference in splicing efficiency in an intron of *RPGR* may be significant. RPGR is a putative guanine nucleotide exchange factor that has been localized to the Golgi region \[[@r55]\], and is thought to be important for maintaining the polarized protein distribution across the connecting cilium of the photoreceptors \[[@r56]\]. Mutations in *RPGR* (*RP3*, Xp21.1) are responsible for up to 70% of X-linked retinitis pigmentosa \[[@r51]\] and also cause X-linked cone-rod dystrophy (*CORDX1*) \[[@r52]\] and atrophic macular degeneration \[[@r53]\]. However, RPGR is ubiquitously expressed, and several alternatively spliced isoforms of RPGR have been found in the eye and other tissues, resulting in product sizes ranging between 5 and 20 kb \[[@r57],[@r58]\]. One of these isoforms contains a retina-specific exon, ORF15, which has been found to be a mutational hotspot in RP patients \[[@r59],[@r60]\]. The intron tested herein is present in this and other isoforms expressed in the eye, but is absent from others also present in the eye. Correct splicing of *RPGR* may be particularly important in the retina; mutations in PRPF31 may have a negative effect on the splicing of *RPGR* pre-mRNA, which could potentially contribute to the RP phenotype in these patients. In the light of this finding it may now be informative to look at splicing efficiency in other ubiquitously expressed genes which are mutated in RP. The one significant difference in splicing between control and PRPF8 mutant cells was found within the U12 intron-containing gene *STK11*. Furthermore, although no other significant differences were observed, there was a consistent trend toward an overall decrease in splicing efficiency of all U12 introns tested. U12 splicing is thought to be the rate-limiting step in the processing of any transcript containing a U12 intron \[[@r14]\]. In the fast-metabolizing retina, with a corresponding high demand placed on the splicing machinery, a reduction in splicing efficiency of U12 intron-containing transcripts could become significant over time, while other tissues remain unaffected. Microarray analysis by Gamundi et al. \[[@r28]\] in PRPF8 mutant lymphoblastoid cell lines showed downregulation of one U-12 type gene, *SLC12A7*, but failed to ascertain a pattern of downregulation of the minor U12-dependent splicing genes as a group. However that study used microarray analysis to examine transcript levels, whereas this study examined intron specific splicing efficiency, potentially a more sensitive method for detecting such an effect. It is also interesting to note that the splicing ratio of *GAPDH* intron 7 was significantly increased in cells from PRPF31-RP patients relative to controls. GAPDH, as well as being a major enzyme of the glycolytic pathway, has several other functions, such as regulation of gene expression \[[@r61]\]. More important, GAPDH has been found to have a role in the apoptosis of retinal cells \[[@r62]\]. The endpoint in many forms of retinal degeneration is apoptosis of the retinal photoreceptors, so this difference could also contribute to the development of an RP phenotype. Given the relatively small number of cell lines investigated and the small number of splice sites chosen, the substantial variation in splicing efficiency between cell lines (control and disease groups) and the inconsistency of differences found between the different genes for RP, these data must be interpreted with caution. Nevertheless, this study has recorded statistically significant differences in splicing efficiency in cell lines from patients with two different mutated genes for splicing-factor RP. This favors the hypothesis that RP may indeed result from a global deficit in splicing, detectable outside the retina, but which only causes pathology in the retina. Further elucidation of the mechanisms underlying splicing-factor RP will play an important role in the search for therapeutic agents. We acknowledge funding from Wellcome Trust (grant 073988 to C.F.I. and grant 067311 to R.J.G.). [^1]: The first two authors contributed equally to this work.

{ "pile_set_name": "PubMed Central" }

Introduction ============ Autism spectrum disorder (ASD) is a heterogeneous group of neurodevelopmental disorders. Before the Diagnostic and Statistical Manual of Mental Disorders-5 (DSM-5), which was the manual that officially labeled this term as a diagnostic category, ASD was used as a common clinical term that referred to the pervasive developmental disorders (PDD), as described in the DSM IV-TR \[[@B1]\] and the ICD-10 (10th International Classification of Diseases). Successive redefinitions of autism diagnostic criteria have not succeeded in constraining complexity and comorbidity in autism. A multidimensional point of view should encompass the issues posed by this categorical approach. From this perspective, child development is shaped by the interaction between several dimensions (e.g., language, motor, cognition, emotion; [@B53]). In the present study, we focus on emotion recognition, taking into account uni- and multimodal stimuli as well as other developmental comorbidities that are not included in ASD criteria but that are frequently associated with ASD, specifically language, fine motor, and neuro-visual skills. Since [@B28] first clinical description of autism, problems related to emotion processing have been seen as a hallmark symptom of the disorder. However, the status of emotion impairments remains, until now, uncertain. Research that examines emotion recognition in ASD has been limited by an over-focus on the visual modality, specifically the recognition of emotion in facial expressions. Several studies have reported that children with ASD display deficits in this ability ([@B19]; [@B46]; [@B24]; [@B48]). However, several studies have failed to find significant differences in emotion recognition tasks when comparing children with ASD and controls ([@B6]; [@B39]; [@B42]; [@B8]; [@B21]; [@B40]; [@B9]; [@B27]). Additionally, the nature of the deficits in emotion processing has been discussed, but with limited agreement among experts. The difficulties that children with ASD have with emotion processing may relate to their inability to recognize specific emotions, such as fear and disgust ([@B26]; [@B50]), anger ([@B2]), or sadness ([@B16]; [@B50]). At the same time, other authors have failed to find any deficits in the recognition of negative emotions ([@B31]). Overall, emotion recognition deficits do not appear to be universal in ASD, as reflected by heterogeneous performance across children and across tasks (for a review, see [@B37]). Studies using eye-tracking technology have examined the scanning of emotional faces in children with ASD and have found that they globally spend less time on core features (i.e., eyes, nose, and mouth) as compared to typically-developing (TD) children (e.g., [@B18]). However, more recent studies have found that deficits in emotion recognition cannot be fully explained by differences in face scanning, as children with ASD showed no particular differences in time fixation on faces ([@B44]) or showed normative pupillary reactions to emotion expressed by familiar people ([@B38]). Overall, the idea of gaze abnormalities in ASD, including less time spent focusing on emotional cues (e.g., facial expressions), remains controversial (for a review, see [@B22]; [@B55]). Understanding emotional states in real life involves processing a variety of cues that include verbal content, non-verbal cues (e.g., postures), non-verbal vocalization, and affective prosody ([@B10]). The recognition of emotion from vocalizations, which appears to be the auditory equivalent of facial emotion recognition, has been less studied, with limited and inconsistent findings. A few studies have found that, compared to typically developing children, children with ASD demonstrate impaired auditory emotion expression recognition skills ([@B41]; [@B11]). On the other hand, others have found no evidence of a fundamental deficit in ASD ([@B3]; [@B27]). According to the "weak central coherence theory," which was conceptualized by [@B23] and corresponds to the difficulties that children with ASD have when synthesizing stimuli into a coherent whole, children with ASD have impairments in multisensory processing ([@B4]; [@B34]; [@B36]; [@B35]; [@B43]; [@B14]). [@B11] noted that both ASD and TD children benefited from exposure to bimodal information, but to a lesser extent in the ASD group than the TD group. [@B49] explored the processing of neutral and emotional human stimuli (in the auditory, visual, and multimodal channels) in children with PDD-Not Otherwise Specified (NOS), when compared to TD children. The PDD-NOS group experienced difficulties with processing emotional stimuli, particularly in the visual modality. They also more easily identified happy, rather than angry or neutral, faces, and vocalizations. The children with PDD-NOS used the multimodal channel to compensate for their unimodal deficits. Similarly, [@B27] studied emotion recognition abilities by using a combination of visual and auditory tasks for two groups of adolescents, with and without ASD. They found that IQ had a large and significant effect on performance, but they also found no evidence of a fundamental deficit with emotion recognition in the adolescents with ASD. The discrepancies in the conclusions of the different studies that have been described above are underpinned by several factors: (1) the use of different experimental designs (e.g., only visual tasks) and data analyses (e.g., only univariate analyses); (2) the sizes and characteristics of the samples; (3) the heterogeneity among ASD patients in terms of developmental course requiring the consideration of developmental age when comparing to TD children; and (4) the heterogeneity among ASD patients in terms of co-occurring impairments that *per se* could affect emotion recognition such as visual-motor impairments or developmental coordination disorder (DCD) for visual tasks and language deficits for auditory tasks. The aim of the current study was to explore unimodal and multimodal emotion processing in children with ASD by comparing them to TD children and taking into account other developmental comorbid factors such as language, fine motor, and neuro-visual impairments (using eye-tracking technology for this last dimension). In order to understand the nature of the heterogeneity and discrepancies in the results between studies mentioned above, we chose to take into account comorbid factors and to assess a potential age effect by including a sample of children with ASD in a wide age range (6--13). We hypothesized that: (1) consistent with the literature, we expect to find significant heterogeneity in children with ASD, with regard to their emotional identification skills; (2) in comparison to the TD group, children with ASD will have difficulties processing emotional information, which will be most prevalent in the unimodal channel (visual or auditory) and partially alleviated in the multimodal condition; (3) because ASD presents atypical eye movements and fine motor skills, we expect impairments in the visual modality to be larger in children with ASD; and (4) similarly, we expect that the subgroup of patients with language disorders will be more impaired in regards to unimodal auditory stimuli. Materials and Methods {#s1} ===================== Participants ------------ A total of 19 children with ASD and 19 typically developing children (14 boys and 5 girls in each group) participated in this study. ASD was used as a common clinical term that referred to the PDD that were described in the ICD-10 classification system ([@B52]). Two trained child psychiatrists (JX and DC) clinically assessed the children with ASD (mental age range = 5.8--13.3 years, *M* age = 7.74 years, *SD* = 2.51). In reference to the PDD diagnostic category, 5 children satisfied the diagnostic criteria for autism, and 14 children satisfied the criteria for atypical autism (i.e., presence of abnormal or impaired development before the age of three and abnormalities in reciprocal social interactions or in communication without fulfilling the full criteria for autism). Children with ASD were recruited via the Child and Adolescent Psychiatry department at Pitié-Salpétrière Hospital and the University Pierre-et-Marie-Curie in Paris, France. **Table [1](#T1){ref-type="table"}** summarizes the socio-demographic and clinical characteristics of the patients. Each child with ASD was individually matched, according to developmental age, with a healthy typically developing child using chronological age (TD: age range = 6--13 years, Mean age = 8.84 years, *SD* = 1.79). The 19 TD children were recruited via a local primary school. The study was conducted in accordance with the hospital's Research Ethics Board regulations. After being fully informed about the study, parents or legal caregivers provided written consent. ###### Main characteristics of children with autism spectrum disorder (ASD; *N* = 19). --------------------------------------------------- --------------- Chronological age, mean (±SD) 9.95 (1.75) Male/Female 14/5 Socio-Economic Status: high/middle/low 10/5/4 *Children's Global Assessment Scale, mean (±SD)* IQ score 78.5 (21.02) Mental age (IQ^∗^ age/100) 7.74 (2.51) ADI-R scores Social impairment 10.11 (5.70) Verbal Communication 7.06 (4.22) Repetitive interest 2.72 (2.56) VABS for Children (socialization domain) Interpersonal relationships 34.59 (8.39) Play and leisure time 21.25 (6.61) Coping skills 15.06 (5.98) *Total score* 70.63 (18.17) *Diagnosis (ICD-10)* Atypical autism 14 Autism disorder 5 Developmental disorder of speech and language 9 Developmental Coordination Disorder (DCD) 12 --------------------------------------------------- --------------- VABS, Vineland Adaptive Behavior Scales; WISC, Wechsler Intelligence Scale for Children. During a 1-week period of clinical testing, each child was given a series of clinical assessments: the Autism Diagnostic Interview-Revised (ADI-R) was used to score autism core symptoms ([@B32]); the Children's Global Assessment Scale (CGAS; DSM-5, 2013) was used to score global severity; and Vineland Adaptive Behavior Scales were used to score psychosocial/adaptive functioning ([@B47]). The cognitive quotient was ascertained by using the WISC-IV (Wechsler Intelligent Scale for Children-IV), the WPPSI (Wechsler Preschool and Primary Scale of Intelligence), or the KABC-II (Kaufman Assessment Battery for Children, second edition). Developmental age was calculated on this basis to match patients with TD children while taking into account possible intellectual deficits. To evaluate the existence of clinical developmental comorbidities, patients also received (i) a speech and language assessment using the ELO Battery (*Evaluation du Langage Oral*: Assessment of Oral Language; [@B29]) by a speech therapist, and (ii) a psychomotor assessment using the M-ABC (Movement Assessment Battery for Children; [@B25]) performed by an occupational therapist. Stimuli ------- As described by [@B33], the stimuli used in the present study were categorized as visual, auditory, or bimodal (**Figure [1A](#F1){ref-type="fig"}**). *Visual stimuli* consisted of pictures of facial expressions that were obtained from the FACES database ([@B20]). The faces of six actors (three men, three women) who expressed six facial expressions (joy, fear, anger, sadness, disgust, and neutral) constituted a set of 36 visual stimuli. *Auditory stimuli* included non-verbal affective vocalizations from the Montreal Affective Voices database ([@B5]), in which actors produced emotional interjections by using the vowel /a/ (cry, laugh, etc.). The voices of six actors (three men, three women) who expressed six vocal expressions (joy, fear, anger, sadness, disgust, and a neutral) constituted a set of 36 auditory stimuli. *Bimodal stimuli* consisted of congruent combinations of an emotional face and an affective vocalization (36 bimodal stimuli). ![**Experimental setting.** **(A)** Examples of facial and vocal emotional stimuli presentations for each of the six primary emotions (e.g., "Joy" with the outline of a prosodic laugh). **(B)** Experimental procedure.](fpsyg-06-01954-g001){#F1} Eye-Tracking Apparatus ---------------------- During tasks with visual and bimodal stimuli, children's gazes were monitored by using an integrated Tobii T120 eye-tracker (Tobii Technology, Danderyd, Sweden). The T120 eye-tracker is a device that is built into a screen (17-inch) and does not require restriction of the children's heads, thus allowing them to look at the pictures freely and naturally. The system tracks both of the children's eyes separately at a rated accuracy of 0.5 degrees and a sampling rate of 120 Hz. A five-point infant calibration was used, and the experiment began after the points were correctly calibrated. By using the Tobii Studio Analysis Software, we evaluated the cumulative number of saccades (as an index of eye movements) and the durations of fixations (as an index of eye stability) on the face. Thus, one area-of-interest (AOI) region was defined for each face image including the whole face. A minimum of 50% valid gaze time was required for the analysis. In the ASD group, seven children were excluded from the eye-tracking analysis, as they did not yield data that were valid for any of the three tasks (e.g., due to excessive movement). A Wilcoxon test comparing the mental age of the 19 vs. 12 subjects remaining was not significant (*W* = 75.5, *p* = 0.122), so we are able to conclude that no bias was introduced by removing the seven patients from the analysis. Procedure --------- Before the experimental procedure, we ensured that all participants were able to understand each basic emotion (i.e., we asked each child to explain with examples of each emotion; for the ASD group this ability was confirmed by their language assessments). Children were tested individually in a single session that lasted approximately 30 min (**Figure [1B](#F1){ref-type="fig"}**). The experiment consisted of three tasks: visual (facial emotions), auditory (vocal emotions), and bimodal. The bimodal (audio--visual) stimuli consisted in the synchronous and congruent facial and vocal presentation for the same emotion. The experiment was run with E-prime software. After the statement of the set, the eye calibration and the familiarization with the four sample items, the experiment began. Each child was seated approximately 60 cm far from the screen of the eye-tracker. Each trial began with the presentation of a fixation cross (500 ms), which was followed by the presentation of the target stimulus during 3 s (i.e., temporal window for the eye-tracking recording); then, labels appeared at the bottom of the screen until the child responded. During the auditory task, no visual stimuli appears on the screen. Participants were asked to select (by clicking with the computer mouse) one label from a list choice that best described the emotion that was being expressed. The order of the three tasks was counterbalanced across children, and the order of trials was pseudo-randomized across each task. There was an inter-trial interval of 700 ms, and a resting pause was offered after every ten trials. Correct answers and the eye-tracking data (the number of saccades and fixation durations on the faces in the visual and bimodal tasks) were recorded. Statistical Analysis -------------------- The data for the present study were Analysed using the statistical program R, version 2.12.2 (R Foundation for Statistical Computing), with two-tailed tests and a 95% confidence level. Due to the repeated measures design and the forced choice paradigm (six possible answers) that was used in our experiment, we used a Generalized Linear Mixed Model (GLMM; lme4 package) to explore the data. A binomial family was specified in the GLMM model to estimate the log-odds ratio for the corresponding factors in the model. With the exception of the eye tracking experiment, we used the entire sample (19 vs. 19). In the general analysis, factors included group (ASD vs. TD), emotion (joy, neutral, anger, disgust, fear, or sadness), task (visual, auditory, or bimodal), developmental mental age (DA) and sex (model formulation: number of successes ∼ Task + Emotion + Group + DA + Sex + Random Participant factor). Then, in tasks where gaze was recorded (unimodal visual and bimodal tasks), we performed new analyses by including fixation duration and number of saccades as additional factors in the precedent model. It is not possible to compute the observed power in the case of mixed models. However, a good estimation is to compute the power for a standard regression and consider the true power to be a bit higher due to the repeated measures design. In our case, if we considered three regressors (group, stimulus, and emotion), a sample size of 38, a type I error rate of 0.05, and a large effect size of 0.35 ([@B13]), the power of the regression was 84%. A medium effect size of 0.15 gave a power of 45%, so even if the true power was \>50%, our experimental protocol did not allow us to detect small differences. A secondary multidimensional analysis was conducted in the entire sample for each task separately (i.e., visual, auditory, and bimodal), and then in each group. In the ASD subgroup, we added the following dimensional factors: Vineland total score (VABS), language disorder and DCD to assess whether such comorbidities affected participant scores during the task. Results ======= General Analysis ---------------- **Figure [2](#F2){ref-type="fig"}** shows the probabilities of correct answers being obtained by children with ASD and TD control children on the three tasks (visual, auditory, and bimodal) and for each emotion. For the two groups, the probability of success on the bimodal task was greater than on the unimodal visual task (estimate = -0.78, *p* \< 0.001), which was greater than that on the unimodal auditory task (estimate = 0.61, *p* \< 0.001). The GLMM model revealed that the rate of correct emotion recognition depended on specific emotions: joy was the most easily recognized by the two groups of children, when compared to the emotions of neutral (estimate = -1.96, *p* \< 0.001), sadness (estimate = -1.51, *p* \< 0.001), anger (estimate = -2.68, *p* \< 0.001), fear (estimate = -1.45, *p* \< 0.001) and disgust (estimate = -1.13, *p* \< 0.001). The ASD group was able to perform the multimodal task. Regardless of the emotion that was involved, their probability to succeed was over 0.7 (**Figure [2A](#F2){ref-type="fig"}**). Neutral and anger were the most difficult emotions to identify for the ASD group (estimate = -1.76, *p* \< 0.001, and estimate = -1.79, *p* \< 0.001, respectively). Regarding all of the tasks, we found no significant difference between the results that were obtained by the two groups (estimate = -0.31, *p* = 0.104), but the scores were strongly and positively associated with the children's developmental age (estimate = 0.27, *p* \< 0.001). ![**Probability of correct emotion recognition, as a function of the emotions and tasks, in ASD **(A)** and TD **(B)** children**.](fpsyg-06-01954-g002){#F2} When taking into account the combination of the two visual stimuli (visual unimodal and bimodal) and the eye-tracking variables, the ASD group performed worse than the control group (estimate = -0.75, *p* = 0.03). In addition, the discrepancy between the scores of disgust and joy (estimate = -1.27, *p* = 0.03) and between the scores of neutral and joy (estimate = -2.13, *p* = 0.001), was greater in the children with ASD than in the TD controls. There were no significant differences between joy and the three other emotions (anger, fear, and sadness) for the ASD group and the control group. Finally, the discrepancy between the results for the unimodal visual stimuli and the bimodal stimuli was less important in the ASD group than in the control group (estimate = 0.61, *p* = 0.033), which suggests that adding audio for TD children was more useful than for children with ASD. Eye-tracking data revealed that the ASD group made more saccades (estimate = 0.61, *p* = 0.002) and had shorter fixation durations (estimate = -0.3, *p* = 0.012) than the TD controls. However, the eye-tracking variables were not associated with the emotion recognition scores (for saccades: estimate = 0.0003, *p* = 0.60; for fixation duration: estimate = -0.37, *p* = 0.054). Multidimensional Analysis by Task --------------------------------- For the *unimodal visual task*, there was a significant correlation between the probability of success and the children's developmental age; the older children had greater scores. However, we found that there was no effect of sex (estimate = 0.24, *p* = 0.50), and we did not find an association between the eye-tracking data (saccades and fixation durations) and the scores of the participants (estimate = -0.0005, *p* = 0.52, and estimate = -0.46, *p* = 0.12, respectively). The children with ASD tended to perform worse than the TD group on this task (estimate = -0.59, *p* = 0.06). Within the ASD group, joy was better identified than the other emotions, including sadness (estimate = -2.01, *p* = 0.0001), anger (estimate = -1.25, *p* = 0.02), fear (estimate = -1.46, *p* = 0.008), disgust (estimate = -1.85, *p* \< 0.001) and neutral (estimate = -1.46, *p* = 0.007). Interestingly, in the ASD group, we found no correlation between the recognition scores and developmental age (estimate = 0.14, *p* = 0.18), the Vineland total score (estimate = 0.009, *p* = 0.52), or the presence of language or coordination disorders (all *p* \> 0.07). For the TD group, joy was better identified than anger (estimate = -2.36, *p* = 0.002), fear (estimate = -2.66, *p* \< 0.001), and sadness (estimate = -3.12, *p* \< 0.001). In contrast, no significant differences were found between the recognition scores for joy and the scores for neutral (estimate = -0.73, *p* = 0.41) and disgust (estimate = -1.46, *p* = 0.07). Regarding *auditory stimuli processing*, the ASD group did not perform significantly worse than the TD group (estimate = -0.08, *p* = 0.64). There was a significant effect of the children's developmental age: the older children had better scores (estimate = 0.3, *p* \< 0.001). For the ASD group, developmental age was strongly correlated with the probability of success in emotion recognition (estimate = 0.24, *p* \< 0.001) and, to a lesser extent, the Vineland total score (estimate = 0.015, *p* = 0.02) and the presence of language impairment (estimate = -0.48, *p* = 0.05). With regard to specific emotions, the two groups obtained the lowest scores, with equivalent values, for anger. Finally, during the *bimodal emotional task*, the ASD group did not perform significantly worse than the TD group (estimate = -0.62, *p* = 0.10). Joy was better recognized than anger (estimate = -1.62, *p* \< 0.001) and neutral (estimate = -0.93, *p* = 0.03). These results were similar for the ASD group. The probability of success was strongly associated with the children's developmental age: the older children had significantly higher scores (estimate = 0.6, *p* \< 0.001). By contrast, we found that there was no significant association between the probability of success in the emotion recognition task and the Vineland total score or the presence of language or coordination disorders (all *p* \> 0.51). Moreover, the probability of emotion recognition was not associated with the eye-tracking data. Discussion ========== Our study aimed to reveal the specific patterns of emotional processing in children with ASD (according to the PDD diagnostic category of the ICD-10), when compared to normally developing children, through a multimodal identification task and the assessment of developmental comorbidities. As a whole, the children with ASD performed well on the emotional multimodal tasks. Unlike [@B49] study, we did not find a significant difference between the scores that were obtained by the two groups. We also did not find a significant difference between the scores of the two groups on the auditory and bimodal tasks. The discrepancy between [@B49] results and our results may be explained by their sample, which included patients with PDD-NOS who also met the Multiple Complex Developmental Disorder criteria (i.e., patients also exhibited emotional/anxiety symptoms; [@B12]; [@B7]; [@B54]). Similar to [@B27], who tested adolescents with and without ASD by using facial and vocal emotional tasks, we found that the probability of correct emotion recognition was strongly correlated with the developmental age of the participants for each of the tasks. Interestingly, these results were confirmed in the ASD group, except for the visual task. In the ASD group, this may be explained by two factors: (1) for the unimodal visual modality, this group tends to perform worse than the TD group, but most of all, (2) the auditory modality seems to be discriminant in this association, with the strongest estimate between emotion recognition scores and developmental age being found for this task (estimate = 0.24, *p* \< 0.001) and being higher than that of the bimodal task (estimate = 0.52, *p* \< 0.001). However, [@B27] described a similar emotion processing style in the ASD and non-ASD groups, except for a difficulty in recognizing surprise. On the contrary, we found, consistent with the literature, an important heterogeneity in our results, which could be considered to be a key issue. For the entire multimodal task, we found a strong contrast between joy recognition (the emotion that was most easily identified) and the neutral emotion or anger (the most difficult to identify) for the ASD group. These results, which are congruent with the data of [@B49], are also found in the bimodal and auditory tasks' secondary analyses. The results concerning the neutral emotion could be due to abnormalities in the recognition of specific emotions in ASD children, based on a greater discrepancy being found between neutral and joy recognition in the ASD group than the TD group. Conversely, for anger in the auditory task, both groups obtained the lowest scores, which had equivalent values. The ASD group performed significantly worse than the TD group in the case of the visual and bimodal stimuli. The lowest scores were obtained by both of the groups in the auditory task; multisensory processing allowed children with ASD to partially compensate for the difficulties that were experienced in the visual modality, which confirms the results of [@B49]. This result seems to be incongruent with the impairments in multimodal processing that are described in individuals with ASD ([@B23]; [@B14]). However, according to [@B11], the discrepancy between the visual and bimodal tasks is significantly more important in the TD group than in the ASD group. These results do not necessarily contradict the "weak central coherence theory" in regard to people with ASD ([@B23]); this partial improvement in accuracy on the bimodal task could be better explained by the addition of redundant (visual + auditory) targets, rather than the multisensory integration of visual and auditory cues into a unified percept. A few studies of young adults have demonstrated that congruent emotional information that is processed via multisensory channels optimizes behavioral responses, which results in enhanced accuracy and a faster response time (RT; see [@B17]; [@B30]; [@B33]). Multisensory enhancement is sometimes explored in behavioral studies that use RT by comparing the observed RT distribution to the distribution that is predicted by a 'race model' (e.g., [@B15]). Multisensory integration occurs when the reaction time for bimodal trials is faster than what is predicted by the race model (e.g., [@B11]; [@B33]). Testing this hypothesis by using RTs in our study was not possible because the RTs were not recorded due to the particular eye-tracking procedure (i.e., forced stimuli exploration over 3 s) and the studied population (i.e., TD children and children with ASD). The eye-tracking data revealed interesting differences between the two groups (i.e., an important number of saccades and shorter fixation durations for the children with ASD than TD children). These results are consistent with studies that describe gaze abnormalities in ASD, including a shorter time being spent on emotional cues (for a review, see [@B22]; [@B55]). However, these different gaze patterns were not associated with the scores of the participants and were not able to explain the discrepancy between the performances of the two groups. It is, therefore, difficult to conclude by stating that autistic children process faces in a holistic fashion. These abnormalities could be, in part, explained by our experimental design and the eye-tracking data analysis (see limitations). Because children with DCD have poor cross-modal integration (see [@B51], for a meta-analysis) and frequent neuro-visual impairments, we expected that children with ASD and comorbid DCD would preferentially fail on the unimodal visual or bimodal visuo-auditory tasks. Our results were not consistent with this hypothesis; no association was found between the presence of a DCD and the scores on the visual or bimodal tasks. We assumed that children in the ASD group who had comorbid language disorders would be particularly impaired concerning unimodal auditory stimuli. Our hypothesis was partially correct: we found a statistical trend in this direction (estimate = -0.48, *p* = 0.05) regarding this modality. Furthermore, we did not find any association between the presence of this comorbidity and the scores for emotion recognition on the other tasks (visual and bimodal; all *p* \> 0.78). Finally, a unique significant correlation between the scores on Vineland and the performances in emotion recognition was found for the auditory stimulus. There are several limitations that warrant consideration. First, different metrics were used to match the ASD and TD groups based on age. Cognitive assessments were performed on the ASD group to assess developmental age but not on the TD group, for which chronological age was used instead. In addition, the lack of some significant effects may be due to the power of the study. Given the small sample size of our groups and considering the clinical heterogeneity of the ASD group, we were only able to detect major effect sizes and unable to detect possible subtle impacts of co-occurring factors (i.e., comorbidities) or potentially clinically meaningful effects (i.e., Vineland Adaptative Coordination Scale). Regarding the eye-tracking data analysis, we consider faces as a whole, rather than more precisely analyzing the gaze upon certain areas of interest (e.g., eyes, nose, and mouth). This choice is a limitation for the fine comparisons between children with ASD and TD children when exploring faces during an emotional task. However, our analysis was based on current, state-of-the-art machine learning methods to recognize facial emotion. Machine learning has shown that facial emotion classifiers use Action Units that are distributed all over the face to achieve the best performance ([@B45]). Conclusion ========== Children with ASD demonstrate rather high performance on emotion recognition, particularly for multimodal stimuli. The difficulties that are experienced for visual stimuli are partially alleviated when using bimodal stimuli. Developmental age plays a major role for TD children, whereas its role is limited to the multimodal task for children with ASD. However, performances in emotion recognition in ASD are heterogeneous and do not simply correlate with comorbidities. The existence of a language disorder seems to have an impact on the performances of the ASD group in the auditory modality. Conversely, in the visual or bimodal (visuo-auditory) tasks, the impact of a DCD or gaze impairments has not been demonstrated. Future studies with larger samples will allow researchers to confirm and refine these data. Given the heterogeneity of the results that are found in the literature, we wonder whether impaired emotion processing may constitute a dimension that should be explored in ASD so that we can define subgroups in this condition that are more homogeneous and tailor interventions. Conflict of Interest Statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. We are grateful to the children and their parents for generously giving their time to participate in this study. [^1]: Edited by: *Rosario Cabello, University of Granada, Spain* [^2]: Reviewed by: *Cheryl Dissanayake, La Trobe University, Australia; Ruth Ford, Anglia Ruskin University, UK* [^3]: This article was submitted to Developmental Psychology, a section of the journal Frontiers in Psychology

{ "pile_set_name": "PubMed Central" }

![](edinbmedsurgj71964-0042){#sp1 .36} ![](edinbmedsurgj71964-0043){#sp2 .37} ![](edinbmedsurgj71964-0044){#sp3 .38} ![](edinbmedsurgj71964-0045){#sp4 .39} ![](edinbmedsurgj71964-0046){#sp5 .40} ![](edinbmedsurgj71964-0047){#sp6 .41} ![](edinbmedsurgj71964-0048){#sp7 .42}

{ "pile_set_name": "PubMed Central" }

Background ========== The distal nephron plays an important role in the maintenance of sodium balance, extra cellular fluid volume and blood pressure \[[@B1]\]. It is well known that inappropriate water and sodium retention is thought to be a key factor in several forms of hypertension, and that aquaporin2 waterchannels (AQP2) play a key role in several water balance disorders \[[@B2],[@B3]\]. Animal models have shown a reduced AQP2 expression in conditions with acquired nephrogenic diabetes insipidus such as lithium treatment and an increased expression in diseases with water retention such as congestive heart failure \[[@B4]\]. Gain-of-function mutations in the epithelial sodium channels (ENaC) cause inappropriate renal sodium retention and consequent increases blood pressure \[[@B5]\]. Because AQP2 and ENaC plays such an important role in water and sodium balance and associated disorders, it is important to identify factors involved in the reabsorption of water and sodium by the kidneys in order to study these disorders in the future. The exact role of AQP2 and ENaC has never been examined under volume expansion in healthy humans. The effect of an acute intravenous volume load, with isotonic and hypertonic saline and glucose, on urinary excretion of AQP2 (u-AQP2), urinary excretion of ENaC gamma subunit fractions (u-ENaCγ) and its relationship to vasopressin (AVP) and kidney function in healthy humans has not been studied by simultaneous measurement of other important regulatory hormones of water and sodium homeostasis such as the renin-angiotensin-aldosterone system (RAAS). In this present study we wanted to study the sodium and water transport in the distal nephron by measuring 1) the excretion of u-AQP2 and u-ENaCγ 2) associated regulating hormones and 3) the renal response after volume expansion in healthy humans. In order to analyse these physiological mechanisms, we performed a randomized, crossover study in healthy subjects. We investigated the effects of infusion with isotonic- and hypertonic saline and isotonic glucose on urinary excretion of AQP2 and ENaCγ corrected for creatinine (u-AQP2~CR~ and u-ENaCγ~CR~), renal function and sodium handling, vasoactive hormones and systemic blood pressure. Methods ======= Participants ------------ ### ***Inclusion criteria*** Healthy non-smoking men and women with age between 18 -- 45 years were included in this study. ### ***Exclusion criteria*** Subjects with clinical signs or history of heart, lung, kidney, endocrine or malignant disease; abnormal findings in ECG, urine dipstick or biochemistry (blood cell count, plasma concentrations of haemoglobin, sodium, potassium, creatinine, albumin, glucose, bilirubin, alanine aminotransferase, alkaline phosphatase and cholesterol); arterial hypertension (ambulatory BP \>130/80 mmHg); medical treatment; alcohol and substance abuse; present smoking; pregnancy; breast feeding; donation of blood within one month prior to the study and obesity (BMI \> 32 kg/m^2^) were excluded from this study. ### ***Withdrawal criteria*** Subjects who developed the condition given in exclusion criteria during the course of the experiment, who withdrew their informed consent, and who had a poor compliance were withdrawn from this study. Ethics ------ This study was approved by the Regional Committee on Health Research Ethics (j. no. M-2011003) and carried out in accordance with the Helsinki Declaration. Written informed consent was obtained from all subjects. Recruitment ----------- Healthy male and female volunteers were recruited through advertisement at public institutions in Holstebro, Denmark. Design ------ The study was conducted as a randomized, placebo-controlled, crossover study. On three different occasions separated by at least two weeks, subjects were randomised to 0.9% isotonic saline (0.9% NaCl), 3.0% hypertonic saline (3% NaCl) or 5% glucose (glucose), which was administered as a sustained infusion over 60 minutes. The amount of fluid used in this study was 0.9% NaCl: 23 ml/kg, 3% NaCl: 7 ml/kg and glucose: 23 ml/kg. Effect variables ---------------- The main effect variables were u-AQP2, as well as u-ENaCγ. Other effect variables were diastolic (DBP) and systolic blood pressure (SBP), plasma concentration of renin (PRC), angiotensin II (AngII), aldosterone (Aldo), vasopressin (AVP), free water clearance (C~H2O~)~,~ glomerular filtration rate (GFR), fractional excretion of sodium (FE~Na~) and potassium (FE~K~), plasma sodium (p-Na), urine osmolality (u-osm), plasma osmolality (p-osm) and plasma albumin (p-alb). Number of subjects ------------------ Using a significance level of 5% and a power of 90% it was calculated that the number of subjects needed were 21, when the minimal relevant difference in U-ENaCγ was 100 ng./min and SD was 95 ng./min. Incomplete voiding during examination days was expected in some subjects, therefore 26 subjects were included in the study. Experimental procedures ----------------------- ### ***Experimental procedure prior to the study day*** Four days prior to each study day, subjects consumed a standardized diet regarding calories, sodium and fluid. The diet consisted of 11,000 kJ/day with an energy distribution of 55% carbohydrates, 30% fat and 15% protein in accordance to general dietary guidelines. The sodium content was 120 mmol pr. day. The subjects were asked to drink exactly 2500 ml/day. No alcohol, coffee, tea or soft drink consumption was allowed while on the standardized diet. Subjects were instructed to keep their physical activity as usual during the experiments and to abstain from hard training. A 24-hour urine collection, ending at 7:00 AM on the examination-day, was used to assess water and sodium balance. ### ***Experimental procedure on the study day*** After an overnight fast, subjects arrived at our facility at 8:00 AM. Two indwelling catheters for blood sampling and administration of ^51^Cr --EDTA and fluid were placed in both cubital veins. Every 30 minutes, starting at arrival, participants received a 175 ml oral water load of tap water. Urine was collected in standing or sitting position. Otherwise subjects were kept in supine position in a quiet temperature-controlled room (22-25°C). Three 30 minutes baseline clearance periods were obtained from 9:30 AM to 11:00 AM. These were followed by one clearance period from 11:00 AM to 12:00 PM during which a sustained infusion of either glucose, 0.9% NaCl or 3% NaCl was administered. The post infusion period consisted of three 30-minute periods from 12:00 PM to 1:30 PM. Blood and urine samples were collected every 30 minutes from 9:30 AM to 1:30 PM and analysed for ^51^Cr-EDTA, electrolytes and osmolality. Analysis of plasma concentrations of PRC, Ang II, Aldo and AVP were conducted from blood samples drawn at 11:00 AM (baseline), 12:00 AM (cessation of fluid infusion), 12:30 PM (30 min after cessation of fluid) and 1:30 PM (90 min after cessation of fluid). For data analysis, the 30-minute periods were subdivided into: baseline (0--90 min), infusion (90--150 min) and post infusion (150--180 min, 180--210 min, 210--240 min). Measurements ------------ ### ***Renal function*** Glomerular filtration rate was measured by the constant infusion clearance technique using ^51^Cr-EDTA as reference substance. A priming dose of ^51^Cr-EDTA was given followed by sustained infusion that was kept stable using a volume-controlled infusion pump. More than 15% variation in GFR between the three baseline periods led to exclusion of analysis. ### ***Blood samples*** Were centrifuged for 10 minutes at 2200 × g at 4°C. Plasma hormone samples were kept frozen at -20°C (AngII) and -80°C (PRC, Aldo, and AVP) until assayed. Renin in plasma was determined using an immunoradiometric assay from and a kit from CIS Bio International, Gif-Sur-Yvette Cedex, France. Minimal detection level was 1 pg./mL the coefficients of variation were 14.5% (interassay) and 4.5% (intra assay). Aldosterone in plasma was determined by radioimmunoassay using a kit from Demeditec Diagnostics Systems Laboratories Inc. (Webster, TX, USA). Minimal detection level was 22 pmol/L. The coefficients of variation were 8.2% (inter-assay) and 3.9% (intra-assay). Arginine vasopressin and Angiotensin II were extracted from plasma with C~18~ Sep-Pak (water associates, Milford, MA, USA) and subsequently measured using radioimmunoassay as previously described. The antibody against angiotensin II was obtained from the Department of Clinical Physiology, Glostrup Hospital, Glostrup, Denmark. Minimal detection level was 2 pmol/L. The coefficients of variation were 12% (inter-assay) and 8% (intra-assay). The antibody against AVP was a gift from Professor Jacques Dürr (Miami, FL, USA). Minimal detection level was 0.2 pmol/L. The coefficients of variation were 13% (inter-assay) and 9% (intra --assay). ### ***Urine samples*** Were kept frozen at -20°C until assayed. U-AQP2 was measured by radioimmunoassay as previously described \[[@B6],[@B7]\]. Antibodies were raised in rabbits to a synthetic peptide corresponding to the 15 COOH-terminal amino acids in human AQP2 to which was added an NH~2~-terminal cystein for conjugation and affinity purification. Minimal detection level was 34 pg/tube/tube. The coefficients of variation were 11.7% (inter-assay) and 5.9% (intra-assay). U-ENaCγ was measured by radioimmunoassay as previously described \[[@B8],[@B9]\]. Antibodies were raised against the synthetic ENaCγ peptide in rabbits and affinity purified as described previously \[[@B10]\]. Minimal detection level was 48 pg/tube. The coefficients of variation were 14% (inter-assay) and 6.7% (intra-assay). ### ***Blood pressure measurement*** Brachial blood pressure was recorded using a semiautomatic oscillometric devise (Omron 705IT, Omron Matsusaka, Japan). Plasma and urine concentrations of sodium, potassium, creatinine and albumin were measured using routine methods at the Department of Clinical Biochemistry, Holstebro Hospital. Fractional excretion of sodium was calculated as \[sodium clearance (*C*~Na~)/GFR × 100%\]. Fractional excretion of potassium was calculated as \[potassium clearance (C~K~)/GFR × 100%\]. Free water clearance was calculated as \[urine output (UO) -- osmolar clearance (*C*~OMS~)\]. *C*~OSM~ was calculated as \[urine osmolarity/plasma osmolarity × UO\]. Statistics ---------- Statistical analyses were performed using IBM SPSS statistics version 20.0.0 (IBM Corp.; Armonk, NY, USA). Single baseline values were obtained by taking the weighed average of the measurements from the three baseline periods. Parametric data are presented as means ± standard deviation (SD) and nonparametric data as medians with interquartile ranges. General linear model (GLM) with repeated measures was performed, with time as within-subject factor and intervention as between subject factor, to test for differences within and between groups. One-way ANOVA was used for comparison of means between groups when differences were found. For non-parametric data related samples Friedman's two-way analysis (FM) was used. For comparison within groups at baseline and post infusion period 210--240 minutes, a paired t-test was used when data were parametric and Wilcoxon's signed rank test was used when data were nonparametric. Statistical significance was defined as p \< 0.05 in all analyses. Results ======= Demographics ------------ A total of 31 healthy women and men were enrolled in the study. Five subjects were excluded due to: abnormal blood samples (1), 24-h BP above 130/80 mmHg (1), non-compliance (1) and withdrawal of informed consent (2). Thus, 26 persons completed the study. Three were not able to void satisfactorily during clearance experiments and were excluded from analysis. One was not able to void in two post intervention periods after 3% NaCl and was excluded in channel analysis only. The remaining 23 males (n = 9) and women (n = 14) had a median age of 26 years (range 18--42) and a mean BMI of 24.4 ± 2.3 kg/m^2^. Mean ambulatory blood pressure was 119/70 ± 8/4 mmHg. Screening blood values were b-haemoglobin 8.5 ± 0.7 mmol/L, p-sodium 139 ± 2, p-potassium 3.9 ± 0.4 mmol/L, p-creatinine 74 ± 9 μmol/L, p-albumin 42 ± 3 g/L, p-glucose 5.1 ± 0.6 mmol/L, p-alanine transaminase 25 ± 9 U/L and p-cholesterol 4.5 ± 0.5 mmol/L. Twenty-four-hour urine collection --------------------------------- Table [1](#T1){ref-type="table"} shows the results of the 24-h urine collection in 23 healthy subjects after 4 days of standardized diet. Mean u-AQP2, u-ENaCγ, urinary sodium, urine osmolarity, C~H2O~ and urine volume were the same in all three examination days indicating that the subjects had kept their supplied diets and fluid intake. ###### **Urine output, urine osmolarity (u-osm), free water clearance (C**~**H2O**~**), urinary AQP2 excretion per minute (u-AQP2), urinary excretion of ENaCγ per minute (u-ENaCγ), urinary sodium excretion (u-Na) and fractional excretion of sodium (FE**~**Na**~**) during 24-hours urine collection with fluid deprivation (12 PM to 8.00 AM) in a randomised, crossover study of 23 healthy subjects** **Examination day:** **P (ANOVA)** ---------------------------- ---------------------- --------------- ------------- ------ **Urine Output (ml/24 h)** 2306(559) 2120(695) 2281(650) 0.56 **u-osm (mosm/24 h)** 848(231) 797(207) 833(239) 0.73 **C**~**H2O**~**(ml/min)** -0.46(0--61) -0.47(0.57) -0.44(0.64) 0.99 **u-AQP2 (ng/min)** 0.93(0.24) 0.91(0.22) 0.93(0.24) 0.89 **u-ENaCγ (pg/min)** 405(163) 389(188) 385(164) 0.92 **u-Na (mmol/24 h)** 124(52) 119(41) 120(56) 0.93 **FE**~**Na**~**(%)** 0.49(0.15) 0.47(0.13) 0.45(0.18) 0.67 Values are means with SD in brackets. One-way ANOVA was used for comparison between groups. Water excretion, u-AQP2, u-osm ------------------------------ Table [2](#T2){ref-type="table"} shows the absolute values of UO, C~H2O~, u-AQP2~CR,~ u-AQP2 excretion rate and u-osm during the baseline period, the infusion period and the post infusion period. ###### **Effect of 0.9% isotonic saline (0.9% NaCl), 3% hypertonic saline (3% NaCl) and isotonic glucose (Glucose) on urinary output (OU), free water clearence (C**~**H2O**~**), urine osmolarity (u-osm), urinary aquaporin2 excretion rate (u-AQP2) and urinary aquaporin2 corrected for creatinine (u-AQP2**~**CR**~**) in a randomized, crossover study of 23 healthy subjects** **Periods** **Baseline** **Infusion** **Post infusion** **p (GLM-within)** --------------------------------- -------------- -------------- ------------------- -------------------- ----------------- ---------- **UO (ml/min)** 0.9% NaCl 6.90(1.33) 8.75(1.76) 7.95(2.31) 7.39(2.36) 8.12(2.26) \<0.0001 3% NaCl 7.12(1.35) 2.75(0.73) 1.81(0.64) 1.98(1.07) 2.05(1.09)\*\* Glucose 6.83(1.58) 11.84(2.28) 16.05(2.82) 15.28(2.47) 13.45(2.47)\*\* p (GLM between) \<0.0001 p ANOVA 0.77 \<0.0001 \<0.0001 \<0.0001 \<0.0001 **C**~**H2O**~ 0.9% NaCl 3.87(1.55) 5.29(1.44) 4.07(2.09) 3.01(2.04) 3.62(1.59) \<0.0001 3% NaCl 3.97(1.21) -0.35(0.73) -1.87(0.83) -2.22(0.88) -2.26(0.96) Glucose 3.84(1.32) 6.38(1.77) 11.60(0.52) 12.47(2.35) 11.17(2.12) p (GLM between) \<0.0001 p ANOVA 0.95 \<0.0001 \<0.0001 \<0.0001 \<0.0001 **U- osm (mosmol/kg)** 0.9% NaCl 140(62) 115(21) 149(58) 176(50) 161(33) \<0.0001 3% NaCl 136(37) 342(96) 606(114) 637(105) 640(111)\* Glucose 136(41) 133(21) 79(21) 52(12) 48(7)\* p (GLM between) p \<0.0001 p ANOVA 0.95 \<0.0001 \<0.0001 \<0.0001 \<0.0001 **u-AQP2 (ng/min)** 0.9% NaCl 1.03 (0.25) 1.02(0.22) 1.22(0.34) \- 1.26(0.32)\* \<0.0001 3% NaCl 1.02(0.22) 0.99(0.31) 1.28(0.30) \- 1.37(0.34)\* Glucose 1.01(0.22) 1.26(0.25) 1.19(0.29) \- 0.78(0.18)\* p (GLM between) 0.274 **u-AQP2**~**CR**~**(ng/mmol)** 0.9% NaCl 102.5(19.9) 110.8(18.1) 119.8(25.4) \- 125.0(21.0)\*\* \<0.0001 3% NaCl 105.9(15.7) 108.6(27.2) 125.1(31.4) \- 132.7(26.5)\* Glucose 105.7(19.1) 137.1(22.3) 118.1(26.5) \- 87.9(15.4)\* p (GLM between) 0.565 Values are mean with SD in brackets. General linear model (GLM) with repeated measures was performed for comparison within the group and intervention as between subjects factor. One way ANOVA was performed when differences were found between interventions. Paired t-test was used for comparison within treatment group at baseline and post infusion period 210 -- 240 minutes. \*p \< 0.0001; \*\*p \< 0.001. UO increased significantly after 0.9% NaCl and glucose. The 3% NaCl infusion induced a significantly and sustained decrease in UO. The relative changes in UO were significantly different between the three interventions. C~H2O~ increased during the infusion with 0.9% NaCl, and decreased slightly, although significant in the postinfusion period. At the end of the examination-day C~H2O~ increased towards baseline levels with an over all relative change of -10%. There was a pronounced increase in C~H2O~ after glucose, whereas C~H2O~ decreased after 3% NaCl and changed from positive values at baseline to negative values after infusion. Thus, indicating a change from free water excretion to water reabsorption (Table [2](#T2){ref-type="table"}). U-AQP2~CR~ increased by 27% (p \< 0.001) in response to 0.9% and by 26% (p \< 0.0001) after 3% NaCl and reached maximum at 240 min after baseline. During glucose infusion (90--150 min) there was a primary increase in u-AQP2~CR~ after which u-AQP2~CR~ decreased and reached a minimum of - 16% (p \< 0.0001) at 210--240 min (Figure [1](#F1){ref-type="fig"}A). The excretion of u-AQP2 divided by gender, showed that u-AQP2~CR~ tended to be higher in women than in men, but there was no statistical significant difference. This was due to a lower creatinine concentration in women's urine (data not shown). U-AQP2 excretion rate followed the same pattern (Table [2](#T2){ref-type="table"}). The relative changes in u-AQP2 did not differ between 3% NaCl and 0.9% NaCl, but both were significantly different from the relative change in u-AQP2 after glucose infusion. ![**Effects of isotonic 0.9% saline (■), hypertonic 3% saline (**♦**) and isotonic glucose (**●**) on urinary excretion of A) u-AQP2 and B) u-ENaCγ adjusted to creatinine, C) plasma concentration of vasopressin (AVP) and D) plasma osmolality in 23 healthy subjects.** Values are means ± SEM. Paired t-test was used for comparison of post infusion period 210--240 min vs. baseline. \* p \< 0.01; \*\* p \< 0.001; \*\*\* p \< 0.0001.](1471-2369-14-202-1){#F1} U-Osm decreased during 0.9% NaCl infusion with minimum after infusion ended at 150 minutes, after which u-osm increased, coherent with the changes seen in C~H2O~. U-osm increased significantly in response to 3% NaCl and lasted throughout the experiment. During glucose infusion u-osm remained constant for 60 minutes until glucose infusion was completed, after which u-osm declined and reached minimum at 210--240 min (Table [3](#T3){ref-type="table"}). ###### **Effect of 0.9% isotonic saline (0.9% NaCl), 3% hypertonic saline (3% NaCl) and isotonic glucose (Glucose) on urinary sodium excretion (u-Na), fractional excretion of sodium (FE**~**Na**~**), urinary potassium wxcretion (u-K), fractional excretion of potassium (FE**~**K**~**), urinary gamma fraction of the epithelial sodium channels excretion rate (u-ENaCγ) and urinary gamma fraction of ENaC corrected for creatinine (u-ENaCγ**~**CR**~**)** **Periods** **Baseline** **Infusion** **Post infusion** **p (GLM within)** ---------------------------------- -------------- -------------- ------------------- -------------------- -------------------- ---------- **u-Na (mmol/min)** 0.9% NaCl 1.24(0.55) 1.83(0.56) 2.51(1.05) 2.78(0.99) 3.07(0.98)\* \<0.0001 3% NaCl 1.38(0.58) 1.86(0.75) 2.65(1.06) 2.94(1.44) 3.17(1.38)\* Glucose 1.23(0.51) 1.38(7.3) 0.94(0.43) 0.83(0.35) 0.72(0.35)\* p (GLM between) \<0.0001 p (ANOVA) 0.599 0.028 \<0.0001 \<0.0001 \<0.001 **FE**~**Na**~ 0.9% NaCl 1.26(0.53) 1.93(0.61) 2.35(0.72) 2.67(0.76) 2.80(0.75)\* \<0.0001 3% NaCl 1.44(0.61) 2.08(0.81) 2.66(1.07) 3.07(1.27) 3.01(1.26)\* Glucose 1.27(0.53) 1.36(0.64) 0.88(0.37) 0.81(0.37) 0.74(0.37)\* p (GLM between) \<0.0001 p (ANOVA) 0.465 0.002 \<0.0001 \<0.0001 \<0.0001 **U-K (mmol/min)** 0.9% NaCl 26.5(9.1) 22.5(7.6) 21.1(8.9) 22.7(12.2) 23.5(10.2)\*\*\* \<0.0001 3% NaCl 25.4(9.2) 17.3(7.3) 16.9(8.1) 20.1(10.5) 21.1(9.9)\*\*\* Glucose 24.9(8.7) 17.2(5.8) 10.8(3.7) 9.6(3.2) 11.23(4.4)\* p (GLM between) \<0.0001 p (ANOVA) 0.827 0.016 \<0.0001 \<0.0001 \<0.0001 **FE**~**K**~ 0.9% NaCl 28.2(9.9) 23.3(7.4) 20.4(7.8) 21.5(8.2) 21.5(7.9)\*\*\* \<0.0001 3% NaCl 26.3(9.0) 18.5(6.8) 17.0(7.5) 20.9(9.3) 20.0(8.9)\*\*\* Glucose 25.7(8.8) 16.7(5.5) 9.8(3.1) 9.3(3.4) 11.5(4.4)\* p (GLM between) \< 0.0001 p (ANOVA) 0.624 \<0.01 \<0.0001 \<0.0001 \<0.0001 **u-ENaCγ (ng/min)** 0.9% NaCl 351.2(155.0) 302.3(108.5) 341.2(123.0) \- 327.2(111.3) \<0.0001 3% NaCl 358.8(160.0) 375.9(173.7) 462.6(203.4) \- 435.9(168.7)\*\* Glucose 338.3(150.0) 301.7(128.8) 335.2(185.1) \- 283.7(164.4)\*\*\* p (GLM between) 0.076 **u-ENaCγ**~**CR**~**(ng/mmol)** 0.9% NaCl 36.25(3.63) 32.98(3.50) 34.40(3.63) \- 32.56(3.57) \<0.001 3% NaCl 37.56(3.70) 43.29(3.58) 45.83(3.72) \- 44.46(3.65)\*\*\* Glucose 35.33(3.62) 32.78(3.50) 31.68(3.63) \- 31.71(3.57) p (GLM between) 0.091 Values are means with SD in brackets. General linear model (GLM) with repeated measures was performed for comparison within the group and intervention as between subjects factor. One way ANOVA was used for comparison of means between subjects when differences were found. Paired t-test was used for comparison within treatment group at baseline and third post infusion period \*p \< 0.0001, \*\*p \<0.001, \*\*\* p \< 0.01. Sodium excretion, u-ENaCγ, u-Na, FE~Na~, u-K and FE~K~ ------------------------------------------------------ Table [3](#T3){ref-type="table"} shows the absolute values of u-Na, FE~Na~, u-K, FE~K~, u-ENaCγ~CR~ and u-ENaCγ excretion rate during the baseline period, the infusion period and the post infusion period. Infusion with 0.9% NaCl and 3% NaCl were accompanied by a significant and similar increase in u-Na and FE~Na~ that lasted throughout the experiment. There were no significant differences between 0.9% NaCl and 3% NaCl infusions. In contrast, U-Na and FE~Na~ decreased after glucose infusion. The relative changes in u-Na and FE~Na~ were significant lower after glucose compared to both saline infusions. U-K and FE~K~ decreased significantly after all three infusions, but with the greatest extend after glucose infusion. In the post infusion period (150--240 min) the excretion of potassium in urine increased slightly more after 3.0% NaCl than 0.9% NaCl, but did not reach baseline levels. U-ENaCγ~CR~ decreased slightly, but non-significantly during 0.9% NaCl and glucose infusions. A significant increase was seen in u-ENaCγ~CR~ in response to 3% NaCl (p \< 0.01) (Figure [1](#F1){ref-type="fig"}B) and the relative increase in u-ENaCγ~CR~ were significantly higher in response to 3% NaCl compared to 0.9% NaCl and glucose. Divided by gender the differences in u-ENaCγ~CR~ showed no statistical significant difference, although u-ENaCγ~CR~ tended to be higher in women due to the lower urine creatinine (data not shown). U-ENaCγ excretion rate followed the same pattern with regard to saline infusions, whereas a significantly lower u-ENaCγ excretion rate occurred after glucose infusion (Table [3](#T3){ref-type="table"}). Vasoactive hormones ------------------- PRC, Ang II and Aldo were suppressed to the same extent in all three parameters in response to 0.9% NaCl and 3% NaCl with no significant difference between interventions. There was a primary decrease during glucose infusion (90--150 min), but when infusion ceased values returned to baseline levels with no overall significant change (Figure [2](#F2){ref-type="fig"}). ![**Effects of isotonic 0.9% saline (■), hypertonic 3% saline (**♦**) and isotonic glucose (**●**) on plasma renin (A), plasma angiontensin II (B) and plasma aldosterone (C) concentrations.** Values are expressed as mean ± SEM. General linear model (GLM) with repeated measures within subjects was significant for all three variables. Paired t-test was used for comparison within treatment groups at postinfusion 240 min vs. basal. \* p \< 0.0001.](1471-2369-14-202-2){#F2} AVP did not change in response to 0.9% NaCl and glucose, but increased significantly after 3% NaCl with a maximum at 150 minutes and a steady fall during the post infusion period (Figure [1](#F1){ref-type="fig"}C). Blood pressure, pulse rate, GFR, p-Na, p-alb and p-osm ------------------------------------------------------ Table [4](#T4){ref-type="table"} shows the absolute values of systolic and diastolic blood pressure, pulse rate, GFR, plasma sodium and plasma albumin during the baseline period, the infusion period and the post infusion period. ###### **Effect of 0.9% isotonic saline (0.9% NaCl), 3.0% hypertonic saline (3% NaCl) and 5% glucose (Glucose) on**^**51**^**Cr-EDTA-clearance, plasma sodium, plasma albumin, systolic blood pressure (SBP), diastolic blood pressure (DBP) and pulse rate in a randomized, placebo-controlled, crossover study of 23 healthy subjects** **Periods** **Baseline** **Infusion** **Post infusion** **p (GLM-within)** ---------------------------------------------------------- -------------- -------------- ------------------- -------------------- ------------------- ---------- ^**51**^**Cr-EDTA-clearance (ml/min/1.73 m**^**2**^**)** 0.9% NaCl 98.1(10.1) 97.2(9.9) 100.0(12.4) 106.4(19.5) 106.1(13.3)\*\* 0.002 3% NaCl 95.7(10.4) 93.0(11.4) 96.6(14.8) 97.0(14.8) 101.8(13.9)\*\*\* Glucose 96.9(10.3) 100.0 (9.9) 105.9(10.5) 104.0(13.9) 99.1(9.7) p (GLM between) 0.244 **p-Sodium (mmol/l)** 0.9% NaCl 138.1(2.1) 139.4(2.1) 139.6(2.4) 139.1(2.4) 138.7(2.3)\*\* \<0.0001 3% NaCl 138.0(2.0) 142.5(1.8) 141.7(1.7) 141.1(2.1) 139.7(2.0)\* Glucose 138.1(2.0) 128.7(3.7) 134.1(3.6) 135.7(3.7) 136.9(2.9)\* p (GLM between) \<0.0001 p (ANOVA) 0.94 \<0.0001 \<0.0001 \<0.0001 =0.001 **p-albumin (g/L)** 0.9% NaCl 40.3(2.6) 40.1(2.4) 35.3(2.0) 35.8(2.1) 36.0(2.2)\* \<0.0001 3% NaCl 40.1(2.6) 39.8(2.6) 35.2(2.2) 36.0(2.3) 36.2(2.7)\* Glucose 40.9(2.2) 40.3(2.8) 36.1(2.5) 39.1(2.7) 39.7(2.9)\* p (GLM between) \<0.05 p (ANOVA) 0.510 0.765 0.328 \<0.0001 \<0.0001 **SBP (mmHg)** 0.9% NaCl 114.5(10.6) 117.7(10.4) 116.6(11.0) 116.9(11.6) 117.7(11.5) 0.439 3% NaCl 114.0(9.5) 117.9(9.1) 117.2(9.2) 115.5(10.0) 116.4(9.7) Glucose 115.5(8.4) 118.2(9.8) 117.7(9.1) 115.5(8.2) 117.0(9.1) p (GLM between) 0.975 **DBP (mmHg)** 0.9% NaCl 62.9(4.3) 64.0(6.1) 62.2(5.7) 63.3(5.6) 63.4(5.1) 0.038 3% NaCl 62.9(4.2) 62.1(5.0) 61.0(5.4) 61.4(4.9) 62.9(4.9) Glucose 64.4(4.2) 65.8(5.6) 65.3(6.3) 65.5(5.3) 63.6(7.5) p (GLM between) 0.12 **Pulse rate** 0.9% NaCl 54.1(11.0) 57.2(11.9) 56.4(11.6) 58.0(12.1) 57.4(11.9)\* \<0.0001 3% NaCl 54.0(10.5) 58.2(11.2) 57.0(11.2) 56.9(11.3) 57.7(11.5)\* Glucose 54.7(10.5) 57.4(12.5) 57.1(12.0) 60.3(11.7) 61.3(12.3)\* p (GLM between) 0.917 Values are mean with SD in brackets. General linear model (GLM) with repeated measures was performed for comparison within the group and intervention as between subjects factor. One way ANOVA was performed for comparison of means when differences were found between intervensions. Paired t-test was used for comparison within treatment groups at baseline and post infusion period 210 -- 240 minutes. \*p \< 0.0001; \*\*p \< 0.001; \*\*\*p \< 0.05. Systolic BP was the same after all three infusions. There was a small difference in diastolic BP pattern during the examination day, but the changes were very small and might be by chance. During the examination day pulse rate increased slightly in response to 0.9% NaCl and 3% NaCl, while the heart rate increased to a higher extent in response to the glucose infusion (Table [4](#T4){ref-type="table"}). The increase in pulse rate did not differ between 0.9% saline and 3% saline, but there was a difference in the relative increase in pulse rate between saline and glucose infusion (p \< 0.01). GFR increased slightly, although significantly, on the examination day. However the changes were very small (Table [4](#T4){ref-type="table"}). P-Na increased in response to both 0.9% NaCl and 3% NaCl with maximum after 150 minutes. In response to glucose p-Na decreased markedly after 150 minutes to a mean of 128.7 mmol/l (Table [4](#T4){ref-type="table"}). The increase was higher after 3% NaCl compared to 0.9% NaCl and accordingly the changes after glucose were lower compared to saline. P-alb decreased significantly in response to 0.9%, 3% NaCl and glucose infusions. The decline was significantly lower and sustained after both saline infusions compared to glucose, which is related to an expected increase in extracellular fluid. P-osm increased slightly during 0.9% NaCl infusion, but remained unchanged at the end of the examination day. P-osm increased significantly in response to 3% NaCl, with a maximum of 293 mosm/kg and decreased significantly after glucose to 280 mosm/kg at 150 min. The changes in p-osm indicated that isotonic, hypertonic and hypotonic conditions were established (Figure [1](#F1){ref-type="fig"}D). Fluid, sodium balance and body weight during the examination days ----------------------------------------------------------------- The average fluid administered intravenous was 1749 ml of 0.9% NaCl (SD 270), 555 ml of 3% NaCl (SD 90) and 1736 ml of glucose (SD 282). The cumulative water input was 3674 ml (SD 270), 2480 ml (SD 90) and 3661 ml (SD 282) respectively, as participants drank an additional 1925 ml of tap water each examination day. During the examination days the average total urine output was 1858 ml (SD 246) in subjects who received 0.9% NaCl, 984 ml (SD 202) in subjects who received 3% NaCl and 2682 ml (SD 351) in subjects who received glucose. The fraction of water excreted after 240 min was 51% when 0.9% NaCl was infused, 40% when 3% NaCl was infused and 73% when glucose was infused. The total amount of sodium infused was 269 mmol (SD 42) of 0.9% NaCl and 285 mmol (SD 46) of 3.0% NaCl. The cumulative sodium output at 240 min was 50 mmol (SD 16) after 0.9% NaCl, 54 mmol (SD 21) after 3% NaCl and 21 mmol (SD 9) after glucose. The fraction of sodium excreted after 240 min was 19% after both 0.9% and 3% NaCl infusions. This was accompanied by a significant increased bodyweight in response to 0.9% NaCl from 73.2 kg (SD 11.3) at baseline to 74.3 kg (SD 11.4) at the end of the study day \[+1.1 kg (SD 0.39); p \< 0.0001\], in response to 3% NaCl from 73.3 kg (SD 11.6) at baseline to 74.1 kg (SD 11.7) at the end of the study day \[+0.8 kg (SD 0.39); p \< 0.0001\] and to a smaller extent in response to glucose from 72.8 kg (SD 11.8) at baseline to 73.1 kg (SD12.0) at the end of the study day \[+0.3 kg (SD 0.5); p \<0.05\]. Discussions =========== In the present study we examined the effect of an acute intravenous volume load of 0.9% saline, 3% saline and isotonic glucose infusions on u-AQP2 and u-ENaCγ in a randomized, crossover study of healthy subjects. The purpose was to evaluate the transport activity via the aquaporin 2 water channels and the epithelial sodium channels in the principal cells in the distal part of the nephron. During infusion and in the period immediately after, adaptive physiological changes take place in renal function and vasoactive hormones. Thus, the main changes in the effect variables could be expected to occur after the infusion. In the present study, we paid special attention to changes in the effect variables in the last post infusion period (Post infusion 210--240), i.e. 60--90 minutes after infusion had ceased. During this period, u-AQP2 increased after hypertonic and isotonic saline infusion and decreased after glucose infusion. At the same time, u-ENaCγ increased after hypertonic saline infusion and remained unchanged after isotonic saline and glucose infusion. U-AQP2 after infusion with hypotonic and isotonic saline and isotonic glucose ----------------------------------------------------------------------------- Aquaporin-2 (AQP2) is located in the collecting duct principal cells \[[@B11]\] and is expressed in the apical plasma membrane \[[@B12]\]. Vasopressin (AVP) regulates AQP2 by binding to V2 receptors in the basolateral membrane, \[[@B11],[@B13]\]. Short term exposure to AVP causes trafficking and insertion of the intracellular vesicles, containing AQP2, to the apical membrane and increases the water permeability and absorption \[[@B11]-[@B14]\]. Long-term regulation occurs over a period of hours to days, and is caused by AVP-regulated gene transcription resulting in an increase in AQP2 whole-cell abundance \[[@B4],[@B13]\]. Experiments in rats showed that infusion of dDAVP increased u-AQP2 \[[@B15]\]. This is consistent with the view that increased delivery of AQP2 channels to the apical membrane results in increased excretion of AQP2 after stimulation with AVP \[[@B6],[@B15]-[@B21]\]. Approximately 3% of AQP2 in the collecting duct are excreted into urine \[[@B20]\], but the underlying mechanisms are unknown. Volume expansion with 3% hypertonic saline increases plasma osmolarity beyond the threshold of the hypothalamic osmoreceptors, triggering release of AVP and a subsequent increase in u-AQP2. Saito et al found a significant relationship between urinary excretion of AQP2 and p-AVP in healthy subjects after 5% hypertonic saline infusion \[[@B17]\]. Pedersen et al found a positive correlation between u-AQP2 and p-AVP during 24 h of water deprivation and after 3% hypertonic saline infusion \[[@B6]\]. Thus, previous studies in humans have demonstrated that the activity of the AQP2 water channels can be determined by measuring u-AQP2 \[[@B6]-[@B8],[@B17],[@B18]\]. Surprisingly, Baumgartner et al found no change in u-AQP2 after infusion of 2.5% NaCl in healthy volunteers, despite a significant rise in both urine osmolarity and AVP \[[@B22]\]. However, the oral water load was 3--4 times higher prior to infusion compared to our study. Thus, the large water load before infusion might have overruled the stimulatory effects of hypertonic saline. As expected, our study showed that u-AQP2 increased after 3% NaCl with a corresponding rise in urine osmolarity and a reduction in C~H2O~. Thus, our findings indicate an increased water reabsorption via the aquaporin-2 water channels in the distal tubules. Prior to the increase in u-AQP2, there was an abrupt rise in p-osm and p-AVP induced by the hypertonic saline infusion. Animal studies have shown that hypertonicity can cause an up regulation of AQP2 expression in the apical membrane comparable with that achieved by AVP alone \[[@B23],[@B24]\]. It cannot be excluded that this might play an active part in the increased excretion of u-AQP2. Most likely, the increased water reabsorption was mediated by an increase in p-AVP. U-AQP2 continued to rise throughout the examination day, suggesting that AQP2 channels remained inserted and active in the apical membrane due to actions of elevated p-AVP. Infusion of isotonic saline depresses the fractional water and salt reabsorption in the proximal tubules in animals \[[@B25]\]. In the present study, infusion with 0.9% NaCl caused the same response in u-AQP2, u-osm and C~H2O~ as 3% NaCl infusion, albeit to a lesser extent. There was a small rise in p-osm to a maximum level of 286 mosmol/kg corresponding to a rise of 0.5%. This increase is below the osmoreceptor threshold, and we did not see, nor expect, any significant change in p-AVP. Therefore AVP could not be the main regulator of AQP2 during 0.9% NaCl. Most likely, the increased water transport via AQP2 is a compensatory phenomenon to antagonize a decrease in the renal water absorption in the proximal tubules, which occurs after isotonic volume expansion. The mechanism might be due to an increased activity in the natriuretic peptide system \[[@B26],[@B27]\]. Infusion of 5% glucose causes a volume expansion distributed throughout the fluid phases in the body with only a very small increase in plasma volume. This is illustrated in the measurements of plasma albumin, where concentrations at 240 min were virtually equal to baseline (Table [4](#T4){ref-type="table"}), indicating no change in extracellular fluid. According to our knowledge, no study has measured u-AQP2 after glucose infusion. A study of healthy subjects showed, that after an oral water load of 20 mL/kg for 15 minutes (mean intake 1605 ml) u-AQP2 decreased 17% after 210 minutes \[[@B21]\]. In a recent study, subjects received an oral water load of 20 ml/kg for 15 min (mean intake 1389 ml) with a subsequent 27% decrease in u-AQP2 after 240 minutes \[[@B28]\]. Both plasma osmolarity and p-AVP decreased. Thus, it has been showed that u-AQP2 is reduced during water diuresis after oral water intake \[[@B21],[@B28],[@B29]\]. In our study, subjects received a mean of 1736 ml glucose IV. In the last post infusion period the expected aquaretic response occurred, with a 16% decrease in u-AQP2~cr~, a decrease in u-osm and an increase in UO and C~H2O~. Plasma osmolarity decreased from 285 mosm/kg to 280 m0sm/kg, i.e. a 2% decline, but with no accompanying reduction in p-AVP. Our findings indicate a reduced reabsorption of water via the aquaporin-2 water channels in the distal tubules after isotonic glucose infusion. The lack of change in p-AVP could firstly be explained by the fact that the subjects had received 1225 ml of oral water load prior to the infusion start, and this could have suppressed AVP in baseline periods beforehand. Secondly, the measurements of p-AVP concentration may not be sensitive enough to detect a small decrease. The recent discovered peptide Apelin, may also play a role. Apelin is colozalized with AVP in magnocellular neurons of hypothalamus \[[@B30],[@B31]\]. In healthy male volunteers decreasing plasma osmolarity by waterloading reduced p-AVP modestly but p-Apelin increased rapidly \[[@B32]\]. Apelin regulation is opposite to that of AVP and data suggests that Apelin, like AVP may participate in regulating water homeostasis \[[@B32]\]. We did not measure p-Apelin, but it could have been of interest to investigate plasma apelin in parallel with p-AVP in conditions of different volume expansions. Thus, in the last post-infusion period, u-AQP2 increased approximately to the same extent after hypertonic and isotonic saline infusions, whereas a marked fall was seen after isotonic glucose infusion. A possible explanation for the delay in changes of u-AQP2 could be that it takes few minutes for changes in AVP to act on the principal cell, either by insertion or removal of AQP2 from the apical membrane, but it takes several minutes before the effect is seen in the excretion of u-AQP2 in the urine. U-ENaCγ after infusion with hypotonic and isotonic saline and isotonic glucose ------------------------------------------------------------------------------ Sodium transport across the collecting duct occurs through the epithelial sodium channel and is responsible for reabsorption of 3--5 % of filtered sodium \[[@B33]\]. ENaC is composed of three distinct subunits: α, β and γ and localized at the apical plasma membrane of principal cells \[[@B34],[@B35]\]. ENaC is a target of aldosterone that acts on the mineralocorticoid receptor. Aldosterone increases sodium transport by redistributing ENaC subunits from intracellular locations to the apical membrane as well as altering gene transcription \[[@B33],[@B36],[@B37]\]. While the action of aldosterone occurs over hours or days, another synergistically pathway involves AVP \[[@B35],[@B37]-[@B41]\]. In the cortical collecting ducts in rats, AVP binds to the V2 receptors, stimulates cAMP and increases sodium reabsorption by promoting trafficking and insertion of ENaC into the apical membrane inducing a rapid change in channel activity \[[@B34],[@B35],[@B40]\]. Recent studies in humans demonstrated that AVP, via V2 receptors, stimulates ENaC mediated sodium reabsorption across principal cells \[[@B39],[@B42],[@B43]\]. Fractions of ENaC are normally excreted into the urine. The amount of ENaC-fractions is supposed to reflect the activity of the sodium transport via the epithelial sodium channels just as u-AQP2 reflects the functional status of the AQP2 water channels. Recently, our group introduced a new method to evaluate sodium reabsorption in the principal cells in the distal tubules. Lauridsen et al demonstrated a significant correlation between changes in urinary sodium excretion and changes in urinary excretion of the beta fraction (u-ENaC~β~) in healthy humans \[[@B44],[@B45]\]. Apparently, u-ENaC~β~ can be used as a biomarker for the transport of sodium via ENaC. In the present study, we measured the gamma fraction of the protein of the epithelial sodium channels to evaluate the up-and down regulation of γ-ENaC expression and sodium transport via ENaC as previously reported from our group \[[@B9],[@B46]\]. The sodium-chloride symporter (NCC) in the distal convoluted tubules (DCT) is as another major sodium reabsorbing pathway. Sodium reabsorption in DCT is essential to define the amount of sodium delivery to the principal cells in the collecting duct. It is widely accepted that NCC is regulated by Ang II and aldosterone \[[@B47],[@B48]\]. Studies have also shown that high AVP increase phosphorylation of NCC and presumably result in greater sodium reabsorption \[[@B49]\]. Experimental animal-studies have demonstrated that isotonic and hypertonic saline IV reduced reabsorption of sodium in the proximal tubules, and thereby increased the amount of sodium in the urine \[[@B25],[@B50]\]. Andersen LJ et al studied the effects of hypertonic and isotonic saline in healthy subjects on a controlled diet. The subjects received an IV sodium load of either 25 ml/kg isotonic saline or 4.5 ml/kg 3% hypertonic saline for 90 minutes \[[@B51]\]. Urinary sodium excretion increased in both isotonic and hypertonic saline, with natriuresis after hypertonic saline exceeding that after isotonic saline. Plasma sodium and plasma osmolarity increased substantially after hypertonic saline, as did p-AVP. Our study showed that 3% NaCl infusion increased u-ENaCγ, FE~Na~, p-Osm, p-Na and p-AVP. Thus, our findings reflect an increased sodium reabsorption via ENaC in the principal cells, and furthermore confirmed the results by Andersen et al \[[@B51]\]. The increased u-ENaCγ could partly be explained by a considerable decrease in the renal sodium absorption proximal in the nephron, compensated for and adjusted by an increase in absorption in the distal part. However, the rise in p-AVP seen immediately after 3% NaCl infusion could also indicate that the increased u-ENaCγ is caused by actions of AVP. An increased sodium movement from the lumen to the cell via ENaC would theoretically drive potassium secretion through the ROMK channels \[[@B52],[@B53]\]. Surprisingly we measured a fall in excretion of potassium in the urine. This could argue against a major role of ENaC mediated sodium transport. If NCC increased sodium reabsorption, both to compensate for a decrease in proximal reabsorption and due to high p-AVP, then less sodium would to be transported by ENaC and thus potassium secretion would not take place. A possible role of NCC after infusion with hypertonic saline is purely speculative as we did not measure the activity of NCC. Perhaps we did not see the positive effect on potassium secretion within our time limits. However, the potassium transport is complex and factors modulating potassium transport, such as altered tubular flow and aldosteron, are many. After volume expansion with isotonic saline the oncotic pressure is slightly reduced, which leads to an immediately increase in GFR and smaller reabsorption of water in the proximal tubule. We measured a small increase in GFR and UO output. Sodium excretion increased, but u-ENaCγ, p-Na, p-osm and p-AVP remained unchanged, thus the findings were as we expected. Regarding NCC, one would not expect any change in NCC mediated sodium reabsorption during isotonic saline. No study has ever evaluated u-ENaCγ during water diuresis. In our study, we measured a trend towards a reduction in u-ENaCγ after glucose infusion reflecting a small reduction of sodium reabsorption via ENaC in the principal cell. As previously mentioned, we measured a 2% fall in p-osmolality after glucose infusion, which theoretically should trigger a decrease in AVP. We did not detect a fall in p-AVP, presumably due to a low p-AVP caused by oral water loading in advance or the fact that the measurements of p-AVP concentration may not be sensitive enough to detect small changes. It could be hypothesized that the decrease in u-ENaCγ could be due to a lack of AVP binding to V2 receptors in the basolateral membrane of the principal cell. Lack of AVP stimuli leads to an increased endocytosis of ENaC channels from the membrane surface into recycling vesicles, there by decreasing reabsorption of sodium \[[@B54],[@B55]\]. Thus, in the last post infusion period u-ENaCγ increased markedly after hypertonic saline infusion, was approximately at the same level after isotonic saline and decreased or tended to decrease in response to glucose infusion. The rise in p-osm and p-AVP was seen immediately after 3% NaCl infusion stopped. The delay and constant level of u-ENaCγ after hypertonic saline could be explained by the fact that it takes few minutes to increase trafficking of intracellular depots of ENaC channels into the apical membrane but several minutes to excrete ENaC into the urine after stimulation with AVP. Vasoactive hormones ------------------- In addition to AVP, the renin-angiotensin-aldosterone system (RAAS) is a key regulator of renal sodium excretion and thereby of body fluid volume. It is well known that sodium depletion activates and that chronic sodium load reduces the RAAS \[[@B56]\]. In vitro and in vivo studies have shown that aldosterone stimulates the mineralocorticoid receptor to an increased transcription of genes-encoding proteins involved in sodium transport i.e. ENaC and Na,K-ATPase \[[@B57]\]. Numerous studies of changes in blood volume have demonstrated that acute changes are associated with inverse adjustments of the renin-angiotensin-aldosterone system \[[@B21],[@B27],[@B51],[@B58]\]. In the present study, volume expansion with 3% and 0.9% saline resulted in a similar and significant reduction in PRC, p-AngII and Aldo consistent with an increase in extracellular volume. This is in agreement with previous studies \[[@B51],[@B58]\]. After glucose infusion, we measured no significant change in PRC, p-AngII or p-Aldo. This was expected, as glucose infusion does not cause any marked change in extracellular volume. Our study was not designed to allow any regulatory effects of aldosterone as the action of aldosterone occurs over hours or days. Therefor other factors must be implicated in the regulation of ENaC. Strengths and limitations ------------------------- The major strength of this study was the design as a randomized crossover study with a homogenous group of healthy young men and women. The test conditions were very well defined regarding diet, sodium and fluid intake. Thus, the results are not confounded by different salt or water balance. This study explored only the acute effects of volume expansion. No doubt, we could have gained further information regarding the long-term effects of volume expansion and the urinary excretion of AQP2 and ENaCγ if the post infusion period had been longer. In addition, the study was not placebo-controlled, by means of infusion with a negligible amount of 0.9% saline. This could have distinguished the effects of volume expansion from the overall variability of water and salt reabsorption. In this study it was not possible to perform ANP measurements. It could have made a positive contribution to our results. Conclusions =========== In conclusion volume expansion with 3% and 0.9% saline clearly increased u-AQP2, while isotonic glucose decreased u-AQP2. Infusion of hypertonic saline increased u-ENaCγ, whereas u-ENaCγ was not significantly changed after isotonic saline and decreased or tended to decrease after glucose. Thus, the transport of water and sodium changed, both via the aquaporin 2 water channels and the epithelial sodium channels, during all three types of volume expansion in order to regulate and maintain water- and sodium homeostasis in the body. Changes in the renin-angiotensin-aldosterone system did not seem to bear a causal relationship with the changes in u-AQP2 or u-ENaCγ. Competing interests =================== The authors declare that they have no competing interests. The authors alone are responsible for the content and writing of the paper. Authors' contributions ====================== All authors have contributed to the manuscript. JMJ, FHM and EBP designed the project. JMJ and FHM performed the experiments and statistical analyzes. JMJ, FHM, JNB, SN and EBP wrote and edited the manuscript. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2369/14/202/prepub> Acknowledgements ================ The authors greatly acknowledge the skilful assistance of our laboratory technicians: Lisbeth Mikkelsen, Anne Mette Ravn, Kirsten Nygaard and Henriette Vorup Simonsen. Funding ======= The study was supported by grants from Region Midt's Research Foundation for Health Science, The Lundbeck Foundation and Helen and Ejnar Bjornows Foundation.

{ "pile_set_name": "PubMed Central" }

1. Introduction {#sec0005} =============== The mediastinum consists of anatomical structures that can give rise to both benign and malignant tumors, or can be the seat of metastases of malignancies that arise elsewhere in the body. The diagnostic approach to patients with mediastinal pathology is not always simple and an improper diagnostic work-up can lead to significant diagnosis delay. This work is reported in accordance with the SCARE criteria \[[@bib0005]\]. 2. Case presentation {#sec0010} ==================== A 34 year-old woman was admitted to the Emergency Department (ED) complaining of thoracic pain lasting from 6 days, dyspnoea, fever, and productive cough. She had a history of smoke and a major thoracic trauma due to a car accident that occurred one year before. A chest X-ray performed on that occasion had ruled out a chest wall injury. At admission to the ED her physical examination showed a painful swelling over the sternum's upper left margin, which had become evident 4 months earlier. A chest X-ray revealed a homogeneous opacity behind the sternum. Blood tests were irrelevant, except for a mildly elevated C-reactive protein (CRP 3.62 mg/dl). To better characterize the lesion, a chest Computer Tomography (CT) scan was performed. This evaluation showed a retrosternal oval lesion (5.5 x 4 cm) with an intra- and extra-thoracic growth and a multifocal lymphadenopathy ranging between 7 mm and 4 cm in the prevascular area, aortopulmonary window, and right paratracheal station ([Fig. 1](#fig0005){ref-type="fig"}).Fig. 1Preoperative CT scans showing the mediastinal mass and the involvement of the sternum.Fig. 1 The new findings, together with the young age of the patient, sped up our investigations, and we performed an ultrasound guide biopsy of the retrosternal mass. At histology, the specimen consisted of "fibroconnectival tissue with necrotic areas, lymphocytic, eosinophilic, and granulocytic infiltrate". No significant results emerged from cultures. All the additional imaging procedures, including abdominal ultrasound and skeletal scintigraphy, proved inconclusive. As anti-inflammation therapy was ineffective on dyspnoea and chest pain, a new CT scan was performed 30 days later. The size and characteristics of the retrosternal lesion had not changed, but the scan revealed a new paracardiac lesion showing similar features, a multifocal bone involvement with osteolytic and osteosclerotic areas and multiple, bilateral, pulmonary nodules, suspicious for malignancy. A bilateral, axillary lymphadenopathy was also described. The retrosternal lesion maintained the same size and features, but there was a deeper bone structural alteration. In the meantime the patient complained of the persistence of pain and dyspnoea, so that a Positron Emission Tomography (PET)/CT scan was performed. It showed a high metabolic activity lesion in the mediastinum, involving the sternum. These data suggested a malignant aetiology for the mediastinal mass, and therefore a surgical biopsy was performed. Pathological analysis of the surgical specimen showed a lymphoid lesion surrounded by dense fibrotic tissue. Within the lesion a mixed infiltrate was evidenced, composed of small lymphocytes, histiocytes, plasmacells, and several eosinophils. Furthermore, several mono- or plurinucleated lymphoid cells were present, reminiscent of Reed-Sternberg cells, sometimes surrounded by a clear halo. At immunohistochemistry these cells showed cytoplasmic (often membranous) positivity for CD30 ([Fig. 2](#fig0010){ref-type="fig"}), and negativity for CLA, CK, CD79 and CD3. The final diagnosis was Hodgkin's lymphoma.Fig. 2Pathological images of the lesion: A) lymphoid nodule surrounded by fibrotic tissue (H&E, 20x), B) intralesional infiltrate composed of small lymphocytes, histiocytes, plasmacells, and numerous eosinophils (H&E, 200x), C) several mono- or plurinucleated lymphoid cells are evident, reminiscents of Reed-Sternberg cells, sometimes surrounded by a clear halo (H&E, 400x), D) these cells showed cytoplasmic (often membranous) positivity for CD30 (CD30, 600x).Fig. 2 The patient underwent chemotherapy in the Haematology-Oncology Department. The chest CT follow-up showed a great improvement: there was a reduction of the size of the primary lesion and the lymphadenopathy, and a great increase in the degree of mineralization of the sternum. 3. Discussion {#sec0015} ============= A sternal swelling associated to thoracic pain and fever can be due to several conditions, such as neoplasms or inflammatory and infective processes. In the present case, the gender, age, and clinical history of the patient, as well as the inflammatory pattern observed, could suggest an inflammatory condition, namely the Tietze's syndrome \[[@bib0010]\]. The Tietze's syndrome was firstly described in 1921 as a benign, non-suppurative, painful swelling of the superior chondrosternal joints \[[@bib0010]\]. It is a quite common pathology \[[@bib0015]\], especially amongst adult women, and it may also follow a thoracic trauma \[[@bib0020]\]. The clinical manifestation is that of a generalized inflammatory state with a huge chondrosternal joints' oedema \[[@bib0025]\]. In particular, the personal history of a thoracic trauma due to a car accident could have explained the sub-acute onset of the symptoms. In fact, the oedema had appeared four months before the ED access. Moreover, the radiological aspects of an aggressive lesion, and the presence of an important mediastinal lymphadenopathy, confirmed by the PET/CT scan, ruled out the Tietze\'s syndrome \[[@bib0030]\]. A malignant tumour with high cellularity \[[@bib0035]\] or a tuberculous osteomyelitis \[[@bib0040],[@bib0045]\] with lung involvement was also suspected. At first, any infectious etiology was excluded, with negative microbiological tests and cultures, including those for *Mycobacteria Tuberculosis*. Unfortunately, the ultrasound guide biopsies were inconclusive in relation to the clinical development, so the surgical biopsy became mandatory \[[@bib0050]\]. Hodgkin's lymphoma (HL), formerly called Hodgkin\'s disease, arises from germinal centre or post-germinal centre B cells \[[@bib0055]\]. HL has a unique cellular composition, containing a minority of neoplastic cells (Reed-Sternberg cells and their variants) in an important acute or chronic inflammatory background \[[@bib0055]\]. Because of the morphological characteristics of HL the diagnosis is not always straight forward \[[@bib0055],[@bib0060]\]. Very few neoplastic diagnostic cells (Reed Sternberg cells) \[[@bib0065],[@bib0070]\] are present within inflammatory tissues containing a variable number of small lymphocytes, eosinophils, neutrophils, macrophages, plasma cells, fibroblasts, and collagen fibers. This case points out how challenging the diagnosis of Hodgkin's disease can be, and the impact that a diagnostic delay can have in the prognosis \[[@bib0075],[@bib0080]\] of a life-threatening condition. The acute and atypical symptoms (pain, dyspnoea, sternal oedema) misled us, letting to the suspicious of a benign disease. Therefore, the correct diagnosis of Hodgkin's lymphoma was delayed as a result. Fortunately, in this case, the response to the specific chemotherapy has been good despite the delay in treatment. 4. Conclusions {#sec0020} ============== In conclusion, mediastinal masses with an aggressive behaviour, confirmed at the chest CT scan or PET CT, should always be considered to be potentially malignant even if needle biopsy proves negative for neoplastic cells. Surgical biopsy can be the only way to correctly diagnose the pathological process, especially in the case of Hodgkin's lymphoma in which few diagnostic cells (Reed-Sternberg cells) are generally embedded in an abundant inflammatory background tissue. Conflicts of interest {#sec0025} ===================== The authors have no conflicts of interest to declare Funding {#sec0030} ======= This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Ethical approval {#sec0035} ================ This study is exempt from ethical approval in our institution. Informed consent {#sec0040} ================ The patient has given informed consent for publication of this manuscript. Author contribution {#sec0045} =================== Alessandro Fois wrote the paper. The other authors collected and analysed data and contributed to the final drafting of the manuscript. Registration of research studies {#sec0050} ================================ Our study does not require registration. Guarantor {#sec0055} ========= Dr. Alessandro Giuseppe Fois.

{ "pile_set_name": "PubMed Central" }

GENOME ANNOUNCEMENT {#s1} =================== Microcystis aeruginosa is one of the most harmful bloom-forming cyanobacteria in freshwater ([@B1]). It is a unicellular colony-forming cyanobacterium distributed worldwide in eutrophic freshwater environments. During summer, Microcystis blooms result in serious environmental problems, such as bad odor releases, bottom-layer anoxia, and production of hepatotoxic cyanotoxins called microcystins. Microcystins are the only cyanotoxins for which the World Health Organization has set drinking and recreational water standards ([@B2]). M. aeruginosa NIES-87 was isolated from Lake Kasumigaura, Japan. This strain is available from the Microbial Culture Collection at the National Institute for Environmental Studies, Japan (<http://mcc.nies.go.jp>). It is an axenic culture and does not produce microcystin; however, it releases a bad odor, the origin of which is unknown. This strain also produces an antialgal peptide kasumigamide ([@B3]). A recent multilocus phylogenetic study suggested that M. aeruginosa is genetically divided into at least eight clades (groups A to G and X) ([@B4]). In that study, the phylogenetic position of NIES-87 was not well resolved. Tominaga et al. ([@B5]) reported that this strain harbors two plasmids with sizes of 2.3 and 5 kb. This strain has been widely used in various studies; however, to date, whole-genome sequencing of the strain has not been conducted. Here, we report the draft genome sequence of M. aeruginosa NIES-87. This information would be useful for unveiling genome evolution in M. aeruginosa. DNA extraction was performed using 10 ml of NIES-87 axenic culture using the DNeasy plant minikit (Qiagen), following the manufacturer's instructions. Subsequently, DNA was fragmented to approximately 550 bp using Covaris M220. Genomic libraries of paired ends and mate pairs were constructed using the TruSeq Nano DNA library prep kit (Illumina) and the Nextera mate pair library preparation kit (Illumina), respectively. Whole-genome sequencing was performed using the MiSeq platform (Illumina) employing the 600-cycle MiSeq version 3 reagent kit. The resulting paired-end and mate pair reads were 1,576,274 and 1,403,390 in number, respectively. Low-quality reads/bases were filtered using Trimmomatic version 0.36 ([@B6]), and *de novo* assembly was performed using SPAdes version 3.7.1 ([@B7]). The resulting genome comprised 206 contigs of 4,947,287 bp. The average genome coverage of the paired-end reads was 95.3×. The maximum contig length was 652,566 bp. Genome annotation was performed using the DDBJ Fast Annotation and Submission Tool (DFAST) ([@B8]). The genome comprised 4,355 predicted protein-coding sequences, one set of rRNAs, and 41 tRNAs. The G+C content of the genome was 42.9%. The genome was scanned using antiSMASH ([@B9]), and microviridin and other secondary metabolite gene clusters were predicted. The 16S rRNA sequence of NIES-87 was subjected to a BLAST search, and the top hits obtained were those of M. aeruginosa NIES-2481 and NIES-2549 (99.8% identity each). We used OrthoFinder ([@B10]) to identify orthologous genes among the three genomes, and 56.9% and 57.1% of the genes in NIES-2481 and NIES-2549, respectively, were identified as orthologous to those of NIES-87. Accession number(s). {#s1.1} -------------------- This whole-genome shotgun project has been deposited in GenBank under the accession no. [BFAC01000001](https://www.ncbi.nlm.nih.gov/nuccore/BFAC01000001) to [BFAC01000206](https://www.ncbi.nlm.nih.gov/nuccore/BFAC01000206). **Citation** Yamaguchi H, Suzuki S, Kawachi M. 2018. Draft genome sequence of *Microcystis aeruginosa* NIES-87, a bloom-forming cyanobacterium from Lake Kasumigaura, Japan. Genome Announc 6:e01596-17. <https://doi.org/10.1128/genomeA.01596-17>. We thank Nobuyoshi Nakajima for technical assistance. This study was supported by the Japan Agency for Medical Research and Development (AMED).

{ "pile_set_name": "PubMed Central" }

Abbreviations used in this paper: DCT, distal convoluted tubule; DVF, divalent-free; TRP, transient receptor potential; WMN, wortmannin. INTRODUCTION ============ Transient receptor potential (TRP) channels are widespread and play many important functions, ranging from thermal, tactile, taste, osmolar and fluid flow sensing, and embryo development to epithelial Ca^2+^ and Mg^2+^ transport ([@bib12]; [@bib22]; [@bib2]). They are classified into TRPC, TRPV, TRPM, TRPP, TRPML, TRPA, and TRPN subfamilies ([@bib12]; [@bib22]; [@bib2]). The TRPV subfamily is named after the first mammalian member of the subfamily, vanilloid receptor 1 ([@bib22]; [@bib2]). TRPV5 and TRPV6 are highly Ca^2+^-selective TRP channels that mediate trans-epithelial Ca^2+^ transport in kidney and intestine ([@bib12]). Intracellular Mg^2+^ is a cofactor for many enzymes and controls inward rectification of ion channels by causing voltage-dependent block of outward currents ([@bib33]; [@bib9]). Recently, intracellular Mg^2+^ has also been reported to regulate the activity of several TRP channels including TRPM6, TRPM7, and TRPV6 via a mechanism different from voltage-dependent block ([@bib23]; [@bib34]; [@bib42], [@bib43]). The intracellular free Mg^2+^ concentration in most mammalian cells is between 0.5 and 1 mM and changes only slightly in response to physiological stimuli ([@bib33]). TRPM7 is permeable to Mg^2+^ ([@bib37]). It is believed that Mg^2+^ ions entering cells through TRPM7 feedback on the channel to regulate its activity ([@bib37]). TRPV6 does not conduct Mg^2+^ in physiological conditions. The physiological role of intracellular Mg^2+^ regulation of TRPV6 is unknown. PIP~2~ directly regulates inward rectifying K^+^ and other channels ([@bib11]). How membrane lipids alter channel function is an area of intense research interests. For many ion channels, direct interaction with PIP~2~ stabilizes channels in certain conformations ([@bib11]). By stabilization in one conformation, PIP~2~ modulates the responses of ion channels to regulators, such as GTP binding proteins, intracellular pH, Na^+^, ATP, etc. ([@bib14]; [@bib38]; [@bib39]; [@bib17]). In this report, we show that intracellular Mg^2+^ causes a fast voltage-dependent block and a slower inhibition of TRPV5. We further show that PIP~2~ activates TRPV5 and decreases sensitivity of the channel to the Mg^2+^-induced slow inhibition. Hydrolysis of PIP~2~ via receptor activation of phospholipase C increases the sensitivity of TRPV5 to the Mg^2+^-induced slow inhibition. These results provide a novel mechanism for regulation of TRP channels by PLC-activating hormones by increasing the sensitivity to intracellular Mg^2+^. MATERIALS AND METHODS ===================== Molecular Biology and Cell Culture ---------------------------------- Nucleotide coding sequence of cDNAs for rabbit TRPV5 in pCDNA3 mammalian expression vector ([@bib47]) was used as template for site-directed mutagenesis using a commercial mutagenesis kit (QuickChange; Stratagene). Mutations were confirmed by sequencing. CHO-K1 clone (from American Type Culture Collection) were cultured in F12-K medium (GIBCO-BRL) containing 10% FCS. Cells (at ∼50% confluence) were cotransfected with cDNA for pEGFP (1 μg) plus cDNAs for wild-type or mutant TRPV5 (10 μg per 35 mM dish for cell-attached and inside-out recording; 3 μg per dish for whole-cell recording) using lipofectamine-plus transfection kits (GIBCO-BRL) as previously described ([@bib47], [@bib48]). About 24--48 h after transfection, cells were dissociated by limited trypsin treatment and placed in a chamber for recording. Transfected cells were identified using epi-fluorescent microscopy for recordings. Electrophysiological Recordings ------------------------------- Cell-attached, inside-out, and whole-cell recordings were performed using an Axopatch 200B patch-clamp amplifier (Axon Instruments) as described previously ([@bib47], [@bib48]). For cell-attached recordings, the pipette and bath solution contained (in mM) 140 NaAsp (sodium aspartate), 10 NaCl, 1 EDTA, 10 HEPES (pH 7.4), and 140 KAsp, 10 NaCl, 1 EDTA, 10 HEPES (pH 7.4), respectively. For cell-attached recordings shown in [Fig. 8](#fig8){ref-type="fig"} A, the pipette solution contained (in mM) 140 NaCl, 1 MgCl~2~, 1 CaCl~2~, and 10 HEPES (pH 7.4). For inside-out recordings, membrane patches were excised into a bath solution containing (in mM) 140 NaAsp, 10 NaCl, 10 HEPES (pH 7.4), 10 EDTA, and various amounts of MgCl~2~ titrated to the desired concentrations of ionized Mg^2+^ as specified. Bath solution was perfused at ∼200 μl per second. The volume of perfusion chamber and connecting tubing is ∼100 μl. Thus, it takes 2--3 s to exchange the bath solution. Concentrations of ionized Mg^2+^ were calculated using a computer program based on algorithms by Fabiato ([@bib6]) ([www.stanford.edu/\~cpatton/webmaxc/webmaxcE.htm](http://www.stanford.edu/~cpatton/webmaxc/webmaxcE.htm)). For whole-cell recordings, the initial pipette and bath solution contained (in mM) 140 NaAsp, 10 NaCl, 10 HEPES (pH 7.4), 10 EDTA, and 140 NaAsp, 10 NaCl, 1 EDTA, 10 HEPES (pH 7.4), respectively. For whole-cell recordings shown in [Fig. 9](#fig9){ref-type="fig"} B, the bath solution contained (in mM) 140 NaCl, 1 MgCl~2~, 1 CaCl~2~, and 10 HEPES (pH 7.4). Intracellular Mg^2+^ in whole-cell recording was altered by intrapipette perfusion of solutions containing (in mM) 140 NaAsp, 10 NaCl, 10 HEPES (pH 7.4), 10 EDTA, and various amount of MgCl~2~ titrated to the desired concentrations of ionized Mg^2+^ as specified. The intrapipette perfusion was performed via a capillary quartz tubing (flamed and pulled manually to outer and inner diameter ∼50 and 20 μm, respectively) placed within ∼200 μm of the tip of the opening ([@bib10]). Voltage protocol for each experiment was described in the individual figure. Currents were low-pass filtered at 1 kHz using an 8-pole Bessel filter, sampled every 0.1 ms (10 kHz) with Digidata-1300 interface and stored directly onto computer hard disk using pCLAMP9 software. Data were transferred to compact discs for long-term storage. Data Analysis ------------- Dose--response curves for inhibition of TRPV5 by intracellular Mg^2+^ and concentration of Mg^2+^ for half-maximal inhibition (IC~50~) were obtained by fitting relative currents (normalized to maximal currents; I/Imax) at different Mg^2+^ concentrations using the Sigma-Plot program according to equation I/Imax = IC~50~/IC~50~ + \[Mg^2+^\]. The fractional electrical distance of the Mg^2+^ binding site from the outer surface of the membrane (δ) was calculated according to equation IC~50~(*V*) = IC~50~(0)exp (2δ*FV*/*RT*) ([@bib26]), where IC~50~(*V*) and IC~50~(0) represent IC~50~ at the test voltage *V* and at 0 mV, respectively, and *R*, *T*, and *F* have their usual meanings. Current traces were fitted to a single or double exponential curve (as indicated) to analyze time constant (τ) for inhibition of currents by Mg^2+^. The on-rate constant (K~on~) for slow inhibition by Mg^2+^ was calculated according to the equation K~on~ = 1/τ\[Mg^2+^\]. Data are shown as mean ± SEM of number of observations. Statistical comparison was made using unpaired Student\'s *t* test. RESULTS ======= Multiple Effects of Intracellular Mg^2+^ on TRPV5: Slow Reversible Inhibition, Fast Voltage-dependent Block, and "Run-down" via Reduction of PIP~2~ --------------------------------------------------------------------------------------------------------------------------------------------------- ### Effect One: Slow Reversible Inhibition We studied TRPV5 channels expressed in Chinese hamster ovary (CHO) cells by cell-attached and inside-out patch-clamp recording ([Fig. 1](#fig1){ref-type="fig"} A). In cell-attached recording, TRPV5-expressing CHO cells exhibited the characteristic strongly inwardly rectifying Na^+^ currents ([@bib27]; [@bib13]; [@bib47]) ([Fig. 1](#fig1){ref-type="fig"} B, C/A). Inward (closed circles) and outward currents (open circles) both increased markedly when inside-out patch membranes containing TRPV5 channels were excised ([Fig. 1](#fig1){ref-type="fig"} B, I/O) into divalent-free (DVF) bath solutions. Exchanging DVF solutions with one that contains 1 mM Mg^2+^ caused inhibition of TRPV5 currents over tens of seconds ([Fig. 1, B and C](#fig1){ref-type="fig"}). Currents recovered upon reintroduction of DVF solutions. TRPV5 currents increased by a similar extent when inside-out membranes were excised into Mg^2+^-free bath solutions containing 0, 200, or 1,000 nM Ca^2+^ (unpublished data). Thus, most (if not all) of the increase in currents upon excision of inside-out membranes into DVF solutions is due to loss of inhibition by intracellular Mg^2+^. We refer to this inhibition by intracellular Mg^2+^ occurring over tens of seconds as "slow reversible inhibition" to distinguish it from the much faster (also reversible) voltage-dependent block described below. ![Multiple effects of intracellular Mg^2+^ on TRPV5. (A) Configuration and voltage protocol of recording. Voltage stimuli were applied every 10 s. Black and gray arrows indicate time points when inward and outward currents shown in [Fig. 1](#fig1){ref-type="fig"} B were taken, respectively. (B) Intracellular Mg^2+^ causes a fast reversible voltage-dependent block, a slow reversible inhibition, and an irreversible run-down of TRPV5 in inside-out membranes. See text for details. Inside-out patches were bathed in either DVF or 1 mM Mg^2+^-containing solution. C/A and I/O indicate cell-attached and inside-out. Application of PIP~2~ and 1 mM La^3+^ (La) in DVF solution are indicated. Time interval between data points (shown in circles) is 10 s. As described previously by us ([@bib47], [@bib48]), expression of TRPV5 in CHO cells is robust. We estimated that under our experimental condition each excised patch contains ∼50--150 channels. PIP~2~ reactivated run-down TRPV5 channels (i.e., currents recovered to \>50% of the level before run-down) in 54 out of 62 recordings. We did not observe significant background currents in inside-out membranes of mock-transfected cells either before or after PIP~2~ (inward \[−100 mV\] and outward \[+100 mV\] currents: −87 ± 32 pA and 117 ± 28 pA, respectively, before PIP~2~ vs. −135 ± 41 pA and 156 ± 48 pA, respectively, after PIP~2~; mean ± SEM, *n* = 6). (C) Ramp I-V curves of currents at indicated time points in B. Scale bars for X (40 ms) and Y axis (0.2 nA) are shown. Dotted line indicates 0 current. In the X axis, the time points corresponding to ramp potentials −100, 0, and +100 mV (from left to right) are indicated by three downward arrows, respectively. Inset shows I-V curves scaled up to the same peak current level.](200509314f1){#fig1} ### Effect Two: Voltage-dependent Block and Distinction from the "Slow Reversible Inhibition" Intracellular free Mg^2+^ also causes reversible voltage-dependent block of TRPV5 occurring over a relatively much faster time course (in millisecond time scale; see [@bib13]; [@bib42]). The distinction between the above slow reversible inhibition occurring over tens of seconds and the voltage-dependent block by Mg^2+^ is rather evident from our results (see ramp I-V curves in [Fig. 1](#fig1){ref-type="fig"} C). First, as mentioned above, inward currents shown in [Fig. 1](#fig1){ref-type="fig"} B were measured at the end of 40-ms hyperpolarization (−100 mV) step (indicated by black arrow in voltage ramp protocol in [Fig. 1](#fig1){ref-type="fig"} A). At these time points, Mg^2+^ ions were completely dissociated (from channels) by the hyperpolarization ([Fig. 1](#fig1){ref-type="fig"} C) and thus did not exert voltage-dependent block on channels. Second, the kinetics of voltage-dependent block distinguishes itself from the slow reversible inhibition. The Mg^2+^-induced voltage-dependent block of TRPV5 was evident by the characteristic unblock during the −100 mV hyperpolarization step (indicated by "a" in [Fig. 1](#fig1){ref-type="fig"} C), block of outward current ("b"), shift of reversal potential toward more negative membrane potential ("c"), and steeper I-V near the reversal potential ("d") in the current--voltage (I-V) relationship curves ([Fig. 1](#fig1){ref-type="fig"} C). The scaled-up I-V curves of currents from 10, 20, and 30 s after application of Mg^2+^ (time point 2, 3, and 4, respectively; time point 1 is immediately before application of Mg^2+^) were superimposable ([Fig. 1](#fig1){ref-type="fig"} C, inset), indicating that voltage-dependent block reached its maximal effect immediately after application of intracellular Mg^2+^ while slow reversible inhibition gradually reached its maximum over 30--40 s ([Fig. 1, B and C](#fig1){ref-type="fig"}). Thus, intracellular Mg^2+^ causes distinct fast voltage-dependent block (occurring in ms time scale) and slower reversible inhibition of TRPV5 (in seconds to tens of seconds) (see [Fig. 6](#fig6){ref-type="fig"} for time constant measurements for both processes). ### Effect Three: "Run-down" via Reduction of PIP~2~ Membrane phospholipid PIP~2~ regulates function of many ion channels and transporters ([@bib11]). In inside-out membranes, a prolonged exposure of the cytoplasmic face to Mg^2+^ activates Mg^2+^-dependent lipid phosphatases and depletes PIP~2~ ([@bib14]; [@bib18]). Depletion of PIP~2~ causes an irreversible loss of activity ("run-down") of channels that are regulated by PIP~2~ ([@bib14]). For TRPV5, though each short exposure to Mg^2+^ solution caused reversible inhibition, repetitive exposures to 1 mM Mg^2+^ solutions led to progressive irreversible run-down of currents eventually ([Fig. 1](#fig1){ref-type="fig"} B). Application of exogenous PIP~2~ to inside-out membranes reactivated TRPV5 after its irreversible run-down by Mg^2+^ ([Fig. 1](#fig1){ref-type="fig"} B), suggesting that irreversible run-down of TRPV5 by intracellular Mg^2+^ is due to loss of PIP~2~ in the membrane and mechanistically different from the reversible voltage-dependent block and "slow inhibition" by Mg^2+^. Activation of a Ca^2+^-sensitive PLC by intracellular Ca^2+^ depletes PIP~2~ and leads to desensitization of TRPM8 channels ([@bib32]). We found that U-37122 (an inhibitor of PLC, 5--10 μM) did not prevent run-down of TRPV5 caused by prolonged exposure to Mg^2+^ (unpublished data), supporting that Mg^2+^-induced run-down is caused by activation of lipid phosphatase(s) rather than by activation of some Mg^2+^-sensitive PLC. A monoclonal antibody against PIP~2~ ([@bib7]) interferes with PIP~2~ regulation of ion channels presumably by sequestering PIP~2~ in the membranes ([@bib14]). Application of anti-PIP~2~ antibody to the cytoplasmic face of inside-out membranes decreased TRPV5 currents ([Fig. 2](#fig2){ref-type="fig"} A). Currents recovered partially after extensive washout of the antibody ([Fig. 2](#fig2){ref-type="fig"} B). Polylysine, with its positively charged residues, has also been used to inhibit the activity of PIP~2~-activated ion channels through competing PIP~2~--channel interaction ([@bib14]; [@bib28]). We found that application of polylysine (average molecular weight 1 kD; 30 μg/ml) to the inside-out membranes inhibited TRPV5 by 89 ± 9% in 3 min (*n* = 5; unpublished data). Similar to application of PIP~2~, application of Mg-ATP after irreversible run-down in Mg^2+^ solutions reactivated TRPV5 ([Fig. 2](#fig2){ref-type="fig"} B). Mg-ATP presumably works by activating lipid kinases, including phosphoinositide 4-kinase to regenerate PIP~2~ in the membranes ([@bib14]; [@bib11]). Consistent with this idea, the reversal by Mg-ATP was blunted by an inhibitor of phosphoinositide 4-kinase wortmannin ([@bib24]) ([Fig. 2](#fig2){ref-type="fig"} B, WMN). Furthermore, Na-ATP did not reactivate TRPV5 (unpublished data). [Fig. 2](#fig2){ref-type="fig"} C shows I-V curves of currents at several indicated time points. These results provide compelling evidence for a critical role of membrane PIP~2~ in the function of TRPV5. ![PIP~2~ is critical for TRPV5 channel activity. (A) Anti-PIP~2~ antibody reduces TRPV5 currents in inside-out patches. Experiments were repeated five times with similar results. The average (mean ± SEM) inhibition by anti-PIP~2~ antibody was 88 ± 13% and 79 ± 14% for outward and inward currents, respectively. (B) A representative experiment showing long exposure of inside-out patches to Mg^2+^ causes run-down of TRPV5, which can be subsequently reactivated by Mg-ATP. // indicates 10 min break. The inhibition by Mg^2+^ shown here appears to be slower compared with that in [Fig. 1](#fig1){ref-type="fig"} B. This variability is likely due to differences in the configuration and accessibility of excised membranes to perfused bath solutions. Closed and open circles are currents at −100 and +100 mV, respectively. Similar results were observed in six separate experiments. In some experiments, Mg-ATP was applied together with WMN (10 μM). Gray triangles illustrate one such recording receiving Mg-ATP + WMN superimposed on one receiving the Mg-ATP alone. Similar results were observed in four separate experiments. Dimethyl sulfoxide (DMSO, vehicle for WMN, 0.5%) did not prevent reactivation by Mg-ATP (*n* = 4, not depicted). Of note is that wortmannin did not prevent recovery of currents from transient inhibition by Mg^2+^ (not depicted), indicating that the slow reversible inhibition is not due to regeneration of PIP~2~ from lipid kinases. (C) I-V curves of currents at indicated time points in B. In this experiment, voltage ramp from −100 to +100 mV was applied without a 40-ms step at −100 mV as indicated in [Fig. 1](#fig1){ref-type="fig"} A. Dotted line indicates 0 current. In the X axis, the time point corresponding to ramp potential 0 mV is indicated by upward arrow. The characteristic voltage-dependent block in trace 3 is due to free (ionized) Mg^2+^ (0.13 mM) present in the Mg-ATP solution.](200509314f2){#fig2} Application of PIP~2~ to Inside-out Membranes Releases Mg^2+^-induced Slow Reversible Inhibition ------------------------------------------------------------------------------------------------ Activation of inward-rectifier K^+^ and other channels by PIP~2~ modulates the regulation of channels by other signaling pathways ([@bib11]). We examined whether PIP~2~ activation of TRPV5 alters its sensitivity to Mg^2+^-induced slow inhibition. We examined dose--response relationships for the intracellular Mg^2+^-induced slow reversible inhibition of TRPV5 in excised inside-out membranes before and after application of PIP~2~. As before, after reaching maximal currents in DVF solutions in inside-out membranes, TRPV5 currents were completely inhibited by a solution containing 1 mM Mg^2+^ ([Fig. 3](#fig3){ref-type="fig"} A). Currents recovered in DVF solutions. The concentration of Mg^2+^ for half-maximal inhibition (IC~50~) of the inward currents of TRPV5 was 0.11 ± 0.03 mM (mean ± SEM, *n* = 7; [Fig. 3](#fig3){ref-type="fig"} C, closed circles). Application of exogenous PIP~2~ to inside-out membranes in DVF solutions increased TRPV5 currents by 2.4 ± 0.3--fold (*n* = 8; see [Fig. 3](#fig3){ref-type="fig"} A for representative experiment), indicating a submaximal concentration of PIP~2~ present in the excised membranes. The application of PIP~2~ reduced the sensitivity of TRPV5 to the slow reversible inhibition by intracellular Mg^2+^. After PIP~2~, the inward currents of TRPV5 were partially inhibited by 1 mM Mg^2+^ but completely inhibited by 10 mM Mg^2+^ ([Fig. 3](#fig3){ref-type="fig"} A). Currents recovered in the DVF solution ([Fig. 3](#fig3){ref-type="fig"} A). The IC~50~ for the slow inhibition by Mg^2+^ was increased by ∼28-fold to 3.11 ± 0.25 mM (*n* = 8; [Fig. 3](#fig3){ref-type="fig"} C, open circles) after application of exogenous PIP~2~. The Mg^2+^-induced voltage-dependent block remains distinguishable from the slow reversible inhibition even after application of exogenous PIP~2~. As shown by the scaled-up I-V curves ([Fig. 3](#fig3){ref-type="fig"} B), voltage-dependent block was complete in 1 mM Mg^2+^ in the presence of PIP~2~ (i.e., I-V curves at time point 2 and 3 are superimposable, [Fig. 3](#fig3){ref-type="fig"} B). Yet, Mg^2+^-induced slow inhibition requires 10 mM Mg^2+^ over 30--40 s to reach its maximum (i.e., more inhibition occurred from time point 2 to time point 3 in [Fig. 3](#fig3){ref-type="fig"} A). ![PIP~2~ decreases the sensitivity to Mg^2+^-induced slow inhibition but not the voltage-dependent block. (A) Application of PIP~2~ in inside-out patches increases TRPV5 currents and reduces the sensitivity to inhibition by Mg^2+^. Voltage ramp protocol is same as in [Fig. 1](#fig1){ref-type="fig"}. Holding potential between ramps is 0 mV. (B) Scaled-up I-V curves of currents at indicated time points in A. (C) Dose--response curves of inhibition by Mg^2+^ in inside-out patches before (no PIP~2~) and after application of PIP~2~ (+PIP~2~). IC~50~ was obtained by fitting relative inward currents (inward currents at −100 mV normalized to maximal currents; I/Imax) at different Mg^2+^ concentrations according to the equation I/Imax = IC~50~/IC~50~ + \[Mg^2+^\].](200509314f3){#fig3} Effect of Intracellular Mg^2+^ on TRPV5 in Whole Cell ----------------------------------------------------- The concentration of PIP~2~ in the excised membranes is likely considerably lower than in the plasma membranes in vivo ([@bib25]). We further examined regulation of TRPV5 by intracellular Mg^2+^ using whole-cell recording ([Fig. 4](#fig4){ref-type="fig"} A), which resembles physiological plasma membrane conditions better than the inside-out membranes. Strongly inwardly rectifying Na^+^ currents were observed immediately after establishment of whole-cell recording in TRPV5-expressing cells (time point 1 in [Fig. 4](#fig4){ref-type="fig"} B and trace 1 in [Fig. 4](#fig4){ref-type="fig"} C) but not in mock-transfected CHO cells ([Fig. 4](#fig4){ref-type="fig"} D). Dialysis of the intracellular space with a DVF solution led to a large increase of whole-cell TRPV5 currents ([Fig. 4, B and C](#fig4){ref-type="fig"}). As reported previously ([@bib13]; [@bib47]), whole-cell TRPV5 currents exhibited slight outward rectification at membrane potentials \>+50 mV in the presence of extracellular EDTA ([Fig. 4](#fig4){ref-type="fig"} C, trace 3). Perfusion of the intracellular space with a solution containing 1 mM Mg^2+^ caused a large reversible inhibition on whole-cell TRPV5 currents ([Fig. 4, B and C](#fig4){ref-type="fig"}). Unlike in the inside-out recordings, run-down of TRPV5 by Mg^2+^ is rather infrequent in whole-cell recordings (\>90% of experiments show \>80% recovery upon changing from Mg^2+^ to DVF solutions) despite an average longer exposure of the cytoplasmic face of plasma membrane to Mg^2+^. Thus, run-down of TRPV5 via activation of lipid phosphatases (as observed in excised inside-out patches) is likely not a physiological action of intracellular Mg^2+^ (as in the conditions of whole-cell recording). Intracellular Mg^2+^ at concentrations ranging from 0.1 to 3 mM caused a dose-dependent inhibition of whole-cell TRPV5 currents ([Fig. 4](#fig4){ref-type="fig"} E). The IC~50~ for intracellular Mg^2+^ inhibition of TRPV5 channels in whole-cell recordings was 0.29 ± 0.02 mM (*n* = 11; [Fig. 4](#fig4){ref-type="fig"} F). This IC~50~ is between that for inside-out membranes before PIP~2~ (IC~50~: 0.11 mM) and after PIP~2~ (IC~50~: 3.11 mM), suggesting that PIP~2~ content in whole cell membranes is intermediate between the latter two inside-out membrane conditions. ![Regulation of TRPV5 by intracellular Mg^2+^ in whole-cell recording. (A) Alteration of intracellular Mg^2+^ in whole-cell recording using intrapipette perfusion. (B) Intracellular perfusion of Mg^2+^ (1 mM) in whole-cell recording inhibits TRPV5 reversibly. The Y axis is current density (nA/pF). Time bar for the X axis represents 4 min. (C) I-V curves of currents at indicated time points in B. (D) Time control of intracellular Mg^2+^ perfusion in mock-transfected CHO cells (transfected with GFP plasmid alone). The Y axis is current density (pA/pF). Time bar for the X axis represents 4 min. (E) Dose-dependent inhibition of whole-cell TRPV5 currents by Mg^2+^ (0--3 mM). Each experiment at a given Mg^2+^ concentration was performed as in [Fig. 4](#fig4){ref-type="fig"} B and inward currents were scaled to the same maximal (current at DVF solution; I/I~max~ = 1) and minimal level (current after La^3+^; I/I~max~ = 0). (F) Dose--response curve of Mg^2+^ inhibition in whole-cell recording (closed squares) vs. in inside-out patches before (closed circles; no PIP~2~) and after PIP~2~ (open circles; +PIP~2~). IC~50~ was calculated according to the equation I/Imax = IC~50~ /IC~50~ + \[Mg^2+^\]. I/Imax = relative currents normalized to maximal currents. The two curves for inside-out patches before and after PIP~2~ are from [Fig. 3](#fig3){ref-type="fig"} C.](200509314f4){#fig4} Regulation of ion channels including inward rectifier K^+^ and TRP channels by PIP~2~ involves direct interaction with positively charged amino acids in the channels ([@bib11]; [@bib30]; [@bib32]). Recently, Rohács et al. reported that arginine-606 of the rabbit TRPV5 (corresponding to arginine-599 of the rat TRPV5) is critical for interaction with PIP~2~ ([@bib32]). We found that R606Q mutant was more sensitive to inhibition by intracellular Mg^2+^ in whole-cell recordings (IC~50~: 0.03 ± 0.01 mM for R606Q \[*n* = 4\] vs. 0.29 ± 0.02 mM for wild type, P \< 0.05; not depicted in [Fig. 4](#fig4){ref-type="fig"}). These results further support that PIP~2~ activation of TRPV5 releases intracellular Mg^2+^ inhibition of the channel. Mg^2+^-induced Slow Reversible Inhibition Is Voltage Dependent -------------------------------------------------------------- To examine membrane voltage sensitivity, whole-cell TRPV5 currents were measured in intracellular DVF solution and in a solution containing 0.3 mM Mg^2+^ at different holding potentials. The fraction of TRPV5 currents in 0.3 mM intracellular Mg^2+^ relative to maximal currents in DVF solution (I/Imax) increased with increasing membrane hyperpolarization ([Fig. 5](#fig5){ref-type="fig"} A). Voltage sensitivity for the intracellular Mg^2+^-induced slow reversible inhibition of TRPV5 is further supported by the finding in experiments using whole-cell recording that hyperpolarization of membrane holding potentials from 0 to −50 mV increased IC~50~ for Mg^2+^ from 0.29 to 0.79 mM ([Fig. 5](#fig5){ref-type="fig"} B). Voltage dependency of inhibition by Mg^2+^ can be viewed as Mg^2+^ binds to inhibit the channel within the membrane electrical field. The fractional electrical distance of the binding site from the outer surface of the membrane (δ) can be calculated according to equation IC~50~(*V*) = IC~50~(0)exp(2δ*FV*/*RT*) ([@bib26]), where IC~50~(*V*) and IC~50~(0) represent IC~50~ at the test voltage *V* and at 0 mV, respectively, and *R*, *T*, and *F* have their usual meanings. Our results reveal that Mg^2+^-induced slow inhibition of TRPV5 is owing to Mg^2+^ binding inside the channel at δ = 0.26. ![Voltage dependency of the Mg^2+^-induced slow reversible inhibition. (A) Whole-cell currents were recorded in DVF and 0.3 mM intracellular Mg^2+^ at different holding potentials (from +50 to −100 mV) as in [Fig. 4](#fig4){ref-type="fig"} B. Relative currents (I/I~max~, current at 0.3 mM Mg^2+^/current at DVF) were plotted against holding potentials (V~h~). (B) Dose--response curve of inhibition by Mg^2+^ at 0 and −50 mV holding potentials. Dose--response curves were performed as in [Fig. 4](#fig4){ref-type="fig"} B.](200509314f5){#fig5} Time Constant Measurement Provides Further Support for Voltage Dependency for Mg^2+^-induced Slow Inhibition and its Distinction from Voltage-dependent Block ------------------------------------------------------------------------------------------------------------------------------------------------------------- Up to now, our studies examine the regulation of TRPV5 by intracellular Mg^2+^ by perfusion of solutions containing different concentration of Mg^2+^. To avoid variability in the rate of perfusion and measure time constants for the Mg^2+^-induced voltage-dependent block and the slow inhibition, we recorded whole-cell currents using voltage jump protocol at fixed intracellular Mg^2+^ concentration (1 mM) ([Fig. 6](#fig6){ref-type="fig"} A). Membrane potential was held at −100 mV. At this hyperpolarized potential, TRPV5 is fully open (not inhibited by intracellular Mg^2+^) and conducts inward current (see [Fig. 1](#fig1){ref-type="fig"}). Stepping from −100 to +100 mV resulted in a fast (i.e., occurring in ms and reaching maximal inhibition in ∼50 ms) time-dependent inhibition of the channel (shown as decrease of outward current in [Fig. 6](#fig6){ref-type="fig"} A). This fast inhibition of currents (known as the Mg^2+^-induced voltage-dependent block) represents the time course of binding of intracellular Mg ^2+^ into the pore and contributes to inward rectification of currents through TRPV5 and TRPV6 channels ([@bib13]; [@bib42]). As shown by others ([@bib42]), the time course of voltage-dependent block could be fitted by two exponential terms. [Fig. 6](#fig6){ref-type="fig"} B shows the two exponential time constants for voltage-dependent block at +25, +50, and +100 mV, respectively. Beyond the fast voltage-dependent block occurring within the first 50 ms, outward currents further decreased in a much slower time course (occurring in seconds and reaching minimal current level in \<60 s) ([Fig. 6](#fig6){ref-type="fig"} C). The mean exponential time constant (τ) for the "Mg^2+^-induced slow inhibition" at +100 mV was 8.3 ± 1.1 s (mean ± SEM, *n* = 4) ([Fig. 6](#fig6){ref-type="fig"} D). The on-rate constant (K~on~) for slow inhibition by Mg^2+^ at +100 mV (K~on~ = 1/τ \[Mg^2+^\]) was estimated at 120 ± 14 s^−1^M^−1^ (mean ± SEM, *n* = 4). Outward currents remained relatively stable beyond the first 50 ms in the absence of intracellular Mg^2+^ (current level at 60 s = 91 ± 8% of that at 100 ms, *n* = 4; not depicted in [Fig. 6](#fig6){ref-type="fig"}). Consistent with voltage dependency of Mg^2+^-induced slow inhibition, time constant at +50 and +25 mV were 13.4 ± 1.2 s and 16.2 ± 0.4 s, respectively ([Fig. 6](#fig6){ref-type="fig"} D, each P \< 0.05 vs. time constant holding at +100 mV). These time constant values at +100, +50, and +25 mV are in agreement with the earlier observation that Mg^2+^-induced slow inhibition occurred over 30--40 s at 0 mV (see [Fig. 1](#fig1){ref-type="fig"}; taking into consideration reduced PIP~2~ content in inside-out patches and probably a few seconds of delay in delivery of Mg^2+^ by perfusion). ![Kinetics of Mg^2+^-induced voltage-dependent block (A and B) and slow inhibition (C and D). (A) Whole-cell recording was performed at fixed intracellular Mg^2+^ (1 mM). Holding potential was −100 mV. Test pulse to +100 mV is illustrated. Time course of outward current (at +100 mV) from peak current (∼1 ms after test potential) to the current level at 50 ms is shown. Dotted line represents double exponential fit to the current. (B) Fast (squares) and slow (circles) exponential time constant of decrease of outward current from peak current (∼1 ms) to 50 ms of test potentials (+25, +50, +100 mV, respectively). (C) Time course of decrease of outward current from 100 ms to 60 s after test potential to +100 mV. Dotted line represents single exponential fit to the current. (D) Exponential time constants of decrease of outward current between 100 ms and 60 s of test potentials (+25, +50, and +100 mV, respectively).](200509314f6){#fig6} Mutation of Aspartate-542 Abolishes Mg^2+^-induced Slow Reversible Inhibition ----------------------------------------------------------------------------- Acidic amino acids located in solvent-accessible region between the second intracellular loop (2^nd^ IC loop) and the proximal region of the intracellular COOH terminus (CT) of TRPV5 may be involved in the binding and slow reversible inhibition by intracellular Mg^2+^ ([Fig. 7](#fig7){ref-type="fig"} A). We mutated seven acidic amino acids within this region, one at a time, to neutral amino acid asparagine or glutamine and tested the sensitivity of each mutant to inhibition by Mg^2+^ in whole-cell recordings. We found that neutralization of aspartate-542 (D542N) in the putative pore region completely abolished the slow reversible inhibition of TRPV5 by intracellular Mg^2+^ ([Fig. 7, B and C](#fig7){ref-type="fig"}). In voltage jump experiment, currents through D542N mutant remained unchanged over 60 s at +100 mV (current level at 60 s = 89 ± 15% of that at 100 ms, *n* = 3; not depicted in [Fig. 7](#fig7){ref-type="fig"}), confirming that aspartate-542 is critical for Mg^2+^-induced slow inhibition. Neutralization of other aspartate or glutamate residues in this region had no significant effects on the slow reversible inhibition of TRPV5 by Mg^2+^ ([Fig. 7](#fig7){ref-type="fig"} D). ![Identification of amino acid of TRPV5 involved in Mg^2+^-induced slow inhibition. (A) Membrane topology of amino acids between the fourth transmembrane (TM4) segment and the proximal COOH terminus (Ct) of TRPV5. Amino acids in the preselectivity filter region likely form a pore helix structure similar to that of KcsA ([@bib5]; [@bib44]). The location of amino acids mutated is shown. (B) A representative experiment of D542N mutant (in whole-cell recording) showing no inhibition by 1 mM Mg^2+^. (C) Dose--response curves of inhibition by Mg^2+^ for D542N mutant and wild-type (WT) TRPV5. (D) Average inhibition by Mg^2+^ for WT and each mutant (mean ± SEM, *n* = 5--11 each). \* indicates P \< 0.05 vs. WT.](200509314f7){#fig7} Aspartate-542 is believed to be a part of the selectivity filter of TRPV5 and a binding site for Mg^2+^-induced voltage-dependent block ([@bib27]; [@bib13]). Our findings indicate that Mg^2+^ binding to aspartate-542 not only causes voltage-dependent block but also the slow inhibition. This conclusion is further supported by the fact that the fractional electrical distance for the slow inhibition we measured in this study (δ = 0.26) is similar to that measured for voltage-dependent block (δ = 0.31) ([@bib26]). Reduction of PIP~2~ Enhances Mg^2+^-induced Slow Reversible Inhibition of TRPV5 without Affecting Mg^2+^ Binding ---------------------------------------------------------------------------------------------------------------- To investigate the mechanism for PIP~2~ modulation of Mg^2+^-induced slow reversible inhibition of TRPV5, we tested whether PIP~2~ affects Mg^2+^ binding. The kinetics of Mg^2+^-induced slow inhibition and voltage-dependent block were studied using voltage jump experiments as in [Fig. 6](#fig6){ref-type="fig"}. Whole-cell recordings were performed in the presence of intracellular Mg^2+^ (1 mM). The extracellular bath solution contained thrombin + wortmannin (WMN) or vehicle. WMN inhibits phosphoinositide 4-kinase (a rate-limiting enzyme for synthesis of PIP~2~ from phosphatidylinositol and phosphatidylinositol-4-phosphate) to allow reduction of membrane PIP~2~ following receptor-mediated hydrolysis of PIP~2~ ([@bib24]). [Fig. 8](#fig8){ref-type="fig"} A shows representative single exponential fit to current traces in response to voltage jump from −100 to +100 mV in the presence of thrombin + WMN (labeled as Thr) or vehicle (Control). As expected for PIP~2~ releasing Mg^2+^-induced slow inhibition, reduction of membrane of PIP~2~ by incubating with thrombin and WMN decreased the time constants for inhibition at depolarized membrane potentials ([Fig. 8](#fig8){ref-type="fig"} B, P \< 0.05, +Thr vs. Control for all potentials). Interestingly, reduction of membrane PIP~2~ had no effects on the time constants for Mg^2+^-induced voltage-dependent block to the channel ([Fig. 8](#fig8){ref-type="fig"} C; time constants measured as in [Fig. 6](#fig6){ref-type="fig"} B). ![Effects of PIP~2~ hydrolysis on Mg^2+^-induced slow inhibition (A and B) and voltage-dependent block (C). (A) Whole-cell currents (between 100 ms and 60 s; at 1 mM intracellular Mg^2+^) in response to voltage jump from −100 mV holding potential to depolarized test potentials were measured as in [Fig. 6](#fig6){ref-type="fig"} C. Cells were incubated with vehicle (Control) or thrombin + WMN (+Thr) in the extracellular bath for 10 min before patch-clamped for whole-cell recordings. Shown are representative single exponential fits to currents for Control (black curve) or +Thr (gray curve). Relative current level corresponding to 1/*e* (*e*, the base of natural logarithms) of the maximal current is shown. (B) Time constants of Mg^2+^-induced slow inhibition at +25, +50, +100 mV for Control and +Thr. \* indicates P \< 0.05 +Thr vs. Control. (C) Time constants of Mg^2+^-induced voltage-dependent block (as in [Fig. 6, A and B](#fig6){ref-type="fig"}) at +25, +50, +100 mV for Control and +Thr.](200509314f8){#fig8} Enhancement of Mg^2+^-induced Slow Inhibition as a Mechanism for PLC-coupled Receptor Regulation TRPV5-mediated Ca^2+^ Entry ---------------------------------------------------------------------------------------------------------------------------- The physiological function of TRPV5 is to allow Ca^2+^ entry through the apical membrane for transepithelial transport of Ca^2+^ in kidney and intestine ([@bib12]). We next examined the effects of PIP~2~ hydrolysis induced by receptor activation of PLC on TRPV5 functioning as a Ca^2+^-permeable channel. In the following cell-attached recordings, the pipette solution contained (in mM) 140 NaCl, 1 MgCl~2~, and 1 CaCl~2~. Because of the anomalous mole fraction behavior, TRPV5 conducts Ca^2+^ exclusively in this physiological solution ([@bib41]). CHO cells express endogenous receptors for thrombin ([@bib4]). Activation of PLC by endogenous thrombin receptors in CHO cells was verified by the release of Ca^2+^ from intracellular stores (unpublished data). Thrombin was applied to the patch membrane via intrapipette perfusion ([Fig. 9](#fig9){ref-type="fig"} A). The bath solution contained vehicle or wortmannin. ![Effect of PIP~2~ hydrolysis on the Mg^2+^-induced slow inhibition. (A) Effect of thrombin plus/minus wortmannin on TRPV5 in cell-attached recording. The top left panel shows cell-attached configuration for recording TRPV5-mediated Ca^2+^ currents. See text for the pipette solution. Voltage stimuli (step to −100 mV for 40 ms and ramp from −100 to +100 mV ramp over 200 ms) were applied every 2 s from +50 mV holding potential. Holding at +50 mV prevents Ca^2+^ entry and the Ca^2+^-dependent inactivation. Thrombin (Thr, 1 U/ml) was applied to the patch membrane by intrapipette perfusion. WMN (10 μM) or vehicle (0.5% DMSO) was applied to the extracellular bath in cells expressing TRPV5 5--10 min before addition of thrombin. The bottom panel shows inward currents (at −100 mV) normalized to the initial current (I/Imax). The top right panel shows ramp I-V curves at indicated time points. (B) Dose--response curve of inhibition by Mg^2+^ before and after membrane PIP~2~ hydrolysis. Whole-cell TRPV5-mediated Ca^2+^ currents were recorded in a bath solution containing (in mM) 140 NaCl, 1 MgCl~2~, 1 CaCl~2~, and 10 HEPES (pH 7.4). Intracellular Mg^2+^ (0 to 3 mM) was altered by intrapipette perfusion. Cells expressing TRPV5 were incubated with a bath solution containing either vehicle (closed circles) or WMN + Thr (open circles) for 5--10 min before patching for whole-cell recording. Voltage protocol and holding potential are as in the cell-attached experiment.](200509314f9){#fig9} In cell-attached recording, application of thrombin (Thr) in the absence of WMN decreased TRPV5 Ca^2+^ currents transiently (by 43 ± 8% in 90 s, *n* = 5, P \< 0.05 vs. control without thrombin) ([Fig. 9](#fig9){ref-type="fig"} A). Presumably due to desensitization of receptors and de novo synthesis of PIP~2~ via phosphoinositide 4-kinase, TRPV5 currents recovered at least partially over 3--4 min in the absence of wortmannin. In the presence of wortmannin, thrombin caused a persistent reduction of TRPV5 currents by 85 ± 5% in 90 s (*n* = 6, P \< 0.05 vs. thrombin alone). As controls, TRPV5 currents did not change significantly without thrombin (decreased by 8 ± 7% over 90 s, *n* = 4; WMN alone in [Fig. 9](#fig9){ref-type="fig"} A). The effects of PIP~2~ hydrolysis on dose-dependent inhibition by Mg^2+^ of TRPV5-mediated Ca^2+^ currents were examined by whole-cell recording ([Fig. 9](#fig9){ref-type="fig"} B). The IC~50~ for Mg^2+^ inhibition of TRPV5-mediated Ca^2+^ currents were 0.61 ± 0.12 mM (*n* = 5) and 0.11 ± 0.08 (*n* = 5) for control cells (treated with vehicle) and cells treated with thrombin and wortmannin, respectively (P \< 0.05). The concentration of the intracellular free Mg^2+^ in the cell-attached recording is estimated at ∼0.5 mM ([@bib33]). These results support the hypothesis that an increase in sensitivity to intracellular Mg^2+^ is important for physiological regulation of TRPV5-mediated Ca^2+^ entry by PLC-activating hormones. DISCUSSION ========== Many TRP channels are downstream of G protein--coupled receptors and PLC (for review see [@bib2]; [@bib31]). Thus, TRP channels may be regulated by G protein--coupled receptors via inositol-1,4,5-trisphosphate and/or diacylglycerol, products of PIP~2~ breakdown catalyzed by PLC ([@bib2]; [@bib31]). Several TRP channels are also directly regulated by PIP~2~ ([@bib3]; [@bib8]; [@bib35]; [@bib19]; [@bib18]; [@bib32]). Therefore, an additional mechanism for PLC to control TRP channels is via reduction of PIP~2~ content ([@bib3]; [@bib35]; [@bib18]). However, the reduction of PIP~2~ in the plasma membrane from physiological stimulation of PLC may not be sufficient to alter the activity of channels if the affinity of channels for PIP~2~ is relatively high ([@bib15]). In the present study, we find that PIP~2~ activates TRPV5 and that activation of TRPV5 by PIP~2~ decreases the sensitivity of TRPV5 to inhibition by intracellular Mg^2+^ (see working model in [Fig. 10](#fig10){ref-type="fig"}). Hydrolysis of PIP~2~ by receptor activation of PLC increases the sensitivity to the inhibition by Mg^2+^. These results provide a novel mechanism for regulation of TRP channels by PLC-activating hormones by increasing the sensitivity to intracellular Mg^2+^. ![A working model showing the relationships of three different effects of Mg^2+^ on TRPV5. Binding of intracellular Mg^2+^ to D542 of TRPV5 causes a fast (in ms time scale) voltage-dependent block. Prolonged binding to the same site by Mg^2+^ (e.g., 10 s) leads to slow inhibition, which is likely due to conformational changes of the channel. The slow inhibition is also voltage dependent. Membrane PIP~2~ stabilizes TRPV5 in open conformation. PIP~2~ interaction with TRPV5 does not affect Mg^2+^ binding but prevents Mg^2+^-induced conformational changes. Mg^2+^ causes run-down by activating lipid phosphatases. However, run-down is probably not a physiological effect of Mg^2+^ as it occurs predominantly in excised inside-out membranes. Hydrolysis of PIP~2~ by activating PLC-coupled receptors sensitizes TRPV5 to inhibition by intracellular Mg^2+^. \* indicates that time constant for Mg^2+^-induced slow inhibition of TRPV5 is shorter in PIP~2~-reduced (e.g., 5 s) than in PIP~2~-repleted state (e.g., 10 s). For comparison, Mg-induced slow inhibition has also been reported for TRPV6, TRPM6, and TRPM7 ([@bib23]; [@bib34]; [@bib42], [@bib43]). Mg^2+^-induced voltage-dependent block contributes to inward rectification of many channels, including inward-rectifier K^+^ channels and TRPV5 and 6. Mg^2+^ would presumably cause run-down on all PIP~2~-regulated ion channels.](200509314f10){#fig10} The cytosolic free Mg^2+^ concentration in most mammalian cells is between 0.5 and 1 mM ([@bib33]). The IC~50~ for the slow inhibition (0.79 mM) at −50 mV (near the physiological resting membrane potential) suggests that intracellular Mg^2+^ is an important physiological regulator of TRPV5 function in vivo. The intracellular free Mg^2+^ concentration changes only slightly in response to physiological stimuli ([@bib33]). However, the dramatic shift in Mg^2+^ sensitivity by PIP~2~ will provide amplification for regulation of TRPV5 by intracellular Mg^2+^ and PLC-activating hormones. One such example is the inhibition of calcium transport in the distal renal tubules by prostaglandin E~2~ through activation of phospholipase C via EP~1~ receptors ([@bib12]). Besides TRPV5 and TRPV6, intracellular Mg^2+^ has recently been recognized as an important regulator of TRPM6 and TRPM7 channels ([@bib23]; [@bib34]; [@bib43]). TRPM7 (also known as magnesium-inhibited cation channel \[MIC\]; [@bib16]; [@bib29]) is a ubiquitous channel important for cellular uptake of magnesium and cell viability ([@bib37]). Magnesium ions entering cells through TRPM7 feedback on the channel to regulate its activity ([@bib37]). TRPM7 is also regulated by PIP~2~ ([@bib35]). Whether PIP~2~ also modulates Mg^2+^ sensitivity for TRPM7 is unknown at present. TRPV5 does not conduct Mg^2+^ in physiological conditions. However, TRPM6 (a close homology of TRPM7) is a Mg^2+^-permeable channel responsible for magnesium transport in epithelial tissues ([@bib36]; [@bib45]; [@bib43]). Both TRPV5 and TRPM6 are present in the distal convoluted tubule (DCT) cells ([@bib12]; [@bib36]; [@bib45]; [@bib43]), which are key cells responsible for transcellular reabsorption of magnesium and calcium in the kidney ([@bib40]). Many diseases with reduced Mg^2+^ transport in the DCT exhibit increased Ca^2+^ transport. These diseases include genetic diseases such as familial hypomagnesemia with hypocalciuria and Gitelman\'s disease ([@bib46]) and acquired diseases such as cis-platinum toxicity ([@bib21]). Our present finding that intracellular Mg^2+^ negatively regulates TRPV5 raises the possibility that a decrease in Mg^2+^ influx through TRPM6 leads to an increase in Ca^2+^ influx through TRPV5/TRPV6 in DCT cells, providing a possible molecular explanation for this disease phenotype. It should be noted that patients with mutations of TRPM6 develop hypocalcemia ([@bib36]; [@bib45]), rather than hypercalcemia as would be expected from an increase in Ca^2+^ reabsorption in DCT cells alone. Hypocalcemia in these patients is believed secondary to parathyroid failure from severe magnesium wasting and hypomagnesemia ([@bib1]). Intracellular Mg^2+^ regulates inward rectification of ion channels including inward rectifier K^+^ channels and TRPV5 and TRPV6 channels via voltage-dependent block ([@bib20]; [@bib13]; [@bib42]). In the present study, we report that intracellular Mg^2+^ regulates TRPV5 additionally via a mechanism likely involving conformational change of the protein ([Fig. 10](#fig10){ref-type="fig"}). We refer to this additional inhibition by Mg^2+^ as slow reversible inhibition to distinguish from the faster voltage-dependent block. A similar dual regulation of TRPV6 by intracellular Mg^2+^ via voltage-dependent block and a slower inhibition has recently been reported ([@bib42]). Aspartate-542 of TRPV5 is a part of the selectivity filter critical for permeation of Ca^2+^ ions and a binding site for Mg^2+^-induced voltage-dependent block of Ca^2+^ and Na^+^ currents ([@bib27]; [@bib13]). Our findings suggest that binding of intracellular Mg^2+^ to aspartate-542 in the selectivity filter inhibits TRPV5 via at least two separate but interrelated mechanisms. The first mechanism is by voltage-dependent block. In this mechanism, Mg^2+^ ions bind and occlude the channel pore, preventing ion permeation. This inhibition by blocking of channel pore occurs as soon as Mg^2+^ enters and binds the selectivity filter driven by membrane depolarization and reverses quickly by membrane hyperpolarization. This is similar to voltage-dependent block of inwardly rectifying K^+^ channels and other channels by Mg^2+^, polyamines, and other charged blockers ([@bib9]). Binding of Mg^2+^ to the selectivity filter of TRPV5 for a longer duration, however, leads to the second mechanism of inhibition: the slow reversible inhibition. The slow reversible inhibition develops over 30--40 s and is not reversed by transient unbinding of Mg^2+^ ions during intermittent membrane hyperpolarization from voltage ramp applied every 10 s The exact mechanism for the slow inhibition remains unknown, but likely involves conformational change(s) of the channel protein. Our results also shed light on the mechanism by which PIP~2~ alters the sensitivity of TRPV5 to slow inhibition by Mg^2+^. PIP~2~ activates many ion channels directly by affecting their structure ([@bib11]). We find that PIP~2~ activates TRPV5 in the absence of intracellular Mg^2+^ ([Fig. 1](#fig1){ref-type="fig"} B). PIP~2~ also regulates D542N, a Mg^2+^ binding site mutant (unpublished data). Thus, activation of TRPV5 by PIP~2~ does not require its regulation by intracellular Mg^2+^. However, activation of TRPV5 by PIP~2~ decreases the sensitivity of TRPV5 for intracellular Mg^2+^-induced slow inhibition. PIP~2~ does so without affecting Mg^2+^ binding to TRPV5 ([Fig. 8](#fig8){ref-type="fig"} C). We suggest that PIP~2~ desensitizes TRPV5 to Mg^2+^-induced slow inhibition by stabilizing channel in an open conformation and prevents the conformational change induced by binding of Mg^2+^. It is known that ion binding may alter the structure of selectivity filter of ion channels. For example, occupancy of K^+^ ions in the selectivity filter slows the rate of C-type inactivation of voltage-gated K^+^ channels ([@bib49]). The crystal structure of the bacterial KcsA K^+^ channel reveals that presence of multiple K^+^ ions in the selectivity filter stabilizes the full open conformation ([@bib50]). Our present study provides an example of conformational changes of ion channels caused by binding of a pore blocker. We thank Dr. D. Hilgemann for discussion and for help in setting up intrapipette perfusion and Drs. Hilgemann, Baum, and Igarashi for critical reading of the manuscript. This work is supported by National Institutes of Health grants DK-20543, DK-54368, and DK-59530 and an Established Investigator Award (0440019N) from American Heart Association. C.L. Huang is holder of the Jacob Lemann Professorship in Calcium Transport of the University of Texas Southwestern Medical Center at Dallas. Olaf S. Andersen served as editor. [^1]: Correspondence to Chou-Long Huang: <chou-long.huang@utsouthwestern.edu>

{ "pile_set_name": "PubMed Central" }

![](transedinobsoc81623-0097){#sp1 .81} ![](transedinobsoc81623-0098){#sp2 .82} ![](transedinobsoc81623-0099){#sp3 .83} ![](transedinobsoc81623-0100){#sp4 .84} ![](transedinobsoc81623-0101){#sp5 .85} ![](transedinobsoc81623-0102){#sp6 .86} ![](transedinobsoc81623-0103){#sp7 .87} ![](transedinobsoc81623-0104){#sp8 .88} ![](transedinobsoc81623-0105){#sp9 .89} ![](transedinobsoc81623-0106){#sp10 .90} ![](transedinobsoc81623-0107){#sp11 .91} ![](transedinobsoc81623-0108){#sp12 .92} ![](transedinobsoc81623-0109){#sp13 .93} ![](transedinobsoc81623-0110){#sp14 .94}

{ "pile_set_name": "PubMed Central" }

1. Introduction =============== Coronaviruses (CoVs) are enveloped viruses with large positive-sense single-stranded RNA genomes. CoVs infect a variety of mammalian and avian species, and can cause serious disease in humans, as exemplified during the 2002--2003 outbreak of the severe acute respiratory syndrome (SARS). The CoV E protein has a well-established role in the assembly of virions where it may induce membrane curvature or aid in membrane scission. Recent studies have expanded the role of CoV E beyond assembly. CoV E has ion channel activity *in vitro*. CoV E also is critical for the efficient trafficking of virions through the secretory pathway, a function that may be related to its ion channel activity. The CoV E protein has recently been shown to inhibit the host cell stress response, implicating it in pathogenesis. New interacting partners for E have been identified that expand the role of the protein during infection. How all of these potential properties of CoV E fit together to impact its function(s) is the focus of this review. 2. Background ============= The E protein, along with N, S, and M, are the major coronavirus structural proteins ([Figure 1](#viruses-04-00363-f001){ref-type="fig"}A,B). The N protein is a soluble protein and packages the RNA genome to form the nucleocapsid. The S protein has a single transmembrane domain, is found in the virion envelope, and serves as the attachment and fusion protein. The M protein has three transmembrane domains, is the most abundant protein in the virion envelope, and directs the assembly process through interactions with the other structural proteins (reviewed in \[[@B1-viruses-04-00363],[@B2-viruses-04-00363]\]). CoV E is a small (76--109 amino acids) integral membrane protein, and has a single predicted hydrophobic domain (HD). It is usually encoded as the second or third message in a bi- or tri-cistronic mRNA \[[@B3-viruses-04-00363],[@B4-viruses-04-00363]\]. The E protein is targeted to the Golgi region in infected cells and also when expressed from cDNA \[[@B5-viruses-04-00363],[@B6-viruses-04-00363],[@B7-viruses-04-00363],[@B8-viruses-04-00363],[@B9-viruses-04-00363]\]. The membrane topology of CoV E is of considerable debate, having been reported as transmembrane or a membrane hairpin (see below) \[[@B5-viruses-04-00363],[@B10-viruses-04-00363],[@B11-viruses-04-00363],[@B12-viruses-04-00363]\]. There are conserved membrane proximal cysteine residues that are targets for palmitoylation \[[@B7-viruses-04-00363],[@B13-viruses-04-00363],[@B14-viruses-04-00363],[@B15-viruses-04-00363]\]. There are also conserved proline residues in the C-terminal tail ([Figure 1](#viruses-04-00363-f001){ref-type="fig"}C,D). Other than these similarities there is large variation in the primary sequence of the E proteins, which differ in size and sequence among virus groups often having \<30% identity ([Figure 1](#viruses-04-00363-f001){ref-type="fig"}C). How or if these differences affect protein function is not known. ![Primary structure of the Coronavirus (CoV) E protein. (**A**) A cartoon depicting a CoV virion. The structural proteins are labeled. (**B**) The three major CoV structural proteins in the virion envelope. Oligosaccharides are shown on S and M. A single topology is shown for E, see below for discussion on E protein topology. (**C**) A multiple sequence alignment of several different CoV E proteins. The hydrophobic domain (HD) is bracketed. The CoV genera (alpha, beta, and gamma) are denoted on the left of the multiple sequence alignment. Positively charged residues are shown in blue, negatively charged residues are shown in red, and polar uncharged residues are shown in yellow. The conserved Cys and Pro residues are labeled with a bracket or an asterisk, respectively. The multiple sequence alignment of CoV E proteins was carried out with ClusalW2 at the European Bioinformatics Institutes server, and Jalview software version 2 was used to generate the figure \[[@B16-viruses-04-00363],[@B17-viruses-04-00363]\]. (**D**) Cartoon depiction of the E protein with the hydrophobic domain shown as a cylinder and the conserved Cys and Pro residues labeled.](viruses-04-00363-g001){#viruses-04-00363-f001} 3. CoV E and Virus Assembly =========================== Unlike many other enveloped viruses, CoVs assemble and bud intracellularly at the ER-Golgi intermediate compartment (ERGIC) ([Figure 2](#viruses-04-00363-f002){ref-type="fig"}A) \[[@B18-viruses-04-00363],[@B19-viruses-04-00363]\]. One of the early discoveries in CoV assembly was that formation of the virion envelope required only expression of M and E and not N. Originally observed for mouse hepatitis virus (MHV) \[[@B20-viruses-04-00363]\], this property has been observed for infectious bronchitis virus (IBV) \[[@B5-viruses-04-00363]\], transmissible gastroenteritis virus (TGEV) \[[@B21-viruses-04-00363]\], and bovine coronavirus (BCoV) \[[@B21-viruses-04-00363]\]. There has been considerable debate about the requirements for SARS-CoV envelope formation, with reports that M and N \[[@B22-viruses-04-00363]\], M and E \[[@B23-viruses-04-00363]\], and even M alone \[[@B24-viruses-04-00363]\] can drive production of released vesicles. These results raise an interesting point about the efficacy of measuring virus-like particle (VLP) production in different cell types and expression systems. It is important to note that membrane proteins that form multimers can be secreted from cells in microvesicles \[[@B25-viruses-04-00363]\]. Thus, overexpression of viral membrane proteins may lead to release in microvesicles, complicating the interpretation of VLP experiments. Additionally, the original reports for MHV, IBV, TGEV, and BCoV all used vaccinia-based expression systems. What has recently become apparent, at least for MHV \[[@B14-viruses-04-00363]\], SARS CoV \[[@B26-viruses-04-00363]\], and IBV \[[@B27-viruses-04-00363]\], is that when using transient transfection to express the proteins from plasmids, the presence of N can greatly increase VLP yield. This result likely means that while not necessarily required for envelope formation, N plays an important role in forming a complete virion. This makes intuitive sense, and explains why empty virions are not readily purified from infected cells. However, the fact that M and E are sufficient for envelope formation is important information when considering the mechanism of assembly. ![CoVs assemble and bud intracellularly at the ERGIC. (**A**) Newly formed virions bud into the lumen of the ERGIC and traverse the secretory pathway for egress. (**B**) Potential roles for E in assembly. The E protein is shown in orange and the M protein is shown in green. CoV E could help to bend membranes or play a role in membrane scission.](viruses-04-00363-g002){#viruses-04-00363-f002} The ability of E and M to drive VLP formation clearly shows that E is important for assembly, but the mechanism is not well understood. The IBV E and M proteins interact via their cytoplasmic tails \[[@B28-viruses-04-00363],[@B29-viruses-04-00363]\]. These interactions may be important for particle assembly \[[@B21-viruses-04-00363]\]. Additionally, mutations introduced into the C-terminal tail of MHV E produced virions that were unstable, elongated, and may have resulted from failed scission events \[[@B30-viruses-04-00363]\]. However, a version of IBV E with a heterologous HD was able to drive VLP formation and produce infectious virus, suggesting that the sequence of the HD was not important for assembly \[[@B29-viruses-04-00363],[@B31-viruses-04-00363]\]. The E protein may also promote membrane rearrangements. MHV E can drive the intracellular formation of electron dense membranes derived from the ERGIC when expressed alone in BHK-21 cells \[[@B6-viruses-04-00363]\]. Additionally, the HD of SARS-CoV E can drive the *in vitro* tubulation of dimyristoyl phosphatidylcholine (DMPC) membranes \[[@B32-viruses-04-00363]\], although a version of the HD that was flanked with lysine residues did not have the same effect \[[@B33-viruses-04-00363]\]. All of these observations point to a pivotal role for E in assembly, possibly in the scission of particles at the ERGIC, or in inducing membrane curvature ([Figure 2](#viruses-04-00363-f002){ref-type="fig"}B). Thus, it was surprising when versions of MHV, SARS-CoV, and TGEV lacking the E gene were shown to assemble virions, albeit to a lower degree than the corresponding wild-type viruses \[[@B34-viruses-04-00363],[@B35-viruses-04-00363],[@B36-viruses-04-00363]\]. Furthermore, the E proteins from IBV, SARS-CoV, and BCoV could functionally replace MHV E in the context of infection, and TGEV E, which cannot substitute for MHV E, needed only a single amino acid change to complement MHVΔE \[[@B37-viruses-04-00363]\]. This suggests that sequence specific protein-protein interactions between M and E are not required for assembly. One final piece of information came from the study of MHVΔE. Characterization of this virus showed that the production of infectious particles was severely compromised. However, after serial passage, revertants arose with a partial duplication of the M gene consisting of the N-terminus and all three transmembrane domains, but lacking most of the C-terminal cytoplasmic tail \[[@B38-viruses-04-00363]\]. How a truncated version of M can functionally replace E is unknown, but it could allow for spacing between M proteins and disrupt lateral interactions with the M tails. Thus, the exact role of the E protein in assembly is not clear. It may be important for membrane curvature and/or scission directly, or it might alter the spacing of the M protein, which in turn is important for these effects. It is also worth noting the varied requirement for E in virion morphogenesis as revealed by the deletion studies in TGEV, MHV, and SARS-CoV \[[@B34-viruses-04-00363],[@B35-viruses-04-00363],[@B36-viruses-04-00363],[@B39-viruses-04-00363],[@B40-viruses-04-00363]\]. The E protein is essential for TGEV production, but not necessary for MHV or SARS-CoV production. Why different viruses have a varied requirement for the E protein is not understood. One possibility is that a specific accessory protein could complement the assembly process in the absence of E for some CoVs. 4. Post-Translational Modifications =================================== The best characterized post-translational modification (PTM) on E is the addition of palmitic acid onto membrane proximal cysteine residues. This modification has been reported for the E proteins of IBV \[[@B13-viruses-04-00363]\], SARS-CoV \[[@B7-viruses-04-00363]\], and MHV \[[@B14-viruses-04-00363],[@B15-viruses-04-00363]\]. It is likely that palmitoylation alters the conformation of the tail in relation to the membrane. Palmitoylation is not important for proper targeting of the E protein \[[@B13-viruses-04-00363],[@B14-viruses-04-00363],[@B15-viruses-04-00363]\]. The functional significance of palmitoylation has been well studied for MHV E, where it is important for assembly as judged both by VLP production and production of infectious virus \[[@B14-viruses-04-00363],[@B15-viruses-04-00363]\]. These studies suggest that palmitoylation of E affects how it interacts with the M protein, possibly by allowing the E protein to gain access to specific lipid microdomains at the site of assembly \[[@B14-viruses-04-00363]\]. Two other PTMs on the E protein have been reported, but not studied in depth. One study demonstrated that transiently expressed SARS-CoV E is N-glycosylated on asparagine 66 \[[@B10-viruses-04-00363]\]; however the functional relevance of this is not known since the residue is on a portion of the protein reported to be in the cytoplasm (see discussion on topology below). Two studies have shown that SARS-CoV E can be ubiquitinated \[[@B41-viruses-04-00363],[@B42-viruses-04-00363]\]. There is currently no known functional role for the ubiquitination of CoV E. 5. CoV E and the Cell Stress Response ===================================== There are varying reports on the role of the E protein in apoptosis. It was shown that overexpression of MHV E and epitope tagged SARS-CoV E induced apoptosis in some cultured cell lines \[[@B43-viruses-04-00363],[@B44-viruses-04-00363]\]. However, a recent study compared the stress response of cells infected with SARS‑CoV to cells infected with SARS-CoVΔE. Using a microarray-based approach, it was shown that the virus lacking E induced a much more robust stress response than the wild-type virus. The virus lacking E also caused a greater degree of apoptosis compared to the wild-type virus \[[@B45-viruses-04-00363]\]. This result, while contrary to the studies using overexpression, shows that SARS-CoV E may be anti-apoptotic during infection. The suppression of the host stress response by E may be important for down regulating the immune response and promoting pathogenesis. 6. Protein-Protein Interactions =============================== While the interaction between E and M has been established (reviewed in \[[@B2-viruses-04-00363]\]), two recent studies have identified novel protein interactions with the SARS-CoV E protein. In the first study, cells were infected with SARS-CoV that had the E gene replaced with a C-terminally tagged version of E. Tandem affinity purification coupled with tandem mass spectrometry was carried out on the infected cells to find interacting proteins. Several candidates were identified, and the interaction of SARS-CoV E with the N-terminal ubiquitin-like domain-1 of SARS-CoV nsp3 protein was characterized. Nsp3 and CoV E colocalize during infection and nsp3 may be responsible for the ubiquitination of SARS‑CoV E \[[@B41-viruses-04-00363]\]. Another report identified the tight junction protein PALS1 as an interactor with the C-terminal domain of SARS-CoV E \[[@B46-viruses-04-00363]\]. The authors speculate that this interaction is important in the pathogenesis of SARS-CoV by promoting disassembly of tight junctions in lung epithelium after primary infection \[[@B46-viruses-04-00363]\]. Understanding the protein-protein interactions of CoV E is an important step in elucidating the functions of the protein. Since the sequences of the E proteins are quite divergent, it will be interesting to determine if different CoV E proteins interact with the same host proteins, or are able to elicit a similar effect through interactions with other proteins. 7. The Ion Channel Activity of CoV E ==================================== Many viruses encode small proteins that have ion channel activity including Hepatitis C virus (HCV) p7 \[[@B47-viruses-04-00363],[@B48-viruses-04-00363]\], influenza M2 \[[@B49-viruses-04-00363]\], and picornavirus 2B \[[@B50-viruses-04-00363]\]. All of these proteins are small (63--97 amino acids) and contain one or two transmembrane domains, making oligomerization a requirement for channel activity \[[@B51-viruses-04-00363]\]. The role of these proteins varies between viruses, but all of them affect the secretory pathway. The M2 protein of influenza virus is the best characterized ion channel. M2 is a 97 amino acid type III membrane protein that forms a tetrameric pH activated proton channel \[[@B49-viruses-04-00363]\]. The first described role for this activity was in the entry of the virus. Influenza virus is endocytosed after binding to a susceptible cell, and upon endosome acidification M2 facilitates the transfer of protons into the virion interior to aid in uncoating of the genome \[[@B52-viruses-04-00363]\]. It was later appreciated that the M2 ion channel is active in infected cells, where M2 has an effect on the secretory pathway by raising the pH of the *trans*-Golgi network \[[@B53-viruses-04-00363]\]. For some strains of influenza virus, this activity prevents the premature activation of the fusion protein \[[@B53-viruses-04-00363],[@B54-viruses-04-00363],[@B55-viruses-04-00363]\]. HCV p7 is not as well studied as M2, but is relevant to the discussion because like CoVs, HCV assembles intracellularly and must navigate the host secretory pathway. HCV p7 is targeted to the ER and Golgi complex \[[@B56-viruses-04-00363]\], where it is thought to oligomerize and act as an proton channel \[[@B48-viruses-04-00363],[@B57-viruses-04-00363]\]. p7 is important for the assembly and release of HCV virions \[[@B58-viruses-04-00363],[@B59-viruses-04-00363]\]. Recently, it was shown that p7 alkalinizes acidic organelles in the secretory pathway, and that M2 can complement a version of HCV lacking p7 \[[@B57-viruses-04-00363]\]. How this activity promotes release of particles is not known. It may create an environment that protects virions from damage during exocytosis, or promote enlargement of Golgi cisternae to accommodate large cargo (see discussion of CoV E and release below). The channel activity of E was first demonstrated for SARS-CoV E in planar lipid bilayers, where it was found that a synthetic peptide corresponding to the protein could permeabilize bilayers to Na^+^ and K^+^, with a 10-fold preference for Na^+^ \[[@B60-viruses-04-00363]\]. Further study generalized this property to other CoV groups, as IBV E, MHV E, and HCoV-229E E all have ion channel activity for monovalent cations \[[@B61-viruses-04-00363]\]. Furthermore, the channels formed by MHV E and HCoV 229E E were inhibited by the broad spectrum Na^+^/H^+^ exchanger inhibitor hexamethylene amiloride (HMA) \[[@B61-viruses-04-00363]\]. Since the E protein only has a single HD, formation of an ion channel would require oligomerization. This possibility was originally addressed using computational modeling, which predicted that the E protein HD could form stable dimers, trimers, and pentamers \[[@B62-viruses-04-00363]\]. This prediction was further bolstered by several spectroscopic studies, which showed that the HD of SARS-CoV E forms a pentamer \[[@B63-viruses-04-00363],[@B64-viruses-04-00363],[@B65-viruses-04-00363]\]. However, this result has not been confirmed for the other E proteins, or for the full length SARS-CoV E protein in a natural membrane. Nonetheless, predictions can be made from these structural data. For one, it appears that Asn15 in SARS-CoV E, and possibly the equivalent residue in other E proteins, is likely important either for oligomerization or ion conductance. Indeed, when alanine is substituted for this residue in a lysine-flanked peptide of the SARS-CoV E HD, conductance is largely inhibited \[[@B63-viruses-04-00363]\]. Additionally, the equivalent residue in IBV E is required for perturbing the secretory pathway, suggesting a role for the ion channel in this effect \[[@B66-viruses-04-00363]\]. The role of the CoV E ion channel in infection is not entirely clear, but studies using inhibitors offer some insight into its function. When HMA is added to the inoculum of either HCoV 229E or MHV, replication is inhibited. Adding more weight to this result, when HMA is added to the inoculum of MHVΔE no further growth defect is observed, showing that the inhibitory effect of HMA is dependent on the presence of the E protein \[[@B61-viruses-04-00363]\]. While these studies suggest an important role for the ion channel activity of the E protein, it is not clear what step of viral replication is blocked by the drug. Two other studies have linked the putative ion channel activity with virus replication and release. A recombinant version of MHV carrying mutations in the HD was shown to have a defect in both assembly and release of particles \[[@B67-viruses-04-00363]\]. A version of IBV carrying an a mutated version of E where the HD was replaced with that of an unrelated protein showed a defect in the release of infectious particles \[[@B31-viruses-04-00363]\]. These results imply that the ion channel activity of E may be important in the release of virions from cells, and are discussed in the section below. 8. Role of CoV E in Virion Release ================================== 8.1. The Mammalian Secretory Pathway and Large Cargo ---------------------------------------------------- Enveloped viruses that assemble intracellularly must navigate through the secretory pathway for release ([Figure 3](#viruses-04-00363-f003){ref-type="fig"}). Whether these viruses simply follow the canonical trafficking pathway, or modify it to handle the flux of large cargo is not known. To better understand this question, it is important to appreciate how the compartments used by CoVs are linked. The ERGIC is the intermediate compartment between the ER and Golgi. Its major function is to sort and concentrate cargo along the biosynthetic pathway. The ERGIC may mature as a unit to form the *cis*-Golgi, or it may be a more stable compartment \[[@B68-viruses-04-00363]\]. From the ERGIC, cargo traffics to the Golgi complex. The Golgi complex is made up of polarized stacks of cisternae that are connected laterally into a ribbon. Golgi cisternae are generally between 700--1100 nm in diameter and \<20 nm thick \[[@B69-viruses-04-00363],[@B70-viruses-04-00363]\]. The polarity of the Golgi from *cis*-, *medial*-, to *trans*-Golgi is important for sequential processing and proper sorting of cargo. The mechanism by which cargo proteins move through the Golgi complex is controversial. It was originally thought this occurred by the maturation of cisternae \[[@B71-viruses-04-00363]\]. This model was challenged when intra-Golgi transport was reconstituted *in vitro* \[[@B72-viruses-04-00363],[@B73-viruses-04-00363]\]. However, some large cargo proteins such as procollagen I are too big to fit into normal transport vesicles. It was shown that procollagen I does not leave the Golgi cisternae during its transport, and instead the cisternae mature, as Golgi resident proteins are moved in a retrograde direction \[[@B74-viruses-04-00363],[@B75-viruses-04-00363],[@B76-viruses-04-00363]\]. However, a different large cargo complex (reversible protein aggregates) is transported between the Golgi cisternae in "megavesicles" \[[@B77-viruses-04-00363]\]. Thus, it is possible that both vesicle transport and cisternal maturation can move cargo through the Golgi complex in mammalian cells. Since virions are large cargo they would likely move through the Golgi complex via a cisternal maturation mechanism or in megavesicles. However, differences between trafficking of small and large cargo are not well understood \[[@B78-viruses-04-00363]\]. Some insight into the trafficking of virions can be gained by examining how cells normally traffic very large cargo. Procollagen I forms 300 nm triple helical bundles in the ER, which move through the secretory pathway. As mentioned above, the procollagen filaments do not appear to move between the Golgi cisternae in vesicles, implying that they move through the stack via a cisternal maturation mechanism \[[@B74-viruses-04-00363]\]. When procollagen bundles are present in the Golgi complex, the cisternae appear distended \[[@B79-viruses-04-00363],[@B80-viruses-04-00363]\]. Additionally, the *trans*-Golgi network (TGN) expands in cells producing collagen, and this expansion is correlated with efficient release of procollagen \[[@B81-viruses-04-00363]\]. The trafficking of lipoproteins induces similar changes in Golgi complex morphology. Chylomicrons are large (400 nm) complexes made up of apoB-48 bound to triacylglycerides \[[@B82-viruses-04-00363]\]. These complexes are formed in the ER of intestinal epithelial cells and traffic through the Golgi complex during their biogenesis \[[@B83-viruses-04-00363]\]. It has long been known that the induction of chylomicron production demands a substantial, but reversible, enlargement of the Golgi complex, as seen in electron micrographs of rat enterocytes after the animals were fed a fatty meal \[[@B84-viruses-04-00363]\]. The changes induced in Golgi complex morphology for both procollagen and chylomicrons seem to underlie a necessary step in their trafficking. How the expansion of the Golgi complex aids in their secretion, and the mechanism used to induce the reversible change(s) are not known. However, it seems likely that enveloped viruses that assemble intracellularly may have evolved to exploit this mechanism. 8.2. Impact of CoV Infection on the Secretory Pathway ----------------------------------------------------- Like the trafficking of large cargo, CoV infection leads to a rearrangement of endomembranes ([Figure 3](#viruses-04-00363-f003){ref-type="fig"}). It was initially observed that during MHV infection the Golgi complex is dispersed from its juxtanuclear position \[[@B85-viruses-04-00363]\]. This result has recently been expanded upon using quantitative immune‑EM, where it was shown that several different types of membrane rearrangements occur during MHV infection. Most relevant to the discussion here, virions were seen in large carriers derived from Golgi/ERGIC membranes, suggesting that the rearrangement of the Golgi complex may be important for virion trafficking \[[@B86-viruses-04-00363]\]. However, the mechanism underlying the change in Golgi complex morphology is not currently known. These changes appear similar to those observed for the large cargo described above, but may constitute a more dramatic change. ![CoVs use the secretory pathway for egress. (**A**) A cartoon depicting virions within the Golgi complex. The Golgi cisternae are enlarged and fragmented during infection, possibly to aid in the trafficking of virions. (**B**) Transmission electron micrographs of uninfected or IBV infected Vero cells. (1) The Golgi complex of an uninfected cell; (2) The Golgi complex of an IBV infected cell with enlarged Golgi cisternae; (3) A putative virion carrier. Arrows denote virions. Scale bar is 500 nm. For images (2) and (3) Vero cells were infected at an moi of 0.1, and samples were fixed and processed 14 hrs post infection as described in \[[@B31-viruses-04-00363]\].](viruses-04-00363-g003){#viruses-04-00363-f003} Besides CoVs, several other enveloped viruses assemble intracellularly. If modifying the secretory pathway is a requirement for virion trafficking, one would predict that an alteration of secretory compartments would be a general feature of cells infected with these viruses. Indeed, many of them do modify the secretory pathway in striking ways. Flaviviruses assemble and bud into the ER \[[@B87-viruses-04-00363]\]. Interestingly, during flavivirus infection Golgi proteins localize to the ER at the site of assembly during infection, but the function of this is not clear \[[@B88-viruses-04-00363],[@B89-viruses-04-00363]\]. Bunyaviruses assemble by budding into the Golgi lumen \[[@B90-viruses-04-00363],[@B91-viruses-04-00363]\]. During bunyavirus infection, the ERGIC and Golgi complex appear fragmented and vesiculated \[[@B92-viruses-04-00363]\], but the underlying mechanism is not known. While it may not be surprising that viruses rearrange cellular membranes for their own use, it is puzzling that these viruses seemingly disrupt the very organelles that they depend on for egress. However, the apparent disruption may represent an exaggeration of the mechanisms used by the large cargo described above. It seems possible, or maybe even plausible, that these viruses are exploiting a pre-existing mechanism to facilitate the release of large cargo from the secretory pathway. 8.3. Effect of CoV E on the Secretory Pathway and Role in Release of Virions ---------------------------------------------------------------------------- The results discussed above suggest a connection between the morphological changes during CoV infection and virion trafficking. However, the relationship between the two has been elusive. Recent studies of CoV E have shed light on this issue, and it is now apparent that a link exists between the E protein, virion trafficking, and morphological changes in the Golgi complex. Several studies have shown that the E protein is important for virion release. The E protein of TGEV is essential for the propagation of the virus \[[@B39-viruses-04-00363],[@B40-viruses-04-00363]\]. However, further characterization of TGEVΔE showed that virions were assembled in infected cells, but appeared arrested in the secretory pathway, apparently unable to traffic properly \[[@B36-viruses-04-00363]\]. Analysis of a version of SARS-CoV lacking the E gene showed that virions accumulated intracellularly with aberrant material, suggesting that they may be sorted improperly in the absence of E \[[@B35-viruses-04-00363]\]. As mentioned above, mutations introduced into the HD of MHV or IBV compromised the release of infectious particles from cells \[[@B31-viruses-04-00363],[@B67-viruses-04-00363]\]. Additionally it was shown that expression of IBV E altered the secretory pathway in a manner dependent on a single residue within its HD, an effect that correlates with the release of particles \[[@B31-viruses-04-00363],[@B66-viruses-04-00363]\]. These results have led to speculation that the E protein acts as an ion channel in the secretory pathway, driving the rearrangement of secretory organelles through the alteration of lumenal environments. This in turn leads to efficient trafficking of virions. This process may be analogous to what occurs for the large protein cargo mentioned above and should be further examined. 9. Topology of CoV E ==================== The topology of the CoV E protein is debated in the literature. The E protein does not have a canonical cleaved signal sequence \[[@B6-viruses-04-00363]\], suggesting that the E protein is likely a type II or type III membrane protein. However, that is where the consensus ends. The C-terminus of TGEV E was detected at the surface of non-permeabilized infected cells, suggesting it adopted an N~cyto~C~exo~ topology ([Figure 4](#viruses-04-00363-f004){ref-type="fig"}(3)) \[[@B12-viruses-04-00363]\]. IBV E has been reported as a transmembrane protein with the opposite topology, N~exo~C~cyto~ ([Figure 4](#viruses-04-00363-f004){ref-type="fig"}(1)) \[[@B5-viruses-04-00363]\] The C-terminus of MHV E is present in the cytoplasm of infected cells, and on the inside of the virion \[[@B6-viruses-04-00363]\]. Later it was shown, using an N-terminally FLAG tagged version of MHV E, that the N-terminus is present in the cytoplasm \[[@B11-viruses-04-00363]\]. These results suggested an N~cyto~C~cyto~ topology, which is very strange for a protein with a single predicted HD, and suggests that the HD forms a membrane hairpin ([Figure 4](#viruses-04-00363-f004){ref-type="fig"}(2)). Multiple topologies have been reported for SARS-CoV E. Using transient expression of an N- or C-terminally FLAG tagged version of SARS-CoV E, both termini were present in the cytoplasm, supporting the membrane hairpin topology \[[@B10-viruses-04-00363]\]. Additionally, this study also found that at least a portion of the expressed protein has its C-terminus modified with N-linked oligosaccharides, which would require the C-terminus to be in the ER lumen as a type II membrane protein \[[@B10-viruses-04-00363]\]. More recently, an untagged version of SARS CoV E was shown to largely adopt an N~exo~C~cyto~ topology in infected cells and in cells transiently expressing the protein \[[@B9-viruses-04-00363]\]. These results, along with those reported in \[[@B66-viruses-04-00363]\] demonstrate the danger of using epitope tags to determine topology. Complicating matters, when prediction programs are used to model the topology of the E protein the results are not consistent with, and often are in direct opposition to, what has been observed experimentally ([Table 1](#viruses-04-00363-t001){ref-type="table"}). ![Topologies of the CoV E protein. The three proposed topologies of CoV E. (1) shows a type III membrane protein; (2) shows a membrane hairpin; and (3) shows a type II membrane protein with a putative N-linked oligosaccharide.](viruses-04-00363-g004){#viruses-04-00363-f004} viruses-04-00363-t001_Table 1 ###### TM Pred \[[@B93-viruses-04-00363]\], HMMTop \[[@B94-viruses-04-00363]\], TMHMM 2.0 \[[@B95-viruses-04-00363]\], MEMSAT3 \[[@B96-viruses-04-00363]\], and MEMSAT-SVM \[[@B97-viruses-04-00363]\] were used to predict the topology of four different CoV E proteins. The location of the N- and C-termini, as well as the number of predicted transmembrane passes are shown. Program IBV E MHV E SARS E TGEV E ------------ ------- ------- -------- -------- ------- --- ------- ------- --- ------- ------- --- TM Pred lumen lumen 2 lumen cyto 1 lumen cyto 1 lumen cyto 1 HMMTop lumen lumen 2 cyto cyto 2 lumen cyto 1 cyto cyto 2 TMHMM 2.0 lumen lumen 2 lumen cyto 1 cyto lumen 1 lumen cyto 1 MEMSAT-SVM lumen lumen 2 lumen lumen 2 lumen lumen 2 cyto lumen 1 MEMSAT3 cyto cyto 2 lumen cyto 1 lumen lumen 2 lumen cyto 1 It seems possible that the E protein may adopt multiple topologies during infection. Certainly the putative ion channel activity of E would require a transmembrane protein, but another function, such as membrane bending in assembly, could require a membrane hairpin. To address this possibility, versions of IBV E were developed that adopted either a transmembrane (by adding a canonical cleaved signal sequence) or a membrane hairpin (by putting a FLAG tag on the N-terminus). The results showed that only the transmembrane protein could alter the secretory pathway (possibly through ion channel activity), but was not as efficient at forming particles as judged by VLP production \[[@B66-viruses-04-00363]\]. This may mean that a small portion of wild-type IBV E protein exists as a membrane hairpin and plays a role assembly. If CoV E is inserted into the membrane both in a transmembrane and membrane hairpin orientation, one question that arises is how a single protein might adopt multiple topologies. Certainly a type II or type III membrane protein could be generated using the signal recognition particle (SRP)-translocon pathway. However, can a membrane hairpin topology be achieved using SRP-translocon mediated membrane insertion? Caveolin is between 18--24 kD depending on the isoform, and has a \~32 amino acid long HD which is inserted as a hairpin \[[@B98-viruses-04-00363]\]. The membrane insertion of caveolin is dependent on Sec61, suggesting that the ER translocon can generate proteins with a membrane hairpin topology \[[@B99-viruses-04-00363]\]. Thus, it is possible that SRP-mediated membrane insertion could produce multiple topologies, including a membrane hairpin and a transmembrane protein. Nonetheless, the E protein is very different than caveolin, both in overall size and in the hydrophobicity of its HD. It is important to consider alternate membrane insertion pathways. Tail-anchored proteins are generally considered to have fewer than 30 amino acids following their transmembrane domain \[[@B100-viruses-04-00363]\]. Many CoV E proteins are just outside of this range, but no studies have investigated the role of the GET pathway (guided entry of tail anchored proteins) in CoV E membrane insertion. Finally, the possibility exists that the E protein may be inserted into the membrane via a novel mechanism of post‑translational insertion \[[@B101-viruses-04-00363]\]. It is interesting to speculate that CoV E may be inserted into the membrane via multiple mechanisms, resulting in distinct topologies. One last point is that the topology of CoV E proteins may differ for different Co*Vs.* In addition to the variability shown in the prediction programs, there are also differences in the overall hydrophobicity and length of the transmembrane segments. How this impacts membrane topology is unknown, but it is worth noting that IBV E is the only E protein predicted to have two transmembrane domains. It is also possible that differences in the hydrophobic domains could affect the ion channel activity of the proteins. 10. Perspectives on CoV E ========================= The CoV E protein is an enigmatic protein. There is a high degree of variability in the behavior of the E proteins from different CoVs, including their requirement for assembly, virion trafficking, ion channel function, and method of expression in the genome. Yet, this protein is present in all known CoVs, suggesting it has a conserved role. It is interesting to speculate that the E protein from different CoVs has evolved to perform different functions. This could be due to cell-type specific requirements for each virus, or the ability of an accessory protein to complement a function normally carried out by E. Understanding how the E proteins from different CoVs vary is a crucial step in elucidating the mechanism of CoV assembly and egress. While the results reviewed here have better characterized the E protein, much work still remains to understand the mechanism of E protein function. The putative ion channel activity of E needs to be further examined. To this end, comprehensive electrophysiological analysis of the various E proteins in their natural setting (inserted into a Golgi membrane with normal post-translational modifications) needs to be carried out. It should also be noted that the E protein from different CoVs show variability in ion specificity, conductivity, and sensitivity to HMA. The reasons for this are not clear. Understanding how the ion channel activity of E affects replication is an important step in elucidating the function(s) of the protein with particular attention to cell types infected in the natural host. Recombinant viruses should be developed that encode other small ion channels along with E proteins that can support assembly but have no ion channel activity to determine if other channels can substitute for E. Additionally, analysis of the ion composition of the disrupted secretory pathway in IBV E expressing cells needs to be carried out to confirm that altered ion balance is responsible for the disruption. The idea that the E protein alters the secretory pathway to allow for virion trafficking is intriguing. If this modification is required for efficient trafficking of virions due to their large size, one would predict that the E protein should facilitate the trafficking of other large cargo such as procollagen I and chylomicrons. To this end, trafficking assays should be developed which allow for the quantitative monitoring of procollagen I and chylomicrons through the Golgi complex in the presence or absence of the E protein. It is also possible that the morphological changes in the Golgi complex are the end result of microenvironment alterations required to protect virions from proteolytic damage during egress. To test this, the lumenal ion concentration of both cells expressing E and infected cells should be investigated using ratiometric imaging. The topology of E appears to be predominately N~exo~C~cyto~; however, the existence of mixed membrane topologies cannot be ruled out. One possibility is that a small portion of the E protein is inserted as a membrane hairpin, and that this conformation facilitates particle assembly, but not efficient trafficking of virions. To address this possibility, recombinant viruses expressing topologically constrained versions of E (like those developed in \[[@B66-viruses-04-00363]\]) should be developed, and particle production as well as infectious particle release monitored. Additionally, *in vitro* membrane translocation assays should be carried out to determine the components necessary for membrane integration of the E protein (SRP mediated, GET mediated, or a novel route). Finally, because there is variability in the functions of E protein in different viruses, experiments to elucidate which functions are conserved and which are not should be performed. The ability of the different CoV E proteins to rescue the E deletions of other CoVs should be tested much the same way as they were for MHV \[[@B37-viruses-04-00363]\]. In addition, membrane associated accessory proteins from different CoVs should be examined in this same assay to determine if any of them can substitute for E. Finally, the tandem-affinity approach used to identify interacting proteins for SARS-CoV \[[@B41-viruses-04-00363]\] should be extended to other CoVs. It is now clear that the CoV E protein has multiple functions during infection, although they are not clearly understood. Also, the role(s) for the E protein in different CoVs may not be identical. A comprehensive model of E protein function will provide a better understanding of CoV biology and the mammalian secretory pathway, and may provide a target for therapeutic intervention during CoV infection. Acknowledgments =============== We would like to thank Andy Pekosz for thoughtful discussion on the E protein. The work in the authors' laboratory was funded by the NIH. TRR was supported by the Isaac Morris and Lucille Elizabeth Hay Graduate Fellowship. Conflict of Interest ==================== The authors declare no conflict of interest.

{ "pile_set_name": "PubMed Central" }

Introduction {#s1} ============ Unstructured Regions Define a New Heterogeneous Structural Reality {#s1a} ------------------------------------------------------------------ One central paradigm of structural biology is that the intricate details of 3-D protein structures determine protein function \[[@pcbi-0030140-b001],[@pcbi-0030140-b002]\]. In the last few years, many studies have shown that often the lack of a unique, native 3-D structure in physiological conditions can be crucial for function \[[@pcbi-0030140-b003]--[@pcbi-0030140-b021]\]. Such proteins are variously called *disordered*, *unfolded*, *natively unstructured*, or *intrinsically unstructured* proteins. A typical example is a protein that adopts a unique 3-D structure only upon binding to an interaction partner and thereby performs its biochemical function \[[@pcbi-0030140-b003]--[@pcbi-0030140-b006]\]. The better our experimental and computational means of identifying such proteins, the more we realize that they come in a great variety: some adopt regular secondary structure (helix or strand) upon binding, and some remain loopy. Some proteins are almost entirely unstructured, and others have only short unstructured regions. The more we can recognize short unstructured regions, the more we realize that the term "unstructured protein" would be misleading, as most unstructured proteins have relatively short unstructured regions. There is no single way to define unstructured regions. Here, we define an unstructured region as that which lacks unique 3-D structure by one of the following experimental techniques: circular dichroism (CD) spectroscopy, nuclear magnetic resonance (NMR) spectroscopy, X-ray crystallography, or proteolysis experiments \[[@pcbi-0030140-b007]--[@pcbi-0030140-b009]\]. Thanks to the outstanding data collection by the Dunker group, we could also describe this as regions that are the minimal common denominator between all proteins collected in DisProt \[[@pcbi-0030140-b010]\]. However, as we learned from prediction methods, DisProt and similar databases cover only a small fraction of all unstructured regions ([Figure 1](#pcbi-0030140-g001){ref-type="fig"}), and as we learned from recent experiments \[[@pcbi-0030140-b011]--[@pcbi-0030140-b013]\], there are many unstructured regions covered neither by these databases nor by existing prediction methods. ![Putative "Map" of Unstructured Regions\ Proteins with unstructured regions are likely to occupy large portions of sequence space \[[@pcbi-0030140-b007],[@pcbi-0030140-b024],[@pcbi-0030140-b027],[@pcbi-0030140-b042]\] as sketched by the light-gray inner rectangle. The space of all proteins with unstructured regions is likely to be considerably larger than what today\'s experimental techniques capture. The rounded darker gray rectangle labeled *experiment* sketches proteins for which some experimental method annotated natively unstructured regions. While most NORS regions (predicted long loops, striped gray ellipse) are likely to be natively unstructured, many unstructured regions are not NORS; i.e., they contain helices and strands even in their native form. Previous methods for the prediction of unstructured regions (left lens) are optimized to somehow reflect today\'s experiments. In contrast, the method introduced here (NORSnet, right lens) is developed based on predictions. This is an advantage because it avoids the bias of today\'s experimental techniques in a field that is just beginning to grasp its own dimensions, and it is a disadvantage because performance on today\'s datasets appears somehow limited.](pcbi.0030140.g001){#pcbi-0030140-g001} Unstructured Regions Can Be Defined and Recognized in Many Ways {#s1b} --------------------------------------------------------------- Methods that predict unstructured regions from sequence are mushrooming. Fast methods identify regions with high net charge and low hydrophobicity \[[@pcbi-0030140-b014],[@pcbi-0030140-b015]\], monitor the differences in amino acid propensities between unstructured and other regions (GlobPlot) \[[@pcbi-0030140-b016]\], or identify motifs associated with regions depleted of regular structure \[[@pcbi-0030140-b017],[@pcbi-0030140-b018]\]. Most methods are based on a different definition of disordered region that has been introduced by the Dunker group \[[@pcbi-0030140-b019]\]: residues for which X-ray structures do not have coordinates are considered as disordered. Methods based on this concept used neural networks \[[@pcbi-0030140-b019]--[@pcbi-0030140-b023]\] or support vector machines \[[@pcbi-0030140-b024]\]. The meetings for the Critical Assessment of Structure Prediction (CASP) have exclusively assessed disorder predictions on subsets of the "noncoordinate" data \[[@pcbi-0030140-b025],[@pcbi-0030140-b026]\]. The major drawback of this approach is that the Protein Data Bank (PDB) is biased toward proteins for which structures can be determined; natively unstructured proteins are underrepresented in the PDB \[[@pcbi-0030140-b005],[@pcbi-0030140-b010],[@pcbi-0030140-b024],[@pcbi-0030140-b027]\]. This may be one reason why most prediction methods tested by Oldfield et al. \[[@pcbi-0030140-b011],[@pcbi-0030140-b012]\] missed a substantial number of the proteins with unstructured regions identified in a large-scale NMR study spinoff from structural genomics. Other sequence features are predictive of disorder. For example, functionally flexible regions are identified from known structures through molecular dynamics simulation and can be generalized through machine learning. The Wiggle method provides predictions that overlap with unstructured regions even though it is focused on a different aspect of protein flexibility \[[@pcbi-0030140-b028]\]. Regions with No Regular Secondary Structure Provide Alternative {#s1c} --------------------------------------------------------------- Our group identified long regions with no regular secondary structure (NORS), which are stretches of 70 or more sequence-consecutive surface residues with few or no predicted helices and strands \[[@pcbi-0030140-b027]\]. NORS regions showed considerable overlap with proteins predicted to have long unstructured regions by various disorder predictors. NORS regions are overrepresented in eukaryotes (over five times more than in prokaryotes), overrepresented in regulatory and interacting proteins \[[@pcbi-0030140-b027],[@pcbi-0030140-b029]\], and share biophysical properties with unstructured regions. In addition, when natively unstructured regions are cocrystallized with their binding partner, they are still enriched in nonregular structure compared with globular proteins; ∼45% and ∼31% of the residues are in coils, respectively \[[@pcbi-0030140-b004]\]. Somewhat surprisingly, the method for predicting regular secondary structure in NORS regions, PROFsec (a profile-based neural network secondary structure predictor) \[[@pcbi-0030140-b030]--[@pcbi-0030140-b032]\], accurately predicts the secondary structure state in unstructured regions \[[@pcbi-0030140-b004]\]. NORS regions capture only one particular aspect of unstructured regions ([Figure 1](#pcbi-0030140-g001){ref-type="fig"}). The major advantages of our focus on NORS regions are that this definition implies a simple structural interpretation, and that we can reliably identify thousands of such regions by scanning entire organisms. The thresholds for the minimal length (70 residues) and for the definition of "largely loop" were optimized in order to minimize the identification of any of these stretches in the PDB \[[@pcbi-0030140-b027]\]. This procedure does not explicitly use any information about a protein other than its prediction of secondary structure and solvent accessibility. Thus, it mainly identifies extreme cases (e.g., highly exposed and long loop regions). Since many unstructured regions are shorter, one of our objectives was to capture much shorter NORS-like regions while ascertaining that we would not confuse long, well-structured loops with unstructured regions. One disadvantage of our focus on NORS was that some unstructured regions contain secondary structure elements (helix or strand) \[[@pcbi-0030140-b004]\]; i.e., not all unstructured regions are captured by NORS ([Figure 1](#pcbi-0030140-g001){ref-type="fig"}). Eukaryotic Disordered Regions Challenge Structural Genomics {#s1d} ----------------------------------------------------------- One goal of structural genomics is the determination of a 3-D structure representative for every protein family \[[@pcbi-0030140-b033],[@pcbi-0030140-b034]\]. Unstructured regions have not impeded structural genomics so far because almost all consortia have focused on bacterial proteins in order to increase the structure-to-clone ratio. However, consortia that focus on eukaryotes, such as the Northeast Structural Genomics (NESG) Consortium, or the Center for Eukaryotic Structural Genomics (CESG) have to carefully exclude such problematic targets \[[@pcbi-0030140-b035],[@pcbi-0030140-b036]\]. More than 10,000 proteins have been cloned and more than 3,000 proteins have been purified by NESG. Many of these did not adopt regular structure, possibly because they have unstructured regions that were not filtered out by our original filter, which discarded targets containing NORS regions \[[@pcbi-0030140-b029]\]. To speed up structure determination we need to increase the sensitivity in identifying unstructured regions \[[@pcbi-0030140-b011]\] (i.e., one goal of the development was to end up with a method that would be complementary to existing methods for the identification of unstructured regions). Our first hypothesis was that NORS regions share commonalities that distinguish such long unstructured loops from well-structured loops. If so, we should be able to somehow distinguish between the two types of loops at least in the sense that all loops predicted to be unstructured by our method ought to have different average features from other loops. We assumed that the neural network would pick up local correlations in amino-acid preferences for the different structural states. Our second hypothesis was that what distinguishes NORS regions from regular loops is exactly what makes regions become unstructured. If so, our method for the identification of NORS regions would also accurately predict unstructured regions. Here, we describe NORSnet, a new method that extends our NORS concept to also detect shorter (30--70 residues) NORS-like regions. The method was developed without ever using proteins with experimentally known unstructured regions. Instead, it was optimized to distinguish predicted NORS from all other regions. This unique approach, unprecedented in any machine learning method competing in a real-life application with other methods, has three important advantages. First, the data used for development and testing do not overlap. Since NORS regions were predicted from sequence, we can identify thousands of such regions. Our dataset was "dirty" in the sense that it contained many false negatives (all residues in PDB were considered to be well-structured during training) as well as some false positives (incorrect NORS predictions). This was the second major advantage: the positives (unstructured regions) sampled entirely sequenced organisms without any major bias with respect to this particular flavor of unstructured regions. Thereby, we identified unstructured regions that were missed by methods trained on more specialized datasets. The third advantage was that the resulting method explicitly focused on one feature of unstructured regions with a structural interpretation, namely that they are loops. Although we could have assessed NORSnet on any existing dataset due to the lack of overlap, we added a new set with experimental data about unstructured regions different from existing data. Note that both sets differed from each other as well as from the set used for development. Our three major results confirmed our hypothesis: (1) training on predictions succeeded in developing a powerful prediction method; (2) long loops are a major component of what is picked up by existing methods predicting unstructured regions; and (3) well-ordered and unstructured loops differ. In conjunction with existing methods, the one that we introduce here will allow the focus on particular structural aspects. Results/Discussion {#s2} ================== Accurate Distinction between Unstructured and Regular Loop Regions {#s2a} ------------------------------------------------------------------ We trained our system on NORS regions that had been predicted by our previous high-accuracy/low-coverage method \[[@pcbi-0030140-b027],[@pcbi-0030140-b029]\] for the identification of very long regions depleted of predicted helices and strands (NORSp; see [Methods](#s3){ref-type="sec"}). Technically, the task was to separate between all residues predicted to be in a NORS region and all residues in the PDB. As we used neural networks for this task, the typical assessment of accuracy usually involves a cross-validation experiment. For the first time in our work, we did not do this. In fact, we completely ignored the performance of the network on the task it optimized. Our hypothesis simply was that the only aspect that consistently separates extreme NORS regions from all residues in the PDB are the building blocks for a particular type of unstructured regions, namely the NORS-like loopy ones. Therefore, we measured performance on rather different datasets and separation tasks. First, we established success by predicting *well-structured loops* and *NORS-like loops* for DisProt, which consists of proteins with experimentally characterized unstructured regions. A total of 88% of the residues predicted by NORSnet were also predicted to be loops by PROFsec, while only 51% of the residues predicted as loops in DisProt also appeared NORS-like. In other words, most regions identified by NORSnet appeared to be in loops. Conversely, many loops were not predicted by NORSnet. Since residues in loops were identified through prediction, this difference may have been caused by prediction mistakes. To rule this out, we collected a set of 45 sequence-unique proteins that had been added to the PDB after we had completed developing our method (September 2005 to June 2006). We found that NORSnet classified only 1% of loop residues (Dictionary of Secondary Structure of Proteins states T, S, L) \[[@pcbi-0030140-b037]\] as natively unstructured regions. In other words, NORSnet largely succeeded for these new proteins. In fact, it predicted only one region in these structures to be unstructured, namely a stretch in the HIV type 1 P6 protein of 52 residues \[[@pcbi-0030140-b038]\], the NMR structure of which indicated depletion of regular secondary structure. This protein has been shown to undergo conformational changes \[[@pcbi-0030140-b038]\], suggesting that our method correctly identified it as unstructured. Very long NORS regions differ statistically from regularly structured or well-ordered loops \[[@pcbi-0030140-b027]\]. In general, unstructured regions that are not NORS-like tend to be more loopy than well-structured regions \[[@pcbi-0030140-b004]\]. Here, we showed that our ability to distinguish between well-ordered and unstructured loops was also successful for much shorter loops. Medium-length (30--70 residues) unstructured loops differed from well-structured loops ([Figure 2](#pcbi-0030140-g002){ref-type="fig"}). ![Regular, Flexible, and Predicted-To-Be Unstructured Loops Differed\ We compared the amino acid compositions between four different subsets representing four types of "loops" (nonhelix/nonstrand): loops from regular, well-ordered structures; i.e., from proteins without natively unstructured regions (states T, S, L from the Dictionary of Secondary Structure of Proteins; in blue); unstructured loops as predicted by NORSnet (in green); "flexible loops" from regular structures (TSL states with normalized B-factors ≥1 \[[@pcbi-0030140-b082]\]; in red); and unstructured regions as predicted by DISOPRED2 (in orange). The sign of the bar corresponds to overrepresentation (positive) or underrepresentation (negative) of amino acids in a subset with respect to the PDB. The NORS and DISOPRED2 residue subsets were taken from the worm genome (from the IntAct database \[[@pcbi-0030140-b067]\]) and were predicted to be unstructured by NORSnet and DISOPRED2. Flexible loops were enriched in amino acids with net charges such as lysine and glutamate (as described before \[[@pcbi-0030140-b016],[@pcbi-0030140-b039]\]). Predicted unstructured regions by NORSnet, however, differed in their composition from regular loops, flexible loops, and from any type of disorder that has been described previously (unpublished data) \[[@pcbi-0030140-b039],[@pcbi-0030140-b044]\]. Cysteines were not overabundant in the unstructured regions predicted by DISOPRED2. Overall, these data suggested that NORSnet captured something other than just "loop" and other than what is captured by methods such as DISOPRED2.](pcbi.0030140.g002){#pcbi-0030140-g002} NORSnet precisely distinguished between unstructured and well-structured loops. Although the amino acid composition of unstructured loops was similar to that in long disordered regions \[[@pcbi-0030140-b039]\], it was unique ([Figure 2](#pcbi-0030140-g002){ref-type="fig"}). For instance, the regions identified by our method contained significantly more cysteines than other PDB proteins and, within these, more than the set of residues unresolved in electron density maps. Thus, methods trained on unresolved residues, such as DISOPRED2, are likely to miss these regions. Furthermore, methods using pairwise energy potentials, such as IUPred, to predict unstructured regions are also likely to miss these regions, as many cysteines typically coincide with many paired cysteine bonds that significantly contribute to protein stability \[[@pcbi-0030140-b040],[@pcbi-0030140-b041]\]. Proteins with Unstructured Regions Accurately Identified {#s2b} -------------------------------------------------------- About 30%--60% of all eukaryotic proteins have been estimated to contain unstructured regions \[[@pcbi-0030140-b024],[@pcbi-0030140-b042]\]. However, DisProt \[[@pcbi-0030140-b010]\], the largest resource of experimentally verified unstructured regions, contains only a few hundred eukaryotic proteins, and thus covers a small fraction of sequence space ([Figure 1](#pcbi-0030140-g001){ref-type="fig"}). Moreover, this small fraction is not representative, as many unstructured regions described experimentally are missing from existing databases and are not identified by prediction methods \[[@pcbi-0030140-b011]\]. NORSnet attempted to solve both problems by sampling sequence space exhaustively (trained on all positives from entirely sequence organisms) and focusing on unstructured loops. To assess the accuracy of NORSnet and to estimate to what extent unstructured loops dominate our current identification of unstructured regions, we investigated two different datasets. The first was built around the DisProt database used previously in the literature; the second originated from careful NMR measurements and has not been used in many previous analyses. ### DisProt dataset. {#s2b1} The first set included proteins with unstructured regions from DisProt as positives and 173 PDB structures from EVA (a server for assessing protein structure prediction servers) as negatives (see [Methods](#s3){ref-type="sec"}). NORSnet correctly identified half of the DisProt proteins without false positives ([Figure 3](#pcbi-0030140-g003){ref-type="fig"}A). DISOPRED2 \[[@pcbi-0030140-b024]\] was ranked as one of the best three methods for predicting residues that are missing in electron density maps from X-ray crystallography at CASP6 \[[@pcbi-0030140-b026]\] and CASP7 (L. Bordoli, unpublished data). Many other studies corroborated the leading role of DISOPRED2 \[[@pcbi-0030140-b022],[@pcbi-0030140-b026],[@pcbi-0030140-b041],[@pcbi-0030140-b043],[@pcbi-0030140-b044]\]. Overall, NORSnet performed almost on a par with DISOPRED2 for the DisProt dataset ([Figure 3](#pcbi-0030140-g003){ref-type="fig"}A). Simply taking the average over the outputs of DISOPRED2 and NORSnet (DISOPRED2 + NORSnet) outperformed both individually. The improvement was particularly important for the realm of very high accuracy ([Figure 3](#pcbi-0030140-g003){ref-type="fig"}A). IUPred predicts unstructured regions based on a statistical potential optimized for this purpose \[[@pcbi-0030140-b041],[@pcbi-0030140-b045]\]. In our hands, IUPred clearly and consistently outperformed the other methods tested, including the averaged DISOPRED2/NORSnet output ([Figure 3](#pcbi-0030140-g003){ref-type="fig"}A). IUPred is optimized to identify all unstructured regions in DisProt \[[@pcbi-0030140-b041],[@pcbi-0030140-b045]\], but it cannot distinguish between unstructured regions dominated by loops and those dominated by regular secondary structure (as are often found in unstructured regions \[[@pcbi-0030140-b004]\]). ![Predictions for DisProt\ (A) ROC-like curve for NORSnet (green), DISOPRED2 (orange), and their combination (through arithmetic average; gray). While the performance of NORSnet and DISOPRED2 were similar, the combined method seemed to outperform both methods. Particularly, at accuracy = 100% (inset), the combined method covers significantly more sequences than each one of the methods individually. IUPred (purple) outperformed all other methods on this dataset. Note that IUPred was optimized on a set similar to the one used in this study. In contrast, NORSnet and DISOPRED2 were optimized on different sets defining disorder differently.\ (B) Venn diagram of overlap between very accurate predictions by NORSnet, DISOPRED2, and the combined method. The numbers in the circles are mutually exclusive; for instance, two proteins were identified only by DISOPRED2 to have an unstructured region, and 17 proteins were identified by both NORSnet and by the combined method to have an unstructured region.](pcbi.0030140.g003){#pcbi-0030140-g003} NORSnet predictions were not superior to those from DISOPRED2. However, the performance of these two methods was surprisingly similar despite the fact that NORSnet was not trained on a single experimentally verified unstructured region. Did the similarity in performance indicate that both methods picked up the same signal, i.e., that DISOPRED2 largely captured unstructured loops? If two prediction methods are based on very different information, their combination typically improves performance over any one of them \[[@pcbi-0030140-b044],[@pcbi-0030140-b046]\]. A more explicit way to demonstrate that methods focus on different aspects is the analysis of their predictions by Venn diagrams. We picked points for which each of the three methods (DISOPRED2, NORSnet, and DISOPRED2 + NORSnet) yielded 100% accuracy and compared the true positives predicted at those thresholds. DISOPRED2 and NORSnet identified the same 73 proteins, but each correctly identified proteins that the other missed ([Figure 3](#pcbi-0030140-g003){ref-type="fig"}B). This agreement supported our initial hypothesis that many unstructured regions are loopy (considerable overlap in true positives). But the most important result was that the two methods complemented each other. At the same 100% accuracy threshold, the combined method (DISOPRED2 + NORSnet) identified more proteins than any of the two individual methods and missed only two proteins that DISOPRED2 correctly identified. Although not surprising given the differences in training set and underlying optimizations, this result highlighted the difference in the types of unstructured regions identified. The combination of DISOPRED2 and NORSnet by averaging their outputs was better than either method alone. This did not work with IUPred and either of the two methods. This might suggest that IUPred covers the same aspects as the other two. However, this notion proved to be incorrect: IUPred missed proteins in the NESG dataset that the others captured ([Figure 4](#pcbi-0030140-g004){ref-type="fig"}B). Therefore, a beneficial combination of different methods predicting unstructured regions may require a more sophisticated algorithm. ![Predictions for NESG Data\ (A) The NESG set contains many proteins with unstructured regions that are not in DisProt and have never been used for method optimization. We compared NORSnet (in green), DISOPRED2 (in orange), their combined method (in gray), and IUPred (in purple) on these proteins. While DISOPRED2 performed better than all other methods in the low accuracy/high coverage region (top left), the combined method, NORSnet, and IUPred individually excelled in the high accuracy/low coverage region (lower right).\ (B) Venn diagram of overlap between very accurate predictions by NORSnet, DISOPRED2, and IUPred. The numbers in the circles are mutually exclusive. Note that five proteins were identified only by NORSnet to have an unstructured region.](pcbi.0030140.g004){#pcbi-0030140-g004} ### Unstructured regions from the NESG dataset. {#s2b2} Many prediction methods were optimized or benchmarked on datasets overlapping with DisProt. In contrast, the dataset from the NESG contained proteins with unstructured regions that have not been used for training existing methods yet. The NESG set was collected with a unified definition of unstructured regions based on 2-D NMR experiments \[[@pcbi-0030140-b047]\]; it included 30 proteins with unstructured regions as positives and 170 regular structures solved by NESG as negatives (Methods and [Table S1](#pcbi-0030140-st001){ref-type="supplementary-material"}). In the high accuracy region, NORSnet captured a considerable fraction of the positives (40% coverage at 100% accuracy; [Figure 4](#pcbi-0030140-g004){ref-type="fig"}A). The performance of DISOPRED2 was clearly lower than that of NORSnet for high accuracy/low coverage ([Figure 4](#pcbi-0030140-g004){ref-type="fig"}A, lower right), while the inverse was true for low accuracy/high coverage ([Figure 4](#pcbi-0030140-g004){ref-type="fig"}A, upper left). False positives from NORSnet (unstructured regions predicted and not observed) were almost equally divided between X-ray and NMR structures, while DISOPRED2′s false positives were predominantly from NMR structures. The most extreme examples for this were the ordered structures of Methanobacterium thermoautotrophicum 1615 \[[@pcbi-0030140-b048]\] and the conserved domain common to the transcription factors TFIIS, elongin A, and CRSP70 \[[@pcbi-0030140-b049]\]. ### Case study: NORSnet differed from other predictions. {#s2b3} As demonstrated above, NORSnet and other predictors give similar predictions with some exceptions. For instance, we applied NORSnet and two other prediction tools (DISOPRED2 and FoldIndex) on the Kappa-casein precursor protein that is found in milk and stabilizes micelle formation by preventing casein precipitation. Raman optical activity and thermal stability experiments revealed the protein as entirely unstructured in isolation \[[@pcbi-0030140-b050]\]. Secondary structure prediction methods such as PROFsec or PSIPRED \[[@pcbi-0030140-b051]\] predicted the protein to be highly enriched in loops ([Figure S1](#pcbi-0030140-sg001){ref-type="supplementary-material"}). We may therefore expect that the prediction of the Kappa-casein precursor as unstructured will be a simple task. However, the distinction between natively unstructured and well-structured loops is not trivial: DISOPRED2 did not identify the long *loopy* segment to be part of a natively unstructured region ([Figure 5](#pcbi-0030140-g005){ref-type="fig"}A). In contrast, NORSnet identified most of this protein to be unstructured in its strictest cutoff (corresponding to 100% accuracy on the DisProt dataset; [Figure 5](#pcbi-0030140-g005){ref-type="fig"}B). FoldIndex, a method that uses only amino acid composition and calculates the hydrophobicity/net charge within a given window, predicted only short segments of this protein to be unstructured ([Figure 5](#pcbi-0030140-g005){ref-type="fig"}C). ![Different Prediction Method Outputs for Kappa-Casein Precursor\ Kappa-casein precursor has been shown to be unstructured by different experiments \[[@pcbi-0030140-b050]\]. Despite its low content in predicted helices and strands, not all prediction methods identify it as unstructured. We compared outputs of DISOPRED2 (A), NORSnet (B), and FoldIndex (C) for this protein. For DISOPRED2 and NORSnet, higher values indicate unstructured regions; for FoldIndex, low values indicate unstructured regions (red). Note that FoldIndex and DISOPRED2 do not use any explicit information about secondary structure. DISOPRED2 disorder probability, however, is somewhat correlated with coil predictions ([Figure S1](#pcbi-0030140-sg001){ref-type="supplementary-material"}). DISOPRED2 was not able to distinguish these loops from structured loops. Only NORSnet clearly picked up the strong signal for unstructured regions for most of the protein.](pcbi.0030140.g005){#pcbi-0030140-g005} This example reveals that NORSnet and DISOPRED2 outputs are rather correlated. However, the signal from NORSnet clearly indicated unstructured regions, while the one from DISOPRED2 did not. One reason for this drastic difference may have been that NORSnet correctly captured some global feature from its global input units (see [Methods](#s3){ref-type="sec"}). ### Natively unstructured loops are elements of domain boundaries. {#s2b4} Although NORSnet was designed to identify all regions in any PDB structure as well-structured, the editor of this manuscript, Phil Bourne, suspected that NORSnet predictions of disorder might more often be in domain boundaries than expected at random and than expected for loop residues in general. To address this, we started with a sequence-unique subset of all PDB proteins considered to be multidomain by SCOP \[[@pcbi-0030140-b052]\] (set taken from \[[@pcbi-0030140-b053]\]). Although a much more comprehensive answer will remain the subject for future investigation, we clearly confirmed this assumption ([Figure S4](#pcbi-0030140-sg004){ref-type="supplementary-material"}); i.e., the regions in otherwise well-structured proteins that most resemble unstructured regions are domain linkers. ### Case study: DFF correctly identified despite being a tough case. {#s2b5} The DNA fragmentation factor (DFF) 45 must bind to DFF40 so that DFF40 can execute its catalytic function required for the onset of caspase-mediated apoptosis \[[@pcbi-0030140-b054]\]. The N-terminal domain (NTD) of DFF45 is natively unstructured: its folding is induced upon binding to DFF40 NTD \[[@pcbi-0030140-b055]\] ([Figure 6](#pcbi-0030140-g006){ref-type="fig"}). Methods that only use amino acid composition to predict unstructured regions are likely to perform better on such proteins than more complex prediction methods, since these proteins often have a high net charge which is neutralized upon binding to the target. For example, FoldIndex \[[@pcbi-0030140-b015]\] identified about a third of DFF45 as unstructured. ![NORSnet Captured Unstructured Regions Related to High Net Charge/Low Hydrophobicity\ DFF45 (white, yellow, and red) becomes structured upon complex formation with DFF40 (purple; \[[@pcbi-0030140-b055]\]). The interface includes a buried hydrophobic patch surrounded by hydrophilic interactions. Usually, charged residues disrupt the formation of tertiary structure; in this case, however, when the complex is formed, the negative charge of the Asp groups in DFF45 is cancelled out, with the positive charges of DFF40 allowing the protein to be folded. Visualization was done using GRASP2 \[[@pcbi-0030140-b085]\]. Since DFF45 has high secondary structure content, it is a relatively hard target for NORSnet prediction. However, NORSnet correctly identified its unstructured region at a rather stringent cutoff.](pcbi.0030140.g006){#pcbi-0030140-g006} Secondary structure-prediction methods, such as PSIPRED and PROFsec, usually predict the secondary structure of these regions the way they appear in substrate-bound form. Therefore, methods that use this type of information might be fooled by the rigidity and stability that are associated with regular secondary structure segments and identify these regions as well-structured. Since NORSnet uses secondary structure predictions as input, it may mispredict unstructured regions that become helices and strands upon binding. However, despite the fact that DFF45 NTD is enriched in regular secondary structure ([Figure S2](#pcbi-0030140-sg002){ref-type="supplementary-material"}), NORSnet identified NTD as an unstructured region at a rather stringent cutoff (the cutoff corresponded to 100% and 97.2% accuracy in the NESG and the DisProt sets, respectively). DISOPRED2 also identified NTD as unstructured, albeit at a less-stringent cutoff (corresponding to 72.2% and 94.2% accuracy). The unstructured regions in DFF45 are correctly identified by many prediction methods. NORSnet, DISOPRED2, and FoldIndex are only three of those. This example was one of 24 proteins with unstructured regions that become structured upon binding and were extensively analyzed in a recent study \[[@pcbi-0030140-b004]\]. NORSnet identified 14 of these proteins to have unstructured regions in its strictest cutoff. Again, this underlines the surprising finding that methods based on loop predictions can capture unstructured regions of this type. DFF45 and similar proteins are just some of many examples for unstructured regions involved in protein--protein interactions. How representative are they? Predicted Unstructured Regions Are Abundant in Protein--Protein Network Hubs {#s2c} ---------------------------------------------------------------------------- The structural plasticity of proteins with unstructured regions may enable its binding to many proteins, i.e., may typify a protein--protein interaction hub (a protein with many binding partners in an interaction network) \[[@pcbi-0030140-b006],[@pcbi-0030140-b056]--[@pcbi-0030140-b059]\]. Several detailed studies have specifically identified unstructured regions in hub proteins that are involved in signaling \[[@pcbi-0030140-b003],[@pcbi-0030140-b005],[@pcbi-0030140-b006],[@pcbi-0030140-b060]--[@pcbi-0030140-b062]\]. Natively unstructured regions are also predicted to be abundant in other regulatory processes (e.g., alternative splicing \[[@pcbi-0030140-b063]\] and transcription \[[@pcbi-0030140-b064]\]) and in cancer-associated signaling proteins \[[@pcbi-0030140-b065]\]. We addressed this point by correlating sustained large-scale datasets of physical protein--protein interactions (see [Methods](#s3){ref-type="sec"}) with predictions for unstructured regions. We applied NORSnet, DISOPRED2, and IUPred to all proteins in the worm *(Caenorhabditis elegans)* proteome and considered only predictions at thresholds corresponding to 100% accuracy. The subset of interacting proteins resulted from the high-throughput experiment by Vidal et al. \[[@pcbi-0030140-b066]\] and from IntAct \[[@pcbi-0030140-b067]\]. Predictions for unstructured regions for all three methods correlated with the average number of interacting partners; in other words, proteins with more unstructured regions had more binding partners ([Figure 7](#pcbi-0030140-g007){ref-type="fig"}). Since we used two different datasets to determine the thresholds for what constituted reliable predictions (DisProt and NESG), we also obtained two different thresholds for each method. For the purpose of fishing for hubs in protein--protein networks, we counted the number of proteins with unstructured regions according to any of those thresholds. Using DisProt to tune thresholds, DISOPRED2 predicted more proteins with unstructured regions than did NORSnet (1279 ± 88 versus 899 ± 76); using the NESG dataset, NORSnet predicted many times more proteins with unstructured regions than did DISOPRED2 (1282 ± 87 versus 321 ± 46; [Figure 7](#pcbi-0030140-g007){ref-type="fig"}). These results agreed with recent studies that estimated hub proteins to be enriched in unstructured regions \[[@pcbi-0030140-b057]--[@pcbi-0030140-b059]\]. However, could NORSnet identify any new unstructured regions in hub proteins? ![Unstructured Regions Overrepresented in Protein--Protein Hubs of the Worm\ We ran both NORSnet and DISOPRED2 on worm proteins that are involved in protein--protein interactions (as identified by yeast two-hybrid \[[@pcbi-0030140-b066]\]). The number of proteins that are predicted to be either unstructured or well-structured is plotted against the number of interacting partners for two different thresholds of reliability of the two methods: (A) and (B) were compiled for thresholds at which both methods maintained 100% accuracy for the NESG data ([Figure 4](#pcbi-0030140-g004){ref-type="fig"}), while (C) and (D) were compiled for 100% accuracy on DisProt ([Figure 3](#pcbi-0030140-g003){ref-type="fig"}). Since the number of observed interaction partners falls off dramatically, we had to group the data into bins of roughly equal sizes (*x*-axes). (A) and (C) show the results for the number of proteins predicted in each bin of interaction partners, while (B) and (D) show the normalized ratios to zoom into the difference between unstructured and structured proteins in each bin. These ratios were compiled as Ratio(bin) = {\#unstructured(bin) / \#structured(bin)} / {\#unstructured(1) / \#structured(1)}. As all ratios are greater than 1, proteins with more than one interaction partner have more unstructured regions than proteins with one partner.\ (A) These graphs were compiled with the reliability threshold at which each method achieved 100% accuracy by the NESG data ([Figure 4](#pcbi-0030140-g004){ref-type="fig"}). Overall, this threshold resulted in NORSnet (filled bars) predicting many more proteins with unstructured regions than DISOPRED2 (hatched bars). The difference was particularly relevant for proteins with more interacting partners.\ (B) NORSnet (filled, dark green) predicted many more unstructured regions in proteins with seven or more interaction partners than did DISOPRED2 (hatched, light green).\ (C) For the thresholds at which both methods achieved 100% accuracy on the DisProt dataset, DISOPRED2 identified more proteins with unstructured regions than did NORSnet. In contrast to the situation for the NESG set (A), the difference was not as significant for promiscuous proteins (ten or more partners).\ (D) Although NORSnet (filled, dark green) predicted as many unstructured as structured regions in hubs (seven or more), this ratio was significantly smaller than the one for proteins with a single interaction partner. In other words, even on this dataset NORSnet picked up a much stronger overrepresentation of unstructured regions in hubs than did DISOPRED2 (hatched, light green).](pcbi.0030140.g007){#pcbi-0030140-g007} We chose the cutoff that yielded the highest number of unstructured regions (NORSnet, 1,279; DISOPRED2, 1,282) for each method and checked whether the two methods predicted unstructured regions in the same hub proteins. Both methods predicted unstructured regions in most (74) of the proteins observed with more than ten partners (140). DISOPRED2 predicted unstructured regions in another 13 of the promiscuous proteins, and NORSnet in another 21 proteins. If the reliable predictions of both methods are correct, 77% of all promiscuous proteins in the worm (74 + 13 + 21 = 108 of 140) have unstructured regions. While these data do not suffice to identify hubs from sequence, we undoubtedly showed that methods such as NORSnet and DISOPRED2 clearly have some capability in the identification of unstructured regions that will adopt 3-D structures upon binding. While this finding was not new, our particular perspective was that the differences between DISOPRED2 and NORSnet resulted from the difference in the focus of the two. NORSnet focuses more on loopy regions than DISOPRED2, and it also identified more hub proteins. Similar results were obtained when we compared NORSnet and IUPred predictions on the same dataset. Again, IUPred identified the hub signal but much less clearly than did NORSnet ([Figure S3](#pcbi-0030140-sg003){ref-type="supplementary-material"}). All these observations suggested that the aspect of unstructured regions most relevant to hubs might actually be the unstructured loops. While NORSnet has some ability to identify unstructured regions that are often involved in binding ([Figure 6](#pcbi-0030140-g006){ref-type="fig"}), it may miss many of these regions due to their enrichment in regular secondary structure (helix, strand) in their bound form. We may therefore wonder why NORSnet identified so many worm hub proteins to have unstructured regions in the first place. Interestingly, many of the hubs had several modules/domains, some of which were predicted not to contain unstructured regions. Some of these modules were DNA-binding domains (such as Homeobox domains) or protein--protein interaction binding motifs (such as EGF repeats). The majority of the unstructured regions predicted by NORSnet in these hubs bridged connections between well-structured domains: these bridges were often on the surface (unpublished data). At first glance, the fact that these regions were predicted to be unstructured might seem biologically unimportant. However, there are several possible biological consequences of the abundance of hubs with unstructured loops. These exposed unstructured/loopy regions might serve as sites for proteolysis, allowing some parts of the protein to undergo proteolytic degradation under different cellular conditions. Such differential degradation could allow different modules of the same protein to be functional under different conditions. Alternatively, these long connecting loops might function as extremely flexible connecting linkers that facilitate the modules to adopt different orientations, thereby allowing the binding of different targets or binding similar targets in different fashion. Each of these alternatives could be at the heart of a different function. These two hypotheses may explain some of the regulatory characteristics of hub proteins. ### Mapping the sequence space of proteins with unstructured regions. {#s2c1} Most likely, unstructured regions and NORS regions occupy slightly different parts in sequence space ([Figure 1](#pcbi-0030140-g001){ref-type="fig"}). Indications for the overlap between NORS and unstructured regions are that both are enriched in proline and both depleted of glycine (\[[@pcbi-0030140-b039]\] and [Figure 2](#pcbi-0030140-g002){ref-type="fig"}). Some experimentally observed unstructured regions have been shown to contain cysteines. For instance, Zinc fingers often become structured only upon binding zinc. Nevertheless, most previous studies of unstructured regions did not find cysteines to be overrepresented with respect to well-structured regions in the PDB. This may be due partially to the fact that in well-structured proteins cysteines often stabilize disulfide bonds. Methods optimized to identify regions missing in electron density maps from X-ray crystallography are therefore likely to miss many of the cysteines in unstructured regions. In contrast, NORSnet captured cysteines in unstructured regions ([Figure 2](#pcbi-0030140-g002){ref-type="fig"}). In addition, the differences between DISOPRED2 and NORSnet that were revealed both by our head-to-head comparison on different sets of proteins with unstructured regions and by our analysis of protein hubs pointed to the different types of unstructured regions that we may have to separate ([Figure 1](#pcbi-0030140-g001){ref-type="fig"}). To complicate matters further, some proteins with unstructured regions may look just like any regular protein, while others may be generically different. Consequently, some of the proteins with unstructured regions may be missed by any prediction method. ### Refining target selection for structural genomics. {#s2c2} One goal of structural genomics projects is to contribute considerably to the increase in the fraction of proteins with known 3-D structures. To achieve this goal, 3-D structures are experimentally determined for representatives of as many large families as possible \[[@pcbi-0030140-b033],[@pcbi-0030140-b034],[@pcbi-0030140-b053],[@pcbi-0030140-b068]--[@pcbi-0030140-b070]\]. In particular, the large structural genomics initiatives financed by the Protein Structure Initiative (PSI) from the US National Institutes of Health (NIH) systematically target the experimental determination of structures for large families without representatives of known structure. Structural genomics also aims at making 3-D structures more readily accessible to nonstructural molecular biology and at reducing the costs and difficulty of determining structures. All of these goals require high-throughput determination of 3-D structures. This implies that experimental high-throughput pipelines have to move on if structure determination fails for some families, and that targets are also chosen with the objective to increase the throughput. This does not imply that PSI consortia "go for the low-hanging fruits." Quite to the contrary, they have succeeded where many small-scale studies have failed. Membrane proteins and proteins with unstructured regions are the two major types of proteins that are not only avoided by conventional small-scale structural biology but also by structural genomics efforts. Due to the fact that proteins with unstructured regions are much more abundant in eukaryotic organisms, consortia that focus on eukaryotes, such as NESG and CESG, have to carefully avoid such difficult targets. In the last six years, thousands of proteins have been cloned, expressed, and purified by NESG. Although the NESG target selection filtered out many domains with strong predictions for the presence of unstructured regions \[[@pcbi-0030140-b035],[@pcbi-0030140-b036]\], many were left for which biophysical data suggested that they contain unstructured regions \[[@pcbi-0030140-b013]\]. We applied NORSnet to 11,587 putative NESG targets that had already passed our previous and cruder NORS filter ([Table 1](#pcbi-0030140-t001){ref-type="table"}). Using two different cutoffs, NORSnet predicted that 13%--20% of the previously filtered targets have unstructured regions. Although NORSnet was not optimized to identify very short unstructured regions (≤30 residues), NORSnet predicted 47%--58% of the proteins to contain such regions. The same filter would not have excluded any of the proteins that succeeded in the experimental pipeline, suggesting that the application of NORSnet could have increased the structure--clone ratio. However, the ultimate proof for this assumption will have to wait until another hundred or so experimentally determined structures are added by NESG to the PDB in the next year(s). ###### NORSnet Predictions for Structural Genomics Targets ![](pcbi.0030140.t001) Conclusions {#s2d} ----------- The intricate details of protein 3-D structures are crucial for their functional role; i.e., structure determines function. Natively unstructured regions do not necessarily contradict this structure--function paradigm. Nevertheless, a variety of proteins require unstructured regions in order to function as domain linkers, filling material, and detergents. For other proteins with unstructured regions, changes in the environment (e.g., pH change, presence of target) or posttranslational modifications can trigger the formation of a regular 3-D structure that will then again determine function. In an evolutionary sense, the required changes/modifications constitute an integral part of the function and are therefore likely to be somehow encoded in the sequence of such proteins. The unusual aspect is that the key structural feature of these proteins is to keep regions natively unstructured or adaptable. The experimental and in silico identification and characterization of proteins with unstructured regions is evolving into an increasingly important challenge for structural biology. In facing this challenge, it becomes increasingly clear that the term "unstructured" describes a rather mixed bag of phenomena from regions that alter between different conformations to those that remain molten globule-like, and from regions that adopt regular helices and strands to those that remain intrinsically loopy. Here, we present NORSnet, a neural network--based method that revisited the task of identifying unstructured regions from a different angle than that taken by other methods. It focuses on the distinction between unstructured and well-structured loops. The success in this undertaking confirmed our initial hypothesis, namely that short unstructured loops resemble very long unstructured loops (NORS regions). Our application of machine learning was rather unconventional in two ways. First, we trained on positives (predicted NORS) that did not contain the feature we sought to predict (short unstructured loops) and on negatives (all regions in the PDB) that contained regions that we wanted the method to predict as positives; i.e., we implicitly hoped that our development would fail for many cases. Second, we did not optimize any parameters on the dataset used for assessing the performance of our method. Due to the difference in our approach, NORSnet complemented existing methods that optimize on previous datasets of unstructured regions. Consequently, NORSnet will enable the application of additional filters for structural genomics. Last, through a comparison between our new and other prediction methods, we confirmed the importance of unstructured regions for protein--protein interactions. Moreover, we specifically touched on the importance of unstructured loops for network complexity. Materials and Methods {#s3} ===================== Dataset of NORS regions. {#s3a} ------------------------ We created our dataset of residues in natively unstructured regions ("positives") in the following way. We grouped all proteins from 62 entirely sequence proteomes into domain-like families using CHOP and CLUP \[[@pcbi-0030140-b035],[@pcbi-0030140-b071],[@pcbi-0030140-b072]\]. We identified proteins with long NORS regions by the application of NORSp; i.e., all residues that are located in a stretch of \>70 consecutive residues with \<12% predicted helix or strand \[[@pcbi-0030140-b027],[@pcbi-0030140-b029]\] by PROFsec \[[@pcbi-0030140-b030]--[@pcbi-0030140-b032]\] and have at least one contiguous segment longer than ten residues predicted to be on the protein surface \[[@pcbi-0030140-b073]\]. The hope was that all residues in this pool have commonalities that we could extract through machine learning, and that will also be shared by proteins with unstructured regions much shorter than 70 residues. Due to the fact that PROFsec is especially accurate for natively unstructured regions \[[@pcbi-0030140-b004]\], the noise in these data that originated from the prediction mistakes was likely very low. To distinguish between proteins with and without unstructured regions, we needed a set of "negatives" (i.e., residues that are well-structured). For this, we chose a sequence-unique subset of globular protein structures from the PDB. Technically, this sequence-unique subset was taken from the EVA server \[[@pcbi-0030140-b074],[@pcbi-0030140-b075]\]. Specifically, the sequence redundancy was removed above HSSP (a measure for sequence-proximity) similarity values of 0 \[[@pcbi-0030140-b076],[@pcbi-0030140-b077]\] (corresponding to \<22% pairwise sequence identity for long alignments). Any pair of sequences between training and testing sets that could be aligned at PSI--BLAST \[[@pcbi-0030140-b078]\] E-values of \<10^−3^ according to our standard procedure of three automated iterations \[[@pcbi-0030140-b079]\] was also removed. To further amplify the signal from well-structured regions in the negative set, we also excluded all loops longer than 30 residues. Our datasets were not fully clean in the sense that our negative set of well-structured PDB proteins certainly contained some residues that did not appear in the X-ray structure (which were implicitly treated as well-structured), and that the positive set (predicted NORS) might contain some regular, ordered regions. However, due to the immense size of both datasets and to our use of neural networks, we did not worry about such outliers. In fact, our particular generation of a prediction-based training set that is more than ten times larger, and certainly more representative, than sets used previously might be the most important difference to all previous methods. In the context of a different problem, we showed how beneficial the use of prediction-based sets with errors might be \[[@pcbi-0030140-b080]\]. Training and testing set. {#s3b} ------------------------- To optimize the parameters of the method, we trained the network on 90% of the sequences and tested it on the remaining 10%. Note that these data were only used for the development of the method. We never reported the performance of the method on these data. The datasets on which we *did* assess NORSnet had no overlap (HSSP-value \<0; i.e., \<22% pairwise sequence identity for 250 aligned residues) with any of the proteins used for development. In particular, NORSnet was not optimized in any way on DisProt and the NESG dataset, as these were solely used to assess its performance. DisProt data. {#s3c} ------------- After optimizing our method to predict NORS regions (as described below in the prediction method section), we assessed NORSnet performance on different sets without any further optimization. In the first benchmark, we used DisProt proteins that have unstructured regions longer than 30 residues as positives and a sequence-unique subset of 173 PDB X-ray structures as negatives. The latter subset was taken from the EVA server \[[@pcbi-0030140-b074],[@pcbi-0030140-b075]\], and did not include sequences that were in the original training set. One particular advantage of testing our method on DisProt was that we did not have to run any additional cross-validation experiment since we used different proteins; respectively, the same proteins with different labels (*all* residues from PDB in DisProt were explicitly treated as "well-structured" by our training procedure). NESG dataset. {#s3d} ------------- To further validate the method, we tested it on a set of proteins from the NESG consortium. The positive set included 30 proteins that were identified to have unstructured regions ("NESG unfolded"), and the negative set included 170 recently determined protein structures. Both sets were identified as such by the NESG consortium. The definition of "unstructured region" was as follows: (1) HSQC (heteronuclear single quantum correlation) was high signal to noise and very low dispersion; and (2) hetNOE (heteronuclear Overhauser effect) data was clean negative (G. T. Montelione, personal communication). Using this set contributed to the removal of two types of biases in DisProt and similar databases. (1) Structure determination method: the negative set was almost equally divided between X-ray and NMR structures. (2) Length bias: while usually sequences selected for NMR structure determination are shorter than for X-ray determination, the NESG consortium reduced this artifact by using both methods in parallel to determine the structures of the same sequences. Thus, the length distribution of the NESG unfolded set is similar to the one of the folded set, in contrast to DisProt database, which consists of some much longer sequences (see [Table S1](#pcbi-0030140-st001){ref-type="supplementary-material"}). Protein--protein interaction set. {#s3e} --------------------------------- For the large-scale predictions of proteins that are involved in protein--protein interactions, we used the IntAct database (<http://www.ebi.ac.uk/intact>). IntAct includes both large- and small-scale experiments for different organisms \[[@pcbi-0030140-b067]\]. Specifically, we used proteins from interactions that were detected in a large-scale yeast two-hybrid screen of C. elegans (worm) proteins \[[@pcbi-0030140-b066]\]. The set included 2,622 proteins that participate in 4,039 interactions. Prediction method. {#s3f} ------------------ We used a standard feed-forward neural network described elsewhere in more detail \[[@pcbi-0030140-b030],[@pcbi-0030140-b032],[@pcbi-0030140-b073],[@pcbi-0030140-b081]\] The crucial novelty for the given task was the choice of input information. This choice was largely influenced by what we found to succeed in different contexts, namely for the prediction of normalized B-values \[[@pcbi-0030140-b082]\] and protein--protein interfaces \[[@pcbi-0030140-b083]\]. Local input information was taken from a sliding window of 13 sequence-consecutive residues (the prediction was for the central residue in that window). For each residue, we used the evolutionary profile (from PSI-BLAST alignments according to our standard protocol \[[@pcbi-0030140-b079]\]), the three-state secondary structure predicted by PROFsec \[[@pcbi-0030140-b030]--[@pcbi-0030140-b032]\], the two-state solvent accessibility state predicted by PROFacc (a profile-based neural network predictor of solvent accessibility) \[[@pcbi-0030140-b073]\], and the two-state flexibility prediction by PROFbval \[[@pcbi-0030140-b082],[@pcbi-0030140-b084]\]. Global input information was represented by the global amino acid composition (20 units), the composition in predicted secondary structure (three units), and solvent accessibility (two units), as well the length of the protein/domain-like fragment (three units as in \[[@pcbi-0030140-b082]\]), and the mean hydrophobicity divided by the net charge as was first suggested by Uversky et al. \[[@pcbi-0030140-b014]\]. DISOPRED2, FoldIndex, and IUPred. {#s3g} --------------------------------- We downloaded the DISOPRED2 package from <http://bioinf.cs.ucl.ac.uk/disopred> and installed it locally. The package included DISOPRED2 V0.2 and PSIPRED Version 2.45 (from November 2003). To assess its performance on our datasets, we ran the program using the default parameters. The prediction for casein precursor in [Figure 5](#pcbi-0030140-g005){ref-type="fig"}A was taken from the DISOPRED2 server. We ran FoldIndex using the server at <http://bip.weizmann.ac.il/fldbin/findex> (in September 2006) with default parameters. We ran IUPred using the server at <http://iupred.enzim.hu/index.html> (in December 2005 and January 2006) with default parameters. Supporting Information {#s4} ====================== ###### PSIPRED Prediction for Kappa-Casein Precursor The protein is predicted to have several long loops (residues 24--42, 89--125, and 130--171). Note that the location of the loops is correlated with high scores predicted by NORSnet and DISORPED2 that use this information. (7.1 MB TIF) ###### Click here for additional data file. ###### Secondary Structure Predictions of the N-Termini Domains of DFF45 Despite the fact that the N-term domain of DFF45 is unstructured, PSIPRED predicts secondary structure elements within that region. (5.0 MB TIF) ###### Click here for additional data file. ###### Unstructured Regions Overrepresented in Protein--Protein Hubs of Worm Similarly to [Figure 7](#pcbi-0030140-g007){ref-type="fig"}, we ran IUPred on worm proteins that are involved in protein--protein interactions. NORSnet data are identical to those presented in [Figure 7](#pcbi-0030140-g007){ref-type="fig"}. The number of proteins that are predicted to be either unstructured or well-structured is plotted against the number of interacting partners for two different thresholds of reliability of the two methods: (A) and (B) were compiled for thresholds at which both methods maintained 100% accuracy for the NESG data ([Figure 4](#pcbi-0030140-g004){ref-type="fig"}), while graphs (C) and (D) were compiled for 100% accuracy on DisProt ([Figure 3](#pcbi-0030140-g003){ref-type="fig"}). (A) and (C) show the results for the number of proteins predicted in each bin of interaction partners, while (B) and (D) show the normalized ratios to zoom into the difference between unstructured and structured proteins in each bin. These ratios were compiled as Ratio(bin) = {\#unstructured(bin)/\#structured(bin)} / {\#unstructured(1)/\#structured(1)}. As all ratios are greater than 1, proteins with more than one interaction partner have more unstructured regions than proteins with one partner. For the thresholds at which both methods achieved 100% accuracy on the DisProt dataset, both IUPred and NORSnet identified unstructured regions in 98 proteins that interact with seven partners or more. IUPred predicted 37 proteins with unstructured regions that NORSnet did not identify, and NORSnet predicted 17 proteins with unstructured regions that IUPred had missed. (7.0 MB TIF) ###### Click here for additional data file. ###### NORSnet Captures Domain Boundaries The domain boundaries of 524 multidomain proteins were marked in a procedure described in Liu and Rost \[[@pcbi-0030140-b053]\]. Due to the fact that NORSnet is optimized to identify unstructured stretches that are longer than 30 (and SCOP domain boundaries are often shorter), we used the raw score by NORSnet rather than the filtered output. NORSnet did considerably better than random (in red) and yielded area under the ROC curve (AUC) of 0.672 (in blue). Morever, according to our gold-standard set, termini residues are never defined as domain borders. In "NORSnet no term" (in green), we treated NORSnet outputs of the 60 termini residues in each protein as negatives, assessing only NORSnet predictions for the middle of the chain. The new method was more accurate in distingushing domain boundaries from other residues (AUC = 0.715). (5.2 MB TIF) ###### Click here for additional data file. ###### Synopsis for Supporting Online Material (624 KB DOC) ###### Click here for additional data file. ###### Dataset of Unstructured Proteins from Northeast Structural Genomics Consortium $A$ NESG id refers to identifiers given by the NESG consortium. $B$ Disorder signal referred to different levels of signal of a protein to be unstructured from NMR experiments. *Largely* marked largely unstructured proteins; e.g., (1) their HSQC has high signal to noise and very low dispersion and (2) their HetNOE data is clear negative. *Partly* marked partly unstructured proteins, which have some local structure but overall obey the same criteria. A total of 20 proteins were identified as largely unstructured and ten proteins were identified as partly unstructured. (63.5 KB DOC) ###### Click here for additional data file. ###### PDB Identifiers Used as a Negative Set in [Figure 3](#pcbi-0030140-g003){ref-type="fig"}A (74.5 KB DOC) ###### Click here for additional data file. Accession Numbers {#s4a} ----------------- The Protein Data Bank (<http://www.rcsb.org/pdb>) accession numbers for the structures discussed in this paper are HIV type 1 P6 protein (2c55_A), Methanobacterium thermoautotrophicum 1615 (1eij), the conserved domain common to the transcription factors TFIIS, elongin A, and CRSP70 (1eo0), and DFF40 (1ibx). The DisProt (<http://www.disprot.org>) accession number for bovine Kappa-casein precursor is DP00192. Thanks to Dariusz Przybylski and Guy Yachdav (Columbia University, United States) for providing preliminary information and programs, to Andrew Kernytsky and Marco Punta (Columbia University) for valuable discussions, and to Kazimierz Wrzeszczynski and Henry Bigelow (Columbia University) for helpful comments on the manuscript. Thanks to Jonathan Ward and David Jones (University College London, United Kingdom) for making DISOPRED2 and PSIPRED available, to Jaime Prilusky and Joel Sussman (Weizmann Institute, Rehovot, Israel) for making FoldIndex available, and to Zsuzsanna Dosztányi and István Simon (Institute of Enzymology, Hungary) for making IUPred available. Particular thanks to Guy Montelione and colleagues (Rutgers University, United States) for creating and providing the NESG datasets. Thanks to the constructive criticism of two anonymous reviewers and to those from the editor, Phil Bourne. Last, not least, thanks to all those who deposit their experimental data in public databases, and to those who maintain these databases, in particular to Keith Dunker and his colleagues for the maintenance of DisProt. The work of BR was also supported partially by grant U54-GM072980 from the US National Institutes of Health. **Author contributions.** AS and BR conceived and designed the experiments, analyzed the data, and wrote the paper. AS and JL performed the experiments. All authors contributed reagents/materials/analysis tools. **Funding.** The work was supported by grants U54-GM074958--01 from the Protein Structure Initiative of the US National Institutes of Health to the Northeast Structural Genomics Consortium and 2R01-LM07329--01 from the National Library of Medicine. **Competing interests.** The authors have declared that no competing interests exist. A previous version of this article appeared as an Early Online Release on June 5, 2007 (doi:[10.1371/journal.pcbi.0030140.eor](10.1371/journal.pcbi.0030140.eor)). DFF : DNA fragmentation factor NMR : nuclear magnetic resonance spectroscopy NORS : no regular secondary structure NTD : N-terminal domain PDB : Protein Data Bank PSI : Protein Structure Initiative

{ "pile_set_name": "PubMed Central" }

Introduction {#s1} ============ Clathrin is a three-legged molecule with a central hub domain from which three ∼190 kDa heavy chains are extended, each ending in an N-terminal seven-bladed β-propeller domain that allows for multiple protein interactions with various specificities between its blades [@pone.0003115-terHaar1]. A single clathrin heavy chain (CHC) molecule contains in addition eight CHC repeat segments, a proximal hairpin, a tripod region believed to be responsible for trimerisation, and a variable C-terminal segment [@pone.0003115-Fotin1]. Each CHC is furthermore associated with a ∼25 kDa clathrin light chain (CLC). This building block of a cage structure is known as a triskelion, and during endocytosis the legs of neighbouring triskelia twist around each other to form a curved lattice that self-polymerizes around invaginated pits, stabilizing them as they bud from the major sites of formation within the cell--plasma membrane, trans-Golgi network and endosomes [@pone.0003115-Crowther1]--[@pone.0003115-Keen1]. Recently, research has focussed on changes in the rates of endocytosis during cell cycle progression and in the distribution of trafficking proteins. This has resulted in some controversy in the literature over whether endocytosis is inhibited during mitosis or is maintained. An original study showed that in a broken assay mitotic cytosol could inhibit endocytosis when compared to interphase material [@pone.0003115-Pypaert1]. Latterly, single-cell imaging has been used to determine that whilst endocytosis is maintained during all phases of the cell cycle, recycling of internalised membrane is inhibited during mitosis [@pone.0003115-Boucrot1], [@pone.0003115-Boucrot2]. Clathrin has also been found at the mitotic spindle both through confocal imaging and proteomic analysis of enriched spindle fractions [@pone.0003115-Andersen1], [@pone.0003115-Okamoto1]. Knockdown of the heavy chain of clathrin in HEK293 and NRK cells using siRNA results in mitotic defects and this has led to the suggestion that clathrin may have a trafficking-independent function in mitosis [@pone.0003115-Royle1]. By contrast, components of the AP-1, AP-2 and AP-3 adaptor complexes did not colocalise with the spindle apparatus [@pone.0003115-Royle1]. Past controversy on changes in the endocytic rates during cell cycle progression suggests that it will prove important to explore the role of clathrin at the spindle in multiple cell-lines using multiple approaches. Consequently, we have used a chicken pre-B lymphoma cell line DT40, which was generated with endogenous alleles for CHC replaced by human CHC under the control of a tetracycline-regulatable promoter, in order to investigate the role of clathrin in the evolutionarily conserved process of clathrin-mediated endocytosis [@pone.0003115-Wettey1], [@pone.0003115-Wettey2]. Following repression of clathrin expression, receptor-mediated and fluid-phase endocytosis were significantly inhibited in a surviving sub-line (DKO-R) [@pone.0003115-Wettey1]. We have now used this well-characterised model of membrane trafficking to quantitatively test for the first time, using flow cytometry, the impact of clathrin knockout on cell cycle progression in a suspension cell-line. We found no difference in the cell cycle distribution of the knockout cells versus the wild-type. Additionally, we observed that the ploidy and recovery kinetics following cell cycle arrest with the microtubule-depolymerising agent nocodazole were unchanged by knocking out clathrin. Consequently, whilst clathrin is an important component of the trafficking machinery and colocalises with the mitotic spindle, in these cells, its contribution to mitosis is not appreciable. Results {#s2} ======= We first monitored the spatial distribution of CHC and alpha-tubulin during cell cycle using confocal microscopy, confirming their colocalization at the spindle apparatus during mitosis in the DKO-R line ([**Figure 1A**](#pone-0003115-g001){ref-type="fig"}). We then treated the line with doxycycline (50 ng/ml) to fully repress CHC expression, which was achieved after 48 hours ([**Figure 1B**](#pone-0003115-g001){ref-type="fig"}). Royle *et al.* had shown an increase in the mitotic index of cells treated with siRNA against CHC using confocal microscopy. To test this more quantitatively, we sought to analyse steady-state cell cycle distribution in the presence and absence of CHC using a propidium iodide flow cytometric assay. Intriguingly, we found no change in the G2/M phase proportion of CHC-depleted cells relative to wild-type, or evidence of changes in their ploidy status ([**Figure 1C**](#pone-0003115-g001){ref-type="fig"}). Nevertheless, the absence of steady-state changes would not preclude differences in the transition rate through cell cycle checkpoints. To investigate this, we added nocodazole (500 ng/ml) for 11 hours to induce full spindle disruption and thus arrest clathrin-expressing and knockout cells at the G2/M checkpoint[@pone.0003115-Sudo1]. We then sampled the cells over a time course (1, 2, 4, 8 and 24 hours) having washed out the drug, and measured the proportions in each phase of the cell cycle by flow cytometry. There were no statistically significant differences in the recovery rates between clathrin depleted and control cells ([**Figure 1D**](#pone-0003115-g001){ref-type="fig"}). Furthermore, an apoptosis assay conducted on these cells over the same time course showed no significant differences in the level of cell death between the recovering populations ([**Figure 2**](#pone-0003115-g002){ref-type="fig"}). Finally, we confirmed that the clathrin-depleted cells are able to assemble mitotic spindles and are mononuclear ([**Figure 3**](#pone-0003115-g003){ref-type="fig"}). On the other hand, the same flow cytometric approach combined with targeting clathrin for knockdown using siRNA in a HEK293 cell-line resulted in an increase in the G2/M population from 15.66% to 24.54% despite the presence of detectable residual clathrin (∼10%) in total cell lysates, as detected using the same clathrin antibody ([**Figure 4**](#pone-0003115-g004){ref-type="fig"}). Consequently whilst the repression of clathrin in DKO-R cells produced a greater reduction in detectable clathrin than siRNA on HEK293 cells, the functional effects of this reduction were only measurable in HEK293 cells. This implies a difference in the mitotic sensitivity of the two cell-lines to changes in clathrin levels. ![A) Clathrin heavy chain colocalizes with alpha-tubulin at the spindle apparatus in mitotic DKO-R cells.\ Confocal immunofluorescence microscopy on an asynchronised population of DKO-R cells grown in the absence of doxycycline (*upper left panel)*. Cells were stained with Hoechst 33258 (blue; Sigma), anti-alpha-tubulin DM1A (green; AlexaFluor 488 donkey anti-mouse conjugate, Invitrogen) and anti-CHC Ab21679 (red; AlexaFluor 594 goat anti-rabbit conjugate, Invitrogen). The asterisk indicates a mitotic cell in which CHC is targeted to the kinetochore fibres. A Z-stack projection of the indicated mitotic cell is shown (*right panel)*. Scale bar, 10 µm. B) Repression of CHC to undetectable levels is achieved after 48 hours treatment with doxycycline. Western blot analysis of DKO-R cells treated with doxycycline for the indicated time periods. β-tubulin was used as a loading control. C) Cell cycle distribution is unaffected by CHC depletion. DKO-R cells were treated with doxycycline for 48--72 hours and then subjected to flow cytometric analysis. DNA content was monitored by PI staining. The cell cycle profiles of CHC-expressing and CHC-depleted cells are shown. The percentage of cells in each cell cycle phase was calculated by ModFit Sofware and is shown in the boxed legend. D) The recovery kinetics from cell cycle arrest at the G2/M checkpoint are unaffected by CHC depletion in DKO-R cells. Time-course analysis of DKO-R cells following release from a nocodazole-induced block. Cells were grown in the presence (CHC-depleted) or absence (CHC-expressing) of doxycycline for 48--72 hours and then synchronized at metaphase by treatment with 500 ng/ml of nocodazole for 11 hours. A representative vehicle-treated asynchronous sample is indicated for each population ('Asynch' bar). At the indicated time points post-release from nocodazole block, cells were harvested and subjected to propidium iodide flow cytometric analysis. Data represent the mean values from three independent experiments with error bars denoting standard errors. The post-arrest recovery kinetics were compared for the clathrin-expressing and clathrin-depleted cells at each phase in the cell cycle using the analysis of variance (ANOVA). A p-value of 0.05 or less would be required to declare a significant difference between the two populations. P-values were 0.20 for G1-phase, 0.16 for S-phase and 0.09 for G2/M phase, indicating that there is no evidence of a difference between the recovery kinetics of the two populations over time following a nocodazole-induced block.](pone.0003115.g001){#pone-0003115-g001} ![Time course analysis of cell death levels after release from a nocodazole-induced metaphase block.\ Asynchronous and nocodazole-arrested CHC-expressing and -depleted DKO-R cells were harvested at the indicated time points and stained with Annexin V and 7-aminoactinomycinD (7-AAD) followed by flow cytometry analysis. Although there is an increase in the levels of cell death during the recovery period, which has been well-established to be an effect of microtubule-targeting agents [@pone.0003115-Jordan1], there are no significant difference in the proportion of dead cells between the two populations. Data represents values from one experiment.](pone.0003115.g002){#pone-0003115-g002} ![CHC-depleted DKO-R cells preserve the ability to assemble a mitotic spindle.\ Confocal immunofluorescence microscopy on an asynchronised population of DKO-R cells grown in the presence of 50 ng/ml doxycycline for 72 hours. Cells were stained with Hoechst 33258 (blue; Sigma), anti-alpha-tubulin DM1A (green; AlexaFluor 488 donkey anti-mouse conjugate, Invitrogen) and anti-CHC Ab21679 (red; AlexaFluor 594 goat anti-rabbit conjugate, Invitrogen). The asterisk indicates a mitotic cell. Scale bar, 10 µm.](pone.0003115.g003){#pone-0003115-g003} ![Clathrin knockdown in HEK293 cells leads to an increase in the G2/M phase content.\ A) Western blot analysis of HEK293 cells transfected with siRNA against CHC17 or with scrambled siRNA. β-actin was used as a loading control and two different clathrin heavy chain antibodies raised against the N-terminal domain (TD.1; Abcam) and the C-terminus (Ab21679; Abcam) were used on cell lysates harvested at the time of cell cycle analysis. B) The cell cycle profiles of scrambled siRNA treated and CHC siRNA treated cells are shown. The percentage of cells in each cell cycle phase was calculated by ModFit Sofware and is shown in the boxed legend.](pone.0003115.g004){#pone-0003115-g004} Discussion {#s3} ========== Given that clathrin does not seem critical for the function of the mitotic spindle in the DKO-R line, an alternative explanation for its localisation at the spindle apparatus remains that clathrin redistribution to the spindle is a by-product of Golgi disassembly which occurs during mitosis. Clathrin has been implicated in a number of nonendocytic functions including the maintenance of basolateral polarity involving the regulation of protein exit from the Golgi [@pone.0003115-Deborde1], and post-mitotic Golgi reassembly [@pone.0003115-Radulescu1]. A number of proteins required for the maintenance of the Golgi stack are known to associate with the spindle [@pone.0003115-Efimov1] and clathrin redistribution may be a marker for their recruitment. Clathrin is a stable protein with a slow turnover [@pone.0003115-Acton1] and consequently degrading and resynthesising this protein within the timeframe of M phase is not a feasible means of disassembling and regenerating subcellular organelles and vesicles during cell cycle progression. Our work strongly suggests that cell type and context are important determinants of the significance of the contribution that clathrin makes to mitosis, as has long been known for the role that it plays in endocytosis. For example, B cell antigen receptor internalisation was shown to be inhibited 70% by clathrin depletion in another DT40 variant cell-line (DKO-S), but to be unimpaired by depletion in the DKO-R line [@pone.0003115-Stoddart1]. Clathrin has also been shown to contribute to the delivery of Lamp1 and lysosomal proteins to the lysosome through the application of siRNA to knockdown expression in HeLa cells [@pone.0003115-Janvier1]. By contrast, no lysosomal defects or abnormalities in the distribution of lysosomal proteins have been observed upon clathrin depletion in the DKO-R cell-line [@pone.0003115-Wettey1]. DKO-R cells are not normally grown in the absence of clathrin, there is merely inducible depletion. However, there remains the possibility of transient adaptive changes to clathrin depletion in these cells over the 72-hour timecourse used in the DKO-R experiments,as there might be in other cells treated with siRNA for up to 72 hours. An expression array study is currently underway to address this point further. Nonetheless, just as in other cell-lines, aspects of receptor-mediated and fluid-phase endocytosis are impaired by clathrin depletion in the DKO-R cell-line. There is additionally a second clathrin heavy chain variant, CHC22, which is predominantly expressed in muscle and is highly similar in sequence to the ubiquitous CHC17 form introduced into the DKO-R cell-line. Whilst there is also a small amount of CHC22 detectable in this cell-line, it is not upregulated in response to doxycycline treatment and small levels are also found in HEK293 and HeLa cells (Dr. A. P. Jackson; Personal communication). Furthermore, CHC22 has not been detected at the mitotic spindle (Professor Frances Brodsky, UCSF; Personal communication). Consequently, we do not feel that compensation or adaptation to clathrin depletion through CHC22 is likely to happen in this case. In conclusion, further detailed study is required to better understand at the molecular level the recruitment of clathrin to the spindle, as this represents a common feature of chicken B-cells, mammalian [@pone.0003115-Okamoto1] and plant cells [@pone.0003115-Tahara1], and yet not of apparently equivalent functional significance for mitosis. Just as clathrin-independent or compensating endocytic mechanisms have been characterised, perhaps in the future both clathrin-dependent and clathrin-independent (or compensating) mechanisms for spindle assembly and chromosome segregation will be unearthed. Materials and Methods {#s4} ===================== Cell Culture {#s4a} ------------ DKO-R cells were cultured in RPMI media (Invitrogen) supplemented with 10% Foetal Bovine Serum (FBS; Invitrogen), 1% chicken serum (Invitrogen) and 10^−5^ M β-mercaptoethanol (Sigma) at 40°C in a humidified incubator supplied with 5% CO~2~ [@pone.0003115-Buerstedde1]. HEK293 cells were cultured in Dulbecco Modified Eagle\'s Medium (DMEM) containing 10% FBS at 37°C in a humidified incubator supplied with 5% CO~2~. Antibodies {#s4b} ---------- All of the antibodies used in this study were obtained from Abcam (Cambridge, UK). Three clathrin antibodies were used; TD.1 (N-terminal domain epitope; Ab24578), X22 (C-terminal domain epitope; Ab2731), and Ab21679 (C-terminal epitope). Loading control antibodies were against beta-tubulin (D66; Ab11307) and actin (AC-40; Ab11003). In addition an alpha-tubulin antibody (DM1A; T9026) obtained from Sigma was used to stain microtubules. siRNA knockdown of clathrin in HEK293 cells {#s4c} ------------------------------------------- The procedure used has been adapted from Motley et al., [@pone.0003115-Motley1]. Briefly HEK293 cells were seeded in full-serum medium into 100 mm dishes, at a density of 3×10^6^ cells per dish. The first transfection was carried out when cells had reached 80% confluency. For each 100 mm dish, 39 µl of DMRIE-C transfection reagent (Invitrogen) were added to 6 ml of serum-free DMEM, to which 70 nM of a set of 4 siRNA CLTC Duplexes (Dharmacon ON-TARGETplus Set of 4, Catalog No. LQ-004001-01) were added. Control siRNA was supplied by Qiagen (AllStars Negative Control siRNA (5 nmol Catalog No. 1027280). The transfection mixture was added to the cells after rinsing them once with serum-free DMEM, and the cells were left in this mixture until they were trypsinized the following day into two 100 mm dishes. 24 hours later a second transfection was carried out in a similar manner, and cells were equally trypsinized the following day and plated onto 100 mm dishes for protein extraction and cell cycle analysis 24 hours later. Western Blotting {#s4d} ---------------- Total cell extracts were prepared by lysing cells in Lysis Buffer containing 50 mM Tris-HCl pH 7.8, 150 mM NaCl, 5 mM EDTA, 15 mM MgCl~2~, 1% IgaPal, 0.75% sodium deoxycholate, 1 mM DTT, complete protease and phosphatase inhibitors (Roche) and 1 mM NEM. The extracts were cleared by centrifugation for 10 min at 14,200 rpm, followed by collection of the supernatant. Protein quantification was performed using Coomassie reagent and a BSA serial dilution for standard curve calculation. Cell lysates containing 15 µg of protein were analysed by 8% SDS-PAGE. Transfer to nitrocellulose membranes was carried out using the i-blot Dry Transfer System (Invitrogen), and membranes were immunoblotted with primary antibodies against clathrin heavy chain and beta-tubulin, followed by the appropriate HRP-conjugated secondary antibodies (Dako). Immunorreactive signals were visualized by ECL Plus (GE Healthcare). Apoptosis detection {#s4e} ------------------- DMSO- (vehicle) or nocodazole-treated wild-type and CHC-depleted DKO-R cells were stained with Annexin V-FITC (Cat\# TA5532, R&D Systems, Inc.) and 7-actinomycinD (7-AAD; Invitrogen) according to the manufacturer\'s instructions, and subsequently analysed by flow cytometry on a four laser LSR II SORP flow cytometer (BD Biosciences, San Jose, CA). The fluorescence emitted by Annexin V-FITC and 7-AAD was collected using a 530/30 and 575/26 bandpass filter. Data collected from the experiment was analysed with FlowJo. V 8.3 (Treestar, Ashland, OR). Immunofluorescence and Confocal Microscopy {#s4f} ------------------------------------------ DKO-R cells were prepared for imaging as described previously[@pone.0003115-Wettey1] with the following exceptions. Coverslips were mounted using ProLong Gold antifade reagent (Invitrogen) with 1 µg/ml Hoechst 33258 (Sigma), and the following primary antibodies were used: anti-alpha-tubulin (DM1A, Sigma) and anti-CHC (ab21679, Abcam). Images were acquired on a Nikon Eclipse C1si confocal microscope system (Nikon UK Limited, Surrey, UK), and processed using Volocity image analysis software (Improvision, Coventry, UK). Flow Cytometric Analysis {#s4g} ------------------------ DMSO- (vehicle) or nocodazole-treated wild-type and CHC-depleted DKO-R cells were stained with Propidium Iodide (Sigma) as described elsewhere[@pone.0003115-Hawkins1], followed by analysis on a four laser LSR II SORP flow cytometer (BD Biosciences, San Jose, CA). The fluorescence emitted by the Propidium Iodide was collected using a 575/26 bandpass filter. Data collected from the experiment was analysed with ModFit LT V3.0 (Verity Software House, Topsham, ME). We would like to acknowledge the wider support of The University of Cambridge, Cancer Research UK and Hutchison Whampoa Limited. The authors would like to acknowledge the support of the National Cancer Research Institute (NCRI) formed by the Department of Health, the Medical Research Council and Cancer Research UK. The authors would also like to acknowledge the helpful input of Heather Zecchini and to thank members of the CRUK Uro-Oncology Research Group and Fanni Gergely for helpful discussions during the preparation of this manuscript. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**JB is the recipient of an EU/Marie Curie Cancer Cure Early Stage Research Training (CANCURE) Fellowship. GV and IGM are supported by core CRUK funding. [^1]: Conceived and designed the experiments: JB IGM. Performed the experiments: JB. Analyzed the data: JB GV IGM. Contributed reagents/materials/analysis tools: APJ. Wrote the paper: JB IGM. Contributed to the preparation of the manuscript: APJ GV.

{ "pile_set_name": "PubMed Central" }

Introduction ============ It is estimated that above 30% of population and 65% of older population have hypertension worldwide ([@B43]). Previous studies suggest that hypertensive individuals have a greater chance of occurring the dementia and physical disability than normotensive individuals ([@B32]; [@B20]; [@B21]). Cognitive functions, especially the executive functions ([@B9]), are widely reported of being impaired in hypertensive individuals ([@B60]; [@B24]; [@B27]). Individuals with hypertension have higher chances of occurring executive dysfunction earlier than individuals with normotension, which indicates a potential vascular-cognitive association ([@B44]). However, there are very limited neuroimaging studies that investigated the neural bases of hypertension in cognitive decline ([@B35]). Former non-neuroimaging case studies further report that the African--Americans bear the greater burden of hypertension in the United States and have earlier onset of hypertension and larger hypertension-associated cognitive symptomatology and mortality than the other racial groups, including the Caucasian--Americans ([@B47]; [@B27]). The neural mechanisms underlying this racial disparity is largely under-investigated so far. A triple brain network model---a model consisting of the brain's default-mode, salience, and central-executive networks and their interactions---has been recently employed in the neuroimaging field to elucidate the differences in the connectivity patterns associated with the different levels of cognitive impairments such as higher versus lower impairment ([@B39]; [@B56]). We therefore use this model to investigate whether there is a difference in the impairment at the nodes of triple network between the African--Americans and the Caucasian--Americans hypertensive cognitively impaired individuals. Resting state functional MRI (rsfMRI) has been used to investigate the functional brain areas or neurocognitive networks ([@B4]; [@B46]). Recent neuroimaging investigations suggest that the neural basis of cognitive activity is related to a dynamically modulating interaction between multiple networks, including the salience, default-mode, and central-executive networks ([@B6]; [@B40]; [@B56]; [@B8]). The key nodes of the default-mode network include the posterior cingulate and the ventromedial prefrontal cortices, the salience network encompasses the insula and the dorsal anterior cingulate cortices, and the central-executive network comprises the posterior parietal and the dorsolateral prefrontal cortices ([@B13]). It has been demonstrated that the insula and dorsal anterior cingulate of salience network are anatomically connected ([@B5]; [@B31]) and consist of a special type of neurons named von Economo neurons that relay information processed within these regions to other brain regions, including the nodes of default-mode and central-executive networks ([@B2], [@B1]; [@B62]; [@B51]). This control signal by the insula and the dorsal anterior cingulate cortex has been suggested to be crucial for cognitive maintenance, including a rest, in cognitively normal individuals ([@B51]; [@B25]; [@B8]). Alternation in insula connectivity has been consistently implicated in diseases, including autism, frontotemporal dementia, and schizophrenia ([@B39]; [@B56]), but it has not been elucidated in cognitively impaired hypertensive patients. Literature suggests that the insula and anterior cingulate cortex---the key regions of salience network---respond as the racially biased brain regions ([@B7]), such as the greater activity of insula to faces of foreign races than faces of the same race of the subject ([@B37]; [@B38]). However, the difference in information flow---a measure from information theory that can be quantified using Granger Causality analysis ([@B17]; [@B10])---in these regions between the African--Americans and the Caucasian--Americans themselves has not been previously investigated. Previous investigations consistently report that the functional changes of brain regions (or networks) are associated with the underlying structural changes of those regions (or networks) with the progression of diseases ([@B63]; [@B40]). Specifically, recent studies suggest that the insula thickness decreases with the progression of cognitive decline in mild cognitive impairment (MCI) patients ([@B28]; [@B41]). However, whether there is a difference in insula thickness of the cognitively impaired hypertensive patients between the African--Americans and the Caucasian--Americans races has not been reported. Here, we seek to examine the difference in information flow using Granger causality ([@B17]; [@B10]) at the default-mode, salience and central-executive nodes between the African--Americans and the Caucasian--Americans hypertensive cognitive impaired individuals. As the African--Americans have higher hypertension-associated cognitive symptomatology and mortality than the other racial groups ([@B47]; [@B27]), we *hypothesized* that (1) the control signal of the insula of salience network over the default-mode and central-executive nodes is more impaired (more negative value) in the African--Americans than in the Caucasian--Americans. We further seek to examine the structural difference that could substrate this racial disparity by comparing the insula thickness between the two racial groups. To test this, we further *hypothesized* that (2) the insula thickness is lower in the African--Americans than in the Caucasian--Americans, and finally (3) lower insula thickness is associated with poorer cognitive performance. Materials and Methods {#s1} ===================== Participants ------------ This study was carried out in accordance with the recommendations of "Institutional Review Board (IRB) of Emory University" with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The study and the protocol were reviewed and approved by Institutional Review Board of Emory University. The informed written consent was provided by the participants before data collection. We recorded magnetic resonance imaging (MRI) data from 78 individuals who had hypertension and MCI. The inclusion criteria were: (a) age ≥ 55 years, (b) hypertension defined by systolic blood pressure ≥ 140 mm Hg or diastolic blood pressure ≥ 90 mm Hg and (c) MCI was assessed based on previously defined Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria ([@B12]; [@B45]): Montreal cognitive assessment (MoCA) ≤ 26, clinical dementia rating score of 0.5, minimal functional limitation as reflected by the functional assessment questionnaire ≤ 7, and cognitive performance at the 10th percentile or below till the 2nd percentile on at least one of four screening tests---trail marking test B ([@B48]; [@B54]), Stroop interference ([@B53]; [@B33]), digit span forward and digit span backward ([@B30]; [@B45]), verbal fluency and abstraction ([@B29]; [@B55]). Trail marking test B screens the participant's executive ability to draw a line from a 'number' to a 'letter' in ascending order such as '1' to 'A', 'A' to '2', '2' to 'B', and so on. Strop interference effect measures the interference of predominant response in the reaction time of a task such as when the name of a word (say 'red') is printed in a different color (say 'blue'), it takes longer time to name the color of that word compared to when the word ('red') matches the name of color ('red'). Digit span forward and digit span backward test consists of two parts: first, the participant listens to and repeats a sequence of numbers, and second, the participant listens to a sequence of numbers and repeats those numbers in reverse order. The former part screens the short-term auditory memory while the latter part screens the participant's ability to manipulate the verbal information based on the auditory information. The verbal fluency screens the participant's fluency such as the participant is asked to tell as many words as the participant can that begin with certain letter (say 'A') in a limited time and the abstraction test screens the participant's ability to deal with ideas such as how an orange and a banana are alike. These tests have been commonly used to screen the MCI patients ([@B12]; [@B45]). The participants exclusion criteria were: (a) systolic blood pressure \> 200 mm Hg or diastolic blood pressure \> 110 mm Hg, (b) renal disease or hyperkalemia, (c) active medical or psychiatric problems, (d) uncontrolled congestive heart failure (shortness of breath at rest or evidence of pulmonary edema on exam), (e) history of stroke in the past 3 years, (f) ineligibility for MRI (metal implants or cardiac pacemaker), (g) inability to complete cognitive test and MRI scan, (h) women of childbearing potential and (i) diagnosis of dementia (self-reported or care-giver reported). In total sample, the mean age was 66.9 years (*SD*: 9.7), 55.1% were women, mean education was 14.9 years (*SD*: 2.6), mean systolic blood pressure 144.4 mm of Hg (*SD*: 22.6), mean diastolic blood pressure 86.4 mm of Hg (*SD*: 12.8), and MoCA ranged from 11 (minimum value) to 26 (maximum value) with mean score of 21.9 (*SD*: 3.1). Out of 78 participants, there were 50 African--Americans and 28 Caucasian--Americans. In the African--Americans, the mean age was 66.3 years (*SD*: 8.9), 60% were women, mean education was 14.9 years (*SD*: 2.5), mean systolic blood pressure 141.8 mm of Hg (*SD*: 21.3), mean diastolic blood pressure 85.4 mm of Hg (*SD*: 12.0), and mean MoCA score was 21.2 (*SD*: 3.2). In the Caucasian--Americans, the mean age was 68.5 years (*SD*: 10.7), 46.4% were women, mean education was 15.0 years (*SD*: 2.9), mean systolic blood pressure 147.6 mm of Hg (*SD*: 24.0), mean diastolic blood pressure 87.0 mm of Hg (*SD*: 12.7), and mean MoCA score was 23.4 (*SD*: 2.4). The age, sex, education year, systolic blood pressure, and diastolic blood pressure were not statistically significant different between the African--Americans and the Caucasian--Americans, but the MoCA score was significantly lower in the African--Americans compared to the Caucasian--Americans (*p* \< 0.003) as shown in **Table [1](#T1){ref-type="table"}**. ###### Mean scores (standard deviations) and statistical comparison between the African--Americans (AA) and the Caucasian--Americans (CA) regarding their age, sex, education, blood pressure (BP), and Montreal cognitive assessment (MoCA) score. Characteristic *N* Total sample 78 AA 50 CA 28 *p*-Value --------------------- ----------------- -------------- -------------- ----------- Age, year 66.9 (9.7) 66.3 (8.9) 68.5 (10.7) 0.44 Sex, women 43 (55.1%) 30 (60%) 13 (46.4%) 0.25 Education, year 14.9 (2.6) 14.9 (2.5) 15.0 (2.9) 0.54 Systolic BP, mm Hg 144.4 (22.6) 141.8 (21.3) 147.6 (24.0) 0.14 Diastolic BP, mm Hg 86.4 (12.8) 85.4 (12.0) 87.0 (12.7) 0.71 MoCA score 21.9 (3.1) 21.2 (3.2) 23.4 (2.4) 0.003 p-Value indicates the level of statistical significance of the comparisons between the African--Americans and the Caucasian--Americans using non-parametric Wilcoxon rank sum for age, education, blood pressure and MoCA, and chi-square test for sex. MRI Acquisition --------------- Magnetic resonance imaging data were acquired on a SIEMENS Trio 3-Tesla scanner available at Center for Systems Imaging of Emory University, Atlanta, GA, United States. Foam padding and ear forms were used to limit head motion and reduce scanner noise to the participants. High-resolution 3D anatomical images were acquired using sagittal T1-weighted magnetization-prepared rapid gradient echo with repetition time = 2300 ms, echo time = 2.89 ms, inversion time = 800 ms, flip angle = 8°, resolution = 256 × 256 matrix, slices = 176, thickness = 1 mm. The rsfMRI were collected axially for 170 volumes during 7.14 min by using an echo-planar imaging (EPI) sequence with repetition time = 2500 ms, echo time = 27 ms, flip angle = 90°, field of view = 22 cm, resolution = 74 × 74 matrix, slices = 48, thickness = 3 mm and bandwidth = 2598 Hz/pixel. We requested the participants to hold still, keep their eyes open and think nothing during the rsfMRI scan. Image Preprocessing and Time Series Extraction ---------------------------------------------- Images were preprocessed for slice-timing correction, motion correction, co-registration to individual anatomical image, normalization to the Montreal Neurological Institute (MNI) template, and spatial smoothing of the normalized images with a 6 mm isotropic Gaussian kernel. The SPM12 (Wellcome Trust Centre for Neuroimaging, London, United Kingdom^[1](#fn01){ref-type="fn"}^) was used to perform those steps. We defined spherical regions of interest with 6 mm radius based on MNI coordinates centered at the posterior cingulate cortex (7, -43, 33) and ventromedial prefrontal cortex (2, 36, -10) of default-mode network, the insula (37, 25, -4) and dorsal anterior cingulate cortex (4, 30, 30) of salience network, and the posterior parietal cortex (54, -50, 50) and dorsolateral prefrontal cortex (45, 16, 45) of central-executive networks (see **Figure [1](#F1){ref-type="fig"}**) similar to the previous studies ([@B51]; [@B8]). We selected the nodes only in the right hemisphere based on most prior neuroimaging studies that report the right-lateralized activations ([@B51]; [@B13]; [@B8]). The MarsBaR software package^[2](#fn02){ref-type="fn"}^ was used to extract the voxel time courses of those nodes. ![Selection of the **(A)** default-mode (VMPFC, ventromedial prefrontal cortex; PPC, posterior cingulate cortex), **(B)** salience (AI, insula; DACC, dorsal anterior cingulate cortex), and **(C)** central-executive (DLPFC, dorsolateral prefrontal cortex; PPC, posterior parietal cortex) networks (P, posterior; A, anterior; L, left; R, right).](fnagi-09-00177-g001){#F1} Granger Causality Analysis -------------------------- Multivariate analysis has become a commonplace to investigate the information flow between the brain areas and to study how such coordinated brain activity disrupts in diseases ([@B16],[@B17]; [@B14]; [@B23]). Here we used Granger causality although other methods such as dynamic causal modeling, directed transfer function, and partial directed coherence provide the similar goals and results ([@B3]; [@B11]). The main benefits of using Granger causality are that it is a data-driven method and therefore computes the information flow based on the data itself at the nodes and networks level, relies on fewer assumptions about the underlying interactions, and does not need computationally intensive time/efforts as opposed to other methods such as dynamical causal modeling ([@B52]; [@B9], [@B10]). Recent studies by our group and by other groups have successfully applied Granger causality to resting state and/or task fMRI data in both health and disease and have produced meaningful results in terms of information flow at the brain nodes and networks ([@B51]; [@B14]; [@B36]; [@B3]; [@B8]). Granger causality can be mathematically expressed by considering simultaneously measured time series. Suppose we have two simultaneously recorded time series represented as, (1) X~1~(1), X~1~(2),\..., X~1~(t),\... and (2) X~2~(1), X~2~(2),\..., X~2~(t). Granger causality analysis in the frequency (f) domain examines the strengths, directions, and frequencies of interactions between dynamic processes. Granger causality from the second time series '2' to the first time series '1' (i.e., from brain region '2' to brain region '1') is computed as ([@B16],[@B17]; [@B9]), M 2 → 1 ( f ) = − ln ( 1 − ( ∑ 22 − ∑ 12 2 / ∑ 11 ) \| H 12 ( f ) \| 2 S 11 ( f ) ) where *H* is a transfer function, f represents a frequency-domain, S is spectral power, and ∑ is noise covariance. The value of Granger causality (*M*) varies between 0 and +∞, representing zero connectivity strength and maximum connectivity strength, respectively. If there are '*N*' numbers of brain areas, the information outflow (*F*) at a node *i* can be calculated as, F i = 1 N − 1 ∑ j N ( M i → j − M j → i ) . In our case, we have six nodes (two key nodes from each network). Therefore, an index *j* can be 1, 2, 3, 4, 5, and 6 nodes. The Granger causality outflow (also referred as the net information outflow) from the first node is *F*~1~ = \[(*M*~1→2~ -- *M*~2→1~) + (*M*~1→3~ -- *M*~3→1~) + (*M*~1→4~ -- *M*~4→1~) + (*M*~→5~ -- *M*~5→1~) + (*M*~1→6~ -- *M*~6→1~)\]/5, where *M*~1→2~ is Granger causality from the first node to the second node, and *M*~2→1~ is Granger causality from the second node to the first node. Similarly, we calculate Granger causality outflows for other nodes. If the net outflow is negative at a node instead of the previously reported positive value in healthy individuals, then that node is said to have impaired directional connections. Cortical Thickness Calculation ------------------------------ FreeSurfer version 5.3^[3](#fn03){ref-type="fn"}^ was used to calculate the cortical thickness from T1-structural images. Briefly, this technique included spatial and intensity normalization, skull stripping, and an automated segmentation of cerebral white matter to locate the gray--white boundary ([@B15]). Cortical thickness was then computed from the distance between the gray--white boundary and the pial-surface ([@B22]). Cognitive Test -------------- We assessed the MoCA ([@B42]) of each participant. The MoCA is a 30-point scale test administered in 10 min and assesses the global cognitive abilities. It encompasses the following sub-tests. The short-term memory recall (five points) consists of two learning trials of five nouns and delayed recall after 5 min. Visuospatial test includes a clock-drawing (three points) and a three-dimensional cube copy (one point). Executive functions test comprises the trail-making time B (one point), a phonemic fluency (one point), and a two-item verbal abstraction (two points). Attention, concentration, and working memory include a sustained attention (one point), a serial subtraction (three points), and digit forward (one point) and digit backward (one point). Language test consists of naming animals (three points), repetition of two syntactically complex sentences and fluency (two points). Orientation test comprises orientation to time and place (six points). Moreover, if participant's formal education is 12 years or less, one point is added to his/her score. Previous studies ([@B42]; [@B26]; [@B18]; [@B49]) suggest that the MoCA is more sensitive screening tool to define the MCI as compared with other existing screening tools such as mini-mental state examination (MMSE). Statistical Analysis -------------------- We first checked whether the data are normally distributed or not using Kolmogorov--Smirnov test. The age, education, MoCA, insular thickness and net flow were negatively skewed and the systolic and diastolic blood pressures were positively skewed (asymptotic *p*-value in the range 6.17 × 10^-71^ to 8.42 × 10^-14^) at the 0.05 significance level. As data variables did not show normal distributions before and after applying the appropriate transformations (square root, log, and reciprocal), we therefore chose the non-parametric alternative. We compared the sample characteristics between the African--Americans and the Caucasian--Americans using non-parametric Wilcoxon rank sum and/or chi-square test for discrete variables (e.g., sex). Net information outflows between the nodes of three networks were compared using non-parametric Wilcoxon rank sum test. The correlation analysis was performed using Spearman's correlation. A *p*-value less than 0.05 was considered a statistically significant. MATLAB (Natick, MA, United States^[4](#fn04){ref-type="fn"}^) was used for analyzing the data. Results ======= Interactions among the Salience, Central-Executive, and Default-Mode Nodes -------------------------------------------------------------------------- We computed the Granger causality between all possible pairs of the salience, central-executive, and default-mode nodes and calculated the net information outflow from each node. Our net outflow calculation showed that the nodes of salience network have significantly lower outflow than that of the central-executive and default-mode nodes (**Figure [2](#F2){ref-type="fig"}**) (Wilcoxon rank sum; *Z* = 10.52; *p* \< 0.05). This negative information outflow of salience network nodes compared to previously reported positive flow in healthy individuals implied the impaired salience network nodes in our overall cohort. We compared the net information outflow between African Americans and Caucasian--Americans and found that the insula of salience network has significantly lower (negative value) in the African--Americans than in the Caucasian--Americans (Wilcoxon rank sum; *Z* = 2.06; *p* \< 0.05) as shown in **Figure [3](#F3){ref-type="fig"}**. ![Net flow (information outflow minus information inflow) of the key nodes of the salience (IN, insula; DACC, dorsal anterior cingulate cortex), central-executive (DLPFC, dorsolateral prefrontal cortex; PPC, posterior parietal cortex) and default-mode (VMPFC, ventromedial prefrontal cortex; PPC, posterior cingulate cortex) networks. The AI and DACC of the salience network had a significantly lower net information flow compared with the central-executive and default-mode nodes (^∗^indicates statistical significance).](fnagi-09-00177-g002){#F2} ![African--Americans (AA) versus Caucasian--Americans (CA) comparison: net flow (out-in) of the key nodes of the salience (IN, insula; DACC, dorsal anterior cingulate cortex), central-executive (DLPFC, dorsolateral prefrontal cortex; PPC, posterior parietal cortex) and default-mode (VMPFC, ventromedial prefrontal cortex; PPC, posterior cingulate cortex) networks. The AI of AA group had a significantly lower net flow (out-in) compared with that of CA group (^∗^indicates statistical significance; ns indicates not significant).](fnagi-09-00177-g003){#F3} Cortical Thickness Comparison and Correlation --------------------------------------------- To evaluate whether structural alterations explain these connectivity differences, we measured the insular cortical thickness and compared the values between African--Americans and Caucasian--Americans as displayed in **Figure [4](#F4){ref-type="fig"}**. Cortical thickness in the African--Americans (median = 2.797 mm) and in the Caucasian--Americans (median = 2.897 mm) was significantly different (Wilcoxon rank sum; *Z* = 2.09; *p* \< 0.05). To investigate how the thickness of impaired insular cortex relates with the performance on global cognitive testing, we performed a correlation analysis between the thickness and the MoCA scores. We found that a lower thickness was associated with lower performance reflected by MoCA scores (*r* = 0.30; *p* \< 0.05). These results are provided in **Figure [5](#F5){ref-type="fig"}**. Furthermore, we found that the lower cortical thickness was correlated with the lower net information outflow at the insular cortex (*r* = 0.31; *p* \< 0.05) as shown in **Figure [6](#F6){ref-type="fig"}**. ![Comparison of insular thickness between African--Americans (AA) and Caucasian--Americans (CA). The thickness had a significantly lower value in AA group than that of CA group (^∗^indicates statistical significance).](fnagi-09-00177-g004){#F4} ![The correlation between the thickness of insula and the Montreal cognitive assessment (MoCA) score in **(A)** both African--Americans (AA) and Caucasian--Americans (CA) (*r* = 0.30 and *p* \< 0.05), **(B)** AA (*r* = 0.19 and *p* = 0.18), and **(C)** CA (*r* = 0.40 and *p* \< 0.05).](fnagi-09-00177-g005){#F5} ![The correlation between the thickness and the net flow (out-in) of insula in both African--Americans (AA) and Caucasian--Americans (CA) (*r* = 0.31 and *p* \< 0.05).](fnagi-09-00177-g006){#F6} Discussion ========== Here, we investigated the pattern of connectivity among the key brain areas of the salience, central-executive, and default-mode networks in hypertensive individuals with MCI and we compared African--Americans to Caucasian--Americans within this group. We found larger impairment in the control signal of the insula (of salience network) in the African Americans than in the Caucasian--Americans, as reflected by negative net information outflow metrics measured by Granger causality analysis. Although, there are very limited neuroimaging studies that investigate the neural bases of hypertension in cognitive decline ([@B35]), our findings of greater information flow impairment in the African--Americans were in line with the previous non-neuroimaging reports that the African--Americans bear a greater risk of hypertension-associated cognitive impairment than the Caucasian--Americans ([@B47]; [@B27]). We further examined the cortical thickness of impaired insula between the African--Americans and the Caucasian--Americans and found that the insula thickness of the African--Americans is significantly lower than that of the Caucasian--Americans. The insula thickness was found to be correlated with the behavior performance and with the net information flow at the insula cortex, respectively. Those results about insula thickness were also consistent with the existing literature that insula thickness decreases with cognitive decline ([@B28]; [@B41]), however the difference in insula thickness between the African--Americans and the Caucasian--Americans has not been previously explored. Former studies consistently reported that the nodes of salience network render modulation effects over the default-mode and central-executive in healthy individuals ([@B51]; [@B25]; [@B8]). The controlling role of salience nodes over the other two networks has been argued to be structurally supported by direct white matter connections between the insula and the dorsal anterior cingulate cortex ([@B5]; [@B31]) and by their unique sharing of cytoarchitecture at neuronal level, i.e., only these regions consist of special type of neurons---von Economo neurons---that relay information processed within those nodes to other nodes, including the default-mode and executive nodes ([@B2], [@B1]; [@B62]; [@B51]). Literature also shows that the insula of salience network is functionally connected to the central-executive network ([@B61]), and has direct white matter connections to the other areas, including the inferior parietal lobe ([@B58]), and temporo-parietal junction ([@B34]). These structural and functional settings show the great involvement of insula in many cognitive processes such as in the evaluation of task performance across varying perceptual and response demands ([@B58]), the reorientation of attention in conscious error perception (or error awareness) ([@B59]), and the switching between available cognitive resources to integrate external sensory information with internal states ([@B57]). The dorsal anterior cingulate cortex of salience network is known for enhanced cognitive control ([@B19]) such as in switching activity in association with the insula during behaviorally harder tasks ([@B8]). The above mentioned neural basis of control signal of the insula and the dorsal anterior cingulate cortex network (salience nodes) has been suggested to be crucial for cognitive maintenance in both task and resting states in cognitively healthy individuals, whereas impairment to such control activity might be caused by the underlying neuroanatomical changes, including the injuries to the highly sensitive/vulnerable von Economo neurons ([@B2], [@B1]; [@B62]; [@B51]; [@B5]). Emerging evidence suggest atypical engagement of the insula of salience network in disease, including frontotemporal dementia, autism, schizophrenia, and Alzheimer's disease ([@B40]; [@B56]). The structural changes of the cortex, including the cortical thinning of insula with disease progression to MCI and/or Alzheimer's disease, is widely reported in elderly people ([@B50]; [@B28]; [@B41]). Previous studies also suggested the link between the underlying structural changes and the corresponding functional changes of brain nodes and networks ([@B63]; [@B40]). Our connectivity findings indicated that the control mechanism of the salience network was impaired (negative value) in both the African--Americans and the Caucasian--Americans, and the insula was more impaired in the African--Americans than in the Caucasian--Americans. Although, the cortical thinning of insula has been consistently reported in cognitive impairment ([@B28]; [@B41]), there are no neuroimaging studies so far to our knowledge that report the racial disparity of insula thickness. Neuroimaging literature suggests that the insula and dorsal anterior cingulate cortex---the key regions of salience network---respond as the racially biased brain regions ([@B7]), especially the greater activity of insula to out-group race than in-group race ([@B37]; [@B38]), but there are no previous reports about the difference in connectivity patterns of those regions between the African--Americans and the Caucasian--Americans groups with hypertension and cognitive impairment. Prior non-neuroimaging studies repeatedly report that the African--Americans bear the greater risk of hypertension-associated cognitive impairment than the Caucasian--Americans ([@B47]; [@B27]). Thus, our findings and existing neuroimaging/non-neuroimaging evidence taken together suggest that the insula is crucial in racial disparity in cognitively impaired individuals with hypertension. This study can be further extended in the future by including the groups of African--Americans and Caucasian--Americans with MCI but without a history of hypertension and by also including the cognitively healthy hypertensive and normotensive individuals to better dissociate the effect of hypertension alone in the brain nodes and networks. In summary, we evaluated the patterns of interactions among the salience, default-mode, and central-executive nodes in the African--Americans and the Caucasian--Americans race groups, who had hypertension and cognitive impairment. We found that the insula of the salience network was functionally impaired greater, and had lower thickness in the African--Americans than in the Caucasian--Americans. Existing literature and our findings taken together thus suggest that the insula is potential biomarker in the cognitive disorders, including the racial disparity of cognitively impaired hypertensive population. It is worth nothing that the future research should direct toward dissociating the role of hypertension alone between the race groups in the insula and the other possible regions that are functionally and/or structurally connected to the insula. Author Contributions ==================== Conceived and designed the experiment: IH, DQ. Performed the experiment: GC, JW, DQ, IH. Analyzed the data: GC, DQ, IH. Wrote the paper: GC, DQ, IH. Participated in the discussion and provided the comments: GC, JW, DQ, IH. Conflict of Interest Statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer AM and handling Editor declared their shared affiliation, and the handling Editor states that the process nevertheless met the standards of a fair and objective review. **Funding.** This research was conducted with NIA/NIH grants RF1AG051633 and R01AG042127 to IH. DQ is supported by NIH grants AG25688, AG42127, AG49752, AG51633, and has received research support from Medtronic and Siemens Medical Solutions. We would like to thank the members of our team and the participants who volunteered for the study. [www.fil.ion.ucl.ac.uk/spm/software/spm12](http://www.fil.ion.ucl.ac.uk/spm/software/spm12) <http://marsbar.sourceforge.net> <https://surfer.nmr.mgh.harvard.edu/> <https://www.mathworks.com> [^1]: Edited by: *Ana B. Vivas, CITY College, International Faculty of the University of Sheffield, Greece* [^2]: Reviewed by: *Anna Emmanouel, CITY College, International Faculty of the University of Sheffield, Greece; Luis J. Fuentes, Universidad de Murcia, Spain* [^3]: ^†^*These authors have contributed equally to this work.*

{ "pile_set_name": "PubMed Central" }

All relevant data are within the paper and its Supporting Information files. Introduction {#sec005} ============ Breast cancer is by far the most commonly diagnosed malignancy and second leading cause of cancer death in women globally. Estimated 232,670 new cases and 40,000 deaths occurred in the United States in 2014 according to American Cancer Society \[[@pone.0143537.ref001]\], which places heavy a burden on both patients and healthcare system. Owing to the breakthrough in the treatments of breast cancer, including surgical resection, adjuvant chemotherapy, radiotherapy, endocrine therapy and targeted therapy, the overall survival improved and tumor recurrence reduced during the last decades \[[@pone.0143537.ref002]\]. However, recurrence and metastasis remain the main challenge for cure \[[@pone.0143537.ref003]\]. Up to now, useful prognostic factors can identify groups of patients with various prognoses. Clinical and pathological parameters have been used as prognostic factors to predict the outcome of breast cancer patients and help to develop appropriate adjuvant treatments for high-risk patients. The American Joint Committee on Cancer (AJCC) tumor-node-metastasis (TNM) staging system and the molecular subtypes classification are important prognostic predictors \[[@pone.0143537.ref004],[@pone.0143537.ref005]\]. Clinically, the breast cancer is classified into several molecular subtypes based on the estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor-2 (HER2) and Ki-67 index status. Patients with HER2 over-expressing subtype (ER-, PR- and HER2+) and triple-negative subtype (ER-, PR- and HER2-) suffer more recurrence, distant metastasis and drug resistance than the luminal subtype \[[@pone.0143537.ref006]\]. In addition to the tumor stage and molecular subtypes, various clinical tools were developed to predict the prognosis of breast cancer patients, such as 21-gene RT-PCR assay recurrence score, Nottingham Prognostic Index (NPI) and Adjuvant Online (Adjuvant!)\[[@pone.0143537.ref007]\]. The NPI combines three prognostic factors: nodal status, tumor size and histological grade \[[@pone.0143537.ref008]\]. Since the three components of the NPI formula reflect the metastatic potential and genetic instability of tumor cells, higher NPI generally indicates worse clinical outcomes. Patients were classified into four subsets with different probability of cancer-related death: excellent, good, moderate and poor groups \[[@pone.0143537.ref009]\]. Recent studies had investigated the association between laboratory parameters and the cancers. Being members of blood lipid profiles, total cholesterol level and triglyceride level were significantly higher in breast cancer patients \[[@pone.0143537.ref010],[@pone.0143537.ref011]\]. Elevated cholesterol level and triglyceride level were risk factors for poor prognosis in lung cancer and prostate cancer \[[@pone.0143537.ref012],[@pone.0143537.ref013]\]. In addition, previous studies demonstrated that fibrinogen participated in the tumorigenesis and progression \[[@pone.0143537.ref014],[@pone.0143537.ref015]\] and elevated serum fibrinogen level predict poor clinical outcomes in several cancers \[[@pone.0143537.ref016],[@pone.0143537.ref017],[@pone.0143537.ref018]\]. In this study, we investigated the laboratory parameters routinely tested prior to the surgical treatment and set up a prognostic index model based on the independent prognostic factors. The prognostic model was assessed and compared with the NPI to estimate the clinical value in breast cancer patients. Patients and Methods {#sec006} ==================== Study population {#sec007} ---------------- We retrospectively reviewed the medical records of consecutive patients with primary breast cancer from January 1, 2002 to December 31, 2008 in Sun Yat-sen University Cancer Center (SYSUCC). Pathological diagnosis was carefully confirmed by pathologists. Other inclusion criteria included: (1) received surgical treatment; (2) female; (3) diagnosed as invasive ductal carcinoma or invasive lobular carcinoma. We excluded patients who (1) received neoadjuvant chemotherapy before surgery; (2) had surgical treatment before admission; (3) with previous or coexisting cancers other than breast cancer; (4) confirmed metastasis. All patients were followed up until December 31, 2014 or date of cancer-related deaths. Routine tests and telephone counseling were performed to understand the patients' condition during follow-up time. Clinical data collection {#sec008} ------------------------ Patient characteristics and clinicopathological factors like age, menopausal status, pathological diagnosis, tumor size, axillary lymph nodal status, histologic grade, hormonal receptor and HER2 status, date of last follow-up or cancer-related death were collected. Laboratory parameters investigated as potential prognostic factors in the current study included albumin, globulin, lactate dehydrogenase (LDH), total bilirubin (TB), uric acid, cholesterol, triglyceride and fibrinogen. The blood samples were collected and measured by the automatic biochemical analyzer in the central laboratory before surgery. The tumor stages were classified according to the AJCC TNM staging system (the 7th edition). The molecular subtypes were as follow: Luminal A (ER +, PR +, HER2 -and Ki-67≤ 14%), Luminal B (ER + and HER2 + or Ki-67\> 14%), HER2 over-expressing (ER -, PR -, HER2 +) and triple-negativer breast cancer (ER -, PR -, HER2 -). HER2 positive was defined as "3+" in immunohistochemical test or "positive" in HER2 fluorescence in situ hybridization test. The NPI was calculated by the formula of lymph node stage (1--3) + histologic grade (1--3) + 0.2×tumor size (cm) and patients were stratified into four groups as excellent (≤ 2.40), good (2.41--3.40), moderate (3.41--5.40) and poor (\>5.4)\[[@pone.0143537.ref008],[@pone.0143537.ref009]\] Statistical analyses {#sec009} -------------------- The primary end point was cancer-specific survival (CSS) calculated from the date of diagnosis to the date of cancer-related deaths or the last follow-up. Means and standard deviations were calculated, and differences were identified by *t* test. Difference between categories were evaluated using chi-square test. Receiver operating characteristics (ROC) curve analysis were performed to assess the prognostic value of each laboratory parameter and to set up the optimal cut-off points for potential prognostic factors. Univariate analysis and multivariate analysis (Cox regression model) were used to confirm the independent prognostic variables associated with CSS. The prognostic index (PI) model was established based on independent variables and enrolled patients were stratified into low-, moderate- and high-risk groups. Kaplan---Meier method was performed for survival analysis and compared by log-rank test. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated from the Cox regression model, and a two-tailed *p* value \<0.05 was considered statistically significant. The comparison between PI and NPI was made using the method of DeLonget al \[[@pone.0143537.ref019]\]. All statistical analyses were performed using SPSS 19.0 (SPSS Inc., Chicago, IL, USA). Ethics Statement {#sec010} ---------------- The study protocol was approved the by independent ethical committee/institutional review board of Sun Yat-sen University Cancer Center (SYSUCC), and written informed consent about the scientific research was obtained from each participant prior to surgery. Patient records were anonymized and de-identified prior to analysis. Results {#sec011} ======= There were 1661 female patients with primary non-metastatic invasive breast cancer enrolled in this study. The mean survival time of all participants was 123.6 months and patient characteristics were shown in [Table 1](#pone.0143537.t001){ref-type="table"}. The mean age was 42.2 years old (range 22--74 years), and 150 (9.0%) patients were under the age of 35. Invasive ductal carcinoma was the predominant type (97.2%). Tumor sizes of T1, T2 and T3 were observed in 698 (42.0%), 876 (52.7%) and 87 (5.2%) of the patients respectively, and 893 (53.8%) patients suffered regional lymph node metastasis. 129 (7.7%), 374 (22.5%), 836 (50.4%) and 322 (19.4%) were classified as excellent, good, moderate and poor NPI groups respectively. Luminal subtype comprised 77.4% of total participants, and 183 (11.0%) and 193 (11.6%) were HER2 over-expressing subtype and triple-negative subtype respectively. 89.3% (1484/1661) of enrolled participants received chemotherapies and all were performed within 45 days after the surgeries. The anthracycline and/or taxane-containing chemotherapy was the main option for adjuvant cytotoxic chemotherapy. Radiotherapies were given sequentially to the patients with more than 3 metastatic lymph nodes or tumor \>50 mm in greatest dimension or breast conserving surgery. Patients with ER/PR positivity received endocrine therapies for more than 4 years. Target therapy was not performed due to the unavailability of trastuzumab at that time. 10.1371/journal.pone.0143537.t001 ###### Clinicopathological characteristics and laboratory parameters of patients. ![](pone.0143537.t001){#pone.0143537.t001g} Characteristic All patients (n = 1661) Low-risk group (n = 1168) Moderate-risk group (n = 410) High-risk group (n = 83) *p* value --------------------------------- ------------------------- --------------------------- ------------------------------- -------------------------- ----------- Age \<0.001 ≤ 35 150(9.0) 126(84.0) 24(16.0) 0(0) \> 35 1511(91.0) 1042(69.0) 386(25.5) 83(5.5) Menopause \<0.001 Yes 433(26.1) 264(61.0) 134(30.9) 35(8.1) No 1228(73.9) 904(73.6) 276(22.5) 48(3.9) Tumor type 0.481 IDC 1615(97.2) 1136(70.3) 400(24.8) 79(4.9) ILC 46(2.8) 32(69.6) 10(21.7) 4(8.7) Histologic grade 0.886 G1 562(33.8) 396(70.5) 135(24.0) 31(5.5) G2 664(40.0) 465(70.0) 165(24.8) 34(5.1) G3 435(26.2) 307(70.6) 110(25.3) 18(4.1) Tumor size 0.805 T1 698(42.0) 799(71.5) 163(23.4) 36(5.2) T2 876(52.7) 607(69.3) 227(25.9) 42(4.8) T3 87(5.2) 62(71.3) 20(23.0) 5(5.7) Lymph node status 0.127 N0 768(46.2) 545(71.0) 188(24.5) 35(4.6) N1 448(27.0) 327(73.0) 98(21.9) 23(5.1) N2 259(15.6) 179(69.1) 70(29.1) 10(3.9) N3 186(11.2) 117(32.9) 54(29.0) 15(8.1) ER 0.550 Positive 1047(63.0) 743(71.0) 256(24.5) 48(4.6) Negative 614(37.0) 425(69.2) 154(25.1) 35(5.7) PR 0.123 Positive 1162(70.0) 831(71.5) 280(24.1) 51(4.4) Negative 449(30.0) 337(67.5) 130(26.1) 32(6.4) HER-2 0.042 Positive 393(23.7) 889(70.1) 324(25.6) 55(4.3) Negative 1268(76.3) 279(71.0) 86(21.9) 28(7.1) NPI 0.652 Excellent group 129(7.7) 91(70.5) 32(24.8) 6(4.7) Good group 374(22.5) 269(71.9) 86(23.0) 19(5.1) Moderate group 836(50.4) 593(70.9) 199(23.8) 44(5.3) Poor group 322(19.4) 215(66.8) 93(28.9) 14(4.3) Cancer-specific survival \<0.001 Alive 1365(82.2) 992(72.7) 319(23.4) 54(4.0) Death 296(17.8) 176(59.5) 91(30.7) 29(9.8) Albumin level (g/L,mean±SD) 43.8±4.1 Globulin level (g/L,mean±SD) 28.3±4.5 LDH (U/L,mean±SD) 161.2±55.0 TB (umol/L, mean+ SD) 13.5±9.3 Uric acid (umol/L, mean±SD) 283.0±80.4 Cholesterol (mmol/L,mean±SD) 5.09±1.05 Triglyceride (mmol/L,mean±SD) 1.37±0.94 Plasma fibrinogen(g/L, mean±SD) 2.95±0.71 Abbreviation: SD *standard deviation*, IDC *Invasive ductal carcinoma*, ILC *Invasive lobula carcinomar*, ER *Estrogen receptor*,PR *Progesterone receptor*, HER2 *Human epidermal growth factor receptor-2*, LDH *lactate dehydrogenase*, TB *total bilirubin* ROC curve analysis was performed and the AUCs were shown in [S1 Table](#pone.0143537.s001){ref-type="supplementary-material"}. The AUCs for albumin, globulin, LDH, triglyceride and fibrinogen are all statistically significant (all *p* \< 0.05), and potential prognostic effects of total bilirubin and uric acid were not proved (*p* = 0.311 and 0.326 respectively). The optimal cut-off points for each potential prognostic factors were set up with the highest Youden index. Enrolled patients were stratified into high- or low-level by various variables. In the multivariate analysis, globulin (\>30.0g/L), triglyceride (\>1.10mmol/L) and fibrinogen (\>2.83g/L) were identified as independent prognostic factors for poor cancer-specific survival in breast cancer patients (all *p*\< 0.05, [Fig 1](#pone.0143537.g001){ref-type="fig"}). ![Cancer-specific survival of patients according to globulin level (A), triglyceride level (B) and fibrinogen level (C).](pone.0143537.g001){#pone.0143537.g001} The preoperative globulin, triglyceride and fibrinogen identified as independent prognostic factors in multivariate analysis were used to establish the prognostic index model (PI). The criteria of prognostic index model were as follows: globulin level \>30.0g/L, triglyceride level \>1.10mmol/L and fibrinogen level \>2.83g/L were allocated 1 point each; globulin level ≤30.0g/L, triglyceride level ≤1.10mmol/L and fibrinogen level ≤2.83g/L were allocated 0 point each. The total score ranging from 0 to 3 was categorized into three prognostic index risk groups defined as: low-risk group, 0 or 1 point; moderate-risk group, 2 point; high-risk group, 3 point. There were 1168 (70.3%) patients allocated to low-risk group, 410 (24.7%) allocated to moderate-risk group and 83(5.0%) patients were categorized as high-risk group. The mean cancer-specific survival time for the low-risk group was 126.8 months (95% CI 124.5--129.2), which was significantly longer than that of 115.5 months (95% CI 111.1--112.0) in moderate-risk group and 99.7 months (95% CI 121.5--125.7) in high-risk group (both *p*\< 0.05). The estimated 10-years survival rates for low-risk group, moderate-risk group and high-risk group were 80.5%, 72.0% and 58.8% respectively ([Fig 2](#pone.0143537.g002){ref-type="fig"}). Univariate analysis and multivariate analysis verified that the PI was significantly associated with CSS (HR = 1.513 for moderate-risk group and 2.481 for high-risk group, both *p*\<0.05). In addition, menopausal status, tumor size, lymph node status, ER and HER2 status were identified as independent factors associated with CSS (all *p*\< 0.05, [Table 2](#pone.0143537.t002){ref-type="table"}). ![Cancer-specific survival of patients by prognostic index risk group.](pone.0143537.g002){#pone.0143537.g002} 10.1371/journal.pone.0143537.t002 ###### Univariate and multivariate analysis of PI for CSS in breast cancer. ![](pone.0143537.t002){#pone.0143537.t002g} Variable Univariate analysis multivariate analysis ---------------------- --------------------- ----------------------- --------- ------- --------------- --------- Age 0.733 0.511--1.053 0.093 0.765 0.521--1.123 0.172 ≤ 35 1 *(reference)* 1 *(reference)* \> 35 0.733 0.511--1.053 0.093 0.765 0.521--1.123 0.172 Menopausal status Yes 1 *(reference)* 1 *(reference)* No 0.755 0.591--0.965 0.025 0.690 0.531--0.896 0.005 Tumor type 0.813 0.384--1.720 0.588 0.742 0.642--1.608 0.449 Histologic grade G1 1 *(reference)* 1 *(reference)* G2 0.711 0.536--0.943 0.018 0.781 0.522--1.170 0.231 G3 1.206 0.915--1.589 0.184 1.097 0.637--1.888 0.739 Tumor size T1 1 *(reference)* 1 *(reference)* T2 1.429 1.110--1.840 0.006 1.100 0.792--1.527 0.570 T3 3.830 2.620--5.601 \<0.001 2.095 1.280--3.429 0.003 Lymph node status N0 1 *(reference)* 1 *(reference)* N1 2.043 1.462--2.854 \<0.001 1.966 1.258--3.075 0.003 N2 3.646 2.600--5.111 \<0.001 3.484 1.907--6.366 \<0.001 N3 7.610 5.508--10.516 \<0.001 6.080 3.410--10.842 \<0.001 ER Negative 1 *(reference)* 1 *(reference)* Positive 0.577 0.460--0.725 \<0.001 0.681 0.514--0.904 0.008 PR Negative 1 *(reference)* 1 *(reference)* Positive 0.612 0.485--0.773 \<0.001 0.902 0.685--1.187 0.461 HER2 Negative 1 *(reference)* 1 *(reference)* Positive 1.961 1.544--2.490 \<0.001 1.335 1.018--1.751 0.036 PI Low-risk group 1*(reference)* 1*(reference)* Moderate-risk group 1.537 1.193--1.980 0.001 1.513 1.169--1.959 0.002 High-risk group 2.696 1.820--3.994 \<0.001 2.481 1.653--3.724 \<0.001 Abbreviation: CCS *Cancer-specific survival*, CI *confidence interval*, ER *Estrogen receptor*,PR *Progesterone receptor*, HER2 *Human epidermal growth factor receptor-2*, PI *prognostic index* Under stratifications based on the molecular subtypes, the PI was independently associated with CSS in patients with Luminal A, Luminal B subtypes and triple-negative breast cancer (all *p*\< 0.05, [Table 3](#pone.0143537.t003){ref-type="table"}). However, no significant difference was observed in HER2 over-expressing among different PI groups. 10.1371/journal.pone.0143537.t003 ###### Multivariate analysis of PI and NPI based on various molecular subtypes. ![](pone.0143537.t003){#pone.0143537.t003g} Molecular subtypes PI NPI[^2^](#t003fn002){ref-type="table-fn"} ------------------------------- ---------------------- ------------------------------------------- ---------------------- --------- Luminal A 1.583 (1.135--2.207) 0.007 2.491 (1.859--3.338) \<0.001 Luminal B 1.885 (1.431--2.482) \<0.001 2.035 (1.557--2.660) \<0.001 HER2 over-expressing 1.352 (0.906--2.015) 0.139 1.260 (0.886--1.834) 0.227 Triple-negative breast cancer 1.579 (1.076--2.318) 0.020 1.631 (1.074--2.479) 0.022 All patients 1.570 (1.316--1.872) \<0.001 2.019 (1.724--2.365) \<0.001 ^1^PI and NPI were processed as a continuous variable ^2^tumor size and lymph node status were in included in NPI multivariate analysis Abbreviation: PI *prognostic index model*, NPI *Nottingham Prognostic Index*, HR *Hazard ratio* CI *confidence interval*, HER2 *human epidermal growth factor receptor-2* As showed in [Table 3](#pone.0143537.t003){ref-type="table"}, NPI was significantly predictive of CSS in all patients and in patients with Luminal A, Luminal B subtypes and triple-negative breast cancer, which was similar with the prognostic implication of PI. The prediction efficiency of PI was compared with that of NPI. The AUC of PI was 0.571 (95%CI:0.534--0.609) and that of NPI was 0.643 (95%CI:0.609--0.677). The difference between the two prognostic models was statistically significant (Z = 3.279, *p* = 0.001). Discussion {#sec012} ========== Breast cancer is the most commonly diagnosed cancers among women in the United States \[[@pone.0143537.ref001]\]. Despite newer therapies in the recent years, recurrence and metastasis remains the main challenge to the cancer management. About 30% to 40% of invasive breast cancer patients will eventually develop into metastatic breast cancer \[[@pone.0143537.ref020],[@pone.0143537.ref021]\], who suffer low 5-year survival rate as 24%\[[@pone.0143537.ref022]\]. Generally, risk factors for poor prognosis include large tumor size, positive lymph node status, negative hormone receptor status and HER2 overexpression \[[@pone.0143537.ref023]\]. Moreover, previous studies have reported several factors related to poor outcomes of breast cancer patients, such as albumin, LDH \[[@pone.0143537.ref024]\], bilirubin and cholesterol \[[@pone.0143537.ref025]\]. In the present study, we retrospectively analyzed the preoperative parameters in breast cancer patients and identified globulin (\>30.0g/L), triglyceride (\>1.10mmol/L) and fibrinogen (\>2.83g/L) as independent unfavourable prognostic factors for CSS. The possible mechanisms of the above factors in tumorigenesis and progression were undefined. Globulins, generally including immunoglobulins and other acute-phase proteins such C-reactive protein (CRP), serum amyloid A, are the major component of serum proteins. An increased globulin level has been reported to reflect the presence of continuous systemic inflammation \[[@pone.0143537.ref026]\], which induces the increased levels of cytokines and promotes the tumor proliferation, progression, invasion and metastasis \[[@pone.0143537.ref027],[@pone.0143537.ref028]\]. It was found that elevated alpha and gamma globulins were linked with poor prognosis in lung cancer and colorectal cancer patients \[[@pone.0143537.ref029],[@pone.0143537.ref030]\]. Alpha-1 globulin could inhibit natural killer-cell activity and T cell-mediated cytotoxicity by suppressing cytotoxic reactions of lymphocytes \[[@pone.0143537.ref031]\]. Furthermore, COOH-terminal fragment, the degradation product of alpha-1 globulin by matrix metalloproteinase, could improve the tumor growth and invasion potential in vivo \[[@pone.0143537.ref032]\]. These biological behaviors regulate host anti-tumour defense mechanisms and promote tumor development \[[@pone.0143537.ref033]\]. Several studies have found a significant association between elevated serum triglyceride level and risk of breast cancer development \[[@pone.0143537.ref034],[@pone.0143537.ref035]\]. Besides the active glycolysis, the lipogenesis is greatly increased in tumor cells \[[@pone.0143537.ref036]\]. The triglyceride/free fatty acid (TG/FFA) cycling plays an important role in multiple signaling pathways \[[@pone.0143537.ref037]\]. The TG/FFA cycling may promote cell survival through the activation of NF-κB and thus improve the expression of anti-apoptotic protein Bcl-2 and Bcl-x \[[@pone.0143537.ref038]\]. Hypoxic condition, which commonly occurred in the centre of solid tumors, could cause an elevation of triglyceride and expression of hypoxia inducible factor (HIF) that promotes metastasis of tumor \[[@pone.0143537.ref039],[@pone.0143537.ref040]\]. The prognostic association between fibrinogen and cancer was reported by several studies. Fibrinogen can deposited around solid tumors and act as a stable framework to combine growth factors, such as fibroblast growth factor-2 (FGF-2) and vascular endothelial growth factor (VEGF), with tumor cells to increase the tumor proliferation and stimulate angiogenesis \[[@pone.0143537.ref014],[@pone.0143537.ref041]\]. Moreover, fibrinogen can improve the adhesion of tumor cells to platelets, protecting tumor cells from the innate immune system and leading to an increase of metastatic cells \[[@pone.0143537.ref042]\]. Based on these three factors, we devised a Prognostic Index (PI) model and classified the enrolled patients into low-, moderate- and high-risk groups. Independently of the tumor size, lymph node status, hormonal receptor status and HER2 status, the PI was significantly associated with the overall postoperative survival, indicating that patients with higher PI experienced poorer prognosis. Moreover, after stratification by molecular subtypes, preoperative PI remained significantly prognostic in Luminal A, Luminal B subtypes and triple-negative breast cancer. Generally, breast cancer patients with luminal subtype, especially Luminal A subtype, had relatively better clinical outcomes. Based on the preoperative PI, clinicians could identify patients with high risk of poor prognosis, and additional adjuvant treatment might be suggested beyond endocrine treatment. Moreover, both the PI and NPI could not effectively predict the prognosis in HER2 over-expressing subtype. That may be due to the carcinogenesis and the proliferation promotion effect of HER2 protein \[[@pone.0143537.ref043],[@pone.0143537.ref044]\], which induce the unique characteristics of HER2 over-expressing breast cancer subtype. The NPI combines three prognostic factors: nodal status, tumor size and histological grade, and the former two were found to be independently associated with survival in the present study. Patients with different NPI were considered to suffer distinguished prognosis. In the current study, the discriminatory ability of PI was compared with NPI, and the AUC value of NPI was significantly higher than that of PI (*p*\< 0.05). Axillary lymph node status is the most important prognostic factor in the staging of the breast cancer patient \[[@pone.0143537.ref045],[@pone.0143537.ref046]\]. The occurrence of axillary nodal metastases has been found to significantly decrease the 5-year survival rate up to 40%\[[@pone.0143537.ref047]\]. In the multivariate analysis, the hazard ratio of lymph nodal status was higher than that of PI, which may explain better discriminatory ability of NPI. Thus PI may act as complement to further improve the discriminatory ability of NPI. The present study is limited to the retrospective nature. Firstly, selection bias cannot be excluded even though consecutive patients were included and eligibility criteria were performed to minimize the bias. Secondly, specific quality control analysis was performed by different quality inspectors since the laboratory parameters were obtained as routine clinical tests before surgery. To sum up, our study identified three independent prognostic factors, globulin, triglyceride and fibrinogen, and set up a prognostic index (PI) based on these factors to stratify breast cancer patients into low, moderate and high risk of poor prognosis. These parameters are low-cost and routinely measured in clinical practice. Thus the PI may assist clinicians to identify high-risk patients and make personalized therapeutic approaches independently or along with other prognostic models. Supporting Information {#sec013} ====================== ###### The optimal cut-off points and multivariate analyses of potential prognostic parameters. (DOC) ###### Click here for additional data file. ###### Relevant data underlying the findings described in manuscript. (XLS) ###### Click here for additional data file. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: JHW XMX. Performed the experiments: FY SJL. Analyzed the data: XJH LY. Contributed reagents/materials/analysis tools: XJH XSX. Wrote the paper: JHW FY. Approved the manuscript: XMX.

{ "pile_set_name": "PubMed Central" }

1. Introduction {#sec1-ijerph-17-05150} =============== According to the World Health Organization (WHO), the number of people over the age of 60 will reach 2 billion by 2050, which is about a fifth of the population worldwide (22%) \[[@B1-ijerph-17-05150]\]. As a country with the largest population, China has the largest number of older adults. By the end of 2018, there were 249 million (17.9%) and 167 million (11.9%) people over the age of 60 and 65, respectively \[[@B2-ijerph-17-05150]\]. However, the aging process in China is so rapid that the social system cannot keep up \[[@B1-ijerph-17-05150],[@B3-ijerph-17-05150]\]. In order to address the aging crisis and ensure older adults' right to live a long and healthy life, "healthy aging" was put forward by WHO and the Healthy China 2030 Strategy, which aimed to create an age-friendly environment through the cooperation of the whole society \[[@B4-ijerph-17-05150],[@B5-ijerph-17-05150]\]. Older adults are in a critical stage of maintaining their health-related quality of life (HRQOL), where the leading health threat is noncommunicable diseases (NCDs) \[[@B6-ijerph-17-05150]\], mainly including cardio-cerebrovascular diseases, cancers, chronic respiratory diseases and diabetes \[[@B7-ijerph-17-05150]\]. Thus, the prevention of NCDs and the maintenance of HRQOL are major difficulties for the achivement of healthy aging. Previous studies have focused on the association between one specific NCD and HRQOL, such as hypertension \[[@B8-ijerph-17-05150]\], type 2 diabetes \[[@B9-ijerph-17-05150]\] and chronic kidney disease \[[@B10-ijerph-17-05150]\]. However, the results were not consistent, since most of these studies were cross-sectional studies. A lot of studies showed a negative effect of one specific NCD on HRQOL, while others did not find a correlation \[[@B8-ijerph-17-05150],[@B9-ijerph-17-05150],[@B10-ijerph-17-05150],[@B11-ijerph-17-05150]\]. Additionally, most studies have ignored the influence of the number of NCDs, while older adults usually have more than one NCD, referred to as multimorbidity \[[@B12-ijerph-17-05150]\]. Some international researchers investigated the HRQOL of people suffering from multiple NCDs in their countries and found a negative relationship between HRQOL and multimorbidity \[[@B13-ijerph-17-05150],[@B14-ijerph-17-05150]\]. However, few studies focused on older adults, especially Chinese older adults \[[@B15-ijerph-17-05150],[@B16-ijerph-17-05150]\]. In fact, the relationship between the HRQOL and NCDs is very complex and numerous factors could make a difference. For instance, a great number of NCDs are due to unhealthy behaviors, especially poor diets \[[@B7-ijerph-17-05150]\]. A few researchers proposed the "modernization theory" to explain the rise of unhealthy diets \[[@B17-ijerph-17-05150]\]. The theory illustrated burgeoning middle classes with more disposable income tended to consume more meat and processed foods that had close associations with weight gain. Therefore, to some extent, it is reasonable to believe people with NCDs could have a good economic status. On the other hand, social demographic determinants exert a tremendous effect on HRQOL \[[@B18-ijerph-17-05150]\]. Among these factors, socioeconomic status has been proven to be an independent predictor of HRQOL \[[@B19-ijerph-17-05150]\]. People with higher socioeconomic status often have a better HRQOL \[[@B20-ijerph-17-05150]\]. Hence, it is reasonable to argue that NCDs might exert little or even no negative effect on HRQOL on account of the good economic status of people with NCDs. The relationship between HRQOL and NCDs should be further clarified through a large-scale survey. In China, due to the rapid process of industrialization and urbanization, citizens' lifestyle and disease spectrum have changed a lot \[[@B5-ijerph-17-05150]\]. Nearly 180 million Chinese older adult people suffer from NCDs, accounting for 75% of older people in China, and the prevalence of multimorbidity is about 35% \[[@B21-ijerph-17-05150]\]. Many studies on the HRQOL and NCDs among older people in China simply classified subjects according to whether they had NCDs or not, while few studies focused on the number of NCDs. Moreover, the samples of those studies were too small to be representative \[[@B15-ijerph-17-05150],[@B22-ijerph-17-05150]\]. Therefore, we conducted the cross-sectional investigation to understand the HRQOL among people with different numbers of NCDs., and to explore the relationship between them. We argue that the HRQOL of Chinese older adults might become worse with the increase in the number of NCDs. The findings are expected to provide theoretical support for the timely advance of NCD management. 2. Materials and Methods {#sec2-ijerph-17-05150} ======================== 2.1. Study Population {#sec2dot1-ijerph-17-05150} --------------------- The cross-sectional study was based on a large-scale survey named China's Health-Related Quality of Life Survey for Older Adults 2018 (CHRQLS-OA 2018) whose chief initiator and executor was the Global Health Institute of Wuhan University \[[@B23-ijerph-17-05150]\]. Interviewers also included teachers and students from other top universities in China (such as Nanchang University). All staff received professional training before the survey started. Conducted in either online or paper format between January and March of 2018, the survey recruited older adults aged 60 years and over across the country. Through convenience sampling, 5442 valid subjects were recruited. The survey mainly collected participants' individual socio-demographic characteristics, social capital, behaviors and lifestyles, health-related quality of life, mental health and coping strategies, etc. This study aims to explore the relationship between the number of NCDs and the HRQOL of Chinese older adults. In the present study, 276 subjects (5.07%) without information on NCDs and/or HRQOL were excluded. Finally, 5166 respondents (94.93%) were included in the analysis. 2.2. Description of the Measures {#sec2dot2-ijerph-17-05150} -------------------------------- ### 2.2.1. General Demographic Characteristics {#sec2dot2dot1-ijerph-17-05150} In the present study, the following general demographic characteristics of the participants were included: age, sex, nationality, body mass index (BMI), years of education, household registration, marital status, average annual household income (CNY), new personal savings and self-rated health status (self-rated health status was based on individuals' subjective feelings). The variable was divided into three groups, which were "Good", "General", and "Poor". "General" was a status between "Good" and "Poor", and was equivalent to "middle" or "moderate". These general demographic characteristics were considered as confounding factors in the analysis of the cross-sectional association between the number of NCDs and HRQOL among Chinese older adults. ### 2.2.2. Assessment of the Number of NCDs {#sec2dot2dot2-ijerph-17-05150} In "China's Health-related older adults Quality of Life Survey 2018", the number of NCDs and the time of diagnosis were self-reported. Diseases investigated included hypertension, diabetes, coronary heart disease, tumor, chronic obstructive pulmonary disease (COPD), intervertebral disc disease, asthma, rheumatoid arthritis, gastroenteritis, cataract, cerebrovascular disease and others (openly filled in). Professional medical diagnoses can reduce certain deviation, and thus the questionnaire refered to the following questions: "where and by whom was the disease diagnosed? How long has the diagnosis been made?" ### 2.2.3. Assessment of the HRQOL {#sec2dot2dot3-ijerph-17-05150} The WHO's HRQOL Research Group has further developed the World Health Organization Quality of Life-Old (WHOQOL-OLD) module for older adults based on the original QOL scale \[[@B24-ijerph-17-05150],[@B25-ijerph-17-05150]\]. The WHOQOL-OLD module can be widely used in a variety of studies concerning crucial issues of the HRQOL in old age, such as population epidemiology, health monitoring, service development and so on. The WHOQOL-OLD module includes 24 items assigned to six aspects (each of the aspects has 4 items): the "Sensory Abilities" (SAB) aspect assesses sensory capacity and the influence of loss of sensory abilities on HRQOL; the "Autonomy" (AUT) aspect means independence in old age; the "Past, Present and Future Activities" (PPF) aspect assesses satisfaction with one's achievements in life and expectations in the future; the "Social Participation" (SOP) aspect refers to participation in daily activities; the "Death and Dying" (DAD) aspect is related to concerns, worries and fears about death and dying; the "Intimacy" (INT) describes the ability to have personal and intimate relationships. The scores of each item range from 1 to 5, thus the scores of each aspect range from 4 to 20. The scores of all the six aspects can be combined into a total ("overall") score for HRQOL in old age. Essentially, high scores represent high HRQOL, low scores represent lower HRQOL. Through the evaluation of the Chinese version of the WHOQOL-OLD module, the results showed good reliability and validity, thus WHOQOL-OLD is suitable for the HRQOL evaluation of Chinese older adults \[[@B26-ijerph-17-05150]\]. 2.3. Statistical Analysis {#sec2dot3-ijerph-17-05150} ------------------------- The Statistical Package for the Social Sciences (SPSS) version 23.0 for Windows (SPSS Inc., Chicago, IL, USA) was applied to conduct all statistical analyses, with a statistical significance level of 0.05. The data analysis consisted of three aspects. First, frequencies and proportions were reported to describe the distribution of NCDs in the population with different demographic characteristics. Second, the difference in scores of all dimensions of the WHOQOL-OLD scale was examined among groups with different numbers of NCDs by ANOVA, and the Student--Newman--Keuls (SNK) test method was used for post-hoc comparisons. Third, multiple linear regression models were used to determine the relationship between the number of NCDs and the total scores and scores of each dimension of the WHOQOL-OLD scale. We established three models in the third step, including the initial model (model 1) and the adjusted models (model 2 and model 3). The final parsimonious model adjusted for confounders that were significantly associated with the dependent variables in model 2. The unstandardized coefficients (B) with a 95% confidence interval (95% CI) obtained from the models were reported. 2.4. Ethical Statements {#sec2dot4-ijerph-17-05150} ----------------------- This study was conducted in accordance with the Declaration of Helsinki, and the study protocol was reviewed and approved by the Institutional Review Board of School of Health Science and Faculty of Medical Sciences, Wuhan University (IRB number: 2019YF2050). Informed consent information was included in each questionnaire and introduced before the surveys. The questionnaires were completed by participants themselves or their caregivers according to older adults's reading and response abilities. Surveys were only conducted when subjects were fully informed of the content and aim of this research project and agreed to participate. The survey was also conducted anonymously, and respondents' information was kept confidential and only for the use of scientific research. 3. Results {#sec3-ijerph-17-05150} ========== 3.1. Descriptions of Sample Characteristics {#sec3dot1-ijerph-17-05150} ------------------------------------------- A total of 5166 older adults was included in the study. The general demographic characteristics are shown in [Table 1](#ijerph-17-05150-t001){ref-type="table"}. In general, there were 51.3% older adults with one or more NCD(s). More than half (50.9%) of men did not have NCDs, while only 46.6% of women did not have NCDs. Most people aged 60--64 years and those aged 85 years or older did not suffer from NCDs (54.0% and 55.3%, respectively). The Han population (50.8%) had a lower proportion of NCDs than minorities (58.5%). People with a normal BMI had the highest percentage of those without NCDs (50.8%). People with NCDs accounted for 58.2% of uneducated participants, which was higher than those receiving education. The proportion of non-agricultural people who did not suffer from NCDs (53.5%) was significantly higher than that of agricultural people (46.6%). People who were married or cohabiting had a lower percentage of having any NCDs (48.6%). Nearly three fifths (59.6%) of the population with an average annual household income of 45,001--60,000 did not suffer from NCDs. About two thirds (62.4%) of people who self-rated their health status as good did not suffer from NCDs. The proportion of people with new personal savings of 100,000 CNY or more who did not suffer from NCDs reached 63.1%. 3.2. Description of the Scores of the WHOQOL-OLD Scale {#sec3dot2-ijerph-17-05150} ------------------------------------------------------ As shown in [Table 2](#ijerph-17-05150-t002){ref-type="table"}, there were differences in total score and scores of the six dimensions of the WHOQOL-OLD scale among people with different numbers of NCDs, and the differences were statistically significant (*p* \< 0.001). Those with the highest scores of SAB, AUT, PPF, SOP, DAD, INT and total score were people without NCDs, while those with the lowest scores were people with two NCDs or three NCDs and above. The multiple comparisons found that as the number of NCDs increased, the scores of each dimension and the overall score gradually decreased. 3.3. Relationship between the Number of NCDs and the Scores of the WHOQOL-OLD Scale {#sec3dot3-ijerph-17-05150} ----------------------------------------------------------------------------------- As shown in [Table 3](#ijerph-17-05150-t003){ref-type="table"},multiple linear regression analysis was used to identify the relationship between the number of NCDs and the scores of the WHOQOL-OLD scale, and the model 2 and model 3 were adjusted by confounding factors. In the crude and adjusted models, the number of NCDs was related to the scores of the WHOQOL-OLD scale. According to the final model 3, with no NCDs as the reference group, having one NCD had a negative impact on the three dimensions of PPF, SOP, and DAD in the HRQOL of older adults (*p* \< 0.05). Suffering from two NCDs had negative effects on SAB, AUT, PPF, SOP, and DAD scores of older adults (*p* \< 0.05). Suffering from three or more NCDs exerted negative effects on all dimensions in the HRQOL of older adults (*p* \< 0.05). In terms of the WHOQOL-OLD total score of people with different numbers of NCDs, as compared with the reference group (with zero number of NCDs), suffering from one NCD (B = --0.87, 95% CI = --1.67 to --0.08, *p* \< 0.05), two NCDs (B = --2.89, 95% CI = --3.87 to --1.90, *p* \< 0.001), and three or more NCDs (B = --4.20, 95% CI = --5.36 to --3.05, *p* \< 0.001), all had negative impacts on the HRQOL of older adults, and as the number of NCDs increased, the HRQOL of older adults became worse, after adjusting for the confounding factors that were significantly associated with HRQOL. 4. Discussion {#sec4-ijerph-17-05150} ============= In the present study, we examined the association between the number of NCDs and HRQOL among Chinese older adults. Suffering from one NCD had a slight negative effect on the HRQOL of older adults, but as the number of NCDs increased, this negative effect also gradually increased. The healthy aging proposed by the WHO is not only the extension of life, but also the improvement of the HRQOL of older adults. At the same time, NCDs serving as a negative factor of the HRQOL of older adults have received increasing attention \[[@B27-ijerph-17-05150]\]. In China, Huang's \[[@B28-ijerph-17-05150]\] survey in Heilongjiang Province found that people with NCDs had a worse HRQOL than people without NCDs. A cross-sectional study of the HRQOL of older adults conducted by Zhu Yaxin \[[@B29-ijerph-17-05150]\] in Liaoning Province concluded that NCDs contributed to the poor HRQOL of older adults. However, these studies did not consider the impact of the number of NCDs on HRQOL. Previous studies from developed countries or regions have confirmed the relationship between NCDs and the HRQOL \[[@B11-ijerph-17-05150],[@B30-ijerph-17-05150]\]. For instance, Wacker \[[@B31-ijerph-17-05150]\] evaluated the HRQOL of 2291 COPD patients, and the results showed that COPD combined with other NCDs significantly reduced HRQOL. These studies have almost proved the harmful effect of NCDs on the HRQOL. However, the relationship between the number of NCDs and the HRQOL is worth discussing. In the results of this study, suffering from one NCD had a negative impact on the HRQOL of older people. However, when compared with two or more NCDs, this negative effect was relatively minor, which was consistent with Xin Yu's survey results in southern China \[[@B16-ijerph-17-05150]\]. In fact, NCDs are generally considered as diseases with an unknown etiology, long duration, and a generally slow progression \[[@B32-ijerph-17-05150]\]. These characteristics may partly explain the smaller impact of having one NCDs on HRQOL. Moreover, related studies have found a positive correlation between economics and NCDs \[[@B33-ijerph-17-05150]\]. Therefore, it is reasonable to speculate that older adults with one NCD have a higher economic level, which may also alleviate the harmful effect of NCDs on HRQOL. By further comparing the dimensions of HRQOL, the results of this study showed that the impact of having one NCD on three dimensions (SAB, AUT, and INT) of HRQOL was not significant. SAB, AUT, and INT evaluate the sensory ability, independent ability and intimacy of older people, respectively. However, NCDs generally refer to medical diseases with a long duration and no infection \[[@B7-ijerph-17-05150]\]. Therefore, the hypothesis that one NCD has a weaker effect on the sensory ability and independent ability of older adults is reliable. Moreover, NCDs are not contagious and have almost no external manifestations, which might explain why there was no difference in intimacy between people without NCDs and with one NCD. Unlike the above three dimensions, having one NCD had significant negative effects on the other three dimensions (PPF, SOP, and DAD) of HRQOL. PPF, SOP, and DAD evaluate the satisfaction with expectations, participation in social activity, and care, worry, and fear of death of older people, respectively. Older adults with NCDs not only experience a change in their physical health, but also bear a huge psychological burden \[[@B34-ijerph-17-05150]\]. This could partly explain why older people lower their expectations for the future. Moreover, because of the existence of NCDs, older people's social participations are likely to be restricted \[[@B35-ijerph-17-05150]\]. Since older adults are at the end of the life cycle, they are already facing the fear of death \[[@B36-ijerph-17-05150]\], which will inevitably increase once they suffer from a disease. All in all, having one NCD had a slight impact on the HRQOL of older adults, which may be a potential hazard. It may cause older adults with one NCD to ignore the prevention and management of the disease, which may lead to multiple diseases and a worse HRQOL. As our research results showed, suffering from two or more NCDs had a more serious negative impact on the HRQOL of older adults. The results are consistent with other studies. Previous research has shown that NCDs can impair the HRQOL of older adults, and the impairment was more severe when suffering from multiple NCDs, which may be related to the additive or synergistic damage caused by various NCDs to the body \[[@B37-ijerph-17-05150]\]. In China, Deng \[[@B38-ijerph-17-05150]\] found that adults with NCDs had a lower HRQOL than adults without NCDs, and their HRQOL had declined more significantly as the number of NCDs increased. Multiple studies in other countries have also demonstrated the inverse relationship between the number of NCDs and HRQOL \[[@B39-ijerph-17-05150],[@B40-ijerph-17-05150]\]. Furthermore, we found that having two NCDs had a significant negative effect on the five dimensions (SAB, AUT, PPF, SOP, DAD) of the HRQOL of older adults. When older people had ≥three NCDs, all dimensions of the HRQOL of older adults were negatively affected. This further confirms the fact that, as the number of NCDs increases, the negative impact on the HRQOL of older adults also increases. Therefore, our study provided strong evidence for the fact that older adult people with multimorbidity might have a worse HRQOL. It is worth noting that NCDs are considered to be complication-prone diseases, especially among older adults \[[@B41-ijerph-17-05150]\]. Although the harm of multimorbidity to older adults' HRQOL has been proven in the previous studies, the incidence of multimorbidity has continued to rise in recent years, and the HRQOL of older adults remains pessimistic \[[@B42-ijerph-17-05150]\]. Therefore, the best option to improve the HRQOL of older adults is to attach importance to the primary prevention of NCDs, which has been proven to be effective \[[@B43-ijerph-17-05150],[@B44-ijerph-17-05150]\]. For older adults with one NCD, more attention should be paid to the treatment and management of the disease to avoid greater harm to the body caused by comorbidities, and fundamentally improves the HRQOL. Generally speaking, implementing primary prevention plays a key role in preventing the occurrence of NCDs and improving the HRQOL of older adults. However, we should also note that there was a close relationship between the number of NCDs and the HRQOL. As the number of NCDs increased, the HRQOL of older adults became worse. Therefore, we should pay attention to the prevention and management of NCDs of older adults to prevent the occurrence of multiple diseases. Some limitations of this study should be taken into consideration. First, the existence of NCDs was self-reported by the respondents, which may lead to the omission of undetected diseases. Second, the types of NCDs listed in the questionnaire were incomplete, which could result in the omission of other NCDs. At the same time, due to the lack of data, the impact of NCD duration on HRQOL was not considered. Obviously, HRQOLs of patients with different types of NCDs are different, and people with the same disease may have a different HRQOL due to different treatment stages. However, considering the complex combination of disease types, in this study, we did not condider the impact of different diseases and their severity on HRQOL. Third, other potential confounding variables, such as smoking and drinking, were not adjusted in the model. Fourth, this cross-sectional survey does not allow us to obtain the dynamic impact of a single NCD or multiple NCDs on HRQOL. 5. Conclusions {#sec5-ijerph-17-05150} ============== This study clearly showed the positive relationship between the number of NCDs and the HRQOL of older people. Compared with older adults without NCDs, suffering from one NCD had a slight negative impact on the HRQOL of older adults; however, having two or more NCDs exerted a greater negative effect on the HRQOL of older adults. As the number of NCDs increased, the HRQOL of older adults was worse. Therefore, we suggest primary prevention as the key to the improvement of HRQOL of older adults by avoiding or delaying the occurrence of one NCD. More importantly, considering the seeming harmlessness of one NCD and the negative effect of multimorbidity on HRQOL, more importance should be attached to joint interventions and programs that target the prevention and control of multimorbidity among older adults with one NCD. The lead agency of China's Health-Related Quality of Life Survey for Older Adults 2018 is the Global Health Institute of Wuhan University. We would like to express our great appreciation to those students from key universities in China for their help in data collection. Conceptualization, S.W.; methodology, J.L.; software, J.L.; validation, W.Y. and S.W.; formal analysis, J.L.; investigation, J.L., W.Y., J.Z., Y.Y., S.C. and S.W.; resources, S.W.; data curation, J.L.; writing---original draft preparation, J.L., and W.Y.; writing---review and editing, J.L., W.Y. and S.W.; supervision, S.W.; project administration, S.W.; funding acquisition, S.W. All authors have read and agreed to the published version of the manuscript. This research was funded by Wuhan University "Double First-Class (World's First-Class University &World's First-Class Disciplines)" development (Special fund \#C). No support was received from industry. All researchers acted independently of the funding bodies. The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results. ijerph-17-05150-t001_Table 1 ###### The number of noncommunicable diseases (NCDs) in different demographic groups. Variables The Number of Noncommunicable Diseases (NCDs) --------------------------------------- ----------------------------------------------- -------------- ------------- ------------- Gender Male 1299 (50.9%) 681 (26.7%) 332 (13.0%) 238 (9.3%) Female 1202 (46.6%) 668 (25.9%) 423 (16.4%) 289 (11.2%) Age 60--64 618 (54.0%) 269 (23.5%) 143 (12.5%) 114 (10.0%) 65--69 581 (48.5%) 321 (26.8%) 179 (14.9%) 118 (9.8%) 70--74 551 (46.2%) 352 (29.5%) 174 (14.6%) 116 (9.7%) 75--79 327 (44.2%) 203 (27.5%) 118 (16.0%) 91 (12.3%) 80--84 257 (45.2%) 141 (24.8%) 100 (17.6%) 71 (12.5%) ≥85 167 (55.3%) 70 (23.2%) 45 (14.9%) 20 (6.6%) Nationality Han 2369 (49.2%) 1255 (26.1%) 690 (14.3%) 498 (10.3%) Minority 102 (41.5%) 76 (30.9%) 44 (17.9%) 24 (9.8%) BMI Underweight (\<18.5) 253 (46.9%) 155 (28.7%) 81 (15.0%) 51 (9.4%) Normal (18.5--23.9) 1601 (50.8%) 806 (25.6%) 438 (13.9%) 306 (9.7%) Overweight (24--27.9) 479 (43.7%) 315 (28.7%) 177 (16.1%) 126 (11.5%) Obese (≥28) 95 (41.3%) 57 (24.8%) 40 (17.4%) 38 (16.5%) Years of education 0 471 (41.8%) 313 (27.7%) 193 (17.1%) 151 (13.4%) 1--5 715 (49.4%) 377 (26.1%) 212 (14.7%) 142 (9.8%) 6--8 461 (49.7%) 241 (26.0%) 139 (15.0%) 87 (9.4%) 9--11 331 (48.0%) 182 (26.4%) 97 (14.1%) 80 (11.6%) ≥12 300 (49.6%) 180 (29.8%) 78 (12.9%) 47 (7.8%) Household registration Agricultural 1626 (46.6%) 948 (27.1%) 534 (15.3%) 385 (11.0%) Non-agricultural 852 (53.5%) 389 (24.4%) 214 (13.4%) 137 (8.6%) Marital status Married/cohabiting 1687 (51.4%) 857 (26.1%) 433 (13.2%) 302 (9.2%) others 815 (43.9%) 492 (26.5%) 323 (17.4%) 225 (12.1%) Average annual household income (CNY) \<15,000 816 (45.0%) 494 (27.3%) 274 (15.1%) 228 (12.6%) 15,000--30,000 552 (43.3%) 352 (27.6%) 215 (16.9%) 156 (12.2%) 30,000--45,000 523 (54.6%) 234 (24.4%) 135 (14.1%) 66 (6.9%) 45,001--60,000 334 (59.6%) 127 (22.7%) 59 (10.5%) 40 (7.1%) \>60,000 254 (53.6%) 135 (28.5%) 59 (12.4%) 26 (5.5%) Self-rated health status Good 942 (62.4%) 347 (23.0%) 163 (10.8%) 57 (3.8%) General 1268 (48.8%) 724 (27.9%) 384 (14.8%) 220 (8.5%) Poor 294 (28.3%) 283 (27.2%) 211 (20.3%) 251 (24.2%) New personal savings (CNY) \<10,000 862 (43.5%) 507 (25.6%) 324 (16.4%) 288 (14.5%) 10,000--30,000 473 (45.1%) 301 (28.7%) 169 (16.1%) 106 (10.1%) 30,000--50,000 355 (53.5%) 169 (25.5%) 94 (14.2%) 46 (6.9%) 50,000--100,000 360 (49.7%) 213 (29.4%) 102 (14.1%) 49 (6.8%) ≥100,000 438 (63.1%) 157 (22.6%) 62 (8.9%) 37 (5.3%) Note: Sample sizes of the demographic characteristic variables may not sum to *n* = 5166 due to missing values. ijerph-17-05150-t002_Table 2 ###### Scores of World Health Organization Quality of Life-Old (WHOQOL-OLD) among the population with different numbers of NCDs. Domains Score Range The Number of Noncommunicable Diseases (NCDs) *F* *p* Multiple Comparisons ------------------------------------------- ------------- ----------------------------------------------- --------------- --------------- ---------------------- -------- -------------- -------------------------- Sensory Abilities (SAB) 4--20 13.60 ± 3.42 13.12 ± 3.56 12.10 ± 3.40 11.21 ± 3.62 89.376 *p* \< 0.001 g 4 \< g 3 \< g 2 \< g 1 Autonomy (AUT) 4--20 13.93 ± 3.34 13.74 ± 3.34 13.21 ± 3.21 12.91 ± 3.39 19.369 *p* \< 0.001 g 4, g 3 \< g 2, g 1 Past, Present and Future Activities (PPF) 4--20 13.50 ± 3.15 13.03 ± 2.94 12.57 ± 2.79 12.25 ± 2.99 36.733 *p* \< 0.001 g 4, g 3 \< g 2 \< g 1 Social Participation (SOP) 4--20 13.37 ± 3.04 12.78 ± 2.92 12.64 ± 2.73 12.10 ± 2.94 35.342 *p* \< 0.001 g 4, \< g 3, g 2 \< g 1 Death and Dying (DAD) 4--20 12.53 ± 3.20 12.09 ± 3.31 11.62 ± 3.57 11.51 ± 4.09 23.346 *p* \< 0.001 g 4, g 3 \< g 2 \< g 1 Intimacy (INT) 4--20 12.95 ± 3.32 12.95 ± 3.32 12.67 ± 3.39 11.74 ± 3.87 20.042 *p* \< 0.001 g 4 \< g 3 \< g 2, g 1 Total score 24--120 79.87 ± 13.25 77.71 ± 13.13 74.80 ± 12.49 71.73 ± 14.56 70.701 *p* \< 0.001 g 4 \< g 3 \< g 2 \< g 1 ijerph-17-05150-t003_Table 3 ###### Relationship between the number of NCDs and WHOQOL-OLD scale scores. Models Sensory Abilities (SAB) Autonomy (AUT) Past, Present and Future (PPF) Social Participation (SOP) Death and Dying (DAD) Intimacy (INT) Total score --------- ----------------------------- ----------------------------- -------------------------------- ----------------------------- ----------------------------- ----------------------------- ----------------------------- Model 1 0 Reference Reference Reference Reference Reference Reference Reference 1 − −0.19 (−041, 0.03) −0.47 (−0.69, −0.27) \*\*\* −0.59 (−0.78, −0.39) \*\*\* −0.45 (−0.67, −0.22) \*\* −0.01 (−0.22, 0.22) −2.17 (−3.04, −1.29) \*\*\* 2 −1.50 (−1.78, −1.22) \*\*\* −0.72 (−0.99, −0.45) \*\*\* −0.94 (−1.18, −0.69) \*\*\* −0.72 (−0.96, −0.48) \*\*\* −0.92 (−1.19, −0.64) \*\*\* −0.28 (−0.56, −0.01) \* −5.08 (−6.15, −4.00) \*\*\* ≥3 −2.39 (−2.71, −2.06) \*\*\* −1.02 (−1.33, −0.71) \*\*\* −1.25 (−1.54, −0.97) \*\*\* −1.26 (−1.54, −0.99) \*\*\* −1.02 (−1.34, −0.70) \*\*\* −1.21 (−1.52, −0.89) \*\*\* −8.15 (−9.39, −6.91) \*\*\* Model 2 0 Reference Reference Reference Reference Reference Reference Reference 1 −0.01 (−0.21, 0.21) 0.12 (−0.19, 0.25) −0.29 (−0.48, −0.10) \*\* −0.37 (−0.56, −0.18) \*\*\* −0.26 (−0.49, −0.03) \* 0.16 (−0.07, 0.38) −0.68 (−1.48, 0.12) 2 −0.85 (−1.11, −0.58) \*\*\* −0.43 (−0.66, −0.16) \*\* −0.58 (−0.82, −0.34) \*\*\* −0.36 (−0.60, −0.13) \*\* −0.83 (−1.12, −0.54) \*\*\* 0.04 (−0.24, 0.32) −2.91 (−3.91, −1.91) \*\*\* ≥3 −1.13 (−1.44, −0.82) \*\*\* −0.56 (−0.86, −0.25) \*\*\* −0.63 (−0.91, −0.35) \*\*\* −0.57 (−0.84, −0.30) \*\*\* −0.62 (−0.95, −0.28) \*\*\* −0.71 (−1.03, −0.38) \*\*\* −4.01 (−5.17, −2.84) \*\*\* Model 3 0 Reference Reference Reference Reference Reference Reference Reference 1 −0.02 (−0.23, 0.19) 0.06 (−0.15, 0.27) −0.33 (−0.52, −0.14) \*\*\* −0.42 (−0.61, −0.24) \*\*\* −0.25 (−0.48, −0.02) \* 0.18 (−0.04, 0.41) −0.87 (−1.67, −0.08) \* 2 −0.81 (−1.07, −0.55) \*\*\* −0.33 (−0.59, −0.07) \*\* −0.58 (−0.81, −0.34) \*\*\* −0.38 (−0.61, −0.15) \*\*\* −0.80 (−1.08, −0.51) \*\*\* 0.04 (−0.23, 0.32) −2.89 (−3.87, −1.90) \*\*\* ≥3 −1.19 (−1.50, −0.89) \*\*\* −0.32 (−0.63, −0.02) \*\* −0.65 (−0.93, −0.37) \*\*\* −0.59 (−0.86, −0.32) \*\*\* −0.62 (−0.95, −0.29) \*\*\* −0.72 (−1.04, −0.39) \*\*\* −4.20 (−5.36, −3.05) \*\*\* Note: \* *p* \< 0.05; \*\* *p* \< 0.01; \*\*\* *p* \< 0.001. Model 1 was the crude model. Model 2 adjusted for all the potential predictors (age, sex, nationality, BMI, years of education, household registration, marital status, average annual household income (CNY), new personal savings and self-rated health status). Model 3 was the final parsimonious model, adjusting for all the predictors that were significantly associated with HRQOL in Model 2. [^1]: These authors contributed equally to this study.

{ "pile_set_name": "PubMed Central" }

1000 Genomes Project data were obtained from a third party, and are available at the following URL: <http://www.1000genomes.org/data>. COGA data are available from the dbGAP database, accession number phs000125.v1.p1. Introduction {#sec001} ============ Consequences of nicotine dependence are the leading cause of preventable death in the USA and it has been noted that some populations experience higher levels of addiction than others \[[@pone.0134393.ref001]\] but the reason for this is not understood. Multiple studies have demonstrated a genetic component to nicotine addiction \[[@pone.0134393.ref002]--[@pone.0134393.ref004]\], but little is known about the role of natural selection in shaping the genetic components of nicotine addiction. Such knowledge could help us understand the genetic and behavioral nature of addiction and ultimately facilitate the design and delivery of appropriate interventions to reduce nicotine addiction. It has been estimated that approximately 10% of the genome has been affected by linkage due to recent selective sweeps \[[@pone.0134393.ref005]\]. However, it is challenging to make direct and clear inference on the phenotypic that is the target of natural selection. This is particularly true when the phenotype being examined has no obvious evolutionary consequences, and has a disease or trait-related association with no clear reproductive consequence today, but where patterns of genetic variation are consistent with a positive selective sweep in recent human history. In this case, additional mechanisms and/or alternative explanations must be sought for the existence of selection on the gene of interest. An example of such a situation occurs in the gene encoding hemoglobin. In homozygous form, the 'sickle cell' allele, HbS, drives the formation of malformed red cells, which aggregate, blocking blood flow to numerous organs including the brain. This results in organ damage and strokes, severely shortening the lifespan of the individual. Nonetheless, the HbS allele is maintained in the gene pool in regions where malaria is endemic because in heterozygous form it provides protection against malaria (for a review see \[[@pone.0134393.ref006]\]. The case of nicotine addiction represents a similar conundrum. Several genetic variants that modify susceptibility to or protection from nicotine dependence have been identified by genome-wide association studies (GWAS) \[[@pone.0134393.ref002],[@pone.0134393.ref004],[@pone.0134393.ref007]\]. Perhaps not surprisingly, the loci identified in these studies mainly include genes encoding neuronal nicotinic cholinergic receptors (*CHRNs*). Neuronal cholinergic nicotinic receptors *(CHRNs)* are a heterogeneous class of cation (positively charged) channels expressed in the central and peripheral nervous system. There are 11 neuronal *CHRN* genes, each of which encodes a receptor subunit. The neuronally expressed nicotinic receptors consist of combinations of alpha and beta subunits, encoded in humans by 8 alpha (α2-α 7, α9-α10) and 3 beta (β2-β4) genes \[[@pone.0134393.ref008]\]. These subunits form homo- or hetero-pentameric subtypes, which are present in various regions throughout the nervous system. To form a receptor, five subunits must be combined within the cell and the specific combination of these subunits defines the receptor subtype. In the body, the opening of these channels is controlled by the endogenous ligand, acetylcholine, a chemical produced by neurons to activate other nearby neurons. Nicotine, the major psychoactive chemical present in tobacco smoke is a chemical present in the environment that can also stimulate the opening of these nicotinic acetylcholine receptor ion channels \[[@pone.0134393.ref009]\]. A number of GWAS studies have been performed that demonstrate an association between the nicotinic receptors and smoking. The strongest association between nicotinic receptors and nicotine addiction is a non-synonymous change (rs16969968, D398N) in the gene encoding the α5 subunit of the nicotinic receptor *(CHRNA5)* on chromosome 15 \[[@pone.0134393.ref007],[@pone.0134393.ref010]--[@pone.0134393.ref014]\]. When cells are made to express nicotinic receptors containing the minor allele form of this SNP (398N), agonists induce less channel opening and cell activation than in cells that express receptors containing the major allele (398D) \[[@pone.0134393.ref011]\]. Thus, the minor allele at this SNP results in a significant functional change in the behavior of this ion channel, causing more nicotine to be needed in individuals with the minor allele to produce the same effect. Additionally, SNPs within the chromosome 15 and chromosome 8 regions have been associated with alocohol and cocaine dependence in addition to their associations with smoking related traits \[[@pone.0134393.ref015]--[@pone.0134393.ref017]\]. Although the underlying functional mechanism underlying the associations at the nicotinic receptors on chromosome 8 is not known, there is nevertheless a genome-wide significant signal with the LD bin \[[@pone.0134393.ref018]\]. Here we test the hypothesis that natural selection has acted on these genes. But if the null hypothesis of neutral evolution or demographic processes is rejected, why would nature seemingly select for this trait, especially given the fact that it is believed that nicotine has not been a part of our evolutionary history long enough, and in large enough quantities, for its effects to be visible in our genomes? One hypothesis is that selection acted on another phenotype and the effect on nicotine addiction was secondary and incidental, a genetic phenomenon termed hitchhiking. Nicotine is known to have an enhancing effect on cognitive performance. It enhances the reorientation of attention in visuospatial tasks \[[@pone.0134393.ref019]\] and alters the neuronal activity responsible for increased attention and arousal \[[@pone.0134393.ref020]\]. Furthermore, several studies have found associations between SNPs in multiple nicotinic receptor genes and cognitive performance \[[@pone.0134393.ref021],[@pone.0134393.ref022]\]. However, effects on cognitive performance at SNPs related to nicotine dependence seem peculiar given the presumed acceleration in cognitive development over recent human history. Therefore, these results suggest that if selection is shaping the genetic landscape of these genes, it may be through their effect on cognitive function in the absence of drug use. Evidence from nicotinic receptor knockout mice also supports a role for nicotinic receptors in memory and learning, as well as anxiety levels. *CHRNA7* knockout mice have impaired reaction times \[[@pone.0134393.ref023]\] and decreased procedural learning \[[@pone.0134393.ref024]\], while, *CHRNA6* knockout mice show that this receptor plays a role in nicotinic modulation of dopaminergic transmission, an important component of learning and memory \[[@pone.0134393.ref025]\]. Based on these observations, we hypothesize that at least some of the nicotinic receptors may have been targets of recent selection and that this selection is related to the role of nicotinic receptors in memory and learning. To test this hypothesis, we use two summaries of genetic variation that have different statistical power to make inferences, depending on the model of selection and associated population demography. Two different methods were used for detecting natural selection at two loci relevant to nicotine dependence, specifically the *CHRNA5-A3-B4* region on chromosome 15q25 and the *CHRNB3-A6* region on chromosome 8p11. We provide strong evidence for selection in the *CHRNB3-A6* region and moderate evidence for selection in the *CHRNA5-A3-B4* region. However, there is only a modest correlation between nicotine dependence and score on the Wechsler Adult Intelligence Scale (WAIS) Digit Symbol test in our dataset. Overall, we suggest that one possible explanation for these results is that SNPs in these regions associated with risk of nicotine dependence are associated with natural selection acting at these loci to improve human memory and learning. Results {#sec002} ======= Tajima's D Test {#sec003} --------------- We calculated Tajima's D over the regions of the nicotinic receptors on chromosomes 8 and 15, as well as lactase on chromosome 2 as the positive control. As the program utilized does not incorporate ancestral information, all results are based on the folded frequency spectrum, i.e. the distribution of polymorphic sites according to the number of chromosomes that carry a given minor allele rather than the number of chromosomes that carry the derived, non-ancestral allele. \[[@pone.0134393.ref026]\][Fig 1](#pone.0134393.g001){ref-type="fig"} shows regional plots of the Tajima's D values for the *LCT*, *CHRNA5-A3-B4* nicotinic receptor gene region and the *CHRNA6-B3* gene region. Within each of these regions, we calculated sliding windows of 10 kb with an increment of 1 kb. We then compared the number of windows within each region with Tajima's D values above 2 or below -2 (this represents the 95% confidence interval of values in our data) to permutations of 10,000 randomly selected regions of the same size across the genome. ![Tajima's D and iHS Results.\ A-CTajima's D andD-F) iHS for the *LCT* region on chromosome 2, the *CHRNA3-B4-A5* gene cluster on chromosome 15 and the *CHRNB3-A6 gene* cluster on chromosome 8 across individuals of European (red), African (Purple) or Asian (green) ancestry. Lines are the 95% confidence intervals as calculated by permutation for iHS and Tajima's D.](pone.0134393.g001){#pone.0134393.g001} We utilized the *LCT* gene region as a positive control. *LCT* encodes the protein lactase and mutations in the region give rise to lactase persistence. The lactase persistence phenotype is due to changes in a regulatory region that enhances the expression of *LCT* located in intron 13 of the neighboring gene, *MCM6* \[[@pone.0134393.ref027]\]. [Fig 1A](#pone.0134393.g001){ref-type="fig"} shows the distribution of Tajima's D values in the sliding windows in the *LCT* region in AFR, ASN and EUR. As expected, both the Asian and European populations show many windows above 2 but very few exist in the African population. All p-values are derived using permutations of 10,000 regions of the same size from across the genome. For Europeans and Asians, the proportion of windows with extreme values was 8.6% and 6.5%, respectively ([Table 1](#pone.0134393.t001){ref-type="table"}). This is highly significant in EUR, when compared to the negative control regions, and modestly significant in ASN (\<p = 2x10^-7^ and p = 0.01 for EUR and ASN, respectively) and appears inconsistent with the null hypothesis of neutral evolution and demography at the *LCT* locus especially in the EUR populations. In contrast, the histogram for AFR shows few windows with extreme Tajima's D values. Indeed, the proportion of windows with extreme values was \<1% and was not significantly different from the negative controls, therefore we are unable to reject the null hypothesis of neutral evolution and demography at this locus in AFR. 10.1371/journal.pone.0134393.t001 ###### Summary of Extreme Tajima's D and iHS Values in Each Region in Comparison with Negative Controls. ![](pone.0134393.t001){#pone.0134393.t001g} Gene(s) Pop \# windows\|TajD\|\>2 Proportion p-value \# windows\|iHS\|\>2 Proportion p-value ---------------- --------- ----------------------- ------------------------- ---------------------------- ---------------------- ------------------------- --------------------------- *CHRNA5-A3-B4* EUR 173/1273 13.60% ***\<2 x 10*** ^***−16***^ 1/181 \<1% NS ASN 91/1273 7.10% ***4 x 10*** ^***−4***^ 3/224 1.3% NS AFR 30/1294 2.30% NS 16/266 5.6% NS *CHRNB3-A6* EUR 13/1066 1.22% NS 26/110 23.6% ***4 x 10*** ^***−4***^ ASN 1/973 \<1% NS 3/116 2.59% NS AFR 23/1097 2.10% NS 93/357 26.1% ***3 x 10*** ^***−4***^ *LCT/MCM6* EUR 98/1136 8.63% ***2 x 10*** ^***−7***^ 111/138 80.4% ***\< 1 x10*** ^***-6***^ ASN 76/1161 6.55% ***0*.*01*** 20/219 9.1% NS AFR 5/1240 \<1% NS 9/278 3.2% NS Proportion of extreme Tajima's D and his values in selected regions. P-values are produced by permutation of 10,000 randomly selected regions of the same size across the genome. We next measured Tajima's D in the *CHRNA5-A3-B4* gene cluster on chromosome 15. In this region, we analyzed a 120 kb region that includes the nicotinic receptor gene cluster as well as \~18 kb upstream of *CHRNA5*. This was done to ensure that we included the large region upstream of *CHRNA5* that has been associated with regulation of the level of mRNA expression for *CHRNA5* \[[@pone.0134393.ref028]\]. [Fig 1B](#pone.0134393.g001){ref-type="fig"} shows the distribution of Tajima's D values in the sliding windows in the *CHRNA5-A3-B4* region in AFR, ASN and EUR. Overall, we observed increased numbers of sliding windows with extreme Tajima's D values in each of the three populations, which is not consistent with a neutral evolution and demography in this region ([Table 1](#pone.0134393.t001){ref-type="table"}). Of note, several SNPs previously shown to be associated with nicotine, alcohol and/or cocaine dependence \[[@pone.0134393.ref007],[@pone.0134393.ref010]--[@pone.0134393.ref014]\] are within regions with high Tajima's D values. Additionally, within the EUR population, these regions also harbor SNPs in an LD bin tagged by rs5887655 previously shown to be significantly associated with *CHRNA5* mRNA expression in the brain \[[@pone.0134393.ref028]\]. [Table 2](#pone.0134393.t002){ref-type="table"} summarizes the Tajima's D values for windows containing associated SNPs in each of these two LD bins. 10.1371/journal.pone.0134393.t002 ###### Tajima's D and iHS values for *CHRNA5-A3-B4* region. ![](pone.0134393.t002){#pone.0134393.t002g} SNP Function EUR Freq Tajima's D EUR iHS EUR ASN Freq Tajima's D ASN iHS ASN AFR Freq Tajima's D AFR iHS AFR ---------------- ------------------- ---------- ------------------ --------- ---------- ------------------ --------- ---------- ------------------ --------- rs72740955 intergenic 0.37 (1.53, 2.07) -0.74 0.03 (0.72, 0.97) -1.25 0.09 (0.88, 1.00) 0.44 rs2036527 intergenic 0.37 (0.11, 0.11) -0.5 0.03 (-0.57, -0.54) -0.85 0.16 (-1.15, 0.20) 1.48 rs55853698 CHRNA5 5\'UTR 0.37 (0.56, 1.93) -0.65 0.04 (-0.02, 1.00) -0.31 0.06 (0.24, 0.56) 1.05 rs17486195 CHRNA5 intronic 0.36 (1.59, 1.64) -0.62 0.03 (-0.17, 0.14) -0.82 0.11 (-0.10, -0.07) 0.48 rs17486278 CHRNA5 intronic 0.36 (0.39, 0.70) -0.62 0.31 (0.04, 0.60) -0.45 0.28 (0.42, 0.47) 0.72 rs72740964 CHRNA5 intronic 0.36 (0.78, 1.63) -0.73 0.03 (0.21, 0.98) -0.57 0.04 (0.08, 0.56) 0.14 rs951266 CHRNA5 intronic 0.36 (-0.22, 0.22) -0.65 0.03 (-0.84, -0.37) -0.5 0.08 (-0.89, -0.74) 0.91 **rs16969968** CHRNA5 missense 0.36 **(2.00, 2.33)** -0.71 0.03 (0.43, 1.01) -0.44 0.02 (0.29, 0.62) -0.01 rs1051730 CHRNA3 synonymous 0.36 (1.40, 1.90) -0.64 0.03 (0.41, 0.78) -0.67 0.09 (-0.29, -0.23) 0.69 **rs1317286** CHRNA3 intronic 0.36 (0.32, 1.27) -0.8 0.09 **(2.00, 2.56)** 0.78 0.24 (0.73, 1.36) -0.43 **rs12914385** CHRNA3 intronic 0.4 **(2.35, 2.55)** -0.39 0.32 (1.85, 2.01) 0.36 0.2 (1.13, 1.35) 0.39 rs114205691 CHRNA3 intronic 0.64 (0.80, 1.65) 1.53 0.68 (0.00, 0.81) 0.15 0.8 (1.75, 1.97) 0.23 rs8040868 CHRNA3 synonymous 0.59 (0.87, 1.37) 1.54 0.62 (1.63, 2.17) 0.25 0.63 (-0.27, 0.13) 0.67 rs55958997 intergenic 0.39 (1.03, 1.66) -1.08 0.06 (0.74, 1.23) 0.23 0.3 (-0.41, -0.38) -0.45 rs72743158 CHRNB4 intronic 0.38 (0.26, 1.27) -0.79 0.02 (0.79, 1.34) -0.83 0.03 (0.20, 0.20) 1.11 rs55988292 CHRNB4 intronic 0.39 (-1.52, -0.24) -0.26 0.02 (-0.35, 028) -0.94 0.14 (-0.93, -0.61) -0.7 **rs4275821** intergenic 0.33 **(2.02, 2.32)** -0.39 0.14 (0.61, 088) -0.02 0.18 (0.74, 0.99) -0.13 **rs588765** CHRNA5 intronic 0.4 **(2.18, 2.22)** -0.32 0.15 (0.00--0.26) -0.2 0.25 (-0.30, -0.30) 0.44 **rs6495306** CHRNA5 intronic 0.4 **(2.07, 2.32)** -0.32 0.15 (-0.42, 0.06) -0.2 0.26 (-0.16, 0.31) 0.41 rs495090 CHRNA5 intronic 0.35 (1.47, 2.06) -0.66 0.2 (1.17, 1.37) 0.48 0.4 (0.77, 1.23) 0.06 rs680244 CHRNA5 intronic 0.4 (-0.77, 0.83) -0.27 0.21 (-0.60, 0.24) 0.62 0.4 (-0.63, -0.06) -0.11 **rs621849** CHRNA5 intronic 0.41 (0.36, 1.35) -0.27 0.21 **(2.07, 2.08)** 0.57 0.4 (0.47, 1.13) -0.21 rs11637635 CHRNA5 intronic 0.65 (1.61, 2.18) 1.51 0.86 (-0.28, -0.13) 0.99 0.78 (-0.60, -0.35) 0.35 rs481134 CHRNA5 intronic 0.6 (0.14, 0.77) 1.1 0.85 (-0.48, 0.19) 1.06 0.75 (-0.87, -0.61) 0.19 rs555018 CHRNA5 intronic 0.4 (-0.84, 0.78) -0.22 0.15 (-0.39, 0.04) 0.01 0.26 (-0.77, -0.23) 0.97 rs647041 CHRNA5 intronic 0.4 (-1.06, -0.14) -0.24 0.18 (-1.05, 0.03) 0.24 0.24 (-1.08, -0.60) 0.77 **rs615470** CHRNA5 3\' UTR 0.65 **(2.54, 2.57)** 1.46 0.83 (-1.20, -0.75) 0.66 0.67 (-1.09, -0.60) -0.08 rs6495307 CHRNA3 intronic 0.6 (-1.17, 0.14) 0.96 0.82 (0.55, 0.84) 0.71 0.66 (-0.83, -0.41) -0.07 rs62010327 CHRNA3 intronic 0.35 (0.15, 1.60) -0.74 0.14 (0.42, 0.85) -0.78 0.08 (0.65, 0.97) -0.99 rs12901300 CHRNA3 intronic 0.4 (-1.37, -0.26) -0.3 0.19 (-0.71, 0.11) 0.26 0.34 (-0.47, 0.32) 0.91 rs3743077 CHRNA3 intronic 0.4 (1.86, 2.62) -0.27 0.18 (-0.25, 0.14) 0.34 0.11 (-0.96, -0.77) 0.33 rs62010328 CHRNA3 intronic 0.34 (1.86, 1.90) -0.8 0.14 (-0.54, -0.25) -0.69 0.07 (-1.17, -0.96) -0.54 **rs2869546** CHRNA3 intronic 0.36 **(2.08, 2.40)** -0.18 0.18 **(2.31, 2.63)** -0.09 0.29 **(3.17, 3.41)** 0.66 rs4366683 CHRNA3 intronic 0.54 (-0.72, 0.13) 0.44 0.45 (1.50, 2.00) -0.96 0.5 (0.70, 1.54) 0.02 rs58643100 CHRNA3 intronic 0.46 (-0.72, 0.13) 0.63 0.26 (1.50, 2.00) 0.84 0.41 (0.70, 1.54) 0.24 All SNPs have r^2^ value of 0.8 or greater with rs16969968 or rs588765. Significant values and SNP names are bolded. For the *CHRNB3-A6* region, we examined a rather larger segment of the genome upstream of the gene cluster. This was done in order to include several upstream SNPs that have previously been shown to exhibit associations with nicotine dependence or cocaine dependence \[[@pone.0134393.ref017]\]. [Fig 1C](#pone.0134393.g001){ref-type="fig"} shows a regional plot of Tajima's D values across the sliding windows in the *CHRNB3-A6* region. In AFR, there were several windows with extreme Tajima's D values upstream of *CHRNB3* and in the intergenic region between *CHRNB3* and *CHRNA6*. Among the three populations tested, none showed a significant increase in the number of windows with extreme Tajima's D values when compared to the permutations. The concentration of windows with extreme Tajima's D values upstream of *CHRNB3* is noteworthy in the context of risk for nicotine addiction. Of note, a recent genome-wide association study found that a SNP in this region, rs1451240, was associated with reduced risk for nicotine dependence, measured using the Fagerstrom Test for Nicotine Dependence (FTND) \[[@pone.0134393.ref018]\]. The LD bin tagged by rs1451240 spans \~66 kb and several other alleles of SNPs in this bin have been associated with reduced risk for nicotine dependence, although no others are significant at the genome-wide level \[[@pone.0134393.ref018]\]. The data in [Table 3](#pone.0134393.t003){ref-type="table"} show that in EUR and AFR, four adjacent SNPs from this LD bin, including rs1451240, were present in sliding windows with extreme Tajima's D values in AFR. These data suggest that these nicotine dependence-associated SNPs may be undergoing balancing selection or positive selection in these two populations. 10.1371/journal.pone.0134393.t003 ###### Tajima's D and iHS values for *CHRNB3-CHRNA6* region. ![](pone.0134393.t003){#pone.0134393.t003g} SNP Function EUR Freq Tajima\'s D EUR iHS EUR ASN Freq Tajima\'s D ASN iHS ASN AFR Freq Tajima\'s D AFR iHSAFR ---------------- ----------------- ---------- ------------------ ---------- ---------- ----------------- --------- ---------- ------------------ -------- rs1979140 intergenic 0.77 (-0.75, -0.34) -1.44 0.82 (-1.21, -1.15) -0.47 0.28 (-1.26, -0.99) -0.15 rs7816726 intergenic 0.77 (0.33, 1.00) -1.44 0.81 (-0.02, 0.70) -0.49 0.28 (0.08, 0.17) -0.56 **rs10958726** intergenic 0.23 (0.1, 0.68) **2.46** 0.18 (-0.85, -0.15) 1.06 0.63 (-0.70, 0.18) 0.65 **rs7842601** intergenic 0.23 (0.17, 0.62) **2.58** 0.18 (-0.33, -0.17) 1.06 0.63 (0.63, 1.33) 0.84 **rs13273442** intergenic 0.23 (-0.39, 0.69) **2.32** 0.18 (-0.60, -0.45) 1.02 0.63 (-0.75, 0.04) -0.21 rs9792277 intergenic 0.77 (1.56, 1.80) -1.45 0.81 (1.06, 1.17) -0.54 0.33 (1.38, 2.32) 0.77 **rs1451239** intergenic 0.77 **(2.41, 2.58)** -1.45 0.82 (1.34, 1.40) -0.46 0.34 **(2.18, 2.51)** 0.89 **rs1451240** intergenic 0.23 **(2.41, 2.54)** **2.48** 0.18 (1.33, 1.41) 1.11 0.67 **(2.18, 2.46)** 0.09 **rs1901281** intergenic 0.77 **(2.37, 2.68)** -1.56 0.81 (1.26, 1.33) -0.54 0.28 **(2.18, 2.19)** 0.37 **rs4736835** intergenic 0.23 **(2.48, 2.68)** **2.6** 0.18 (1.17, 1.26) 1.11 0.66 (1.47, 2.63) -0.08 rs1955185 intergenic 0.77 (1.49, 1.67) -1.58 0.82 (0.64, 0.69) -0.52 0.28 (1.08, 1.59) 0.54 **rs13277254** intergenic 0.23 (0.89, 1.27) **2.69** 0.18 (0.64, 0.97) 1.32 0.63 (1.11, 1.29) -0.6 rs13277524 intergenic 0.77 (0.64, 0.89) -1.63 0.82 (0.24, 0.97) -0.65 0.28 (0.43, 1.29) 0.61 rs6474412 intergenic 0.77 (1.09, 1.43) -1.63 0.82 (0.53, 0.66) -0.65 0.34 (1.01, 1.49) 1.25 rs6474413 intergenic 0.77 (1.86, 2.19) -1.66 0.82 (0.13, 0.22) -0.65 0.28 (1.15, 1.33) 0.62 **rs7004381** intergenic 0.23 (0.68, 1.86) **2.72** 0.18 (-0.15, 0.22) 1.32 0.63 (0.29, 1.15) -0.72 rs6985052 intergenic 0.77 (-0.37, 0.68) -1.66 0.82 (-0.85, -0.16) -0.65 0.29 (-1.04, -0.29) 0.71 **rs4950** CHRNB3 5\'UTR 0.23 (-0.39, 0.12) **2.71** 0.18 (-0.44, -0.16) 1.31 0.8 (0.38, 1.55) 0.25 rs9643891 CHRNB3 intronic 0.77 (-0.34, 0.13) -1.85 0.82 (0.60, 0.72) -1.01 0.15 (-0.65, -0.16) -0.15 rs9643853 CHRNB3 intronic 0.77 (-1.04, -0.34) -1.89 0.82 (-0.60, 0.60) -1.01 0.15 (-1.55, -0.65) -0.1 **rs13280604** CHRNB3 intronic 0.23 (0.47, 1.68) **2.9** 0.18 (0.07, 0.50) 1.94 0.8 (0.04, 0.83) 0.12 rs6997909 CHRNB3 intronic 0.77 (0.70, 0.70) -1.88 0.82 (-0.25, 0.11) -1.19 0.15 (-1.00, -1.00) 0 rs6474414 CHRNB3 intronic 0.77 (0.40, 0.40) -1.88 0.82 (-0.60, -0.21) -1.19 0.15 (-0.76, -0.12) 0 **rs6474415** CHRNB3 intronic 0.23 (1.26, 1.35) **3.07** 0.18 (-0.12, 0.22) 1.97 0.85 (-0.85, -0.52) 0.84 rs4236926 CHRNB3 intronic 0.77 (-0.04, -0.03) -1.71 0.81 (-0.14, -0.11) -1.41 0.15 (-0.54, -0.54) 0.14 rs16891561 CHRNB3 intronic 0.77 (-1.28, -0.90) -1.67 0.81 (-0.89, 0.72) -1.26 0.15 (-1.08, -0.48) 0.2 **rs55828312** CHRNB3 intronic 0.24 (-(0.89, -0.74) **2.86** 0.19 (-1.13, -0.31) 2.13 0.83 (-0.85, 0.67) 1.12 All SNPs have r^2^ value of 0.8 or greater with rs1451240. Significant values and SNP names are bolded. Integrated Haplotype Score (iHS) Analyses {#sec004} ----------------------------------------- [Fig 1](#pone.0134393.g001){ref-type="fig"} shows regional plots of iHS scores for the *LCT*, *CHRNA5-A3-B4* nicotinic receptor gene region and *CHRNA6-B3* nicotinic receptor gene region. Within each of these regions, we calculated the iHS score for all SNPs for which we could unambiguously determine the chimp ancestral allele. We then compared the number of SNPs within each region with iHS values above 2 or below -2, to the 10,000 equally sized permutated regions that were chosen at random. We again used the *LCT/MCM6* region as a positive control. [Fig 1D](#pone.0134393.g001){ref-type="fig"} shows the regional plots of iHS values in the *LCT/MCM6* region for EUR, ASN and AFR populations. As expected, the AFR population shows few extreme values (only 3.2%) and this does not differ significantly from the negative control. In EUR and ASN, the overall average proportion of extreme values for this region is 80.4% and 9.1%, respectively ([Table 1](#pone.0134393.t001){ref-type="table"}). The clustering of extreme iHS values in the genic areas of this region is consistent with what is known about large-scale positive selection at this locus in the EUR population and to a lesser extent in the ASN population \[[@pone.0134393.ref027]\]. This demonstrates the validity of this approach for identifying genes undergoing selection. [Fig 1E](#pone.0134393.g001){ref-type="fig"} shows the histogram of iHS values across the *CHRNA5-A3-B4* locus on chromosome 15. The summary statistics are given in [Table 1](#pone.0134393.t001){ref-type="table"}. None of the three populations showed a proportion of extreme iHS values that was significantly different than predicted by permutation. In addition, none of the SNPs with extreme values included any of the SNPs previously found to be associated with nicotine dependence ([Table 2](#pone.0134393.t002){ref-type="table"}). The presence of a significant enrichment of extreme Tajima's D values in this region, however, suggests that if positive selection in this region occurred, it may have occurred long enough ago such that the long haplotypes required for iHS have broken down by recombination. [Fig 1F](#pone.0134393.g001){ref-type="fig"} shows a histogram of the iHS values for the *CHRNB3-A6* cluster on chromosome 8. Summary statistics are shown in [Table 1](#pone.0134393.t001){ref-type="table"}. Both the AFR and EUR populations show an excess of extreme iHS values (26.1% and 23.6%, respectively). By contrast, in the ASN population, there were few windows of extreme iHS scores and the overall proportion was not significantly different from the negative control. Several SNPs with extreme iHS values in these populations are contained within bins previously shown to be associated with either nicotine or cocaine dependence phenotypes \[[@pone.0134393.ref017],[@pone.0134393.ref018]\]. [Table 3](#pone.0134393.t003){ref-type="table"} lists the SNPs from the LD bin showing genome-wide significance for reduced risk for nicotine dependence and provides the iHS value for rs1451240, the tag SNP. While there are several extreme values in AFR in the middle of this gene cluster, these values do not overlap with known SNPs related to nicotine dependence in this region. In EUR, 13 SNPs in this LD bin, including the tag SNP rs1451240, have extreme iHS values. All have positive values, indicating the presence of unusually long haplotypes containing the ancestral allele suggesting that the ancestral allele, which is associated with a greater risk of nicotine dependence is being favored by selection. An LD bin in the *CHRNB3-A6* region bin that has been shown to have SNPs significantly associated with increased risk for cocaine dependence \[[@pone.0134393.ref017]\] also contained an abundance of SNPs with extreme iHS values ([S1 Table](#pone.0134393.s001){ref-type="supplementary-material"}). This bin is fairly large and spans the entire *CHRNB3-A6* cluster. It contains rs4952 and rs4953, two low frequency synonymous variants in *CHRNB3* that have previously been reported to be associated with lower risk for nicotine dependence \[[@pone.0134393.ref010]\]. All SNPs in the bin are present at around 10% in AFR and 4% in EUR but absent in ASN, possibly explaining the lack of extreme iHS values in the ASN population. Overall, 40% of the SNPs in this bin showed extreme values in AFR and 15% of the SNPs in this bin showed extreme values in EUR. As these SNPs are absent from the ASN population, none showed extreme iHS values. The dense clustering of extreme iHS values in AFR and EUR is rarely expected under a neutral model, and is consistent with the hypothesis of the action of recent selection. In EUR, all of the SNPs in this LD bin with extreme iHS values had positive values and all were shared with AFR. Thus, in both populations, the ancestral allele associated with increased risk for nicotine dependence and decreased risk for cocaine addiction is being favored. Nicotine addiction and cognitive function {#sec005} ----------------------------------------- The Tajima's D analysis and integrated haplotype score both indicate that the *CHRNB3-A6* cluster is undergoing selection and in particular, the iHS scores suggest that it is the risk allele for nicotine dependence on chromosome 8 that is under positive selection. As it seems unlikely that risk for nicotine dependence is the phenotype undergoing selection, and because nicotinic receptors are involved in memory and learning, we hypothesized that a phenotype related to memory or learning, such as attention, might be the phenotype being selected. To test this possibility, we obtained genotype and cognitive phenotype data from the Collaborative Study of the Genetics of Alcoholism (COGA). Using this dataset, we tested the association between genotypes and three of the most relevant phenotypes, namely scores on the Wechsler Adult Intelligence Scale (WAIS) Block Design, WAIS Digit Symbol and WAIS Information tests. These tests were designed to measure aspects of perceptual organization, processing speed and verbal comprehension, respectively \[[@pone.0134393.ref029]\]. [Table 4](#pone.0134393.t004){ref-type="table"} summarizes our findings for the top SNPs associated with WAIS Digit Symbol test. No other neurocognitive phenotypes besides WAIS digit symbol had SNPs with significant values in the *CHRNB3-A6* region and none of the three neurocognitive phenotypes had a significant association with SNPs in the region of *CHRNA5-A3-B4* on chromosome 15 (not shown). Of the 17 SNPs in the *CHRNB3-A6* region on chromosome 8, one SNP--rs7017612---passed multiple test correction (p≤0.003) for association with the score on the WAIS Digit Symbol test (β = 0.43, p = 0.003). rs7017612 lies in the intergenic region between *CHRNB3* and *CHRNA6*. This SNP is highly correlated with rs6474413 (r^2^ = 0.75; D' = 0.95), a SNP tagging the genome-wide significant bin for decreased risk for nicotine dependence. These data suggest a modest association between genotype at these SNPs one measure of cognitive function. 10.1371/journal.pone.0134393.t004 ###### Association results with SNPs in *CHRNB3-A6* region and Scores on the WAIS Digit Symbol test. ![](pone.0134393.t004){#pone.0134393.t004g} SNP β SE p-value ---------------- ---------- ---------- --------------- **rs7017612** **0.43** **0.19** ***0*.*003*** **rs6982753** **0.39** **0.20** ***0*.*009*** **rs10958725** **0.25** **0.18** ***0*.*035*** **rs13273442** **0.28** **0.18** ***0*.*036*** **rs4950** **0.24** **0.18** ***0*.*038*** **rs1530848** **0.24** **0.18** ***0*.*038*** **rs6474413** **0.25** **0.18** ***0*.*038*** rs10107450 0.10 0.19 0.064 rs16891620 0.27 0.25 0.066 rs2196128 0.22 0.20 0.066 rs1530847 0.23 0.20 0.104 rs16891530 0.19 0.42 0.414 rs4952 0.19 0.42 0.414 rs7815274 0.19 0.42 0.426 rs10109429 0.07 0.31 0.494 rs13270610 -0.16 0.35 0.632 rs16891604 0.23 0.42 0.826 Beta, Standard Error (SE), and P-value for all SNPs in the *CHRNB3-A6* region of the COGA family GWAS and their association with scores on the WAIS Digit Symbol test. Covariates used were age, sex and FTND score. For all tests, 492 individuals were used for the association test. A second SNP in the *CHRNB3-A6* region--rs6982753- had a nominal p-value with the WAIS Digit Symbol phenotype before multiple test correction and almost passed the multiple test correction (p = 0.009). Interestingly, this SNP has an r^2^ of 0.91 with rs892413 (β = 0.39, p = 0.008), a SNP that has previously been associated with increased risk for cocaine dependence \[[@pone.0134393.ref017]\]. Discussion {#sec006} ========== Many studies have demonstrated that risk for nicotine addiction has a genetic component. We performed two tests of selection on chromosomal regions containing the genes encoding five nicotinic receptor subunits and each of these analyses indicate that selection likely occurred at the *CHRNB3-A6* locus. Both the Tajima's D test and iHS point to an ongoing sweep in humans on chromosome 8. In the case of the *CHRNB3-A6* locus, all of the extreme values in the Tajima's D analysis were positive. High positive Tajima's D values occur when there is an excess of variants in a region with intermediate allele frequencies. This can occur in either balancing selection or ongoing positive selection. We also found extreme iHS values in the *CHRNB3-A6* locus. This region fulfills the criteria for a sweep laid out by Voight and colleagues \[[@pone.0134393.ref030]\], i.e. clustering of extreme iHS values. Extreme values of iHS are unlikely under simple demographic models, and thus can indicate the action of an ongoing selective sweep. Despite the fact that none of the populations had significantly different p-values from the negative control in the chromosome 8 region, a few key SNPs did have extreme values. Together, these data imply that the Tajima's D analysis is picking up on ongoing positive selection rather than balancing selection. Several SNPs in the *CHRNB3-A6* locus on chromosome 8 have previously been associated with a decreased risk of nicotine dependence \[[@pone.0134393.ref018]\]. One of these, rs1451240, was present in a window that showed extreme values in both the Tajima's D test and iHS. The extreme positive iHS value in the window including rs1451240 indicates that the haplotype containing the ancestral allele is being positively selected. As the derived allele provides protection from nicotine addiction, this suggests that it is the allele that is associated with a greater risk of nicotine dependence that is being selected. Since highly concentrated sources of nicotine were not present in the ancestral environment, it seems likely that this phenotype of nicotine dependence would have hitchhiked along with a more beneficial phenotype. One challenge with this region is that it is approximately 1,500,000 bp away from the centromere of chromosome 8. This could be affecting the results by some unknown mechanism. However, the region including the nicotinic receptors on chromosome 8 was among the top 5% of iHS scores among all regions tested in the genome. Selective pressures in our ancestral environments were likely not on addiction, but rather on behaviors that were biologically rewarding (i.e. mate or food finding, avoidance of harmful stimuli). Given the role of nicotine in neurological function, it is possible that, in the case of nicotine addiction, the phenotype on which natural selection was working was related to enhancements in memory or cognition. The addiction phenotype would have hitchhiked along because it acts through the same or related mechanisms. The addiction phenotype was likely not selected against in ancestral environments because the availability and opportunity for prolonged use of purified drugs was negligible. To test this possibility, we assessed the association of SNPs in the *CHRNB3-A6* locus with scores on WAIS tests of memory and cognitive function. Our analysis of the individuals in the COGA dataset suggests that one SNP, rs7017612, which lies in the intergenic region between *CHRNB3* and *CHRNA6*, is associated with increased score on the WAIS Digit Symbol test. This test is thought to largely measure processing speed, but also, to some extent, memory. rs7017612 itself has not been reported to be associated with nicotine dependence. However, it is in moderately high LD (r^2^ = 0.75) with rs6474413, a SNP tagging the genome-wide significant bin for decreased risk for nicotine dependence. Thus, our data are consistent with the possibility that improved performance on this particular cognitive test is modestly associated with a *decreased* risk for nicotine dependence and that alleles of SNPs in these regions have effects on a subset of cognitive pathways best captured here by the WAIS digit symbol test. It is possible, however, that a function other than addiction or cognition is the true phenotype undergoing natural selection at these loci. Genetic studies of nicotine addiction have identified an inverse relationship between the risk for nicotine addiction and the risk for cocaine addiction. For instance, the minor allele of rs16969968, a missense variant in *CHRNA5*, increases risk for the development of nicotine dependence, and independently decreases risk for cocaine dependence \[[@pone.0134393.ref015]\]. One hypothesis would be that the true underlying selective pressure is on cocaine related phenotypes or a characteristic that affects cocaine related reward pathways in the brain and that the alleles' effect on nicotine dependence is merely an accidental consequence. However, caution must be used when interpreting this information, given that all drugs of addiction similarly affect the dopaminergic reward pathways. Another alternative hypothesis is that the selective pressure at this locus was on social behavior. Cocaine addiction is characterized by a dampened reward response to social interaction, meaning that it inhibits the positive emotions that accompany social interaction or feelings of belonging. A recent study demonstrated that cocaine users process social gaze (joint attention on an object) differently than controls, resulting in a reduced activation of the reward system during social interactions \[[@pone.0134393.ref031]\]. Using fMRI, these authors showed that cocaine users had decreased activation of the medial orbitofrontal cortex, a region of the brain central for reward processing. If alleles that alter cocaine dependence risk alter an individuals' natural reward system during social interactions, these observations could explain why alleles that protect against cocaine dependence could have provided advantage to carriers in the ancestral environment. Since nicotine sensitizes the animal to the effects of cocaine, which blunts the reward of social interactions, alleles that reduced the ability of nicotine to enhance the effects of cocaine would have undergone positive selection. In this scenario, the nicotine and cocaine dependence phenotypes are not hitchhiking with memory or learning, but rather with phenotypes protecting against antisocial and therefore maladaptive behavior. There was also moderate evidence for selection at the *CHRNA5-A3-B4* locus. In particular, rs16969968, the SNP that encodes the missense mutation in α5, that is strongly associated with risk of nicotine dependence, lies in a sliding window exhibiting a high Tajima's D score. The iHS analysis of this locus did not provide evidence for selection. This could indicate that the selective pressure exerted on this locus is older that that seen for the *CHRNB3-A6* locus and as such has allowed extended haplotypes to be broken down by recombination. In this scenario, Tajima's D would still be extreme while iHS scores in the region might be less so. This may be particularly true if the selective pressure being exerted on the *CHRNB3-A6* locus is ongoing. Here we have used two statistical tests of selection and uncovered evidence of positive selection at the nicotinic receptors on chromosome 8 chromosome 15. Multiple drug-related phenotypes are associated with SNPs in or near these loci, however for several reasons it is unlikely that these phenotypes are the direct targets of this selective pressure. We have proposed two possible explanations 1) phenotypes related to memory and learning and 2) phenotypes related to social behavior. We were only able to discover a modest association with memory-related phenotypes, likely due to the small sample size. We also are as of yet unable to test this second hypothesis because we do not have data in our sample for a phenotype measuring sociality. However this work is the first to explicitly describe signs of natural selection acting on loci underlying substance dependence phenotypes. Materials and Methods {#sec007} ===================== To determine whether the nicotinic receptor loci are under selection, we used Tajima's D, and integrated haplotype score (iHS) to examine the landscape of natural selection at three loci previously demonstrated to harbor genetic variants contributing to the risk of nicotine dependence. These tests have different but complementary strengths. Tajima's D test functions best on recently completed selective sweeps. There are many variables that contribute to how far in the past a sweep can be detected, such as how extreme the sweep was in the first place. Both the mutation rate and the recombination rate affect it as well and vary widely across the genome making generalizations difficult. By contrast, integrated haplotype score iHS functions best for detecting sweeps in progress with alleles at intermediate frequencies, mainly in the range of or after the separation of European, Asian and African populations, during the agricultural phase of human evolution. We utilized 1000 Genomes data for these analyses. Data was obtained from the 1000 genomes website (<http://www.1000genomes.org/data>), a third party source for population whole genome sequence data. The populations were grouped into EUR (GBR, TSI, CEU, FIN), ASN (CHS, CHB, JPT), and AFR (YRI, LWK, ASW). All methods were calculated for the same regions: the *CHRNA5-A3-B4* region on chromosome 15q25, the *CHRNB3-A6* region on chromosome 8p11, the *LCT* region as a positive control on chromosome 2q21, and ten intergenic negative control regions where applicable. Tajima's D test {#sec008} --------------- Tajima\'s D is a method of addressing the frequencies of variant sites, based on the expectation that under neutrality, different estimates of expected diversity (θ) should be equal. Tajima's D tests for a skew in the frequency spectrum by comparing two estimates of θ -- the number of segregating sites (S), and pairwise nucleotide diversity (π) \[[@pone.0134393.ref032]\]. Extreme positive values can indicate either balancing selection or population subdivision, and extreme negative values can indicate positive selection or population growth \[[@pone.0134393.ref033]\]. If the same skew is detected across the genome, the effect is likely due to demography, whereas if the skew is localized to a few loci, selection is more likely to be occurring. In a review, Garrigan & Hammer \[[@pone.0134393.ref034]\] have combined published data for Tajima's D values from 65 autosomal loci. They find the mean value for Africans is slightly negative (-0.20) and for non-Africans is slightly positive (0.13). Overall, the values range from approximately -2 to 2. As such, we have taken Tajima's D values above 2 or below -2 to count as extreme values, as this represents the 95% confidence interval of values in our data. Tajima's D was calculated using the program Variscan \[[@pone.0134393.ref035]\]. After an exploratory data analysis of window size, we used a sliding window size of 1000 bp, and window increments of 100 bp for the analysis. Smaller window sizes resulted in too few SNPs in a window to calculate Tajima's D, while larger windows made it much harder to narrow down specific SNPs that may be the ultimate target of selection. Variscan outputs a file giving the Tajima's D value for every window of the specified bp size on the sliding scale \[[@pone.0134393.ref035]\]. These values were then superimposed onto graphs of the regions. Integrated Haplotype Score {#sec009} -------------------------- iHS is a measure of whether a SNP is on an unusually long haplotype carrying the ancestral or derived allele. It compares the rate of haplotype decay between haplotypes carrying either the ancestral or derived allele at a given site, referred to as the core SNP. Haplotypes whose core SNP is under selection will be unusually long compared to those evolving neutrally. Long haplotypes with derived alleles are indicated by negative iHS values and those with ancestral alleles are indicated by positive iHS values. Under neutrality, extreme scores are distributed throughout the genome, however under selection, they are clustered across the selected region \[[@pone.0134393.ref030]\]. iHS is a good method for detecting directional selection, especially in a sweep that is in its early phases. We used the program WHAMM to calculate this statistic \[[@pone.0134393.ref030]\]. The haplotype decay is calculated until the extended haplotype homozygosity (EHH) reaches 0.05. EHH is defined as "the probability that two randomly chosen chromosomes carrying the core haplotype of interest are identical by descent for the entire interval from the core region to point x" \[[@pone.0134393.ref036]\] (p. 833). Long haplotypes with derived alleles are indicated by negative iHS values and those with ancestral alleles are indicated by positive iHS values. Under neutrality, extreme scores are distributed throughout the genome, however under selection, they are clustered across the selected region \[[@pone.0134393.ref030]\]. First, we extracted the desired regions from the 1000 Genomes dataset. We then selected known SNPs within each region, and extracted a region of plus or minus 2000 SNPs around that SNP, except in the case of *CHRNB3-A6* where we selected plus or minus 2500 SNPs. We constructed recombination maps using cM maps provided by the SHAPEIT2 program \[[@pone.0134393.ref037]\]. Ancestral alleles were determined using the latest version of Seattleseq (<http://snp.gs.washington.edu/SeattleSeqAnnotation137/>). Phased haplotypes were coded as number of copies of the derived allele. All positions in which the derived allele could not be determined unambiguously (i.e. C/G or A/T SNPs) as well as those without known chimp alleles were removed from further analyses. All analyses were run on each population separately. As iHS is greatly influenced by SNP allele frequency, iHS values from WHAMM were standardized using the average and standard deviation of all SNPs on chromosome 15 and 8 binned by allele frequency such that the average iHS value for each bin after standardization was identical. We excluded SNPs with a minor allele frequency less than 5% because low frequency SNPs are difficult to normalize accurately. After removing these SNPs, extracting just the desired gene regions, and removing those with MAF of \<0.05, there were \~150--350 SNPs per region, depending on the population. Standardization was done separately for each population using population specific averages and standard deviations. iHS values were then superimposed onto graphs of the regions. The haplotype on which a beneficial allele resides tends to be significantly longer than the other haplotypes at the same frequency in the population when adjusted for the recombination background. However, long haplotypes tend to occur in regions with low recombination, and these can be confused with genuine genomic signals of positive selection \[[@pone.0134393.ref038]\]. This is why WHAMM attempts to control for recombination by requiring the input of a cM map. The map we used here was the cM map for imputation available on the website for the program SHAPEIT2. Candidate regions of positive selection were defined as genomic regions containing an uncharacteristic clustering of SNPs with high iHS statistics. This was quantified as the proportion of SNPs with \|iHS\| \> 2 in the four regions of interest. Candidate regions of positive selection were identified as containing any SNP with an iHS score of \|iHS\| \> 2, as this corresponds to the top \~5% of all scores. The iHS value at a SNP "measures the strength of evidence for selection acting at or near that SNP" however does not provide a formal significance test (Voight et al. 2006). Association Analyses {#sec010} -------------------- COGA recruited subjects in Hartford, Connecticut; Indianapolis, Indiana; Iowa City, Iowa; New York City, New York; San Diego, California; St Louis, Missouri; and Washington, DC. For inclusion in SAGE, cases had to meet lifetime criteria for DSM-IV alcohol dependence, the majority of cases were recruited from alcoholism treatment centers. Control subjects, were both biologically unrelated to cases, and had consumed alcohol but never experienced any significant alcohol or drug-related problems, according to the Semi-Structured Assessment for the Genetics of Alcoholism (SSAGA) \[[@pone.0134393.ref018]\]. The COGA sample utilized in this study consisted of family GWAS data from 2102 European-Americans \[[@pone.0134393.ref014]\]. De-identified data from the Collaborative Studies on the Genetics of Alcoholism (COGA) were used. All participants in COGA provided written informed consent for genetic studies and agreed to share their DNA and phenotypic information for research purposes. The Washington University Human Research Protection Office granted approval for data to be used for this study. COGA administered a variety of neuropsychological tests to its subjects including the three used here: Wechsler Adult Intelligence Scale-Revised (WAIS-R) Block Design, WAIS Digit Symbol, and WAIS Information. In total there were 1247 European-Americans with these neuropsychological phenotypes. However, the overlap between this number and those with family GWAS data was 492. Therefore, our analyses were performed on 492 subjects. In WAIS Block Design, the subject replicates models or pictures of two-color designs with blocks. In WAIS Information, the subject answers a series of questions about factual information. In WAIS Digit Symbol, the subject writes down as quickly as possible the symbols that correspond to a series of numbers. SNPs in the region of the nicotinic receptor clusters on chromosomes 8 and 15 were tested for association with the scaled scores of neuropsychological phenotypes in European-Americans from the COGA study using an additive linear mixed effects (LME) model with the lmepack.batch function as implemented in the GWAF package in R using age, sex and FTND score as covariates \[[@pone.0134393.ref039]\]. Neither alcohol nor cocaine symptom count were significant covariates in the analysis and were thus not included in the analysis. The GWAF package enables association testing with the ability to include individuals from families by correcting for relatedness specified in a pedigree file. Supporting Information {#sec011} ====================== ###### Tajima's D and iHS values for SNPs correlated with rs9298626 within the *CHRNB3-A6* region. All SNPs were highly correlated (r^2^\>0.9) with the SNP rs9298626. (PDF) ###### Click here for additional data file. The authors would like to thank Dr. Benjamin Voight at University of Pennsylvania for his edits and advice with regard to the use of WHAMM as well as data interpretation. The Collaborative Study on the Genetics of Alcoholism (COGA), Principal Investigators B. Porjesz, V. Hesselbrock, H. Edenberg, L. Bierut, includes ten different centers: University of Connecticut (V. Hesselbrock); Indiana University (H.J. Edenberg, J. Nurnberger Jr., T. Foroud); University of Iowa (S. Kuperman, J. Kramer); SUNY Downstate (B. Porjesz); Washington University in St. Louis (L. Bierut, A. Goate, J. Rice, K. Bucholz); University of California at San Diego (M. Schuckit); Rutgers University (J. Tischfield); Texas Biomedical Research Institute (L. Almasy), Howard University (R. Taylor) and Virginia Commonwealth University (D. Dick). Other COGA collaborators include: L. Bauer (University of Connecticut); D. Koller, S. O'Connor, L. Wetherill, X. Xuei (Indiana University); Grace Chan (University of Iowa); S. Kang, N. Manz, M. Rangaswamy (SUNY Downstate); J. Rohrbaugh, J-C Wang (Washington University in St. Louis); A. Brooks (Rutgers University); and F. Aliev (Virginia Commonwealth University). A. Parsian and M. Reilly are the NIAAA Staff Collaborators. We continue to be inspired by our memories of Henri Begleiter and Theodore Reich, founding PI and Co-PI of COGA, and also owe a debt of gratitude to other past organizers of COGA, including Ting-Kai Li, currently a consultant with COGA, P. Michael Conneally, Raymond Crowe, and Wendy Reich, for their critical contributions. The authors thank Kim Doheny and Elizabeth Pugh from CIDR and Justin Paschall from the NCBI dbGaP staff for valuable assistance with genotyping and quality control in developing the dataset available at dbGaP (Accession \#: phs000125.v1.p1). [^1]: **Competing Interests:**Dr. AM Goate is listed as an inventor on the patent "Markers for Addiction" (US 20070258898) covering the use of certain SNPs in determining the diagnosis, prognosis, and treatment of addiction. [^2]: Conceived and designed the experiments: BS GH AG. Performed the experiments: BS GH. Analyzed the data: BS GH. Contributed reagents/materials/analysis tools: HE JT AB JK MS JN. Wrote the paper: BS.

{ "pile_set_name": "PubMed Central" }

INTRODUCTION ============ Human beings have become more sensitive to the welfare of farm animals. For each individual animal, the quality of life contributes to productivity, with better welfare promoting increased efficiency compared to poor welfare. Animal well-being can however be depressed during aversive human-animal interactions \[[@b1-ajas-18-0151]\] that are done in the pens and crushes. These handling facilities are designed for management practices such as vaccinations that enhance animal welfare. However, they are usually associated with the isolation of an individual animal from its cohorts, for 'aversive', stressful and fearful handling \[[@b2-ajas-18-0151]\]. It is therefore maintained that in handling facilities such as the race and crush, cattle lose control and this in itself can be stressful. Hence, race and crush scores are some of the indices that have since been used to determine the extent to which animals can actively resist or be willing to be restrained. Eccentric behaviour, such as stupor, restlessness, kicking, shaking the facility and attempt to escape can be displayed by animals that are reluctant to be handled and restrained \[[@b2-ajas-18-0151]\]. Behavioural responses such as avoidance, and the distance maintained from handlers can be used to determine fear of cattle towards humans. The avoidance distance (AD) test has since been used to determine the distance at which animals withdraw from an approaching human \[[@b3-ajas-18-0151]\]. Further, excretory behaviour such as urination and defaecation can be frequent especially in fearful animals \[[@b4-ajas-18-0151]\]. Repeated exposure of an animal to the same stressor is however associated with reduced subsequent responses \[[@b5-ajas-18-0151]\], such that aggressive animals that regularly go through the same handling procedures may progressively become docile. Animal handling and restraint on the other hand trigger physiological responses such as the activation of the hypothalamic-pituitary-adrenal (HPA) axis exhibited in increased release of glucocorticoids (GCs) such as cortisol. Cortisol is therefore used to quantify stress \[[@b6-ajas-18-0151]\]; hitherto, blood cortisol is often used as a reliable stress biomarker, to determine how animals respond to different degrees of stress \[[@b7-ajas-18-0151]\]. Cook \[[@b8-ajas-18-0151]\] however highlighted that handling may have confounding effects on the measurements, which can be caused by restraint and venepuncture. In addition, Rushen \[[@b9-ajas-18-0151]\] underlined that data related to the HPA is inconsistent. To mitigate erratic measurements, less-invasive procedures which make use of saliva and faecal samples are being used. For more than a decade, the quantification of faecal GCs has become popular as a non-invasive tool in the study of adrenocortical activity during stressful situations. Faecal glucocorticoid metabolites (FGM) are used for the measurement of anthropogenic disturbances on animals \[[@b10-ajas-18-0151]\] and in cattle, the use of dung to reflect faecal GCs is validated \[[@b11-ajas-18-0151]\]. To our knowledge, the relationship between FGM and stress-related behaviour is illustrated in some zoo and wild animals such as the marmoset \[[@b12-ajas-18-0151]\], as well as the domestic horse \[[@b13-ajas-18-0151]\], but is not yet determined in cattle. Observation of animal behaviour in relation to handling stress is less-invasive. Nevertheless, accurate assessment of how cattle react to handling and restraint must employ both behavioural and physiological parameters \[[@b14-ajas-18-0151]\]. The juxtaposition of behaviour and physiological parameters can be used to assess animal welfare. For instance, rectal temperature, a more reliable parameter for measuring core body temperature, increased in cattle that exhibited adverse behavioural reactions during restraint in the chute \[[@b15-ajas-18-0151]\]. Understanding of behaviour and physiological parameters can help reduce animal stress and ultimately increase animal production efficiency. High GC values can be expected in animals that lack experience in handling stress. However, as the number of handling encounters increase, the magnitude of GC levels can be anticipated to subsequently decline \[[@b5-ajas-18-0151]\]. Salivary cortisol reflects the biologically active fraction of the total plasma steroid hormone with a positive relationship between the two media \[[@b16-ajas-18-0151]\]. Moreover, the association between the media is linear over a wide range of concentrations \[[@b8-ajas-18-0151],[@b16-ajas-18-0151]\]. Saliva samples can be collected both prior to and after an imposed stress even at fixed time intervals \[[@b17-ajas-18-0151]\], to indicate the activity of the HPA axis in response to different stressors. The correlation of plasma and saliva cortisol has been established in some domesticated ruminant animals such as goats \[[@b18-ajas-18-0151],[@b19-ajas-18-0151]\] and sheep \[[@b20-ajas-18-0151],[@b21-ajas-18-0151]\]. Studies that validate the use of saliva cortisol as an alternative to blood cortisol in cattle are however still limited. Hitherto, a hormonal challenge was done based on sampling intervals of at least 15 min \[[@b17-ajas-18-0151],[@b22-ajas-18-0151]\]. The study by Hernandez et al \[[@b23-ajas-18-0151]\] was characterised by short sampling intervals of 10 min, and, the hormonal challenge was not done. Notwithstanding, plasma cortisol levels can change due to non-aversive procedures. Thus, cattle handling prior to sampling can affect the characterisation of the relationship between plasma and saliva cortisol. Our objective was to determine the relationship between plasma and salivary cortisol levels in ACTH induced beef cows that were handled prior to sampling. Acclimation to handling was done prior to the characterisation of the relationship. MATERIALS AND METHODS ===================== Ethical approval of animal use and location ------------------------------------------- All procedures involving animals were reviewed and approved by the University of Fort Hare Animal Research and Ethics Committee (Reference Number: MUC291SDZV01). The study was conducted at a research farm under the coordinates 32°48′S (Latitude) and 26°53′E in Alice, South Africa during mid-spring (mean temperature 23°C±4.1°C). Description and management of animals ------------------------------------- Twenty-one clinically healthy non-pregnant Nguni cows aged between 48 and 144 months and weighing between 426 and 436 kg were selected at random, from the different camps at the farm. They were of different age groups, group 1 with n = 7 (5 to 7 yr), group 2 with n = 6 (8 to 10 yr), and group 3 with n = 8 (11 to 13 yr) and these had average weights of 436 kg, 429 kg and 426 kg for each respective age group. The cows were identified by ear-tagging and markings on the flanks according to the respective farm's management system of identification. The identification system made it feasible to use the same cows throughout the trial. The cows were handled separately, however, they were all exposed to handling through routine paddock rotations, health and welfare related check-ups. All the cows had access to grazing on natural pastures and water *ad libitum*. Measurements and data collection -------------------------------- ### Experiment 1 Data was collected from each individual animal unsystematically between 9:00 and 11:00 am from July to September. There was a seven-day interval per observation and sampling per animal. Maximum and minimum temperatures during sampling days are indicated in [Figure 1](#f1-ajas-18-0151){ref-type="fig"}. Different animal groups were separately brought into the holding pens 14 h prior to observations and samplings. The AD test was done on each randomly selected individual animal in a handling pen, based on modifications to the studies by Dodzi and Muchenje \[[@b3-ajas-18-0151]\]. The handler who wore the same clothing (green work-suit) on every occasion, approached an individual stationary cow at a rate of one step every 5 s from a distance of 3 m, whilst avoiding direct stares to the animal. The test ended when the cow showed an avoidance reaction (retreat). Scores 1 to 3 as indicated in [Table 1](#t1-ajas-18-0151){ref-type="table"} were allotted accordingly. The lower end indicated that the animal was less willing to be approached by the handler than the one which was assigned a Score 3. For cows which could not be immobile enough to be approached for the AD test, an AD Score 1 was assigned. Each individual animal was then restrained in the chute for 30 s and its behaviour observed and scored based on scores modified from Goldhawk et al \[[@b24-ajas-18-0151]\]. Thereafter, for 15 s the cow was restrained in the crush pen with its head fixed with a head gate. Immediately after fixation, behaviour was observed and evaluated with the scoring system in [Table 1](#t1-ajas-18-0151){ref-type="table"}. Rectal temperature readings were then collected on the assumption that, human contact affects cattle behaviour and this may stimulate different biological variables including the former \[[@b15-ajas-18-0151]\]. A digital thermometer (GLA M500, GLA Agricultural Electronics, San Luis Obispo, CA, USA) was used and readings recorded as done by Sánchez-Rodríguez et al \[[@b25-ajas-18-0151]\]. Ambient temperature readings were also taken. Fresh faecal samples were then extracted directly from the rectum of each cow, using a gloved hand and placed in sterile vials. The samples were placed on ice and thereafter frozen at −20°C until analysis. During handling, vocalisation, excretion of watery faeces and urine were observed. Only occurrences of at least three times during the sampling times were considered. The parameters were recorded and assigned scores as either present (1) or absent (0), thereby giving vocalisation, defecating and urinating scores accordingly. From when avoidance behaviour was observed to when the animals were released from the crush pen, total handling time for each animal was on average 120±10 seconds. #### Faecal sample analysis Faecal samples were defrosted at room temperature (20°C to 23°C) over an average period of 4 h. A well-mixed (stirred) wet faecal sample (1 g) was dispersed in 5 mL of 80% methanol and vortexed for 16 h. Of the dispersion, 50 μmL was pipetted into Eppendorf tubes, centrifuged at 1,000×g for 10 min at 4°C and mixed into a methanol 4:1 distilled water solution. The FGM concentrations were assayed using a commercial double-antibody 125I-corticosterone radioimmunoassay kit for animal testing (MP Biomedicals, 07120103, Carlsbad, CA, USA; Lot No. RCBK163) at half volume \[[@b26-ajas-18-0151]\]. The supernatant was aspirated and the precipitate counted on a PerkinElmer Wizard2 Gamma Counter. The intra-assay coefficient of variation was 7.1%. ### Experiment 2 i. Animals: Six clinically healthy non-pregnant Nguni cows were randomly selected from animals used for experiment 1. The cows had a mean weight of 407 kg (range: 336 to 506 kg) and each cow was used as an experimental unit. Samples were collected between 08:30 and 11:30 h at ambient temperatures of not more than 30°C ([Figure 1](#f1-ajas-18-0151){ref-type="fig"}). ii. Adrenocorticotropic hormone administration: A standardised dose (1 μg/kg), of ACTH (Synacthen Depot, tetracosactide 1 mg/mL, Lot S1358, Novartis, South Africa) was administered once to each of the six cows at time 0 \[[@b27-ajas-18-0151]\]. iii. Extraction of blood samples: Blood samples (approximately 6 mL) were collected whilst the animals were restrained in a crush pen with the head in a head gate through jugular venepuncture into tubes containing SST gel. A nose grip was also used for stable restraint. Baseline samples were collected 10 min prior to ACTH administration. Thereafter, samples were drawn every 10 min for 1 h \[[@b28-ajas-18-0151]\], to obtain a total of eight samples per animal. Samples were placed on wet ice (4°C) and then centrifuged for 10 min at 3,550×g (22°C). The serum was then transferred to red-topped tubes (4 mL) with a Clot Activator and stored at −20°C until analysis for cortisol. iv. Extraction of saliva samples: Saliva samples were collected immediately after blood sample collection using cotton based swabs (Salivette cortisol; Sarstedt, Nümbrecht-Rommelsdorf, Germany), which provided a method for easy and safe collection. Cotton balls were also used to collect drooling saliva to maximise the amount of saliva collected. The cotton swabs were inserted at an angle of the lips into the mouths of the cows with the help of a nose grip until well soaked \[[@b29-ajas-18-0151]\]. Each swab and the corresponding cotton ball were then placed in the salivette tube which was placed on ice (4°C). The samples were centrifuged at 1,000×g for 2 min at 20°C. The swabs and cotton balls were removed together with the inner tube of the salivettes. The saliva which collected into the outer tubes was immediately stored at −20°C until analysis. Salivary cortisol was determined by competitive enzyme-linked immunosorbent assay (ELISA). v. Cortisol analysis in plasma and saliva: Blood plasma samples were defrosted and vortexed (Vortex Genie-2, Scientific Industries Inc, New York, USA) at room temperature (24°C). Quantification of plasma cortisol was then done by ELISA using a Cortisol ELISA kit (IBL International, GmBH, Hamburg, Germany, RE52611) as described by Olbrich and Dittmar \[[@b30-ajas-18-0151]\]. Saliva samples were defrosted and centrifuged at 7°C, 1,200 ×g for 10 mins and cortisol levels were measured as explained for plasma cortisol. The intra-assay coefficient of variation (CV) ranged from 3.1% to 6.1% for saliva, and 5.9% to 9.9% for plasma. The inter-assay CV ranged from 4.2% to 17.0% for saliva and 13.0% to 20.0% for plasma. The detection limits of the saliva and plasma assays were 0.083 nmol/L and 0.14 nmol/L, respectively. For both saliva and blood plasma, optical density measurements were done at 450 nm using a photometer Multiskan Ascent with a Genesis Lite microplate computer programme (Labsystem, Finland). Statistical analysis -------------------- ### Experiment 1 Data on subjective scores were square root transformed. Thereafter, all data on FGM concentrations, rectal temperature; avoidance, chute, crush, urinating, defecating and vocalising behaviour scores, were analysed using a repeated measure PROC general linear model procedure of Statistical Analysis System (SAS Inst. Inc., Cary, NC, USA). A repeated measures analysis of covariance with weight as a covariate, was run to determine the differences within variables. Week of sampling and age were fixed factors and animal was a random variable. Significant differences among means were tested by use of the Fisher's least significant differences method at p\<0.05. The model was differences were considered statistically different at p\<0.05. The model used was as follows: Y~ijk~ = μ+α~i~+β~j~+γ(W\*) +ɛ~ijk~; where; Y~ijk~ is the response variable (physiological and subjective parameters); μ is the mean; α~i~ is the effect of sampling week (i = 1, 2, 3, 4, 5); β~j~ is the effect of group (1, 2, 3); γ(W\*) is the adjusted covariate mean and ɛ~ijk~ is the standard error. Pearson's correlation coefficients were also determined using SAS. ### Experiment 2 Data were analysed by means of a repeated measures general linear model of SAS. The individual cow was a random variable and time was a repeated factor. Data for plasma cortisol was adjusted for normality using the log-transformation. Significance was set at p\<0.05 and values are given as means±standard error of the mean (SEM). To assess the relationship between plasma and saliva cortisol the CORR procedure of SAS was used. RESULTS AND DISCUSSION ====================== Experiment 1 ------------ The FGM were assayed and rectal temperatures were measured in addition to observing stress-related behaviour. The FGM and rectal temperatures were different throughout the sampling weeks ([Table 2](#t2-ajas-18-0151){ref-type="table"}). Repeated handling, however affected the physiological parameters, avoidance and crush behaviour as well as excretory behaviour. The FGM can be estimated to quantify the level of stress an animal is subjected to. We anticipated that with each handling encounter, FGM values would decrease. From our findings, the values were erratic, however, they are generally comparable to those reported by Xavier et al \[[@b31-ajas-18-0151]\]. When there is an overrepresentation of faeces of some individuals, FGM values can be inconsistent \[[@b32-ajas-18-0151]\]. Our study was however characterised by individual identification of each faecal sample, making the source of variability relatively indeterminable. In the same way, we recorded variable rectal temperature values, though they were analogous to those reported by Gruber et al \[[@b15-ajas-18-0151]\] (range: 38.3°C to 40.8°C). Restlessness as a result of movements associated with the experimental procedures could have triggered irregular blood flow thereby stimulating erratic rectal temperatures. In addition, variable responses of the different individuals could have contributed to erraticism. Furthermore, Grandin \[[@b14-ajas-18-0151]\] highlighted that such inconsistencies in the physiological parameters can be attributed to fear, a psychological stressor. Fear can be regarded as a reaction to imaginary danger \[[@b33-ajas-18-0151]\], whereby an animal is fearful of that which is perhaps absent or non-existent. As cattle are gregarious animals, they are vulnerable to exhibit fear of novelty. Accordingly, the general isolation of each individual animal from its respective group for the avoidance test and restraint in the crush could have elicited such kind of fear, especially in the first weeks \[[@b4-ajas-18-0151]\]. Fear of novelty or the human-animal interaction could have simultaneously triggered excretory behaviour in the first weeks \[[@b34-ajas-18-0151]\]. Perhaps, this could be the reason why the Pearson's correlation test, indicated a significant positive relationship between AD test and urination, though it was moderately weak ([Table 3](#t3-ajas-18-0151){ref-type="table"}). However, we could not establish why excretory behaviour increased as animal behaviour became more docile while the animals were restrained in the crush. On the other hand, results imply that animals that avoid human approach at longer distances (\<3 m away but ≥2 m away) have a tendency of being aggressive in the chute and the crush. Results for the avoidance test and restraint in the crush showed some degree of animal adaptation to human approach and handling ([Table 2](#t2-ajas-18-0151){ref-type="table"}). Animals can habituate if they can control or predict the stressor. The animals might have become familiar with the handler and the handling procedure, thereby becoming more at ease with being approached as indicated by a decrease in the ADs. In addition, cows could have learned that after restraint in the crush, they were to be released, thus elucidating calmer behaviour during the last weeks of handling. We anticipated that group 1 cows would be more excitable during handling than their cohorts, assuming that they have the least human-animal and restraint experience \[[@b5-ajas-18-0151]\]. Such behaviour was only observed whilst the cows were handled in the chute. Their behaviour was however not maintained through the other procedures and therefore we assume that cows can for one reason or the other, react variably even when subjected to uniform environmental conditions. On the other hand, group 2 cows excreted urine less than (p\<0.05) the other cows ([Table 2](#t2-ajas-18-0151){ref-type="table"}). This could imply that cows in this group were generally less stressed than their cohorts. However, resolute conclusions could be made if the other parameters showed a similar trend. Experiment 2 ------------ [Figure 2](#f2-ajas-18-0151){ref-type="fig"} illustrates results for plasma and saliva cortisol. Data are presented as means±SEM. Both plasma and saliva cortisol concentrations were expected to escalate after the hormonal challenge and decrease thereafter to baseline values. Basal plasma cortisol concentrations did not significantly increase (p\>0.05) after ACTH administration ([Figure 2](#f2-ajas-18-0151){ref-type="fig"}). Endogenous secretion of ACTH could have occurred at its peak due to the handling involved in the study. In contrast, saliva cortisol concentrations significantly increased (p\<0.05) 30 min after the challenge, suggesting that there is an association between the HPA axis and saliva cortisol. Similar findings for saliva were reported for dairy cows \[[@b17-ajas-18-0151],[@b22-ajas-18-0151]\]. These early studies however observed lower baseline values for both plasma and saliva cortisol than those reported in this study. Schwinn et al \[[@b22-ajas-18-0151]\] assayed saliva cortisol concentrations of about 2.79 to 19.45 nmol/L (converted) and 6.32 to 60.31 nmol/L for plasma cortisol whereas Negrão et al \[[@b17-ajas-18-0151]\] reported baseline values of approximately 5.52 to 8.28 nmol/L for plasma cortisol. It was however stated by Dunn \[[@b34-ajas-18-0151]\] that extreme stress would cause plasma cortisol values of 256.59 nmol/L (93 ng/mL) in cattle, suggesting that values obtained in our study (552.12±106.617 nmol/L, plasma cortisol and 8.75±15.869 nmol/L, saliva cortisol) were generally high. Activation of the HPA axis begins during restraint and handling \[[@b7-ajas-18-0151]\]. From the time when the animals were restrained in the pen to when they were in the crush pen, both plasma and saliva cortisol levels could have risen. Restraint itself stimulates a dramatic increase in cortisol \[[@b5-ajas-18-0151]\], especially with respect to plasma, and this could have occurred in our study. The cows exhibited high avoidance behaviour as they at many occasions retreated from entering the crush pen; therefore, fear a psychological stressor might have contributed to elevated baseline values. High cortisol values are expected in animals that do not willingly enter a crush \[[@b5-ajas-18-0151]\]. These could be reasons why basal plasma cortisol values were already high before hormonal stimulation. Repeated handling thereafter might have then led to unrestrained stress levels. It is also important to note that adaptation to stress depends on the intensity and type of the handling experience \[[@b3-ajas-18-0151]\], hence, repeated venepuncture could have contributed to confounding effects on cortisol levels \[[@b8-ajas-18-0151]\]. Total plasma cortisol increase is an indicator that an animal is exposed to stress. However, the free plasma cortisol fraction accounts for the exponential rise, thus defining the biological activity of plasma cortisol. Saliva sampling is done to determine cortisol levels as a reflection of the free plasma state \[[@b16-ajas-18-0151]\]. In saliva cortisol, only the biologically active fraction is measured \[[@b7-ajas-18-0151]\] and the steroid concentrations only reflect those in the free fraction of plasma. Results for this study, suggest that plasma cortisol could be a reflection of the total plasma cortisol fraction. This could have been coupled with its volatility making baseline values difficult to determine. In addition, response to stress is governed by the interplay of aggressive previous handling experiences and genetic factors such as temperament. Thus, cattle that are naturally excitable, have difficulty in adapting to handling procedures. These excitable cattle are associated with increased physiological responses \[[@b14-ajas-18-0151]\]. This tallies with observations that cows used in this study displayed aggression such as shaking the chute during sampling. Salivary cortisol is associated with a pronounced activation of the HPA axis and is correlated to plasma cortisol \[[@b22-ajas-18-0151]\]. Our study, to a large extent, was consistent with the assertion indicated by the pattern for saliva cortisol. However, due to varying patterns in saliva and plasma cortisol upon ACTH challenge, we observed a weak insignificant correlation (r = 0.0131, p = 0.9310) between saliva and plasma cortisol concentrations. Thus, our results for the correlation were inconsistent to findings by other researchers \[[@b17-ajas-18-0151],[@b22-ajas-18-0151],[@b23-ajas-18-0151]\]. Weak correlations between the two media were however reported in dairy cattle subjected to feeding and drinking actions \[[@b22-ajas-18-0151]\]. In the same way, the delay to reach a steady state between plasma and saliva cortisol as suggested by Hernandez et al \[[@b23-ajas-18-0151]\] was not observed in our study, consistent with the reports by Schwinn et al \[[@b22-ajas-18-0151]\]. While the fact that the breed used is an aggressive breed, the predictor variables of the relationship between plasma and saliva cortisol could be a combination of independent and mediating factors such as causal activity. However, how animals respond and cope with handling or restraint, cannot be easily understood due to the complex physiological and biochemical responses involved \[[@b15-ajas-18-0151]\], something to be considered in our study. In conclusion, no significant relationship was observed between serum and salivary cortisol levels in beef cows that were handled prior to sampling. Cows that are fearful of being approached by handlers, can exhibit excitable behaviour in the crush, thereby influencing physiological response. This study can therefore be regarded as a preliminary for more studies in beef cows. **CONFLICT OF INTEREST** We certify that there is no conflict of interest with any financial organization regarding the material discussed in the manuscript. We thank the Nguni cattle Project (University of Fort Hare) for funding this project. The authors would like to thank Dr Neville Pitts at the Witwatersrand University for his assistance with ELISA and RIA analyses for this study. The University of Fort Hare Research farm staff and manager, Mr D. Potgieter as well as Mr D. Pepe are highly acknowledged for their support. Authors thank the National Research Foundation (NRF) of South Africa for providing a scholarship for the primary author. ![Line graphs showing the average maximum ( ![](ajas-18-0151i1.jpg)) and minimum ( ![](ajas-18-0151i2.jpg)) environmental temperatures during the five sampling days for Experiment 1. The temperatures increased then decreased as indicated by the patterns of graphs and error bars represent standard error of means.](ajas-18-0151f1){#f1-ajas-18-0151} ![Cortisol concentrations of cows subjected to a hormonal challenge using a standardised dose (1 μg/kg) of adrenocorticotropic hormone at time 0. Line graphs represent plasma cortisol ( ![](ajas-18-0151i1.jpg) PlCortisol) and salivary cortisol ( ![](ajas-18-0151i2.jpg) SaCortisol) concentrations. Line graphs represent cortisol concentrations which either increased or decreased as depicted by the down-up or up-down lines respectively and the error bars represent the standard error of means. Insignificant differences were observed for plasma concentrations in contrast to the saliva concentrations which were variable (p\<0.001).](ajas-18-0151f2){#f2-ajas-18-0151} ###### Avoidance distance, chute and crush scores used for this study Category Score --------------------------------------------------------------------------------------- ----------------- Responses to an approaching experimenter Avoidance score Animal avoided the approaching person at a distance \<3 m away but ≥2 m away 1 Animal avoided the approaching person at a distance \<2 m but ≥1 m away 2 Animal avoided the approaching person at a distance \<1 m but ≥0 m away 3 Behaviour displayed in the chute Chute score No resistance to movement 1 Moves entire body to and fro rhythmically at least twice 2 Uneasy, head, body, tail and feet movements, hides head under another cow's abdomen 3 Active escape behaviour 4 Rears and needs encouragement to move forward 5 Stupor/refusing to move requiring force to probe forward 6 Behaviour exhibited in the crush (modified from Geburt et al \[[@b28-ajas-18-0151]\]) Crush score Cow is calm with no movement 1 Cow is slightly excited with minimal movement 2 Cow is nervous and occasionally shakes the crush 3 Cow is agitated and continuously moves and shakes the crush 4 Cow is very agitated and exhibits escape behaviour, animals may kneel/fall 5 ###### Effect of sampling week and age category on the physiological and behavioural parameters[1)](#tfn2-ajas-18-0151){ref-type="table-fn"} Variable Sampling and observation week SEM Group p-value ---------- ---------------------------------------------------- --------------------------------------------------- ---------------------------------------------------- ---------------------------------------------------- ---------------------------------------------------- ------- -------------------------------------------------------- -------------------------------------------------------- -------------------------------------------------------- ---------- -------- -------- FGM 23.84[bc](#tfn3-ajas-18-0151){ref-type="table-fn"} 19.08[c](#tfn3-ajas-18-0151){ref-type="table-fn"} 30.57[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 26.40[ab](#tfn3-ajas-18-0151){ref-type="table-fn"} 23.44[bc](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.776 23.84±1.249 25.10±1.444 25.06±1.464 0.0008 0.5978 0.6350 Rect T 37.88[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 38.76[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 38.30[ab](#tfn3-ajas-18-0151){ref-type="table-fn"} 37.78[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 38.12[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 0.207 38.09±0.146 38.19±0.168 38.22±0.171 0.0052 0.8071 0.2400 ADS 1.20[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.21[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.38[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.47[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.51[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 0.053 1.28±0.038 1.35±0.043 1.43±0.044 \<0.0001 0.0841 0.1250 Chute 2.00 1.88 1.84 1.75 1.77 0.073 1.99[a](#tfn3-ajas-18-0151){ref-type="table-fn"}±0.052 1.78[b](#tfn3-ajas-18-0151){ref-type="table-fn"}±0.060 1.78[b](#tfn3-ajas-18-0151){ref-type="table-fn"}±0.060 0.1540 0.0084 0.3539 Crush 1.85[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.87[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.63[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.56[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.47[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 0.067 1.67±0.047 1.58±0.054 1.77±0.055 \<0.0001 0.0514 0.1829 Urin 1.02[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.13[ab](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.10[ab](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.21[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.21[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 0.040 1.11[b](#tfn3-ajas-18-0151){ref-type="table-fn"}±0.028 1.22[a](#tfn3-ajas-18-0151){ref-type="table-fn"}±0.033 1.08[b](#tfn3-ajas-18-0151){ref-type="table-fn"}±0.033 0.0065 0.0045 0.8710 Def 1.14[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.13[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.17[b](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.32[a](#tfn3-ajas-18-0151){ref-type="table-fn"} 1.24[ab](#tfn3-ajas-18-0151){ref-type="table-fn"} 0.044 1.19±0.031 1.23±0.036 1.18±0.036 0.0085 0.4836 0.1801 Vocal 1.09 1.10 1.17 1.15 1.17 0.045 1.10±0.031 1.21±0.036 1.10±0.036 0.5806 0.0531 0.9798 SEM, standard error of means; W×G, interaction of sampling and observation week and group; FGM, faecal glucocorticoid metabolites concentrations (ng/g); Rect T, rectal temperature (°C); ADS, avoidance score; Chute, chute score; Crush, crush score, Urin, urinating score; Def, defecating score; Vocal, vocalisation score. Data are least squares (LS) means. Means within a row without a common superscript significantly differ (p\<0.05). ###### Pearson correlation coefficients for parameters Variable FGM RT Avoidance Chute Crush Urinating Defecating ------------ -------- -------- ------------------------------------------------------- ---------------------------------------------------- ------------------------------------------------------- ----------- ------------ RT 0.102 \- \- \- \- \- \- Avoidance −0.003 −0.024 \- \- \- \- \- Chute −0.011 0.135 −0.372[\*\*](#tfn6-ajas-18-0151){ref-type="table-fn"} \- \- \- \- Crush −0.125 −0.017 −0.351[\*\*](#tfn6-ajas-18-0151){ref-type="table-fn"} 0.241[\*](#tfn5-ajas-18-0151){ref-type="table-fn"} \- \- \- Urinating −0.022 −0.004 0.280[\*](#tfn5-ajas-18-0151){ref-type="table-fn"} −0.132 −0.335[\*\*](#tfn6-ajas-18-0151){ref-type="table-fn"} \- \- Defecating 0.099 −0.035 0.188 −0.117 −0.319[\*](#tfn5-ajas-18-0151){ref-type="table-fn"} 0.156 \- Vocalising −0.011 −0.077 0.067 −0.198 −0.077 0.127 0.027 FGM, faecal glucocorticoid metabolites concentrations; RT, rectal temperature; Avoidance, avoidance distance test score; Chute, chute score; Crush, crush score; Urinating, urinating score; Defecating, defecating score; Vocalising, vocalisation score. p\<0.05, p\<0.01.

{ "pile_set_name": "PubMed Central" }

Introduction {#Sec1} ============ Governments and health organizations worldwide have published dietary guidance for sugar intake \[[@CR1]\]. Despite access to the same published literature, recommendations vary greatly and create confusion for health practitioners and consumers. Since 1980, Dietary Guidelines for Americans (DGA) has recommended we "avoid too much sugar", yet dietary advice has typically recommended foods high in sugar, such as fruits and dairy products. As a way to clarify the types of sugar to avoid, the terms added sugars and free sugars are used. Added sugar recommendations have been in existence since 2002, with recent recommendations becoming progressively more restrictive over the years \[[@CR1]\]. This paper addresses current and proposed added sugar recommendations and assesses their practicality within the United States. Definition of added sugars {#Sec2} ========================== No universally accepted definition for added sugars exist (Table [1](#Tab1){ref-type="table"}). The Food and Drug Administration (FDA) classifies added sugars as, mono and disaccharides added to foods during production including sugars, syrups, fruit juice concentrates, honey, etc. This would not include sugars that naturally exist in foods, such as sugars in fruits or dairy products \[[@CR2]\]. A common point of contention between institutions is whether or not fruit juice should be included as added sugars. The proposed revisions to the Nutrition Facts and Supplements Label published by the FDA in 2014 classifies fruit juice concentrate added to food products as added sugar, while juice not from concentrate as not added sugar. In comparison, the USDA recommendations do not specify that fruit juice from concentrate contributes to added sugar totals \[[@CR2], [@CR3]\].Table 1Added sugar definitions and distinctions from various agenciesFood and Drug Agency (FDA) Proposed Nutrition Facts Label \[[@CR3]\]Mono- and disaccharides added to foods during production including: sugars, syrups, fruit juice concentrates, honey, etc.United States Department of Agriculture (USDA) Choose MyPlate \[[@CR2]\]Sugars added during processing and preparation. Includes: sugars, syrups, honey, nectars, etc.Excludes: fruit juice and fruit juice concentratesWorld Health Organization (WHO) Free Sugar Guidelines \[[@CR4]\]Mono- and disaccharides added to foods during processing or by the consumer during preparation. Includes: Sugars, syrups, honey, fruit juice and fruit juice from concentrate.Scientific Report of the 2015 Dietary Guidelines Advisory Committee \[[@CR8]\]Sugars, syrups, isolated naturally occurring sugars (ex: fruit juice concentrate) and other caloric sweeteners In addition to various definitions for the term "added sugars", the World Health Organization (WHO) utilizes the term "free sugar". Free sugar is similar to added sugars, as it includes all sugars and syrups added to foods; however, free sugar also includes sugars naturally present in fruit juices and fruit juice concentrates \[[@CR4]\]. Free sugar includes sugars naturally found in fruit juice that is consumed as a beverage as well as fruit juices added to food products. Assessing added sugar intake and compliance with recommendations would be extremely difficult without a clear and established definition of the term "added sugar" and, specifically, how fruit juice should contribute to added sugar values. Function of added sugar {#Sec3} ======================= Added sugars are chemically identical to sugar that naturally occurs in food products \[[@CR1]\]. The body cannot distinguish the source of the nutrient and processes the sugar in the same way. Sugar may be added to food products for many reasons, the most obvious reason being adding sweetness and enhancing the palatability of foods. Although this function of sugar is often opposed and criticized, many American consumers would not find a number of "healthy" foods palatable without added sugar. Some examples include cranberries, yogurt and oatmeal. Nutrition professionals often encourage clients to consume these foods as part of a healthy diet, even with some added sugar \[[@CR5]\]. Another function of sugar within food products is texture enhancement. Sugar produces a tender texture in baked products, and inhibits ice crystallization in frozen products. Sugar provides body to products and, when removed, has to be substituted with bulking agents to achieve a similar mouth feel \[[@CR6]\]. Carmelization and maillard browning are both reactions specific to sugar and provide an appearance expected in food products. Sugar also plays a role in food safety by inhibiting the growth of microorganisms at high concentrations. By binding with water molecules, sugar can maintain moisture contents in products lengthening the shelf life \[[@CR6]\]. Overall, it is important to remember that sugar functions in many capacities beyond just flavor. Added sugar intake in the American diet {#Sec4} ======================================= Added sugar intake is on average 13 % of total energy intake in adults and 16 % in children, consistently decreasing with age \[[@CR7]\]. Added sugar consumption has declined in all age groups from NHANES data taken in 2001--2004 to data from 2007--2010. Meanwhile, rates of obesity did not mimic the decline over the same time period \[[@CR8]\]. According to NHANES data from 2009--2010, 47 % added sugars in the American diet come from beverages, 31 % from snacks and sweets, 8 % from grains, and 14 % from the categories of dairy, mixed dishes, condiments, fruits and fruit juice and vegetables combined \[[@CR8]\]. While there is room for improvement in the American diet, this decrease in added sugar intake is encouraging and understanding the main sources of added sugars provides a direction to focus our efforts. Added sugar recommendations in America {#Sec5} ====================================== In 2002, the Institute of Medicine (IOM) Dietary Reference Intakes recommended that less than 25 % of total energy should come from added sugars. The recommendation is based on the concept that foods containing high amounts of added sugars are typically high in calories and low in micronutrients \[[@CR9]\]. The idea that added sugars are "empty calories" is a commonly cited reason that added sugar recommendations are necessary. Diets containing a large amount of energy as "empty calories" can lead to micronutrient malnutrition or over consumption of calories. Consuming the daily recommendation of all nutrients within an individual's estimated energy requirement is challenging when the individual is consuming a large portion of his or her calories as empty calories. Repeated consumption of empty calories without compensation from other nutrients can lead to weight gain. The current 2010 Dietary Guidelines for Americans, includes solid fats and added sugars (SoFAS) in their recommendation of 5--15 % of total energy from solid fats and added sugars \[[@CR10]\]. Minimizing SoFAS consumption is encouraged to reduce excess calorie consumption and to replace foods high in added sugars with foods lower in added sugars and greater nutrient density. SoFAS consumption above the recommendation is considered to be incompatible with the USDA Food Patterns, likely exceeding calorie limits or obtaining inadequate micronutrient intake \[[@CR10]\]. The USDA Food Patterns were created to assist the public in following Dietary Guideline recommendations, providing amounts of food from each food group to achieve optimal nutrient intake \[[@CR11]\]. The USDA Food Patterns groups added sugars and solid fats together and recommends adult females and adult males to limit "empty calorie" intake to 120--250 calories per day and 160--330 calories per day, respectively, depending on caloric needs \[[@CR8]\]. Consumption of empty calories is typically above the current recommendations in all age groups; almost 90 % of Americans exceed the USDA food pattern recommendations \[[@CR8]\]. The evolution of the concept of discretionary calories (2005 DGAs) to empty calories (2010 DGAs) is explained by Nicklas and O'Neil \[[@CR12]\]. The authors also explain that the reduction of solid fats and added sugars is to remove calories from the diet, not because solid fats and added sugars are linked to negative health outcomes \[[@CR12]\]. The World Health Organization not only cites the effects of excess calories, but also the impact that sugar can have on dental health. The current World Health Organization recommendation of fewer than 10 % of total calories from free sugars was set in 2003 \[[@CR13]\]. However in 2015, WHO set a conditional recommendation suggesting that less than 5 % of total energy should come from free sugars \[[@CR4]\]. This conditional recommendation proposed by WHO is based on a positive association between free sugar intake and dental caries among children \[[@CR4]\]. Sugar consumption has been positively associated with risk of dental disease. According to a meta-analysis published in 2014, there is moderate evidence indicating that a free sugar intake less than 10 % of total calories was associated with decreased risk of dental caries \[[@CR14]\]. Further decrease in caries was seen in Japanese surveys, taken between 1959 and 1960, when free sugar intake approached 5 % of total calories \[[@CR15]\]. The area surveyed had low fluoride exposure so it may not be an accurate model to extrapolate to areas with good fluoride exposure in the United States. Although, the WHO states that all populations, regardless of fluoridation, could possibly see improvement in dental caries with decreased free sugar intake \[[@CR4], [@CR15]\]. Additionally, the sugar consumption data was calculated by looking at sugar consumption per capita, added sugar intake compared to incidence of dental carries for each individual was not known \[[@CR15]\]. The limitations of the Japanese studies prevented the WHO from setting a strong recommendation to consume fewer than 5 % of calories from free sugars \[[@CR4]\]. However, because dental caries occur throughout the lifespan, consuming fewer free sugars is estimated to have a cumulative effect and result in decreased dental problems later in life and no evidence of harm was seen in diets containing fewer than 5 % energy from free sugars \[[@CR4]\]. Dietary Guidelines for Americans 2015 {#Sec6} ===================================== The release of the Scientific Report of the 2015 Dietary Guidelines Advisory Committee (DGAC) in February 2015 brought further attention to added sugars. The DGAC report placed a large focus on added sugars, making it one of the five "cross-cutting topics" \[[@CR8]\]. The Committee reexamined the evidence surrounding the potential health effects of added sugars. The DGAC assessed the evidence that added sugar negatively impacts the health risks for obesity, type II diabetes, cardiovascular disease and dental carries. The DGAC determined, based on the available evidence, there was a strong correlation between added sugars and negative health risks. Most of the cited evidence examines the association between sugar sweetened beverage (SSB) consumption and the health risk rather than the consumption of added sugar from variety of foods \[[@CR8]\]. It is easier to count consumption of SSBS with food frequency instruments used in epidemiologic studies than to estimate total added sugar intake since few databases included information on added sugars. While SSB consumption may be the best method available for added sugar estimates, it is not without its limitations including possible confounding variables within the population. According to a recent study of over 12,000 participants, individuals reporting to consume one or more SSB per day were significantly more likely to smoke, consume fewer fruits and vegetables and report a sedentary lifestyle \[[@CR16]\]. No discussion of if these confounding variables were considered in the DGAC report \[[@CR8]\]. After an examination of the evidence and diet modeling, the DGAC suggested an appropriate intake of calories from added sugars to be between 4--6 % and set a maximal intake of 10 % total energy from added sugars \[[@CR8]\]. After a period of time allowing for comments from the general public, the USDA and Department of Health and Human Services will assess evidence behind the recommendation from the USDA to set the Dietary Guidelines for Americans 2015 added sugar recommendation \[[@CR8]\]. With this suggested restriction on added sugars, the DGAC recognizes that the logical consequence of removing added sugars from the diet and food products would be replacing the added sugars with low calorie sweeteners. However, the DGAC report advises against this replacement due to the minimal evidence regarding long-term effect of low calorie sweeteners. Instead, the DGAC encourages the replacement of sugar-sweetened beverages with water and does not suggest a replacement in food products. Removal of sugar from products will change the taste, texture and shelf-life of products due to the functions of sugar previously discussed \[[@CR6]\]. The sugar must be replaced with other ingredients and, if not low calorie sweeteners, what would be a better alternative? Evidence exists to support the use of low-calorie sweeteners in weight reduction \[[@CR8]\] and many consumers utilize this approach to support weight loss. The FDA recognizes artificial and low-calorie sweeteners as safe for consumption, and the Academy of Nutrition and Dietetics advises that non-nutritive sweeteners can fit into a healthy diet \[[@CR17]\]. Identifying alternative sweeteners or ingredients to produce comparable food and beverage products is essential in changing the consumption patterns in Americans. Taste is consistently the most important buying factor for most Americans, and without great tasting alternatives consumers are not likely to make dietary changes \[[@CR18]\]. Proposed addition of "Added sugar" to nutrition facts panel {#Sec7} =========================================================== Currently, there is no easy way for consumers, researchers or health professionals to track added sugar consumption and assess compliance with recommendations. Very few databases exist that calculate added sugars, and, due to the various added sugar definitions, the information obtained from these databases may result in a range of added sugar values. In March 2014, the FDA proposed changes to the Nutrition Facts Panels to assist consumers in making more educated food choices that would lead to a healthy diet consistent with Dietary Guidelines for Americans. The recent proposal to update the Nutrition Facts Panel advocates for the addition of an "Added sugars" category below the "Sugars" category, that would provide a way to track and compare added sugars \[[@CR19]\]. The proposed amendments to the food labels suggest displaying added sugar in grams. The DGAC report supports such changes to the food labels and recommends displaying added sugar values in grams, teaspoons and percent daily value \[[@CR8]\]. A supplemental proposed rule regarding the Nutrition Facts Panel was published in July of 2015. The FDA proposed to establish a less than 10 % Daily Reference Value (DRV) and to include the percent Daily Value (DV) on the Nutrition Facts Panel \[[@CR20]\]. The supplemental proposed rule cites the 2015 DGAC report as their basis for instituting an added sugar DRV. The proposed rule states that the 2015 DGAC showed a "strong association between a dietary pattern of intake characterized, in part, by a reduced intake of added sugars and a reduced risk of cardiovascular disease" \[[@CR20]\]. Traditionally, DRVs and %DVs have been established for nutrients where an average dietary requirement can be determined from available scientific evidence \[[@CR21]\]. The data used to determine the \<10 % DRV for added sugar was based primarily on diet modeling conducted for the 2015 DGAC. No DRV is has been proposed for total sugars at this time due to lack of available evidence for a reference intake \[[@CR8], [@CR20]\]. The purpose of the FDA's changes to the Nutrition Facts Panel is to help consumers make choices leading to healthier diets, however; the addition of the "added sugar" category may not provide much novel knowledge to consumers. According to NHANES 2009--2010, nearly 80 % of added sugars come from sugar-sweetened beverages (47 %) and snacks and sweets (31 %) \[[@CR8]\]. The proposed changes to the food labels would require food companies to invest their resources to calculate the added sugar in their products, when the majority of added sugar consumed comes from obvious sources of sugar. Just over 20 % of added sugars consumed by Americans come from non-obvious forms where the consumer would benefit from the knowledge of added sugars on the food labels, if they choose to read the label \[[@CR8]\]. A study presented by the International Food and Information Council showed that the addition of the category "Added Sugars" to the Nutrition Facts Panel reduced the consumer comprehension of the food label. The percent of participants able to accurately identify the total grams of sugar dropped from 92 to 55 % when the added sugars category was included, with more than half the participants adding added sugars with the sugar category \[[@CR22]\]. A similar study was later conducted by the FDA, finding consistent results. Ability to accurately identify the grams of sugar per serving decreased from 81 % to 65 % when the label was updated to the proposed format \[[@CR23]\]. Other research supports that consumers are interested in added sugar labeling. Kyle & Thomas report that consumers believe Nutrition Facts labeling for added sugar will be more helpful than confusing \[[@CR24]\]. A study in European Union found that consumers expect that a reduction in free sugars in a product will be linked to a reduction in the calorie content of the food \[[@CR25]\]. Nevertheless, a consumer study with cereals found that participants rated cereals containing "fruit sugar" as healthier than cereals containing "sugar", although there were no differences in nutrient content between the cereals \[[@CR26]\]. Total sugar analysis is challenging enough. When sugar content of commercial foods targeted to infants and children was conducted by a blinded laboratory analysis of accepted chemical methods, nutrient label data underestimated or overestimated actual sugar content routinely. The authors suggest that more effort should be made to standardize methods for sugar labeling of foods, especially foods targeted to children \[[@CR27]\]. Health Canada recently removed the added sugars category from their proposed nutrition facts table and included a 20 % DV for total sugar \[[@CR28]\]. Consumer research by the Canadian government found that information about carbohydrates and total sugars was confusing when the table included added sugars. It also found the % DV approach to be useful and easy to understand. They state: "*the proposal to declare the amount of added sugars was popular among consumers and health stakeholders (including health professionals). However, industry stakeholders questioned the scientific basis of requiring the declaration of added sugar given that the body metabolizes naturally occurring and added sugars in the same way. Similarly, the inability of analytical method to distinguish between naturally occurring and added sugars would contribute to significant compliance and enforcement challenges."* Because added sugars are not chemically different from intrinsic sugars, there is no way to analytically determine the amount of added sugar in a food product \[[@CR1]\]. Food manufacturers would have to calculate the added sugars based on the recipe in order to determine the added sugar content in each product every time the product is reformulated. Without a clear definition of added sugars the resultant labeling will likely be inconsistent. The FDA would require food companies to document, maintain and provide records on product composition to verify the published value of added sugars \[[@CR3]\]. Due to competition within the food industry and the proprietary nature of the formulations, food manufacturers would be very resistant to release such information. Moreover, each of these steps will require additional time, money and an acquired skill set that smaller food companies may not have the resources to comply with. Conclusion {#Sec8} ========== Excess calorie consumption can lead to weight gain and increased risk of obesity and obesity-related co-morbidities \[[@CR8]\]. Empty calories which include solid fats and added sugars play a role in this when consumed in abundance. Added sugars are low in nutrient density and calories from added sugars can add up quickly if the individual is not conscious of their diet. However, there is no evidence suggesting that excess calories from added sugars specifically are worse than excess calories from any other food source. Much of the evidence linking added sugars to chronic disease is done measuring sugar sweetened beverages rather than percent calories from all added sugars \[[@CR8]\]. With nearly half of added sugar consumption in America being attributed to sweetened beverages, perhaps encouraging healthy beverage alternatives to sugar sweetened beverages should be the focus, rather than zeroing in on all added sugars. Regulatory attempts to tax sugar sweetened beverages in countries, such as Mexico and communities in the US, including Berkeley, CA may increase tax revenue, but whether these more aggressive approaches can limit calorie intake and/or improve health outcomes await clinical trial results. Recommendations as low as 5 % total energy from free sugars are likely too restrictive for most Americans to achieve \[[@CR29]\]. Added sugars should be consumed at a minimum as they are often a source for surplus calories in the American diet; however, stringent recommendations and mandatory food labeling are likely not the most effective ways to reduce added sugar and excess calorie consumption. Education on healthy beverages, snack choices and portion sizes may be a better starting point for reducing empty calorie intake. DGA : Dietary Guidelines for Americans DGAC : Dietary Guidelines Advisory Committee DRV : Daily reference value DV : daily value FDA : Food and Drug Administration IOM : Institute of Medicine SoFAS : solid fats and added sugars SSB : sugar sweetened beverage WHO : World Health Organization **Competing interests** Jennifer Erickson, is a PhD student in Nutrition at the University of Minnesota working with Dr. Joanne Slavin. Joanne Slavin is a professor in the Department of Food Science and Nutrition, University of Minnesota. In the past 5 years, she has given 150 scientific presentations in 13 countries. Many of these meetings received sponsorship from companies and associations with an interest in carbohydrates and nutritive sweeteners. Her research group works mostly on satiety and gut health outcomes with dietary exposures including dietary fiber and protein. Her research funding for the past 5 years has included grants from General Mills, Inc., Tate and Lyle, Nestle Health Sciences, Kellogg Company, USA Rice, USA Pears, Minnesota Beef Council, Minnesota Cultivated Wild Rice Council, Barilla Company, USDA, American Egg Board, American Pulse Association, MNDrive Global Food Ventures, International Life Science Institute (ILSI), and the Mushroom Council. She serves on the scientific advisory board for Tate and Lyle, Kerry Ingredients, Atkins Nutritionals, Midwest Dairy Association and the Alliance for Potato Research and Education (APRE). She holds a 1/3 interest in the Slavin Sisters Farm LLC, a 119 acre farm in Walworth, WI. **Authors' contributions** JE and JS contributed to the planning, research and writing of this paper. Both authors read and approved the final manuscript.

{ "pile_set_name": "PubMed Central" }

1. Introduction {#sec1-materials-13-02906} =============== Ferromagnetic shape memory alloys (FSMA) are recently one of the most extensively studied group of modern smart materials \[[@B1-materials-13-02906],[@B2-materials-13-02906],[@B3-materials-13-02906]\]. Among them, NiMnGa-based Heusler compounds stand out as the most complex alloys due to their unique magnetomechanical properties, such as magnetic field induced strains \[[@B4-materials-13-02906],[@B5-materials-13-02906],[@B6-materials-13-02906]\], pseudoelasticity/superelasticity \[[@B7-materials-13-02906],[@B8-materials-13-02906],[@B9-materials-13-02906]\], magnetoresistance \[[@B10-materials-13-02906],[@B11-materials-13-02906]\] or magneto- and mechanocaloric effects \[[@B12-materials-13-02906],[@B13-materials-13-02906],[@B14-materials-13-02906],[@B15-materials-13-02906],[@B16-materials-13-02906]\]. All above mentioned properties are associated with the stress-, magnetic field- or temperature-driven first-order reversible phase transition, and undergo a change from a high symmetry austenitic phase to low symmetry martensitic phase \[[@B17-materials-13-02906]\]. Furthermore, a low temperature martensitic phase is characterized by strong magnetocrystalline anisotropy, which helps to induce and control the twin variant reorientation within the martensitic phase \[[@B18-materials-13-02906]\]. The strong correlation between the microstructure and the magnetic properties of the austenitic and martensitic phases leads to the abrupt drop of magnetization in the vicinity of the martensitic transformation. This significant difference in magnetization strongly influences the majority of magnetomechanical properties in NiMnGa-based materials. When considering potential multifunctional applications of FSMA, it is necessary to manipulate the magnetic behavior of the martensite and austenite phases. It is well documented that the magnetism of NiMnGa-based alloys varies significantly with composition in and near stoichiometric Ni~2~MnGa samples depending on the Ni \[[@B19-materials-13-02906],[@B20-materials-13-02906],[@B21-materials-13-02906]\], Mn \[[@B22-materials-13-02906],[@B23-materials-13-02906]\] or Ga \[[@B24-materials-13-02906],[@B25-materials-13-02906]\] concentration. However, it has been recently reported that an addition of the fourth alloying element may cause a significantly change in the magnetostructural properties of the existing phases \[[@B13-materials-13-02906],[@B26-materials-13-02906],[@B27-materials-13-02906]\]. The introduction of doping elements changes the stability of the austenite and martensite phases, which shifts the structural transformation temperatures and changes the temperature dependent magnetic behavior of doped NiMnGa-based materials. Moreover, alloying elements have different magnetic moments than primary Ni, Mn and Ga, which also influence the ferromagnetic properties of the material. Crucially, even a small amount of doping element can also notably shift the temperature of the martensitic transformation \[[@B28-materials-13-02906],[@B29-materials-13-02906],[@B30-materials-13-02906],[@B31-materials-13-02906],[@B32-materials-13-02906],[@B33-materials-13-02906],[@B34-materials-13-02906]\]. The present work aimed to investigate the influence of Ti and Gd substitution for Ga in polycrystalline Ni~50~Mn~25~Ga~20~X~5~ (X = Ti or Gd) FSMAs with a magnetic behavior of the martensite and austenite phases before, during, and after phase transitions. Due to the fact that magnetocrystalline anisotropy changes during structural transformation, we estimated the temperature dependence of anisotropy energy from the series of high field magnetization curves based on the law of approach to magnetic saturation \[[@B35-materials-13-02906]\]. 2. Materials and Methods {#sec2-materials-13-02906} ======================== The polycrystalline bulk samples with the nominal composition of Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ (at.%) were produced from high purity elements via an arc-melting method under a protective argon atmosphere. The melting procedure was repeated several times to ensure a high homogeneity of the samples. Despite the fact that Mn is a highly volatile element, the weight loss after the melting procedure was less than 1% for each sample. The produced ingots were subsequently vacuum-sealed in a quartz ampule and annealed at 1173 K for 5 h, followed by water quenching. Our previous studies \[[@B30-materials-13-02906]\] showed that Gd-doped NiMnGa-based FSMA require a higher annealing temperature, so to ensure a good homogeneity in the Ni~50~Mn~25~Ga~20~Gd~5~ sample, the alloy was further annealed at 1430 K for 3 h and then quenched in water. The nominal compositions of the Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ alloys were confirmed with the help of a Scanning Electron Microscope JEOL JSM-5800LV (JEOL Ltd., Akishima, Tokyo, Japan) supported by an energy-dispersive detector (EDS). The crystal structure identification was done on the basis of powder X-ray diffraction measurements at room temperature performed in the X-ray Diffractometer Rigaku MiniFlex 600 (Rigaku, Tokyo, Japan)) with CuKα radiation, followed by subsequent Rietveld analysis. The magnetic properties characterization was carried out using the Vibrating Sample Magnetometer (VSM) module from the VersaLab System (Quantum Design, San Diego, CA, USA). In order to reduce the influence of the demagnetizing field, the thin, needle-like samples (with a length to diameter ratio of more than 10) were used for the magnetic measurements. The temperature dependence of magnetization under an external magnetic field of 250 mT was measured from 50 K up to 400 K using zero-field cooled (ZFC) and field cooled (FC) protocols with heating and cooling rates of 10 K/min. Then, the series of magnetic hysteresis loops was also recorded in the same temperature range (50--400 K) under an external magnetic field of up to 2 T. 3. Results and Discussion {#sec3-materials-13-02906} ========================= 3.1. Microstructure Analysis {#sec3dot1-materials-13-02906} ---------------------------- [Figure 1](#materials-13-02906-f001){ref-type="fig"} depicts the observed and calculated XRD patterns obtained at room temperature for the investigated Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ alloys. In order to clearly show the slight deviation of the peak locations in each pattern, only three peaks with the highest intensity are presented in [Figure 1](#materials-13-02906-f001){ref-type="fig"}. The presented reflections were assigned to a standard cubic L2~1~-type austenitic structure with an Fm3m space group (No. 225). In this case Ni, Mn and Ga occupied the 8*c* (0.25, 0.25, 0.25), 4*a* (0, 0, 0) and 4*b* (0.5, 0.5, 0.5) Wyckoff atomic positions, respectively. The lattice parameters calculated from the Rietveld analysis are as follows: *a* = 5.826 Å for Ni~50~Mn~25~Ga~25~, *a* = 5.848 Å for Ni~50~Mn~25~Ga~20~Ti~5~ and *a* = 5.832 Å for Ni~50~Mn~25~Ga~20~Gd~5~ alloys. It can be seen that an addition of doping elements (Ti or Gd) leads to the slight expansion of the austenitic unit cell. Moreover, in the case of the Ni~50~Mn~25~Ga~20~Gd~5~ sample, additional peaks were observed in the XRD pattern. These reflections were assigned to a residual hexagonal Ga~1.5~Gd~1~Ni~3.5~ phase (space group P6/mmm, No. 191) precipitated during the annealing process. The lattice parameters calculated for this phase are *a* = 4.980 Å, *c* = 4.107 Å and *γ* = 120°. 3.2. Temperature Dependence of Magnetization {#sec3dot2-materials-13-02906} -------------------------------------------- [Figure 2](#materials-13-02906-f002){ref-type="fig"} shows the temperature dependence of magnetization for the investigated Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ alloys. A notable change of magnetization occurred in the vicinity of the phase transition in every investigated sample, both during heating and cooling procedures. The substantial hysteresis between the temperature of the martensitic transformation (during cooling) and the austenitic transformation (during heating) was very characteristic for the first-order temperature-driven transformation, especially in polycrystalline materials where the transition takes place in every single grain. Moreover, the martensitic transformation temperature was strongly dependent from the Gibbs free energy difference between the austenite and martensite phases and the addition of alloying elements may have significantly changed these energies \[[@B36-materials-13-02906],[@B37-materials-13-02906],[@B38-materials-13-02906]\]. In order to compare the transformation temperatures between the studied alloys, the average value of all temperatures (*M~s~*, *M~f~*, *A~s~*, *A~f~*) was calculated as:$$T_{M} = \frac{1}{2}\left( {\frac{M_{s} + M_{f}}{2} + \frac{A_{s} + A_{f}}{2}} \right),$$ where *T~M~* is structural transformation temperature, *M~s~* and *M~f~* are martensitic transformation start and finish temperature, and *A~s~* and *A~f~* are austenitic transformation start and finish temperature. As was expected, both Ti and Gd substitutions for Ga notably shifted the phase transformation temperatures (*T~M~*). Ti slightly decreased the *T~M~* from 193 K to 173 K for Ni~50~Mn~25~Ga~25~ and Ni~50~Mn~25~Ga~20~Ti~5~ alloys respectively. On the other hand, the same atomic amount of Gd substantially increased the *T~M~* to 279 K for the Ni~50~Mn~25~Ga~20~Gd~5~ alloy. These results suggest that the addition of Ti decreases the overall free energy difference between the martensite and austenite phases, which subsequently results in a decrease of the martensitic transformation temperature. In a similar manner, the addition of Gd notably increases the free energy difference between the austenite and martensite phases, which leads to a considerable increase in the martensitic transition temperature. Moreover, in the case of the Gd-doped sample, the temperature of the phase transition was very close to the room temperature; this may be an advantage in potential future applications of this material. Furthermore, the influence of Ti and Gd on magnetic transformation is significantly different than it is on structural transitions. The Curie temperature (calculated as an average value of the peak location on the first derivative's *dM*/*dT* curves established during the heating and cooling protocol) dropped substantially in the case of the Ti-doped sample (*T~C~* = 373 K and 312 K for the Ni~50~Mn~25~Ga~25~ and Ni~50~Mn~25~Ga~20~Ti~5~ alloys, respectively). When it comes to the Ni~50~Mn~25~Ga~20~Gd~5~ alloy, the Curie temperature remained almost unchanged (*T~C~* = 374 K). 3.3. Law of Approach to Magnetic Saturation {#sec3dot3-materials-13-02906} ------------------------------------------- The well-known law of approach (LoA) to magnetic saturation was used to analyze the magnetic behavior of all of the studied alloys and to estimate the temperature dependence of their effective magnetocrystalline anisotropy \[[@B35-materials-13-02906]\]. However, in order to properly use the LoA model, the authors made some important assumptions. First of all, only the high field magnetization curves (*H* ≫ *Hc*) were taken into account as initial curves. Next, after the first calculations were made considering the fitting procedure accuracy, the high field magnetization curves were limited to the last 5% of magnetization (*M* \> 0.95*M~s~*). Further, in this applied range of magnetization, recorded hysteresis loops were closed so that all hysteretic processes could be neglected, since further increase of magnetization was mainly induced by rotational processes of magnetic domains. Properly selected magnetization curves were then analyzed in terms of the law of approach to magnetic saturation \[[@B35-materials-13-02906]\]:$$M = M_{s}\left( {1 - \frac{a}{H} - \frac{b}{H^{2}}} \right) + \chi H,$$ where *M~s~* is saturation magnetization, coefficient *a* is related to the to the structural inhomogeneity of the material, non-magnetic inclusions or internal microstresses, and coefficient *b* is directly connected with the existence of magnetocrystalline anisotropy. The last term of the equation *χH* describes the high field spontaneous magnetization. Considering the fact that coefficient *a* plays a significant role only in the lower fields \[[@B35-materials-13-02906],[@B39-materials-13-02906]\] and its contribution at high magnetic fields could be neglected, Equation (2) may be simplified to:$$M = M_{s}\left( {1 - \frac{b}{H^{2}}} \right) + \chi H,$$ Equation (3) was used to fit to the selected part of the magnetization curves, which is depicted in [Figure 3](#materials-13-02906-f003){ref-type="fig"}. The open circles show the measured data and the solid lines represent the best fittings according to Equation (3). The significant difference between the magnetization behavior of the low temperature martensite phase and the high temperature austenite phase is observable for the all studied alloys. As expected, even though martensite is characterized by a higher magnetization at high values of an applied magnetic field, it saturates much slower than notably softer austenite. This reflects the fact that high magnetocrystalline anisotropy of the martensite phase decreases the initial permeability of the material at low magnetic fields. Coefficient *b* calculated from Equation (3) was used to estimate the effective magnetic anisotropy constant (*Keff*) from the following relation:$$b = c\left( \frac{K_{eff}}{\mu_{0}M_{s}} \right)^{2},$$ where *c* is a constant dependent from the crystal structure of the material and equals *c* = 8/105 for cubic anisotropy \[[@B35-materials-13-02906]\]. Thus, *K~eff~* can be finally estimated by utilizing the following formula:$$K_{eff} = \mu_{0}M_{s}\sqrt{\frac{105c}{8}},$$ [Figure 4](#materials-13-02906-f004){ref-type="fig"} presents the calculated effective magnetic anisotropy constant for all three Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ alloys. A significant drop of anisotropy (about 60%) in the vicinity of the phase transition was evident for every studied material. It was also clearly seen that the anisotropy energy decreased with the increase of temperature in both low temperature martensitic phases and high temperature austenitic phases (up to the Curie temperature where the material became paramagnetic). Moreover, the obtained values of anisotropy constant for the martensitic phase were the following: 9.6--8.5 × 10^5^ J/m^3^ for Ni~50~Mn~25~Ga~25~, 2.6--1.9 × 10^5^ J/m^3^ for Ni~50~Mn~25~Ga~20~Ti~5~ and 5.0--3.4 × 10^5^ J/m^3^ for Ni~50~Mn~25~Ga~20~Gd~5~ alloys; for the austenitic phase: 3.4--1.1 × 10^5^ J/m^3^ for Ni~50~Mn~25~Ga~25~, 0.7--0.3 × 10^5^ J/m^3^ for Ni~50~Mn~25~Ga~20~Ti~5~ and 1.4--0.7 × 10^5^ J/m^3^ for Ni~50~Mn~25~Ga~20~Gd~5~ alloys are in good agreement with the results reported for other NiMnGa-based materials \[[@B40-materials-13-02906],[@B41-materials-13-02906],[@B42-materials-13-02906],[@B43-materials-13-02906]\]. The relatively high values of magnetocrystalline anisotropy in the martensitic phase were essential for magnetic field induced strains. The obtained results suggested that the additions of Ti and Gd to NiMnGa-based compositions significantly decreased the effective anisotropy constant. This was particularly apparent in the case of the Ni~50~Mn~25~Ga~20~Ti~5~ alloy in which *K~eff~* decreased about three times in comparison to the undoped Ni~50~Mn~25~Ga~25~ sample. [Figure 4](#materials-13-02906-f004){ref-type="fig"} also depicts the temperature dependence of a coercive field obtained upon heating for Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ alloys. In this case, the structural martensitic transition was also visible as a distinct drop in coercivity, which further showed the evident differences between the magnetic softness of the austenitic and martensitic phases. What is more in general, the coercive field was strongly coupled with the magnetic anisotropy of the material. This phenomenon is clearly observable in [Figure 4](#materials-13-02906-f004){ref-type="fig"}, as the temperature dependence of the coercive field followed the same trend as the temperature dependence of effective anisotropy. Another comparison of the temperature dependence of anisotropy is presented in [Figure 5](#materials-13-02906-f005){ref-type="fig"}a, where an effective anisotropy constant was normalized by dividing it by a maximum value of *K~eff~* measured at 50 K; alternatively, the temperature axis was normalized by using the phase transformation temperature of each alloy. Thus, the dashed line at *T*/*T~M~* = 1 on [Figure 5](#materials-13-02906-f005){ref-type="fig"}a corresponds to the structural transformation. The presented normalization procedure revealed that despite the significant differences in *K~eff~* and *T~M~*, the overall temperature dependence of effective anisotropy behaved very similarly in all studied materials and was strongly connected with the structural phase transition. Moreover, [Figure 5](#materials-13-02906-f005){ref-type="fig"}b compares the normalized *K~eff~* with the normalized coercive field (*H*/*H~c(50K)~*). It was previously shown in [Figure 5](#materials-13-02906-f005){ref-type="fig"} that *K~eff~* is significantly related to the coercivity; this is even more evident now, in [Figure 5](#materials-13-02906-f005){ref-type="fig"}b. The two characteristic regions for magnetically softer austenite and significantly harder martensite were easily distinguishable for all studied alloys. 4. Conclusions {#sec4-materials-13-02906} ============== The influence of elemental doping with Ti and Gd on the temperature dependence of magnetic behavior in the Ni~50~Mn~25~Ga~20-x~Z~x~ (x = 0 or 5, Z = Gd, Ti) ferromagnetic shape memory alloys were studied in detail. The XRD analysis confirmed the cubic L2~1~ austenitic structure in all investigated materials and revealed that an addition of Ti or Gd elongates the lattice parameters of the austenite cell. Thermomagnetic measurements showed the strong influence of chemical composition on both martensitic and magnetic transformations. Following this fact, Ti addition to the Ni-Mn-Ga alloy significantly reduces the martensitic transformation temperature and Curie temperature in comparison to the Ni~50~Mn~25~Ga~25~ precursor. In the Gd-doped sample, the temperature of magnetic transition remained almost the same as in the Ni~50~Mn~25~Ga~25~ alloy, whereas the temperature of phase transition increased significantly, and came close to room temperature. Comprehensive studies in high magnetic fields based on the law of approach to magnetic saturation theory revealed the close correlation between magnetic and structural behavior in investigated samples. The abrupt drop of magnetocrystalline anisotropy at phase transformation temperatures and significant differences of anisotropies between martensitic and austenitic phases confirmed the completely different magnetic nature of the two existing phases in NiMnGa-based FSMA. What is more, the strong connection between the temperature dependence of the coercive field and the temperature dependence of magnetocrystalline anisotropy was also emphasized in this study. Conceptualization, A.Ł. and M.H.; methodology, A.Ł. and M.H.; formal analysis, A.Ł. and M.H.; investigation, A.Ł., M.H. and J.K.; data curation, A.Ł.; writing---original draft preparation, A.Ł.; writing---review and editing, A.Ł., M.H. and J.K.; supervision, J.K. All authors have read and agreed to the published version of the manuscript. This research received no external funding. The authors declare no conflict of interest. ![XRD patterns of the three peaks of the L2~1~ austenitic phase (**a**) (2 0 0), (**b**) (4 0 0) and (**c**) (4 2 2) obtained for the Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ alloys. The dotted line represents the measured values and the solid line shows the calculated pattern obtained from the Rietveld analysis. The additional peaks observed only in the Ni~50~Mn~25~Ga~20~Gd~5~ alloy are related to the residual Gd-rich Ga~1.5~Gd~1~Ni~3.5~ phase (space group P6/mmm).](materials-13-02906-g001){#materials-13-02906-f001} ![Temperature dependence of normalized magnetization for the (**a**) Ni~50~Mn~25~Ga~25~, (**b**) Ni~50~Mn~25~Ga~20~Ti~5~ and (**c**) Ni~50~Mn~25~Ga~20~Gd~5~ alloys measured under an external magnetic field of 250 mT. The arrows represent the austenitic transformation's start/finish temperatures (*A~s~*/*A~f~*) and the martensitic transformation's start/finish temperatures (*M~s~*/*M~f~*).](materials-13-02906-g002){#materials-13-02906-f002} ![High field parts of magnetization curves (*M* \> 0.95*M~s~*) for the (**a**) Ni~50~Mn~25~Ga~25~, (**b**) Ni~50~Mn~25~Ga~20~Ti~5~ and (**c**) Ni~50~Mn~25~Ga~20~Gd~5~ alloys measured from 50 K up to the Curie temperature. Open circles represent recorded data and solid lines show the best fitting to the law of approach to magnetic saturation (see Equation (3)).](materials-13-02906-g003){#materials-13-02906-f003} ![The temperature dependence of effective anisotropy constant *K~eff~* (circles) estimated from the law of approach to magnetic saturation (see Equation (5)) and temperature dependence of coercive field (squares) for the (**a**) Ni~50~Mn~25~Ga~25~, (**b**) Ni~50~Mn~25~Ga~20~Ti~5~ and (**c**) Ni~50~Mn~25~Ga~20~Gd~5~ alloys.](materials-13-02906-g004){#materials-13-02906-f004} ![The normalized effective anisotropy constant (*K~eff~*/*K~eff(50K)~*) (**a**) as a function of normalized temperature (*T*/*T~M~*); the dashed line at *T*/*T~M~* = 1 shows the structural transformation, (**b**) as a function of the normalized coercive field (*H*/*H~c(50K)~*) for the Ni~50~Mn~25~Ga~25~, Ni~50~Mn~25~Ga~20~Ti~5~ and Ni~50~Mn~25~Ga~20~Gd~5~ alloys.](materials-13-02906-g005){#materials-13-02906-f005}

{ "pile_set_name": "PubMed Central" }

INTRODUCTION {#S1} ============ Psoriasis is a chronic immune-mediated inflammatory skin disease, affecting 1% to 2% of the Caucasian population.^[@R1]^ A pathogenic cross-talk between innate and adaptive cells, including keratinocytes, dendritic cells, and T-cells, underpins a dysregulated immune response leading to abnormal epidermal proliferation.^[@R2]^ Although psoriasis was initially considered as a Th1-mediated disease, mounting evidence from preclinical results, genetic data and clinical trials have demonstrated a key contribution of IL-17-producing T-cells to psoriasis in humans.^[@R3]^ Expression of IL-17A, IL-17F, and IL-17C is elevated in psoriatic lesional tissue compared with non-lesional tissue.^[@R4]^ IL-17A, IL-17F and IL-17C act directly on keratinocytes to stimulate the production of a number of molecules known to be elevated in psoriasis lesional tissue such as cytokines; antimicrobial peptides (AMPs); and chemokines such as IL-8, CCL20 and CCL2, thereby enabling IL-17 to bridge the innate and adaptive immune systems to sustain chronic inflammation.^[@R5],\ [@R6]^ The IL-17-signaling pathway utilizes a ubiquitin ligase signaling adaptor, Act1 (NF-κB activator 1), which is also known as CIKS (connection to IKK and SAPK/JNK), to propagate downstream signaling events.^[@R7]--[@R9]^ The binding of IL-17 to the heterodimeric IL-17 receptor (IL-17R) leads to Act1 recruitment through a homotypic SEFIR domain-dependent interaction. This results in incorporation of TNF receptor-associated factor 6 (TRAF6) into the signaling complex and the subsequent downstream activation of the NF-κB and MAPK pathways, leading to the induction of target genes in keratinocytes, epithelial cells, and fibroblasts stimulated with IL-17 family cytokines.^[@R9]--[@R11]^ The role of Act1 as a signal transducer is not confined to IL-17-responsive cells, as earlier studies found that Act1 is recruited to CD40 (TNF receptor superfamily member 5) and BAFFR (TNF receptor superfamily member 13C) upon stimulation of murine B-cells with CD40 ligand (CD40L, TNF superfamily member 5) and BAFF (tumor necrosis factor superfamily member 13b), respectively.^[@R12]^ One of the most robust genetic associations with psoriasis identified to date involves a variant in the *TRAF3IP2* gene (rs33980500 C/T, which specifies a glutamic acid (D) to asparagine (N) change in Act1 (pD10N). This variant is unusual among complex disease susceptibility signals^[@R13]^ in that it specifies an amino acid change that appears to be functionally relevant. Moreover, this variant is associated with both cutaneous psoriasis and psoriatic arthritis^[@R14]--[@R17]^. Although this genetic association has been independently replicated, the manner in which the Act1 pD10N variant predisposes patients to psoriasis remains to be fully elucidated. An enigmatic feature of this association is that Act1 D10N appears to behave as a loss-of-function variant in acute signaling responses through IL-17R and CD40, whereas it appears to be pro-inflammatory in more broadly biological contexts. This enigma extends to mouse models in which Act1 has been silenced (see Discussion). In order to better understand the functional consequences of the Act1 D10N variant, we have undertaken an approach in which individuals of known Act1 genotype obtained through our previous GWAS studies of psoriasis^[@R14]--[@R17]^ were re-contacted and asked to provide biological samples for further functional analysis of Act1 genetic variation. Short-term responses, such as signal transduction responses, are desirable for such an analysis, because longer-term cellular responses involve the interaction of increasing numbers of proteins, whose functional variants cannot easily be controlled for due to random segregation of unlinked genes. To this end, we profiled peripheral blood mononuclear cells (PBMC) for Act1 expression and responsiveness to IL-17, making use of phospho-flow cytometry to measure short-term signaling responses. However, although multiple PBMC subsets expressed Act1, we were unable to identify any robust signaling responses to IL-17. Because Act1 has also been implicated in signaling events downstream of CD40^[@R12]^, we assessed the impact of the Act1 D10N variant on CD40L-stimulated B-cell signaling events in PBMCs from individuals homozygous, heterozygous, or nullizygous for the Act1 D10N allele. RESULTS {#S2} ======= To identify candidate assays for functional genetic testing of the Act1 D10N variant, we assessed Act1 protein levels in different PBMC subsets by flow cytometry. CD19^+^ B-cells, CD3^+^CD4^+^ ("helper") T-cells, CD3^+^CD8^+^ ("cytotoxic") T-cells and CD14^+^ monocytes were gated as shown in [Figure 1a](#F1){ref-type="fig"}. The percentages of Act1-positive cells were determined within each PBMC subset ([Figure 1b](#F1){ref-type="fig"}). We found the highest percentage of Act1-positive cells in monocytes (90.9% of CD14^+^ cells), with lower percentages in T-cells and B-cells (41.5% of CD4^+^ T-cells, 36.3% of CD8^+^ T-cells, and 54.9% of CD19^+^ B-cells, [Figure 1b](#F1){ref-type="fig"}). We also assessed Act1 expression levels by subtracting the median fluorescence intensity (MFI) of the isotype control mAb from that of the anti-Act1 Ab for each cell population ([Figure 1c](#F1){ref-type="fig"}). As assessed by MFI, all cell types expressed Act1, with CD14^+^ monocytes again expressing the highest levels, followed by CD4^+^ T-cells, CD8^+^ T-cells, and B-cells ([Figure 1c](#F1){ref-type="fig"}). The transcription factor NF-κB is a central mediator of immune and inflammatory responses.^[@R18]^ NF-κB is reported to be a downstream target of both IL-17R and CD40 signaling.^[@R12],\ [@R19],\ [@R20]^ To explore whether IL-17R and/or CD40 triggering leads to NF-κB activation in PBMC subsets, we combined surface marker phenotyping with intracellular assessment of phospho-NF-κB p65 (pp65) to assess IL-17A or CD40L-induced signaling by phospho-flow cytometry. As shown in the upper left panel of [Figure 2](#F2){ref-type="fig"}, pp65 was increased after CD40L stimulation for 15 minutes in B-cells but not in T-cells or monocytes. In contrast, no increase in pp65 was observed after 15 min of stimulation with 100 ng/ml or 200 ng/ml IL-17A in any of the IL-17A-treated PBMC subsets that we analyzed ([Figure 2](#F2){ref-type="fig"}, middle panels). In contrast, the levels of pp65 increased in all subsets when PBMCs were treated with phorbol myristate acetate (PMA) for 15 min as a positive control ([Figure 2](#F2){ref-type="fig"}, lower panels). Previous studies in Act1-deficient mice demonstrated a general increase in the numbers of peripheral B-cells.^[@R12],\ [@R21]^ However, Act1-deficient mice on a different background did not show this phenotype.^[@R22]^ To ask whether the same phenomenon might be observed in humans as a function of the Act1 D10N variant, we assessed the percentage of B-cells in individuals homozygous, heterozygous, or nullizygous for the D10N allele. The frequency of CD19^+^ B-cells did not differ between the Act1 D10N homozygotes and nullizygotes, although it was modestly higher in Act1 D10N heterozygotes ([Figure 3b](#F3){ref-type="fig"}). As shown in [Figure 3a](#F3){ref-type="fig"}, the frequency of CD3^+^ T-cells also did not differ significantly as a function of Act1 D10N genotype. We next investigated the functional consequences of the SNP-D10N on CD40 signaling in B-cells. The downstream pathways activated by CD40 triggering include not only NF-κB but also mitogen-activated protein kinases (MAPKs).^[@R23],\ [@R24]^ We thus sought to assay the phosphorylation state of the MAPKs p38 and Erk as potential functional measures of CD40 signaling. PBMCs from individuals homozygous, heterozygous, or nullizygous for the D10N allele were stimulated with CD40L for 15 minutes, and phosphorylation of p65, p38 and Erk was assayed by flow cytometry in the CD19^+^ B-cell subset. CD40L stimulation of PBMCs from all three genotype groups resulted in increased phosphorylation of p65, p38 and Erk in CD19^+^ B-cells ([Figure 4a](#F4){ref-type="fig"}). We found that the increase of pp65, pp38 and pErk (expressed as fold-change in MFI relative to unstimulated cells) determined across individuals were significantly correlated with each other ([Figure 4b](#F4){ref-type="fig"}). To determine the impact of the D10N variant on CD40 signaling in B-cells, we compared the increases of pp65, pp38 and pErk induced by CD40L among the three genotype groups. As shown in [Figure 4c](#F4){ref-type="fig"}, CD40L-mediated increases in pp65, pp38 and pErk were significantly attenuated in CD19^+^ B-cells from Act1 D10N homozygotes, relative to D10N nullizygotes. Moreover, the pp38 and pErk responses were significantly smaller in D10N homozygotes, relative to D10N heterozygotes. While responses did not differ significantly in heterozygotes relative to nullizygotes, there was a significant linear trend for decreasing phosphorylation of p65, p38 and Erk as a function of decreasing SNP-D10N variant counts across genotypes (pp65: *P* \< 0.01; pp38: *P* \< 0.01; pErk: *P* \< 0.01). DISCUSSION {#S3} ========== Genetic association studies of psoriasis have implicated a number of genes that function in NF-κB, IL-23R, and IL-17R signaling.^[@R14]--[@R16],\ [@R25]--[@R27]^ Act1, encoded by *TRAF3IP2*, is of particular interest due the fact that a disease-associated *TRAF3IP2* SNP (rs33980500 C/T, encoding Act1 pD10N) appears to have functional impact^[@R15],\ [@R28]^, which may be due at least in part to its effect on the IL-17R signaling cascade.^[@R9]--[@R11]^ However, the direction of its effect seems paradoxical, in that the D10N mutation confers disease risk (i.e., is found at higher frequency in cases than in controls), yet it leads to loss of interaction with TRAF6^[@R15]^ and behaves as a hypofunctional allele in short-term signaling and gene expression assays.^[@R28]^ Reminiscent of the D10N mutation, Act1 knockout mice manifest many of the short-term signaling defects in IL-17 responses that are seen in Act1 D10N mice,^[@R7]^ and yet have been reported to manifest a skin inflammation phenotype characterized by epidermal hyperplasia and cell infiltration.^[@R12]^ Reports considering the effects of the Act1 D10N variant in the context of human cells are limited, with one previous report finding that fibroblasts from a D10N homozygous individual responded weakly to IL-17A.^[@R29]^ In the present study, we assayed Act1 expression and signaling responses to IL-17 treatment in different PBMC subsets by flow cytometry. As measured by both percentages of positive cells and by MFI, Act1 was expressed at the highest levels in monocytes. Lower levels of Act1 expression were observed in CD4^+^ T-cells, CD8^+^ T-cells and B-cells, though the rank order of expression depended upon the assay utilized ([Figure 1](#F1){ref-type="fig"}). However, despite the presence of Act1 in each cell type, PBMC-derived monocytes, as well as T-cells and B-cells, failed to respond to IL-17A stimulation of PBMC, as assessed by flow cytometric determination of NF-κB phospho-p65 ([Figure 2](#F2){ref-type="fig"}). This result was initially surprising, as a previous report indicated that human monocytes highly express both IL-17RA and IL-17RC, and that monocyte migration is stimulated by IL-17, in part due to PI3K and Erk-dependent induction of CCL2.^[@R30]^ However, that study involved longer-term assays using purified monocytes (a two-hour *in vitro* Boyden chamber chemotaxis assay, and a 24-hour *in vivo* sponge migration assay). While phosphorylation of p38 did increase after 15 and 30 minutes of IL-17 stimulation, three other IL-17-stimulated monocyte signaling events (pERK1/2, pJNK, and pAKT) were not significantly induced at 15 or 30 minutes, reaching maximal levels only after 60 to 180 minutes. In another paper, Krueger and colleagues^[@R31]^ reported gene expression profiling experiments in purified monocytes, which again involved longer IL-17 stimulation times (24 hours). In both of these studies, CD14^+^ monocytes were purified using magnetic beads, which might increase the sensitivity of monocytes to IL-17A treatment, whereas we utilized freshly-isolated PBMC. Future studies could utilize more sensitive detection methods, purified PBMC subsets, and/or assess longer-term responses to IL-17. Based on the aforementioned outcomes, we turned to B-cells, in which we reproducibly observed robust short-term responses to CD40L stimulation as assessed by p65, p38, and ERK phosphorylation ([Figures 2](#F2){ref-type="fig"} and [4](#F4){ref-type="fig"}). In addition to its key role in IL-17R signaling, Act1 was identified as an important adaptor molecule for CD40 signaling in B-cells.^[@R12]^ Ligation of CD40 by CD40L expressed on activated T-cells stimulates B-cell survival, proliferation, differentiation, isotype switching, and upregulation of surface molecules contributing to antigen presentation.^[@R32],\ [@R33]^ The CD40 receptor utilizes adapter molecules (TNF receptor associated factors, or TRAFs) that recognize binding sites on the CD40 cytoplasmic tail upon binding of CD40L. TRAFs then recruit TRAF-interacting kinases and together influence a number of well-characterized signal transduction pathways, including NF-κB and MAPKs.^[@R34]^ In the process, Act1 is activated by its recruitment to CD40 through its interaction with TRAF proteins.^[@R12],\ [@R20],\ [@R24]^ It would be worthwhile to further examine the impact of D10N mutation on the physical interaction of Act1 with CD40 and TRAFs. Engineering of the D10N variant into murine Act1 abolished its interaction with CD40 and BAFFR, suggesting that the D10N variant might phenocopy an Act1 knockout.^[@R28]^ However, previous studies in germ-line Act1 knockout mice revealed increased numbers of B-cells relative to WT littermates.^[@R12]^ These studies also found that activation of CD40 by treatment with a cross-linking anti-CD40 antibody increased survival of splenic B-cells, with stronger phosphorylation of IκBα and ERK.^[@R12]^ Taken together, these studies lead to the paradoxical conclusion that CD40 ligation activates short-term downstream signaling events in an Act1-dependent fashion, and yet in the long-term physiological context of transgenic animals, Act1 silencing increases overall B-cell numbers and function. A similar dichotomy has emerged from studies addressing the role of Act1 in IL-17 signaling in T-cells, as one strain of Act1 knockout mice develops inflammatory skin hyperplasia.^[@R28]^ It is possible that these paradoxical observations could be due to developmental influences and/or counter-regulatory feedback from other cell types, as the CD40L-dependent survival phenotype was much less evident in B-cells isolated from B-cell-targeted Act1−/− mice.^[@R12]^. Physiologically, CD40L is expressed primarily by activated T lymphocytes, therefore, B-cell signaling following CD40L-CD40 ligation represents a crucial component of the process known as T-cell "help" for antibody production. However, in addition to their role as antibody-producing cells, it has become apparent that B-cells markedly influence immunity through the secretion of cytokines.^[@R35]^ In particular, a subset of IL-10 producing B-cells has been found to have anti-inflammatory activity by diminishing both acquired and innate immune responses.^[@R36],\ [@R37]^ In humans, evidence is accumulating that B-cells can exert similar suppressive functions.^[@R38]^ Among many potential explanations for the "Act1 paradox" that remain to be tested, it is possible that the Act1 D10N variant might compromise the CD40-mediated function of B regulatory cells. In summary, we found that the Act1 D10N variant attenuated, rather than enhanced CD40L-dependent phosphorylation of NF-κB p65, p38, and Erk and MAPKs in B-cells ([Figure 4](#F4){ref-type="fig"}). However, we did not find a significant effect of the D10N variant on overall B-cell numbers ([Figure 3](#F3){ref-type="fig"}), as was previously observed in one strain of Act1 knockout mice.^[@R12]^ A limitation of our study is that our short-term signaling assays may not have probed more subtle levels of homeostatic regulation that are likely to be present in longer-term cellular responses and at the organismal level. In any case, our results establish a functional effect of the Act1 D10N variant in human cells for the first time, and as such, provide important albeit indirect evidence for the biological relevance of this variant to the immunopathogenesis of psoriasis. MATERIALS AND METHODS {#S4} ===================== Study sample and genotyping {#S5} --------------------------- All human subjects provided written informed consent and were enrolled according to the protocols approved by the institutional review board of the University of Michigan Medical School, in adherence with the Declaration of Helsinki principles. Genomic DNA was isolated from peripheral blood by phenol-chloroform extraction followed by ethanol precipitation. Sixty-three individuals of Caucasian descent (21 males, 42 females; 8 psoriasis cases, 55 controls, mean age 44 years, range 19 -- 83 years) were identified based on Act1 genotypes that were either measured directly or obtained from genome-wide or Immunochip-based association studies of psoriasis.^[@R17],\ [@R25],\ [@R27]^ and confirmed using a Taqman assay (Assay ID C_2473124_10, Applied Biosystems, Foster City, CA). The typing of these 63 subjects resulted in 13 Act1 D10N homozygotes, 20 Act D10N heterozygotes, and 30 Act1 D10N nullizygotes. Isolation of PBMCs {#S6} ------------------ For the isolation of PBMCs, anticoagulated blood was mixed with an equal volume of phosphate-buffered saline. The diluted blood was slowly layered over Histopaque^®^-1077 (Sigma-Aldrich, St. Louis, MO) solution by gently pipetting the diluted blood down the side of the tube, and was then centrifuged for 20 min at 800 g at 22°C, with no brake. The mononuclear cell layer was transferred into 10 ml phosphate-buffered saline and centrifuged for 10 min at 400 g at 4°C. After two washings with RPMI-1640, PBMCs were gently resuspended in freezing medium containing 90% FBS and 10% DMSO as described^[@R38]^ and cryopreserved with gradual cooling followed by storage in liquid nitrogen. Flow cytometric analysis of ACT1 expression {#S7} ------------------------------------------- PBMCs were thawed, washed by centrifugation, and 5 × 10^5^ cells were aliquotted to each tube. Dead cells were excluded from the analysis by staining with LIVE/DEAD Fixable Near-IR Dead Cell Stain (Catalog\#: L10119, Life Technologies, Waltham, MA). Antibodies to cell-surface markers were mixed and added to the cells, including PE-Cy5-coupled anti-CD3 (5 μl per 10^6^ cells; Catalog\#: 344808, BioLegend, San Diego, CA), PE-coupled anti-CD19 (Catalog\#: 12-0199-41, eBioscience, Santa Clara, CA), V450-coupled anti-CD14 (Catalog\#: 560349, BD Biosciences, San Jose, CA), PE-Cy7-coupled anti-CD4 (5 μl per 10^6^ cells; Catalog\#: 300512, BioLegend) and V500-coupled anti-CD8 (Catalog\#: 560774, BD Biosciences). After incubation for 30 min at 4°C;, the cells were washed twice with BD Pharmingen^™^ Stain Buffer (Catalog\#: 554656). The cells were then fixed and permeabilized for 60 min at 4°C; using the Fixation/Permeabilization diluent and concentrate (Catalog\#: 00-5123-43 and 00-5223-56, eBioscience). Cells were then stained with eFluor 660-coupled anti-Act1 (Catalog\#: 50-4040, eBioscience) or eFluor 660-coupled isotype control (eBioscience). The analysis was performed using flow cytometer (BD LSR II, BD Biosciences) and FlowJo 7.6.2 (TreeStar, Ashland, OR) software. Flow cytometric analysis of phosphorylated proteins {#S8} --------------------------------------------------- For determination of phosphorylated proteins after CD40 stimulation, PBMCs were thawed and rested for 2 h and then either left unstimulated or stimulated with trimeric CD40L (100 ng/ml) (Enzo Life Sciences, Farmingdale, NY) for 15 min, performed concurrently with decoration with V450-coupled anti-CD19 (Catalog\#: 560353, BD Biosciences). After incubation with BD Cytofix^™^ Fixation Buffer (Catalog\#: 554655) for 10 minutes at 37°C;, followed by permeabilization in BD Phosflow^™^ Perm Buffer III (Catalog\#: 558050) for 30 minutes on ice, cells were then stained with PE-Cy5-coupled anti-CD3 (BioLegend), FITC-coupled anti-pp38 (Catalog\#: 4551, Cell Signaling Technology, Danvers, MA), PE-coupled anti-pp65 (Catalog\#: 558423, BD Biosciences), or PE-coupled anti-pErk (Catalog\#: 612566, BD Biosciences). The analysis was performed using an LSR II flow cytometer (BD Biosciences) and FlowJo 7.6.2 software (TreeStar, Ashland, OR). Statistical analysis {#S9} -------------------- For flow cytometry experiments, CD3^+^ T-cell and CD19^+^ B-cell percentages and MFI values obtained for each donor belonging to the three genetic groups were assessed for normal Gaussian distribution using the Kolmogorov-Smirnov test and then analyzed by one-way analysis of variance, followed by post hoc least significant difference (LSD) test or Dunnett's test by using SPSS version 21.0 (IBM SPSS Statistics, New York). Pearson's coefficient was used to assess two-way correlations between pp65 and pp38, pp65 and pErk, and pErk and pp38. All results are shown as mean and the standard error of the mean (mean ± SEM). In this exploratory analysis, a *P* value \< 0.05 was considered to be significant. We thank the volunteers who provided blood samples for this study, and Dr. Andrew Johnston for advice on flow cytometry. This research was supported by National Institutes of Health (NIH) R01 grants AR042742, AR050511, AR054966, AR062382, and AR065183 to James T. Elder. We also acknowledge generous support from the Dawn and Dudley Holmes Memorial Fund and the Babcock Endowment Fund to the Department of Dermatology at the University of Michigan. James T. Elder is supported by the Ann Arbor Veterans Affairs Hospital. **CONFLICT OF INTEREST** The authors declare no conflict of interest. ![Act1 is expressed in different subsets of PBMCs. PBMCs were stained with LIVE/DEAD Fixable Near-IR Dead Cell Stain, PE-Cy5-coupled anti-CD3, PE-coupled anti-CD19, V450-coupled anti-CD14, PE-Cy7-coupled anti-CD4 and V500-coupled anti-CD8 and eFluor 660-coupled anti-Act1 or eFluor 660-coupled isotype control. CD19^+^ B-cells, CD4^+^ T-cells, CD8^+^ T-cells and CD14^+^ monocytes were identified. (**a**) Cell subset distributions from a representative flow cytometry experiment (one of four experiments). (**b**) Act1^+^ cells percentages within each PBMC population from a representative flow cytometry experiment (one of four experiments). Mean Act1^+^ percentages from all experiments (n = 4). Bars represent mean ± SEM. \**P* \< 0.05, as assessed by one-way analysis of variance followed by least significant difference post hoc test. (**c**) Act1 expression levels for each PBMC population from a representative experiment (one of four experiments). Median fluorescence intensities (MFI) from all experiments (n = 4). Error bars represent SEM. \**P* \< 0.05, \*\**P* \< 0.01, as assessed by one-way analysis of variance followed by least significant difference post hoc test.](nihms902935f1){#F1} ![CD40L induces p65 phosphorylation in B-cells. Various PBMC subsets were assayed for phosphorylated p65 (pp65) after 15 minutes of stimulation with CD40L (100 ng/ml, upper panels), IL-17A (100 ng/ml, middle panels) or PMA (100 nM, lower panels). Results from one of four representative experiments are shown. Blue histograms demonstrate pp65 response after stimulation with CD40L, IL-17A, or PMA, as compared to unstimulated cells (red curves).](nihms902935f2){#F2} ![Percentages of T-cells and B-cells in PBMC, as a function of Act1 D10N status. PBMCs were collected from D10N homozygous (red), heterozygous (blue), and nullizygous (green) individuals. Percentages of CD3^+^ T-cells (**a**) and CD19^+^ B-cells (**b**) within the lymphocyte gate (defined by forward and side scatter) are shown. Each symbol represents three experiments averaged for one donor. Bars represent mean ± SEM. \**P* \< 0.05, as assessed by one-way analysis of variance followed by least significant difference post hoc test.](nihms902935f3){#F3} ![D10N impairs CD40L-induced p65, p38 and Erk phosphorylation in PBMC-derived B-cells. PBMCs were cultured with (blue curves) or without (red shadow) 100 ng/ml CD40L for 15 minutes, followed by phospho-flow cytometric analysis of pp65, pp38 and pErk within the CD19^+^ gate. (**a**) Representative histograms (from one of two experiments performed for each genotype) are presented for PBMC-derived B-cells from Act1 D10N nullizygous (left panels), heterozygous (middle panels), and homozygous (right panels) individuals. (**b**) Correlations between MFI values for pp65, pp38 and pErk in stimulated CD19^+^ B-cells for all tested individuals (n = 63), Data are expressed as fold-change increases. Pearson correlations between pp65 and pp38, pp65 and pErk, pErk and pp38 were performed. Each symbol represents two experiments averaged for one donor (**c**) Phosphorylation of p65, p38 and Erk in B-cells from nullizygous, heterozygous, and homozygous groups in response to CD40L were expressed as fold MFI. Each symbol represents one individual donor. Summary data represented by horizontal lines represent mean ± SEM. \**P* \< 0.05, \*\**P* \< 0.01, as assessed by one-way analysis of variance followed by Dunnett's post hoc test.](nihms902935f4){#F4} [^1]: These authors contributed equally to this work

{ "pile_set_name": "PubMed Central" }

INTRODUCTION ============ Home bleaching is a popular and convenient treatment to improve the appearance of vital teeth.[@b1-dent5_p0157] During this procedure, whitening gel is deposited on moulded acrylic trays that allow the solution to act indistinctly over both enamel and pre-existing aesthetic restorations. Many investigations have shown that low-concentration bleaching substances may damage these structures[@b1-dent5_p0157]--[@b3-dent5_p0157] and thus increase their susceptibility to staining and bacterial adhesion.[@b4-dent5_p0157]--[@b6-dent5_p0157] Major colour improvements are sometimes enthusiastically desired by patients, whereas the adverse effects of whitening procedures may be unintentionally downplayed by professionals. Recently, manufacturers released a flood of high-concentration bleaching gels in the market. These gels are claimed to increase the effectiveness as well as longevity of at-home treatments. As expected, initial *in vitro* studies reported significant alterations to enamel[@b7-dent5_p0157],[@b8-dent5_p0157] or composite resins.[@b2-dent5_p0157],[@b9-dent5_p0157] However, these results were not confirmed by the scant evidence originating from clinical investigations.[@b10-dent5_p0157],[@b11-dent5_p0157] Since the effects of whitening substances have been evaluated separately and in a great diversity of regimens, it is difficult to extend those incompatible evidences to aesthetically restored teeth. Thus, this prospective clinical study aimed to assess simultaneously the effects of 16% carbamide peroxide on the external topography of enamel and composite resin. MATERIALS AND METHODS ===================== Preparation of specimens ------------------------ This study was initiated only after obtaining suitable approval from the Local Ethics Committee, which is in conformity with the Declaration of Helsinki (DoH). A total of five human impacted upper third molars were included in this study. All molars included fit the following criteria: 1) arising from different donors and 2) without structural damage, signs of hypomineralisation (white spots) or prior contact with saliva. After cleaning, teeth were individually positioned in a sectioning machine equipped with a water-cooled 300-μm thick diamond saw (Minitorm, Struers A/S, Copenhagen, Denmark). Roots were removed at the cemento-enamel junction. After that, the remaining crows were equidistantly sectioned in the mesiodistal and buccolingual directions to provide four enamel fragments (4 mm x 4 mm and 2 mm thickness) each. The pieces obtained were moulded into a light regular-set silicon (Virtual Light-Body, Ivoclair Vivadent, São Paulo, SP, Brazil), and the moulds were filled with increments of a micro-hybrid composite resin (EA2 Opallis, FGM Dentscare Ltda, Joinville, SC, Brazil). Moulds were sequentially polymerised over 20 seconds using a light-emitting diode unit set between 460 and 480 mW/cm^2^ (Ultra-Blue i5, DMC Vasconcellos, São Carlos, SP, Brazil) in order to produce enamel-like composite resin specimens. The resulting 40 specimens (20 of composite resin) were then individually placed at the end of polyvinyl chloride cylinders (3/4 inch diameter and 20 mm high), which were filled with chemically-activated acrylic resin (Technovit 4000, Heraeus Kulzer, Wehrheim, Germany). After 24 hours, the exposed side of each embedded fragment was flattened and polished by sequential use of wet 400-, 600- and 1200-grit silicon carbide papers (Norton®,Guarulhos, SP, Brazil) adapted in a slow-speed polishing machine set at 150 rpm (Ballerup, Struers A/S, Copenhagen, Denmark). The finished specimens were then removed from blocks and ultrasonically cleaned for 5 minutes before being sterilised and stored in distilled water (37±0.5ºC) until use. Ten volunteers (20--25 years old) were recruited to participate in this investigation. All volunteers presented the following inclusion criteria: absence of caries and gingivitis, no history of xerostomia, no previous bleaching therapy and willingness to understand and sign a consent form. Participants had their maxillary right teeth cleaned with a rotary toothbrush and dentifrice before isolation with a rubber dam and stainless steel clamps. The buccal sides of the right upper second premolar and first molar were etched with 37% phosphoric acid (Dentsply, Petrópolis, RJ, Brazil) for 30 seconds and immediately washed with running water for 60 seconds. To fix the enamel and composite resin randomly on these surfaces, a two-step adhesive (Adper™ Scotchbond, 3M, Sumaré, SP, Brazil) was brushed and light-polymerised for 15 seconds before an increment of low-viscosity composite resin (Natural Flow, DFL, Jacarepagua, RJ, Brazil) was used to glue the specimens on teeth using light-polymerisation for 40 seconds. These same procedures were performed on the adjacent lower right teeth to fix the enamel and composite resin on specimens not undergoing the bleaching process. These specimens thus served as paired controls (n=10). Bleaching process ----------------- An upper jaw impression was taken of each participant with alginate materials (Jeltrate, Dentsply, Petrópolis, RJ, Brazil), and impressions were poured with type II dental stone (Pasom, São Paulo, SP, Brazil) to fabricate a master cast. A 0.5 mm-thick wax layer was applied on the buccal face of each cast 1.5 mm away from the gingival margin to fabricate reservoirs for the bleaching gel. Soft trays (Cristal, Bio-art, São Carlos, SP, Brazil) were made using a heat/vacuum tray-forming machine (Plasvac P7, Bio-art, São Carlos, SP, Brazil). The trays were tested and trimmed to fit the gingival margin of each volunteer. The participants were instructed to use the tray at the upper arch and to keep the bleaching product (Whiteness Perfect 16%, FGM Dentscare Ltda, Joinville, SC, Brazil) in contact with teeth exactly 6 hours per day for 8 consecutive days. No restrictions on dietary or hygiene habits were imposed, but the necessity to remove the residual solution after each whitening session was emphasised. Every 48 hours, the volunteers were questioned about sensitivity of teeth or soft tissue. When hypersensitivity was confirmed, desensitising treatment was performed with 2% potassium nitrate (Desensibilize KF 2%, FGM Dentscare Ltda, Joinville, SC, Brazil). The four subjects were instructed to place the desensitizing gel in their tray and wear it for 20 minutes once a day, as recommended by the manufacturer. Surface analysis ---------------- After detachment from teeth, the specimens were left to dehydrate for 96 hours before being gold sputter-coated to permit analysis in a scanning electron microscope (DSM-940 A, Carl Zeiss, Oberkochen, Germany). Digital SEM photomicrographs were taken at four different areas over the surface of each fragment at 5000x magnification. Enamel and resin surface alterations were classified qualitatively by increasing order of scores ranging from 0 (no observable alterations) to 4 (heavy erosion with deep depressions). Prior to blind analysis of the 160 images, a single examiner was calibrated by viewing 30 additional SEM micrographs. A second assessment was repeated 4 weeks after the first evaluation, and the data obtained were considered as a whole for statistical comparison. Statistical analysis -------------------- Statistical differences between the experimental and appropriate control groups were executed with the Mann-Whitney Test adjusted to the 95% confidence interval. The intra-observer agreement was determined using Cohen's Kappa statistic. Statistical analyses were performed with SPSS software v.11.0 for Windows (SPSS Inc., Chicago, IL, USA). RESULTS ======= Representative SEM images are shown in [Figures 1](#f1-dent5_p0157){ref-type="fig"}--[4](#f4-dent5_p0157){ref-type="fig"}, and the results of the Mann-Whitney test are presented in [Table 1](#t1-dent5_p0157){ref-type="table"}. Following application of 16% carbamide peroxide, the enamel surface in the experimental group exhibited extensive mild erosion with shallow depressions and destruction of interprismatic matrix. This appearance differed significantly from that of non-bleached enamel fragments, which exhibited a smooth and amorphous aspect. The bleached composite resin specimens displayed a flat appearance, with slight erosion and some striation due to the grinding procedure. This appearance was statistically similar to that observed in the control group (P\<.05). Intra-observer agreement was 0.89. DISCUSSION ========== This study was the first to adhere human enamel and resin fragments to the buccal side of regular teeth to reproduce, as closely as possible, an actual home-bleaching situation. This manoeuvre permitted controlled and continuous exposure of specimens to saliva, beverages and oral hygiene habits, a procedure completely different from prior investigations in which removable appliances were adopted to carry the specimens.[@b1-dent5_p0157],[@b9-dent5_p0157]--[@b11-dent5_p0157] Therefore, such biases make previous studies difficult to interpret and compare. Scanning electron microscopy is a simple and effective method for identifying surface morphology alterations on enamel and composite resins submitted to home bleaching.[@b1-dent5_p0157],[@b12-dent5_p0157],[@b13-dent5_p0157] Nonetheless, dehydration and metal coating necessary to SEM analysis are able to change part of the specimens structure;[@b14-dent5_p0157] thus, the appearance of enamel observed may not correspond precisely to that of bleached enamel in its natural condition. Porosity and erosion resulting from demineralisation were demonstrated by SEM after *in vitro* and *in vivo* intermittent exposure of enamel to 10% carbamide peroxide.[@b1-dent5_p0157],[@b15-dent5_p0157],[@b16-dent5_p0157] In the present study, the time of contact between the enamel and whitening gel (48 hours in total) was shorter than those adopted in other investigations[@b7-dent5_p0157],[@b10-dent5_p0157],[@b15-dent5_p0157]--[@b17-dent5_p0157] and sufficient to produce an eroded appearance similar to that of acid-etched surfaces. This finding has already been observed in vitro by Adebayo et al[@b8-dent5_p0157] but not *in vivo* by Metz et al,[@b11-dent5_p0157] likely because the enamel specimens used in the latter research had prior contact with fluoride. The damage detected on enamel could be justified by the ability of urea, derived from the reaction of carbamide peroxide with water, to denature protein structures and thus cause structural and morphological changes through the degradation of organic molecules.[@b14-dent5_p0157] In addition, the slightly acidic pH of the gel tested (6.04)[@b18-dent5_p0157] may have contributed secondarily to these external modifications.[@b12-dent5_p0157],[@b17-dent5_p0157] However, a possible increased susceptibility of unerupted third molars' enamel to the action of chemical agents such as carbamide peroxide should be considered before this result can be generalized. Since these findings were detected *in vivo,* it may be suggested that the expected dissolving and buffering effects of saliva on the whitening gel[@b19-dent5_p0157] or the attenuation provided for a peroxide-consuming enzyme on the surface of teeth[@b20-dent5_p0157] were not able to minimise the deleterious impact of a highly concentrated bleaching gel on human enamel. Alterations to enamel topography caused by low concentrations of bleaching gels may undergo repair over time by precipitation of mineral phases derived from saliva into the existing porosities.[@b1-dent5_p0157],[@b16-dent5_p0157],[@b21-dent5_p0157] Tooth remineralisation is a slow process, and the high concentration of bleaching gel used in this study probably smothered the saliva recovery properties. Clinically, however, erosion reversal should be expected as soon as whitening gel application is discontinued. In recent laboratory investigations, the addition of fluoride to carbamide peroxide gel resulted in less demineralisation and shorter periods for enamel hardening recovery following bleaching.[@b22-dent5_p0157],[@b23-dent5_p0157] The absence of significant alterations on the surface of the composite resin bleached was already described for nano-hybrid and packable resins after extended periods of *in situ* and *in vitro* exposure to 15% carbamide peroxide.[@b6-dent5_p0157],[@b11-dent5_p0157] Since morphological alterations were noticed on enamel specimens subjected to the same home-whitening conditions, we inferred that salivary attenuation of the highly concentrated peroxide carbamide substance did not completely respond for resin resistance to bleaching detrimental effects.[@b19-dent5_p0157],[@b20-dent5_p0157] Enamel and composite are completely different materials, and an incompatibility between the solvents present in the whitening substances and the components in the polymer matrix of the resin-based material may justify the absence of deleterious effects on resin surfaces.[@b2-dent5_p0157],[@b11-dent5_p0157],[@b13-dent5_p0157],[@b24-dent5_p0157] Another possible explanation for the absence of statistical difference between groups relates to the subjective and qualitative evaluations necessary for the SEM micrographs and also to the low number of composite resin specimens used; it is possible that major changes in the surface of bleached resin have been underrated, despite the effort to calibrate the examiners, thus reducing the numerical difference between groups. Although not widespread as SEM, atomic force microscopy (AFM) enables quantitative and independent measurement of specimens' surface morphology in their natural condition.[@b17-dent5_p0157] Thus, the possibility of getting different results, if AFM had been used, cannot be disregarded. Colour improvements have been described separately for enamel[@b25-dent5_p0157] and composite resin.[@b26-dent5_p0157] Since these structures were bleached together in this investigation, we provide evidence that these substrates responded differently to 16% carbamide peroxide. This phenomenon may justify the common observation of poor colour matching of pre-existing restorations after home whitening.[@b11-dent5_p0157] Nonetheless, given the diversity of composite resins available, the susceptibility of their surfaces to this or other vital bleaching regimens and formulations may be different. To date, it is uncertain how the surface changes that occur during bleaching are related to colour improvement and/or increases in staining susceptibility.[@b5-dent5_p0157],[@b6-dent5_p0157] However, the complete repolishing of aesthetically restored teeth to recover smoothness and delay staining of composite resins[@b9-dent5_p0157],[@b27-dent5_p0157] may both be unnecessary, at least for the material studied, and thus avoid the irreversible loss of enamel structure. CONCLUSIONS =========== Within the limitations of this clinical study, we concluded that the occurrence of surface topography changes during home bleaching of aesthetically restored teeth with 16% carbamide peroxide was restricted to enamel. Dr. Elliot Watanabe Kitajima, from the Nucleus of Research Support, Electron Microscopy Center (NAP/MEPA-ESALQ/USP). We also thank all colleagues and volunteers who contributed to this research. ![Representative SEM micrograph of bleached enamel at original magnification 5000x.](DENT-5-160-g1){#f1-dent5_p0157} ![Representative SEM micrograph of unbleached enamel at original magnification 5000x.](DENT-5-160-g2){#f2-dent5_p0157} ![Representative SEM micrograph of bleached resin composite at original magnification 5000x.](DENT-5-160-g3){#f3-dent5_p0157} ![Representative SEM micrograph of unbleached resin composite at original magnification 5000x.](DENT-5-160-g4){#f4-dent5_p0157} ###### Median values and ranges of scores attributed to enamel and resin composite micrographs **Substrate** **Treatment** **Median** **P value** ----------------- ---------------- --------------------------------------------------- --------------------------------------------------- Enamel bleached 3.4 (3.0--4.0) 0.004[\*](#tfn1-dent5_p0157){ref-type="table-fn"} unbleached 1.4 (1.0--2.0) 0.004[\*](#tfn1-dent5_p0157){ref-type="table-fn"} Composite resin bleached 2.3 (1.0--3.0) 0.091 unbleached 1.6 (1.0--2.0) 0.091 statistically significant.

{ "pile_set_name": "PubMed Central" }

1. Introduction {#sec1} =============== A plasmacytoma is a tumor of plasma cells which often develops as a systemic spread of multiple myeloma, in bone as a solitary plasmacytoma of bone, or in soft tissue as extramedullary plasmacytoma (EMP). Extramedullary plasmacytomas are rare, comprising about 3% of all plasma cell neoplasms with bone marrow plasma cell infiltration \<5% of nucleated cells and demonstrating no evidence of myeloma \[[@B1]\]. While solitary bone plasmacytomas reveal single tumor inside the bone comprising of abnormal plasma cells. The majority (60%--80%) of extramedullary plasmacytomas are found in the head and neck region, particularly in the upper respiratory tract, and to a lesser extent in gastrointestinal tract. Cases of primary cutaneous plasmacytoma are comprised of 2--4% of extramedullary plasmacytomas. It is unusual to have a primary cutaneous plasmacytoma present as a lip lesion, as only 5 cases have been reported in the literature \[[@B2]\]. The first case reported was by Volk in 1936 \[[@B2]\]. Primary cutaneous plasmacytomas present as relatively slow growing papules or plaques or an erythematous-violaceous nodule either existing in solitary tumor (62%) or possibly involving multiple sites (38%). Patients usually present with a benign past medical history. Symptoms are dependent on location and size of lesion. Rarely, patients present with fever, malaise, night sweats, or weight loss. Other features that may be present are lymphadenopathy and hepatosplenomegaly. The lesion may be ulcerated and necrotic. The skin lesions range in diameter from 1 to 5 cm and the shape may vary. Histologically, it shows a nonepidermotropic dermal infiltrate of plasma cells at different stages of maturation. The plasma cells often show atypical, binucleation, and increased mitotic activity. The epidermis is generally spared. The mean age at diagnosis is 60 years with a male to female predominance of 3 : 1 and it is commonly reported in Asian men. The incidence rate increased with advancing age. Treatment for solitary or a few lesions can be treated with radiotherapy or surgical excision. Chemotherapy or intralesional corticosteroid therapy may be provided for multiple skin lesions. Due to this rarity, we feel the need to present this case in order to increase the awareness of the disease to better serve our patients. The clinicopathological features, prognostic factors, and treatment options with its unusual presentation of lip involvement is being discussed in this report. 2. Case Description {#sec2} =================== A 65-year-old male presented to the Ears, Nose, and Throat clinic in September 2015 with complaints of a nonhealing lower lip sore for the past 3 years. The slow growing lesion was described as a burning sensation with scabbing and occasional bleeding. It was slowly increasing in size. The patient had a greater than 50-year history of smoking tobacco (not chewing) which he quit about 7 months prior to his visit. He also complained of unintentionally losing 3 lbs in the last 2 weeks. He denied any fevers or chills. He had no other complaints. Physical examination was significant for a single lesion on the lower lip 1.5 × 1 cm. A biopsy was performed and sent to pathology. The patient was worked up thoroughly for evidence of multiple myeloma/light chain disease. Laboratory testing was significant for no anemia, normal serum calcium, normal serum albumin, normal total protein, and normal creatinine. There was no Bence Jones protein in urine. Bone survey was done which was negative. Bone marrow biopsy showed \<5% plasma cells, ruling out plasmacytoma of the bone. No monoclonal gammopathy was seen, which precluded immunofixation order. Kappa to lambda ratio was normal. Hematoxylin-eosin staining of the lip biopsy revealed a dense plasmacytic infiltration in the deep dermis (see [Figure 1(a)](#fig1){ref-type="fig"}). Immunohistochemical panel included CD3, CD20, and CD138 immunohistochemical stains as well as kappa and lambda light chain in situ hybridization. The superficial dermis and the epidermis were not seen to be involved. Immunohistochemistry was positive for CD138 (see [Figure 1(b)](#fig1){ref-type="fig"}) and revealed that most plasma cells were lambda positive and with rare kappa-positive plasma cells (see Figures [1(c)](#fig1){ref-type="fig"} and [1(d)](#fig1){ref-type="fig"}). Based on the ISCL/EORTC Proposed on TNM Classification of Cutaneous Lymphoma our patient was staged at T1aN0M0. The patient was treated in November 2015 with radiation therapy as follows: 3000 cGy in 15 fractions using 3D conformal technique to his lower lip and level I lymph nodes and he received 2000 cGy boost bringing his total dose to 5000 cGy in 25 fractions. He responded well to treatment. At his last follow-up, there was no evidence of multiple myeloma or light chain disease. 3. Discussion {#sec3} ============= Primary cutaneous plasmacytoma is a rare diagnosis that represents approximately 4% of extramedullary plasmacytoma cases. Of the 4% cases, only a few cases have been reported to be associated with lip lesions. In the latest comprehensive literature review by Tsang et al. in 2016, there are 5 case reports of primary cutaneous plasmacytomas of the lip \[[@B2]\]. Primary cutaneous plasmacytomas without underlying multiple myeloma is included in the 2005 WHO/EORTC classification among the primary cutaneous marginal zone B-cell lymphoma due to considerable overlapping features \[[@B3]\]. Primary cutaneous marginal zone B-cell lymphoma is considered as part of extranodal marginal zone lymphoma of MALT type \[[@B3]\]. Due to the infrequent and nonspecific presentation of cutaneous plasmacytomas, a clinical diagnosis is not sufficient. Diagnosis is based on clinical, histopathological, and immunohistochemical findings, and multiple myeloma must be ruled out by laboratory, radiologic, and bone marrow studies \[[@B4]\]. After ruling out multiple myeloma, the only findings will be a histopathologic dense monomorphic dermal plasmacytic infiltrate and a monoclonal immunohistochemistry. Consistent with published literatures, our male patient in this case, at the age of 65, presented with a slow growing lesion and insignificant systemic review. Tissue biopsy immunohistochemically shows cytoplasmic expression of immunoglobulins that are restricted to one immunoglobulin chain. Histopathologic finding of diffuse or nodular dermal and subcuticular infiltrate of plasma cells, displaying varying degrees of maturation and atypia but no epidermotropism, are findings that are consistent with our patient. The immunophenotype demonstrates plasma cells positive for CD138, CD38, and CD79a, but generally not CD20 and leukocyte common antigen, and shows monotypic cytoplasmic immunoglobulin light chain expression on paraffin sections \[[@B3]\]. On the other hand, cutaneous marginal zone B cells express CD 20. Our patient was positive for CD138 and negative for CD20. While 30% of all marrow-based myelomas are lysozyme positive, lysozyme is not expressed by cutaneous plasmacytomas. Laboratory, radiological, and bone marrow investigations must be done to exclude multiple myeloma. Laboratory studies include but are not limited to complete blood count, blood smear, blood sedimentation rate, renal function, electrolyte (calcium), serum and urine protein electrophoresis, quantitative Ig determination in serum, immunoelectrophoresis in serum and urine, and B-2 microglobulin determination in serum. All laboratory findings will be normal except for quantitative Ig elevation. Radiological studies include a full skeletal X-ray to rule out any osteolytic lesions that are consistent with multiple myeloma. A bone marrow biopsy will demonstrate plasma cell infiltration less than 5% of all nucleated cells. Plasma cells infiltration is common in reactive lesion and must be ruled out with staining for kappa or lambda light chain. In contrast to multiple myeloma, extramedullary plasmacytoma showed absence of cyclin D1 and infrequent expression of CD56. Other differential diagnoses included secondary cutaneous plasmacytoma in the setting of systemic disease, mucosal extramedullary plasmacytoma with secondary skin involvement, marginal zone B-cell lymphoma, plasma cell granuloma, non-Hodgkin lymphoma, high-grade malignant immunoblastic lymphoma, and poorly differentiated neoplasms, as well as infectious diseases that cause infiltration of plasma cells. Currently, there are approximately 50 cases of primary cutaneous plasmacytomas that have been reported in the literature \[[@B2]\]. Due to its rarity, little is known about its treatment. Treatment options include surgery, radiation, chemotherapy, and intralesional corticosteroid therapy. Type of treatment depends on the numbers (solitary versus multiple) and location of lesions. Most patients with solitary primary cutaneous plasmacytoma received radiotherapy as a component of treatment. Radiotherapy with curative intent (\>4000 cGy) is the standard of care and provides promising result since the tumor is radiosensitive. The optimal dose for local control is 40--50 Gy in 20 fractions delivered over 4--6 weeks. However, caution should be considered when using radiotherapy as local recurrence or lymph node metastases could develop thereafter \[[@B2]\]. In addition to high dose radiotherapy and fractions, adjuvant chemotherapy may also be considered for lesions with high-grade histology or tumors \>5 cm or for patients with multiple lesions. Palliative irradiation may also be used for the most symptomatic and large lesions. At the time when this case was written, the patient was in remission with no evidence of multiple myeloma or light chain disease. As reported in previous literatures, patient with solitary lesions received better outcomes than patient with multiple lesions at presentation. They tend to have less local regional or distant dissemination and remained confined to the skin. As a note, large solitary lesions might have a higher risk of progression to metastatic disease, although a size cut-off had not been determined \[[@B2]\]. The main prognostic factor seems to be the clinical presentation with more aggressive cases associated with multiple lesions, larger size, and IgA secretion by neoplastic cells. Multiple myeloma has been estimated to develop in one-third of patients \[[@B5]\]. It is found that progression to multiple myeloma is lower in soft-tissue plasmacytomas compared to osseous plasmacytomas \[[@B2]\]. Patients may also develop local recurrence and distant skin recurrence. Patients with multiple lesions had a higher likelihood of progression to multiple myeloma and subsequent death \[[@B2]\]. From previous case reported, recurrent-free survival increases from 1.4 years to 11 years and overall survival rate increases from 4.3 years to 17 years in patients with solitary primary cutaneous plasmacytomas compared to those with multiple primary cutaneous plasmacytomas lesions \[[@B2]\]. Currently, the pathophysiology is unknown; however there are a few developing hypothesis. Of all up to date cases of primary cutaneous plasmacytoma, five have been described in which local triggering stimulus from trauma may have contributed to the development of plasma cell proliferation. One of these cases reported a primary cutaneous plasmacytoma localized to the lower lip, which had been affected with recurrent herpes simplex virus-1 for 15 years. It was postulated that chronic stimulation of keratinocytes by herpes simplex virus-1, possibly through toll-like receptors, may have favored the release of interleukin-6 able to induce plasma cell proliferation, transformation, and survival \[[@B6]\]. Interleukin-6 may also have some oncogenic effects via enhancement of C-MYC and is antiapoptotic \[[@B7]\]. In the second hypothesis, it is thought that plasma cells expressing CXCR4 cytokine receptor can interact with the inflammatory cascade released due to tissue injury, facilitating the migration of neoplastic plasma cells into the skin \[[@B8]\]. Genetically, extramedullary plasmacytomas share many features of other plasma cell disorders, such as recurrent losses in chromosome 13, chromosome arm 1p, and chromosome arm 14q, as well as gain in chromosome arms 19p, 9q, and 1q \[[@B9]\]. According to the International Society for Cutaneous lymphoma (ISCL) and the cutaneous lymphoma task force of the European Organization of Research and Treatment of Cancer (EORTC), TNM classification system (see the list below) had been proposed to primarily document disease extension and to a lesser extent for prognostic guide \[[@B10]\]. *ISCL/EORTC Proposed TNM Classification of Cutaneous Lymphoma Other Than Mycosis Fungoides and Sezary Syndrome* (T) *⁡*(T1) Solitary skin involvement(T1a) A solitary lesion \<5 cm diameter(T1b) A solitary \>5 cm diameter(T2) Regional skin involvement of multiple lesions limited to 1 body region or 2 contiguous body regions(T2a) All-disease encompassing in a \<15 cm diameter circular area(T2b) All-disease encompassing in \>15 and \<30 cm diameter circular area(T2c) All-disease encompassing in a 30 cm diameter circular area(T3) Generalized skin involvement(T3a) Multiple lesions involving 2 noncontiguous body regions(T3b) Multiple lesions involving equal to or greater than 3 body regions(N) *⁡*(N0) No clinical or pathologic lymph node involvement(N1) Involvement of 1 peripheral lymph node region that drains an area of current or prior skin involvement(N2) Involvement of 2 or more peripheral lymph node regions or involvement of any lymph node region that does not drain an area of current or prior skin involvement(N3) Involvement of central lymph nodes(M) *⁡*(M0) No evidence of extracutaneous non-lymph node disease(M1) Extracutaneous non-lymph node disease present In order to stage the disease, patients should undergo the following testing: a complete history and physical examination, complete blood count with differential count, chemistry panel with LDH, flow cytometry on peripheral blood, and imaging with computed tomography scan of the chest, abdomen, and pelvis with contrast or positron emission-tomography/CT as it can visualize metabolically active disease in relatively small lesions \[[@B10]\]. According to this classification, our patient was stage T1aN0M0 disease. As for treatment management, clinicians are recommended to follow the WHO-EORTC cutaneous lymphoma classification to understand the clinical behavior and prognostic variables in the heterogeneous group of cutaneous lymphomas in order to provide optimal care. Due to its shortcomings, the proposed TNM system will be reexamined as more data are obtained. 4. Conclusion {#sec4} ============= With few cases in the literature reported, this case is an example of a primary cutaneous lesion in an unusual location with no systemic disease. Follow-up care includes continual monitoring for progression to multiple myeloma. This consists of M-protein measurements and complete blood counts at six-week intervals for the first 6 months. This rarer and nonspecific presentation of a primary extramedullary plasmacytoma is important to keep in mind when evaluating lip lesions. Competing Interests =================== The authors declare that they have no competing interests. ![(a) Dense plasma cell infiltrate, H&E, 40x; (b) CD138 immunostain labeling plasma cells, IHC 40x; (c) Kappa in situ hybridization with only rare plasma cell labeling, ISH, 40x; (d) Lambda in situ hybridization with most plasma cells labeling, ISH, 40x.](CRIONM2017-4296802.001){#fig1} [^1]: Academic Editor: Ossama W. Tawfik

{ "pile_set_name": "PubMed Central" }

Related literature {#sec1} ================== For the biological and pharmacological properties of benzopyrans and their derivatives, see Brooks (1998[@bb2]); Hatakeyama *et al.* (1988[@bb7]); Hyana & Saimoto (1987[@bb8]); Tang *et al.* (2007[@bb13]). For a detailed account of the importance of 4*H*-chromenes, see Liu *et al.* (2007[@bb9]); Wang, Fang *et al.* (2003[@bb14]); Wang, Zhang *et al.* (2003[@bb15]). For hydrogen-bonding interactions and motifs, see: Bernstein *et al.* (1995[@bb1]); Desiraju (1989[@bb4]); Desiraju & Steiner (1999[@bb5]); Etter (1990[@bb6]). Experimental {#sec2} ============ {#sec2.1} ### Crystal data {#sec2.1.1} C~21~H~18~O~4~*M* *~r~* = 334.35Orthorhombic,*a* = 6.299 (2) Å*b* = 15.798 (6) Å*c* = 17.429 (6) Å*V* = 1734.3 (11) Å^3^*Z* = 4Mo *K*α radiationμ = 0.09 mm^−1^*T* = 293 (2) K0.35 × 0.32 × 0.29 mm ### Data collection {#sec2.1.2} Bruker SMART APEX CCD diffractometerAbsorption correction: multi-scan (*SADABS*; Sheldrick, 1998[@bb10]) *T* ~min~ = 0.969, *T* ~max~ = 0.97513979 measured reflections2055 independent reflections1793 reflections with *I* \> 2σ(*I*)*R* ~int~ = 0.062 ### Refinement {#sec2.1.3} *R*\[*F* ^2^ \> 2σ(*F* ^2^)\] = 0.046*wR*(*F* ^2^) = 0.098*S* = 1.012055 reflections226 parametersH-atom parameters constrainedΔρ~max~ = 0.10 e Å^−3^Δρ~min~ = −0.10 e Å^−3^ {#d5e448} Data collection: *SMART* (Bruker, 2007[@bb3]); cell refinement: *SMART*; data reduction: *SAINT* (Bruker, 2007[@bb3]); program(s) used to solve structure: *SHELXS97* (Sheldrick, 2008[@bb11]); program(s) used to refine structure: *SHELXL97* (Sheldrick, 2008[@bb11]); molecular graphics: *PLATON* (Spek, 2003[@bb12]); software used to prepare material for publication: *SHELXL97*. Supplementary Material ====================== Crystal structure: contains datablocks I, global. DOI: [10.1107/S1600536808039743/fb2126sup1.cif](http://dx.doi.org/10.1107/S1600536808039743/fb2126sup1.cif) Structure factors: contains datablocks I. DOI: [10.1107/S1600536808039743/fb2126Isup2.hkl](http://dx.doi.org/10.1107/S1600536808039743/fb2126Isup2.hkl) Additional supplementary materials: [crystallographic information](http://scripts.iucr.org/cgi-bin/sendsupfiles?fb2126&file=fb2126sup0.html&mime=text/html); [3D view](http://scripts.iucr.org/cgi-bin/sendcif?fb2126sup1&Qmime=cif); [checkCIF report](http://scripts.iucr.org/cgi-bin/paper?fb2126&checkcif=yes) Supplementary data and figures for this paper are available from the IUCr electronic archives (Reference: [FB2126](http://scripts.iucr.org/cgi-bin/sendsup?fb2126)). AN thanks Dr S. Kannan and Dr B. S. Krishnamurthy, School of Chemistry, Bharathidasan University, Tiruchirappalli, and Organica Aromatics Pvt Ltd Bangalore, India, for providing laboratory facilities. Comment ======= Chromenes (benzopyrans) and their derivatives have numerous biological and pharmacological properties (Tang *et al.*, 2007) such as antisterility (Brooks, 1998) and anticancer activity (Hyana & Saimoto, 1987). In addition, polyfunctional chromene units are present in numerous natural products (Hatakeyama *et al.*, 1988). 4*H*-chromenes are important synthons for some natural products (Liu *et al.*, 2007). As a part of our structural investigations on 4*H*-chromene derivatives, the single-crystal X-ray diffraction study on the title compound was carried out. The chromene ring is almost planar: The puckering amplitude of the chromene ring is 0.097 (3) Å. In the related chromene derivatives (Wang, Zhang *et al.*, 2003; Wang, Fang *et al.*, 2003), the chromene ring is also planar. In the title structure, the interplanar angle between the chromene ring and the 2-phenyl ring is 6.1 (1)° thereby indicating their almost coplanar arrangement (Fig. 1). The propenyloxy substituents at both C5 and C7 are coplanar with the chromene ring with the respective interplanar angles 1.7 (2)° and 8.8 (2)°. The crystal structure is stabilized by the interplay of C--H···O and C--H···π interactions (Fig. 2, Table 1; Desiraju & Steiner, 1999; Desiraju, 1989). Each of C15--H15A···O12, C19--H19A···O16 and C25--H25···O1 interactions are involved in the S(5) motifs (Bernstein *et al.*, 1995; Etter, 1990). Experimental {#experimental} ============ A suspension of chrysin (3.93 mmol, 1.00 g) and potassium carbonate (11.81 mmol, 1.64 g) in dimethyl formamide (10 ml) were added into a round bottom flask. The reaction mixture was heated to 383 K for 2--3 h. The reaction mixture was then cooled to 333 K and allyl bromide (15.74 mmol, 1.90 g) was slowly added to the reaction mixture with the help of a dropping funnel. The reaction mixture was maintained for 8--9 h at 333 K and monitored by high pressure liquid chromatography (HPLC). After completion of the reaction, the content was quenched with water and stirred for 30--45 min at 303 K. T he obtained crude solid was filtered and washed with plenty of water followed by methanol and dried under vacuum at 343 K. The compound was purified by column chromatography using ethyl acetate/n-hexane (1:1) as eluent. All highly pure column fractions were concentrated in a rota evaporator. The dried compound was dissolved in dichloromethane/hexane (1:1) mixture (10 ml). The clear solution was kept for a week and the resulting needle shaped crystals of average size 0.3 mm were washed with n-hexane. The crystals were dried over high vacuum at 343--348 K. Yield: 90% Refinement {#refinement} ========== In the absence of significant anomalous scattering effects, 1488 Friedel pairs have been merged. All the H-atoms were observed in the difference electron density map. However, they were situated into idealized positions with C--H = 0.93 and 0.97 Å for aryl and methylene H, respectively, and constrained to ride on their parent atoms with *U*~iso~(H) = 1.2*U*~eq~(C). Figures ======= ![The title molecule showing the displacement ellipsoids depicted at the 50% probability level for all non-H atoms. The hydrogen atoms are drawn as spheres of arbitrary radius.](e-65-000o1-fig1){#Fap1} ![The molecular packing viewed down the c-axis. Dashed lines represent weak C--H···O interactions.](e-65-000o1-fig2){#Fap2} Crystal data {#tablewrapcrystaldatalong} ============ ------------------------------- -------------------------------------- C~21~H~18~O~4~ *D*~x~ = 1.281 Mg m^−3^ *M~r~* = 334.35 Melting point = 434--437 K Orthorhombic, *P*2~1~2~1~2~1~ Mo *K*α radiation, λ = 0.71073 Å Hall symbol: P 2ac 2ab Cell parameters from 576 reflections *a* = 6.299 (2) Å θ = 1.8--26.0° *b* = 15.798 (6) Å µ = 0.09 mm^−1^ *c* = 17.429 (6) Å *T* = 293 K *V* = 1734.3 (11) Å^3^ Needle, colourless *Z* = 4 0.35 × 0.32 × 0.29 mm *F*(000) = 704 ------------------------------- -------------------------------------- Data collection {#tablewrapdatacollectionlong} =============== --------------------------------------------------------------- -------------------------------------- Bruker SMART APEX CCD diffractometer 3543 independent reflections Radiation source: fine-focus sealed tube 1793 reflections with *I* \> 2σ(*I*) graphite *R*~int~ = 0.062 Detector resolution: 0.3 pixels mm^-1^ θ~max~ = 26.4°, θ~min~ = 1.7° ω scans *h* = −7→7 Absorption correction: multi-scan (*SADABS*; Sheldrick, 1998) *k* = −19→19 *T*~min~ = 0.969, *T*~max~ = 0.975 *l* = −21→20 13979 measured reflections --------------------------------------------------------------- -------------------------------------- Refinement {#tablewraprefinementdatalong} ========== ------------------------------------- ------------------------------------------------------------------------------------------------- Refinement on *F*^2^ Primary atom site location: structure-invariant direct methods Least-squares matrix: full Secondary atom site location: difference Fourier map *R*\[*F*^2^ \> 2σ(*F*^2^)\] = 0.046 Hydrogen site location: difference Fourier map *wR*(*F*^2^) = 0.098 H-atom parameters constrained *S* = 1.01 *w* = 1/\[σ^2^(*F*~o~^2^) + (0.0343*P*)^2^ + 0.1512*P*\] where *P* = (*F*~o~^2^ + 2*F*~c~^2^)/3 2055 reflections (Δ/σ)~max~ \< 0.001 226 parameters Δρ~max~ = 0.10 e Å^−3^ 0 restraints Δρ~min~ = −0.10 e Å^−3^ 72 constraints ------------------------------------- ------------------------------------------------------------------------------------------------- Special details {#specialdetails} =============== ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Geometry. All e.s.d.\'s (except the e.s.d. in the dihedral angle between two l.s. planes) are estimated using the full covariance matrix. The cell e.s.d.\'s are taken into account individually in the estimation of e.s.d.\'s in distances, angles and torsion angles; correlations between e.s.d.\'s in cell parameters are only used when they are defined by crystal symmetry. An approximate (isotropic) treatment of cell e.s.d.\'s is used for estimating e.s.d.\'s involving l.s. planes. Refinement. Refinement of *F*^2^ against ALL reflections. The weighted *R*-factor *wR* and goodness of fit *S* are based on *F*^2^, conventional *R*-factors *R* are based on *F*, with *F* set to zero for negative *F*^2^. The threshold expression of *F*^2^ \> σ(*F*^2^) is used only for calculating *R*-factors(gt) *etc*. and is not relevant to the choice of reflections for refinement. *R*-factors based on *F*^2^ are statistically about twice as large as those based on *F*, and *R*- factors based on ALL data will be even larger. ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Fractional atomic coordinates and isotropic or equivalent isotropic displacement parameters (Å^2^) {#tablewrapcoords} ================================================================================================== ------ ------------- -------------- --------------- -------------------- -- *x* *y* *z* *U*~iso~\*/*U*~eq~ O1 0.4396 (4) 1.06971 (12) 0.13245 (11) 0.0650 (6) C2 0.5860 (6) 1.0832 (2) 0.0765 (2) 0.0639 (8) C3 0.5782 (6) 1.0403 (2) 0.0102 (2) 0.0745 (9) H3 0.6761 1.0541 −0.0278 0.089\* C4 0.4277 (6) 0.9746 (2) −0.0056 (2) 0.0696 (9) C5 0.0891 (6) 0.90891 (19) 0.04852 (18) 0.0608 (8) C6 −0.0584 (6) 0.90372 (18) 0.10630 (18) 0.0622 (8) H6 −0.1728 0.8668 0.1020 0.075\* C7 −0.0358 (6) 0.95419 (19) 0.17160 (18) 0.0590 (8) C8 0.1332 (5) 1.00863 (19) 0.17997 (18) 0.0610 (9) H8 0.1495 1.0414 0.2239 0.073\* C9 0.2778 (5) 1.01254 (18) 0.12026 (18) 0.0557 (8) C10 0.2641 (5) 0.96435 (18) 0.05336 (17) 0.0559 (8) O11 0.4392 (5) 0.93093 (17) −0.06419 (14) 0.1049 (9) O12 0.0781 (4) 0.86278 (13) −0.01724 (12) 0.0748 (6) C13 −0.0923 (6) 0.8053 (2) −0.02818 (18) 0.0778 (10) H13A −0.2267 0.8353 −0.0279 0.093\* H13B −0.0942 0.7635 0.0125 0.093\* C14 −0.0582 (8) 0.7632 (2) −0.1044 (2) 0.0860 (11) H14 −0.1579 0.7232 −0.1196 0.103\* C15 0.0989 (7) 0.7778 (2) −0.15084 (19) 0.0906 (12) H15A 0.2021 0.8173 −0.1378 0.109\* H15B 0.1080 0.7487 −0.1971 0.109\* O16 −0.1951 (4) 0.94477 (13) 0.22389 (13) 0.0729 (7) C17 −0.1985 (6) 1.0013 (2) 0.28825 (18) 0.0773 (11) H17A −0.1929 1.0594 0.2704 0.093\* H17B −0.0753 0.9912 0.3204 0.093\* C18 −0.3953 (7) 0.9877 (2) 0.3335 (2) 0.0815 (11) H18 −0.4206 1.0255 0.3733 0.098\* C19 −0.5312 (6) 0.9305 (3) 0.3237 (2) 0.1011 (13) H19A −0.5136 0.8910 0.2846 0.121\* H19B −0.6496 0.9277 0.3555 0.121\* C20 0.7460 (6) 1.1464 (2) 0.0988 (2) 0.0670 (9) C21 0.9252 (7) 1.1595 (2) 0.0537 (2) 0.0818 (11) H21 0.9425 1.1292 0.0084 0.098\* C22 1.0767 (7) 1.2170 (3) 0.0757 (3) 0.0980 (13) H22 1.1961 1.2253 0.0452 0.118\* C23 1.0540 (8) 1.2625 (3) 0.1423 (3) 0.1075 (15) H23 1.1572 1.3015 0.1568 0.129\* C24 0.8787 (8) 1.2503 (3) 0.1873 (3) 0.1096 (14) H24 0.8634 1.2806 0.2327 0.131\* C25 0.7248 (6) 1.1932 (2) 0.1657 (2) 0.0887 (11) H25 0.6052 1.1859 0.1962 0.106\* ------ ------------- -------------- --------------- -------------------- -- Atomic displacement parameters (Å^2^) {#tablewrapadps} ===================================== ----- ------------- ------------- ------------- -------------- -------------- -------------- *U*^11^ *U*^22^ *U*^33^ *U*^12^ *U*^13^ *U*^23^ O1 0.0626 (14) 0.0716 (14) 0.0608 (14) −0.0033 (13) 0.0010 (14) 0.0035 (11) C2 0.059 (2) 0.066 (2) 0.067 (2) 0.008 (2) 0.003 (2) 0.0157 (18) C3 0.071 (2) 0.084 (2) 0.069 (2) −0.004 (2) 0.016 (2) 0.007 (2) C4 0.067 (2) 0.079 (2) 0.063 (2) 0.003 (2) 0.006 (2) 0.0007 (19) C5 0.072 (2) 0.0608 (19) 0.050 (2) 0.007 (2) −0.003 (2) 0.0017 (16) C6 0.066 (2) 0.0620 (18) 0.058 (2) −0.0021 (18) 0.003 (2) 0.0036 (17) C7 0.062 (2) 0.0673 (19) 0.048 (2) 0.0043 (19) 0.0072 (19) 0.0059 (17) C8 0.067 (2) 0.0646 (19) 0.0509 (19) 0.0001 (18) 0.001 (2) 0.0015 (16) C9 0.054 (2) 0.0572 (17) 0.056 (2) 0.0028 (18) −0.0023 (19) 0.0097 (16) C10 0.059 (2) 0.0577 (17) 0.051 (2) 0.0056 (19) 0.0027 (18) 0.0042 (15) O11 0.107 (2) 0.128 (2) 0.0792 (18) −0.029 (2) 0.0327 (17) −0.0308 (16) O12 0.0831 (16) 0.0798 (14) 0.0614 (15) −0.0099 (15) 0.0049 (14) −0.0105 (13) C13 0.086 (3) 0.076 (2) 0.071 (2) −0.013 (2) 0.002 (2) −0.0011 (19) C14 0.110 (3) 0.078 (2) 0.070 (3) −0.008 (3) −0.013 (3) −0.010 (2) C15 0.118 (3) 0.094 (3) 0.060 (2) 0.011 (3) −0.004 (3) −0.007 (2) O16 0.0753 (17) 0.0844 (14) 0.0591 (14) −0.0090 (13) 0.0117 (13) −0.0065 (12) C17 0.088 (3) 0.088 (2) 0.057 (2) −0.004 (2) 0.010 (2) −0.0027 (19) C18 0.092 (3) 0.092 (3) 0.060 (2) 0.012 (3) 0.000 (3) 0.003 (2) C19 0.082 (3) 0.125 (3) 0.096 (3) 0.003 (3) 0.010 (3) −0.010 (3) C20 0.061 (2) 0.0622 (19) 0.078 (3) 0.002 (2) −0.004 (2) 0.0161 (19) C21 0.070 (3) 0.084 (2) 0.092 (3) 0.000 (2) −0.002 (3) 0.028 (2) C22 0.072 (3) 0.095 (3) 0.127 (4) −0.012 (3) −0.002 (3) 0.044 (3) C23 0.092 (3) 0.083 (3) 0.147 (5) −0.021 (3) −0.015 (4) 0.012 (3) C24 0.096 (3) 0.100 (3) 0.132 (4) −0.022 (3) 0.000 (3) −0.019 (3) C25 0.077 (3) 0.082 (2) 0.107 (3) −0.010 (2) 0.005 (3) −0.010 (2) ----- ------------- ------------- ------------- -------------- -------------- -------------- Geometric parameters (Å, °) {#tablewrapgeomlong} =========================== --------------------- ------------ ----------------------- ------------ O1---C2 1.360 (4) C14---H14 0.9300 O1---C9 1.378 (3) C15---H15A 0.9300 C2---C3 1.340 (4) C15---H15B 0.9300 C2---C20 1.471 (5) O16---C17 1.434 (3) C3---C4 1.433 (5) C17---C18 1.485 (5) C3---H3 0.9300 C17---H17A 0.9700 C4---O11 1.233 (3) C17---H17B 0.9700 C4---C10 1.465 (4) C18---C19 1.256 (4) C5---O12 1.360 (3) C18---H18 0.9300 C5---C6 1.373 (4) C19---H19A 0.9300 C5---C10 1.410 (4) C19---H19B 0.9300 C6---C7 1.397 (4) C20---C25 1.387 (4) C6---H6 0.9300 C20---C21 1.391 (5) C7---O16 1.364 (4) C21---C22 1.373 (5) C7---C8 1.376 (4) C21---H21 0.9300 C8---C9 1.384 (4) C22---C23 1.373 (5) C8---H8 0.9300 C22---H22 0.9300 C9---C10 1.395 (4) C23---C24 1.368 (6) O12---C13 1.419 (4) C23---H23 0.9300 C13---C14 1.501 (4) C24---C25 1.377 (5) C13---H13A 0.9700 C24---H24 0.9300 C13---H13B 0.9700 C25---H25 0.9300 C14---C15 1.299 (5) C2---O1---C9 119.6 (3) C15---C14---H14 117.2 C3---C2---O1 121.0 (3) C13---C14---H14 117.2 C3---C2---C20 126.6 (4) C14---C15---H15A 120.0 O1---C2---C20 112.4 (3) C14---C15---H15B 120.0 C2---C3---C4 123.8 (3) H15A---C15---H15B 120.0 C2---C3---H3 118.1 C7---O16---C17 117.8 (3) C4---C3---H3 118.1 O16---C17---C18 109.7 (3) O11---C4---C3 121.7 (3) O16---C17---H17A 109.7 O11---C4---C10 124.1 (3) C18---C17---H17A 109.7 C3---C4---C10 114.2 (3) O16---C17---H17B 109.7 O12---C5---C6 123.5 (3) C18---C17---H17B 109.7 O12---C5---C10 115.0 (3) H17A---C17---H17B 108.2 C6---C5---C10 121.5 (3) C19---C18---C17 126.9 (4) C5---C6---C7 119.6 (3) C19---C18---H18 116.5 C5---C6---H6 120.2 C17---C18---H18 116.5 C7---C6---H6 120.2 C18---C19---H19A 120.0 O16---C7---C8 124.5 (3) C18---C19---H19B 120.0 O16---C7---C6 114.0 (3) H19A---C19---H19B 120.0 C8---C7---C6 121.5 (3) C25---C20---C21 118.3 (4) C7---C8---C9 117.2 (3) C25---C20---C2 121.2 (3) C7---C8---H8 121.4 C21---C20---C2 120.5 (3) C9---C8---H8 121.4 C22---C21---C20 120.3 (4) O1---C9---C8 113.6 (3) C22---C21---H21 119.8 O1---C9---C10 122.1 (3) C20---C21---H21 119.8 C8---C9---C10 124.3 (3) C21---C22---C23 120.7 (4) C9---C10---C5 115.9 (3) C21---C22---H22 119.7 C9---C10---C4 118.9 (3) C23---C22---H22 119.7 C5---C10---C4 125.2 (3) C24---C23---C22 119.7 (4) C5---O12---C13 119.6 (3) C24---C23---H23 120.2 O12---C13---C14 107.1 (3) C22---C23---H23 120.2 O12---C13---H13A 110.3 C23---C24---C25 120.2 (4) C14---C13---H13A 110.3 C23---C24---H24 119.9 O12---C13---H13B 110.3 C25---C24---H24 119.9 C14---C13---H13B 110.3 C24---C25---C20 120.8 (4) H13A---C13---H13B 108.5 C24---C25---H25 119.6 C15---C14---C13 125.6 (4) C20---C25---H25 119.6 C9---O1---C2---C3 −1.5 (4) O11---C4---C10---C9 174.5 (3) C9---O1---C2---C20 179.6 (2) C3---C4---C10---C9 −5.1 (4) O1---C2---C3---C4 −4.0 (5) O11---C4---C10---C5 −5.6 (5) C20---C2---C3---C4 174.8 (3) C3---C4---C10---C5 174.8 (3) C2---C3---C4---O11 −172.6 (3) C6---C5---O12---C13 −0.4 (4) C2---C3---C4---C10 7.1 (5) C10---C5---O12---C13 −179.7 (3) O12---C5---C6---C7 −179.3 (3) C5---O12---C13---C14 −178.8 (3) C10---C5---C6---C7 −0.1 (5) O12---C13---C14---C15 −0.3 (5) C5---C6---C7---O16 178.4 (3) C8---C7---O16---C17 6.5 (4) C5---C6---C7---C8 −0.8 (5) C6---C7---O16---C17 −172.7 (3) O16---C7---C8---C9 −177.9 (3) C7---O16---C17---C18 173.6 (3) C6---C7---C8---C9 1.2 (4) O16---C17---C18---C19 6.9 (5) C2---O1---C9---C8 −176.4 (2) C3---C2---C20---C25 173.0 (3) C2---O1---C9---C10 3.2 (4) O1---C2---C20---C25 −8.1 (4) C7---C8---C9---O1 178.8 (2) C3---C2---C20---C21 −8.2 (5) C7---C8---C9---C10 −0.9 (4) O1---C2---C20---C21 170.7 (3) O1---C9---C10---C5 −179.5 (2) C25---C20---C21---C22 0.4 (5) C8---C9---C10---C5 0.1 (4) C2---C20---C21---C22 −178.4 (3) O1---C9---C10---C4 0.3 (4) C20---C21---C22---C23 −0.1 (5) C8---C9---C10---C4 179.9 (3) C21---C22---C23---C24 0.2 (6) O12---C5---C10---C9 179.7 (3) C22---C23---C24---C25 −0.6 (7) C6---C5---C10---C9 0.4 (4) C23---C24---C25---C20 1.0 (6) O12---C5---C10---C4 −0.1 (4) C21---C20---C25---C24 −0.9 (5) C6---C5---C10---C4 −179.4 (3) C2---C20---C25---C24 178.0 (3) --------------------- ------------ ----------------------- ------------ Hydrogen-bond geometry (Å, °) {#tablewraphbondslong} ============================= --------------------- --------- --------- ----------- --------------- *D*---H···*A* *D*---H H···*A* *D*···*A* *D*---H···*A* C15---H15A···O12 0.93 2.35 2.691 (4) 101 C19---H19A···O16 0.93 2.42 2.749 (5) 101 C25---H25···O1 0.93 2.39 2.714 (4) 101 C17---H17B···O11^i^ 0.97 2.51 3.229 (4) 131 C14---H14···Cg1^ii^ 0.93 3.22 4.081 154 --------------------- --------- --------- ----------- --------------- Symmetry codes: (i) −*x*+1/2, −*y*+2, *z*+1/2; (ii) −*x*−1, *y*+3/2, −*z*+1/2. ###### Hydrogen-bond geometry (Å, °) *D*---H⋯*A* *D*---H H⋯*A* *D*⋯*A* *D*---H⋯*A* --------------------- --------- ------- ----------- ------------- C17---H17*B*⋯O11^i^ 0.97 2.51 3.229 (4) 131 Symmetry codes: (i) .

{ "pile_set_name": "PubMed Central" }

INTRODUCTION {#s1} ============ Regulation of reactive oxygen species (ROS) homeostasis plays key roles in living organisms \[[@R1], [@R2]\]. A moderate increase in ROS promote cell proliferation and differentiation, while the excessive amounts of ROS causes oxidative damage to lipids, proteins and DNA and induces cells undergoing apoptosis \[[@R3]\]. Cancer cells are characterized by persistently higher levels of ROS than non-transformed cells due to increased metabolic activity and the dysregulation of redox balance, which renders cancer cells more vulnerable to oxidative stress and a high dependency on antioxidant enzymes to detoxify from ROS \[[@R4]\]. As such, modulation of ROS homeostasis, via an increase in ROS levels or impairing antioxidant capacity, has been considered as an important strategy for cancer therapy \[[@R5]--[@R7]\]. Superoxide dismutase 1 (SOD1), which is involved in the conversion of toxic superoxide anions into molecular oxygen and hydrogen peroxide, is an important member in the intracellular ROS-scavenging system \[[@R8]\]. Active, mature SOD1 is a homodimeric protein containing two zinc (Zn^2+^) and two copper (Cu^2+^) ions for its stability and activity. The association with the copper chaperone for SOD (CCS) is essential for the activation of copper/zinc SOD, although an additional minor CCS-independent pathway has been reported in mammals \[[@R8]\]. CCS specifically delivers Cu to SOD1, which allows the formation of an intrasubunit disulfide bond between SOD1^Cys-57^ and SOD1^Cys-146^, and results in an enzymatically active homodimers of SOD \[[@R9], [@R10]\]. Thus far, CCS binding remains the most dominant mechanism for the regulation of the enzymatic activity of SOD1. Apart from CCS association, increasing evidence has indicated that diverse post-translational modifications, including nitration \[[@R11]\], phosphorylation \[[@R12]\], glutathionylaion \[[@R13]\] and glycation \[[@R14]\], are involved in the regulation of the dismutase activity of SOD1. Post-translational modifications have emerged as an important aspect in fine-tuning the signal process of SOD1 involved redox homeostasis. In the meanwhile, we have noticed that recent global proteomic profiling has identified lysine acetylation as a frequently occurred modification for cytoplasmic proteins, in particular metabolic enzymes including SOD1 \[[@R15]--[@R17]\], but the cellular functions of these modifications are still unknown. This study started from the validation of occurrence of SOD1 acetylation in cancer cells, and focused on the investigation of the biological significance of SOD1 acetylation. Our findings provided first evidence revealing the role of acetylation in modulating the SOD1 activity. The study highlighted a SOD acetylation mediated positive feedback loop in strengthening oxidative stress caused by genotoxic anticancer agents, and suggested the translational value of SOD1 acetylation for camptothecin-based chemotherapy. RESULTS {#s2} ======= SOD1 is acetylated at lysine 71 {#s2_1} ------------------------------- A number of mass spectrometry-based proteomic studies have suggested the occurrence of acetylation on SOD1 \[[@R15]--[@R17]\], but there lacks evidence to support acetylation of endogenous SOD1, and the biological significance of this modification remains unclear. We firstly validated the acetylation of SOD1 using a pan-specific anti-acetylated lysine antibody in cancer cells with ectopically expressed SOD1. Acetylation was detected on flag-tagged SOD1 enriched from HCT116 colon cancer cells. Treatment of protein deacetylase inhibitors, namely nicotinamide (NAM) and Trichostatin A (TSA), resulted in an increase in the acetylation of SOD1 (Figure [1A](#F1){ref-type="fig"}). ![SOD1 is acetylated on K71 residue\ **A.** Acetylation of exogenous SOD1. Ectopically expressed SOD1 was immunopreciptated and acetylation of SOD1 was examined using a pan-acetyl-lysine antibody. Cells were pretreated with TSA (500 nM) or/and NAM (10 mM) for 12 hr prior to immunoprecipitation; **B.**, **C.** Impact of lysine mutant on SOD1 acetylation. **B.**, wildtype (WT) SOD1 or indicated lysine mutants was transfected into cells and acetylation of each mutant was examined as described in **A.**. EV, empty vector; **C.**, Acetylation of ectopically expressed SOD1WT, SOD1K71R, SOD1K71Q with or without TSA (500 nM) and NAM (10 mM) treatment was analyzed; **D.** Specificity of anti-acetyl-K71 antibody. Flag-tagged SOD1 WT or K71R mutant was transfected into HCT116 cells and acetylation of each purified protein was measured by immunoblotting using an anti-Ac SOD1 (K71) antibody; **E.** Endogenous SOD1 is acetylated at K71. Cells treated with NAM and TSA was analyzed by anti-Ac SOD1 (K71) antibody.](oncotarget-06-20578-g001){#F1} We next determined the main lysine sites where the acetylation occurred. SOD1 contains 11 lysine (K) residues, which are K4, K10, K24, K31, K37, K71, K76, K92, K123, K129 and K137. As lysine lysine (K)-arginine (R) replacement is widely used to generate acetylation-deﬁcient mutants \[[@R18]--[@R20]\], each of the lysine was individually mutated to a nonacetylatable arginine, and the impact on SOD1 acetylation was examined. Among the 11 mutants, only the K71R mutation largely abolished SOD1 acetylation (Figure [1B](#F1){ref-type="fig"}) and the treatment of deacetylase inhibitors failed to increase the detectable signaling of acetylation (Figure [1C](#F1){ref-type="fig"}), indicating the acetylation of SOD1 occurred at K71. Meanwhile, alignment of SOD1 protein sequence revealed that K71 was evolutionarily conserved across diverse species ([Supplemental Figure S1](#SD1){ref-type="supplementary-material"}). Further, we generated an antibody that specifically recognized SOD1 bearing acetylation at K71. The antibody was able to detect the acetylation of ectopically expressed wild-type SOD1 but not the K71R mutant (Figure [1D](#F1){ref-type="fig"}), demonstrating the selectivity of the antibody. This antibody enabled the first detection of the acetylation of endogenous SOD1, which was significantly increased by treatment of deacetylase inhibitors NAM plus TSA (Figure [1E](#F1){ref-type="fig"}). Meanwhile, SOD1 depletion by two independent siRNAs significantly decreased the SOD1 acetylation enriched by deacetylase inhibitors, further supporting the specificity of this antibody ([Supplemental Figure S2](#SD1){ref-type="supplementary-material"}). These results together demonstrated the acetylation of SOD1 at K71, which intrigued us to explore the biological significance of SOD1 acetylation. Acetylation inactivates the dismutase activity of SOD1 {#s2_2} ------------------------------------------------------ We asked whether acetylation of SOD1 affected its enzymatic activity. The dismutase enzymatic activity of SOD1 was measured using a specific in-gel enzymatic activity assay using the native polyacrylamide gel electrophoresis. Treatment with deacetylase inhibitors NAM or TSA, similar to SOD1 inhibitor DDTC, resulted in the reduction of SOD1 activity while the SOD1 protein level was not affected in parallel (Figure [2A](#F2){ref-type="fig"}), suggesting that acetylation of SOD1 negatively regulates the SOD1 activity. For further confirmation, we compared the enzymatic activity of wild type SOD1, K71R mutant and acetylation mimetic K71Q mutant. Flag-tagged wild type or mutant constructs was transfected into HCT-116 cells, and the enzymatic activity of endogenous and exogenous SOD1 was differentiated by their diverse migration in the native polyacrylamide gel electrophoresis. K71R mutant behaved similar to wildtype SOD1 in the activity assay, whereas the K71Q mutant showed a significant decrease in the catalytic activity (Figure [2B](#F2){ref-type="fig"}). These results suggested acetylated SOD1 as an inactive form of SOD1. ![SOD1 acetylation inactivates SOD1\ **A.** SOD1 activity is decreased by NAM/TSA treatment. Cells were treated by TSA (500 nM) and NAM (10 mM) for 12 hr and SOD1 activity was determined using an in-gel SOD1 activity assay; **B.** Mutations of the acetylation lysine residues affect SOD1 activity. Flag-tagged SOD1 WT, K71R or K71Q mutant was transfected into HCT116 cells and SOD1 activity was measured as in **A.**. endo, endogenous SOD1, exo, Flag-tagged SOD1; **C.** Sufficient supply of copper rescues the activity of K711Q mutant; **D.** The interaction between SOD1 and CCS is decreased by NAM/TSA treatment. HCT-116 cells transfected with Flag-tagged SOD1 were treated with NAM/TSA for 12 hr. The presence of CCS in the immunoprecipitated protein complexes was assessed by immunoblotting; **E.** Acetylation mimetic mutation at K71 decreased the interaction between SOD1 and CCS; **F.** The level of SOD1 homodimers is reduced under NAM/TSA treatment. HCT116 cells were treated with NAM/TSA for 12 hr. Cells lysis in the presence or absence of DMP (20 mM) were assessed by immunoblotting using SOD1 antibody; **G.** Acetylation mimetic mutation at K71 decreased the level of SOD1 homodimers. HCT116 cells stably knock-down SOD1 (shSOD1) were transfected with Flag-tagged SOD1 WT, SOD1K71R or SOD1K71Q. SOD1 homodimers were detected as described in **C.**.](oncotarget-06-20578-g002){#F2} Acetylation of SOD1 disrupts its interaction with CCS {#s2_3} ----------------------------------------------------- We then asked how acetylation affected the SOD1 activity. To address this question, we inspected the multi-step process of SOD1 maturation, which involves zinc binding, copper loading by CCS, and homodimerization prior to turning into an active homodimeric enzyme. We firstly examined whether the impaired SOD1 activity was due to the impaired zinc or/and copper loading, which initiates the process of SOD1 maturation. To this end, the acetylation mimetic K71Q mutant was incubated with increasing amount of zinc or copper to examine whether the deficient SOD1 activity could be rescued by sufficient zinc/copper supplies. Indeed, we observed that copper incubation instead of zinc incubation was able to reverse the enzymatic activity of K71Q mutant to the similar level of wildtype SOD1 (Figure [2C](#F2){ref-type="fig"}). This data largely excluded the possibility of impaired zinc loading of the K71Q mutant, and led us to speculate that acetylation of SOD1 probably affected its interaction with CCS, a SOD1 binding partner specifically responsible for copper delivery. As such, flag-tagged SOD1 was transfected into HCT-116 cells and the interaction between SOD1 and CCS were assessed using co-immunoprecipitation assay. It was found that treatment with NAM and TSA, which effectively enriched cellular SOD1 acetylation, largely disrupted the interaction between SOD1 and CCS (Figure [2D](#F2){ref-type="fig"}). Further, we compared CCS binding ability between SOD1 wildtype and the two acetylation relevant mutants. The interaction between CCS and the acetylation defective K71R mutant was enhanced compared with that of wildtype SOD1, whereas K71Q mutant exhibited decreased association with CCS (Figure [2E](#F2){ref-type="fig"}). CCS binding and it mediated copper loading are required for the subsequent homodimerization and ultimate activation of SOD1. Therefore, we also examined the impact of acetylation on SOD1 homodimerization. The dimerized SOD1 was visualized by the treatment with cross-linking reagent dimethyl pimelimidate (DMP) prior to denatured polyacrylamide gel electrophoresis. SOD1 dimmers were recognized as a subset with shifted molecular weight. In agreement with our results above, deacetylase inhibition using NAM and TSA decreased the proportion of endogenous SOD1 homodimers (Figure [2F](#F2){ref-type="fig"}). Further, SOD1 wildtype, K71R or K71Q mutant was transfected into HCT-116 cells, where endogenous SOD1 was depleted using short hairpin (shRNA) to eliminate the interference of endogenous band for recognizing ectopically expressed proteins. Compared with the wildtype SOD1, acetylation mimetic mutant SOD1 K71Q showed a considerable decrease in the proportion of homodimers, whereas K71R barely effected SOD1 dimmers formation (Figure [2G](#F2){ref-type="fig"}). Together, we concluded that SOD1 acetylation at K71 disrupted the interaction between SOD1 and CCS, which impaired formation of SOD1 homodimers, and in turn attenuated the enzymatic activity of SOD1. SIRT1 deacetylates SOD1 acetylation {#s2_4} ----------------------------------- Protein acetylation is critically regulated by deacetylases, which are often dysregulated in cancer cells and result in aberrant acetylation status in cancer cells \[[@R21], [@R22]\]. To identify responsible deacetylases may help understand the physiological significance of SOD1 acetylation. HDACs and sirtuins represent the two major classes of protein deacetylases and can be respectively inhibited by the pan-inhibitor TSA and NAM. According to our results in Figure [1E](#F1){ref-type="fig"}, NAM treatment considerably increased the level of SOD1 K71 acetylation whereas TSA treatment only had a marginal effect, suggesting that sirtuin family members are primarily involved in SOD1 deacetylation. To further identify the involved sirtuin(s), endogenous sirtuin 1 (SIRT1) to 7 was individually depleted using specific siRNAs followed by the detection of SOD1 acetylation. Among the 7 sirtuins, only SIRT1 depletion dramatically increased endogenous level of SOD1 acetylation at K71 (Figure [3A](#F3){ref-type="fig"} and [Supplemental Figure S3](#SD1){ref-type="supplementary-material"}). SIRT1 depletion enriched the acetylation of the wildtype SOD1 but not that of the K71R mutant (Figure [3B](#F3){ref-type="fig"}), suggesting a dominant role of SIRT1 in the regulation of SOD1 acetylation at K71. Further, the enhanced SOD1 acetylation by SIRT1 knockdown was reversed by the ectopic expression of wildtype SIRT1 but not the catalytically inactive H363Y mutant. The alteration in the SOD1 acetylation was associated with the interaction change between SOD1 and CCS (Figure [3C](#F3){ref-type="fig"}). The impact of SIRT1 on SOD1 acetylation was also reflected by the enzymatic activity of SOD1. Reconstitution of wildtype SIRT1 in SIRT1 stably depleted cells partially rescued the suppressed SOD1 activity by SIRT1 downregulation, but the effect of H363Y mutant was indiscernible (Figure [3D](#F3){ref-type="fig"}). In support of all these observations, we detected the interaction between SOD1 and SIRT1, as shown by the detection of SIRT1 in the immunocomplex pull-downed by Flag-tagged SOD in HCT-116 cells (Figure [3E](#F3){ref-type="fig"}). We also tested whether the SOD1 and SIRT1 was able to interact endogenously. Immunoprecipitation of endogenous SOD1 using an anti-SOD1 antibody revealed the interaction of SOD1 with endogenous SIRT1 in HCT-116 cells (Figure [3F](#F3){ref-type="fig"}). ![SIRT1 deacetylates SOD1\ **A.** SIRT1 knockdown increased SOD1 acetylation. Cells were transfected with siRNA against SIRT1 to 7 and SOD1 acetylation was detected by anti-Ac SOD1 (K71) antibody; The Ac-SOD1 band was indicated by the solid triangle. **B.** SIRT1 deacetylated SOD1 acetylation on K71. HCT116 ShSIRT1 cells were transfected with flag-tagged SOD1 wildtype (WT) or K71R mutant. SOD1 acetylation was detected by using a pan-acetyl-lysine antibody following anti-Flag immunoprecipitation; **C.** Catalytic activity of SIRT1 is required for the deacetylation of SOD1. HA-tagged SOD1 was co-transfected with Flag-tagged SIRT1 wildtype (WT) or catalytically inactive mutant H363Y into SIRT1-depleted HCT-116 cells. Acetylation was detected as in **B.**, **D.** Catalytic activity of SIRT1 is required for the regulation of SOD1 activity. Endogenous SOD1 activity was measured in shSIRT1 cells reconstituted with SIRT1 WT or H363Y mutant; **E.**, **F.** SOD1 interacts with SIRT1. E, HA-tagged SOD1 was co-transfected with Flag-tagged SIRT1 into HCT-116 cells. Co-immunoprecipitation was performed using anti-Flag agarose. The presence of SOD1 in the immunoprecipitate was assessed by immunoblotting. F, interaction of endogenous SOD1 and SIRT1. Endogenous SOD1 was immunoprecipitated with a rabbit polyclonal antibody against SOD1. Rabbit Immunoglobulin G (IgG) was used as a negative control. The presence of SIRT1 in the immunoprecipitate was assessed by immunoblotting with an antibody against SIRT1.](oncotarget-06-20578-g003){#F3} These findings collectively indicated that SIRT1 deacetylates SOD1 at K71, which promotes its interaction with CCS, and enhances the enzymatic activity of SOD1. Genotoxic agents induce SOD1 acetylation via ROS generation {#s2_5} ----------------------------------------------------------- SIRT1 is essentially involved in coping with various stress including oxidative stress, while SOD1 plays a key role in scavenging cellular ROS \[[@R23]\], a natural byproduct of the normal oxygen metabolism but dramatically increased in environmental stress such as chemotherapy. Indeed, mounting evidence has suggested that cytotoxic anticancer agents induced oxidative stress contribute to the anticancer efficacy of these agents \[[@R24]--[@R29]\]. These information together implicate a possible involvement of SOD1 acetylation in cytotoxic agents caused oxidative stress. We then treated HCT-116 cells with various genotoxic anticancer agents including cisplatin (CDDP), camptothecin (CPT), etoposide (VP16) and bleomycin (BLM) to test the possible impact on SOD1 acetylation. Treatment with these agents all considerably increased SOD1 acetylation (Figure [4A](#F4){ref-type="fig"}). Interestingly, the impact on the increase of SOD1 acetylation was correlated with the level of ROS accumulation caused by these agents ([Supplemental Figure S4A](#SD1){ref-type="supplementary-material"} and [S4B](#SD1){ref-type="supplementary-material"}). Moreover, pretreatment with ROS scavenger N-acetylcysteine (NAC) or inhibition of NADPH oxidase using apocynin (APO), apparently reversed DNA damaging agents induced SOD1 acetylation (Figure [4B](#F4){ref-type="fig"}), implicating that genotoxic stress associated ROS generation accounted for the enhanced SOD1 acetylation. ![Genotoxic agents promote SOD1 acetylation and inactivate SOD1\ **A.**, Genotoxic agent induces SOD1 acetylation. HCT-116 were treated with camptothecin (CPT, 100 nM), etoposide(VP-16, 10 μM), bleomycin (BLM, 10 μM) or cisplatin (CDDP, 20 μM) for 12 hr and SOD1 acetylation was examined using an anti-Ac SOD1 (K71) antibody; **B.** CPT-induced SOD1 acetylation is mediated by ROS. Cells were pretreated with or without NAC or APO for 2 hr before exposure to CPT (100 nM, 12 hr); **C.**, **D.** The impact of CPT treatment on SOD1 acetylation and enzymatic activity. Time-dependent **C.** and dose-dependent **D.** induction of SOD1 acetylation and corresponding alteration in SOD1 activity was measured using immunoblotting or an in-gel enzymatic assay; **E.** The impact of CPT treatment on the interaction between SOD1 and CCS. Flag-tagged SOD1 WT, K71R, K71Q was transfected into HCT116 cells treated with or without 100 nM CPT for 12 hr. The presence of CCS in the immunoprecipitated protein complexes was assessed by immunoblotting. **F.** The impact of CPT on the formation of SOD1 homodimers. Flag-tagged SOD1 WT, K71R, K71Q was transfected into HCT116 cells treated with or without 100 nM CPT for 12 hr; **G.** The impact of CPT on the interaction between SOD1 and SIRT1. HA-tagged SOD1 were co-transfected with Flag-tagged SIRT1 into HCT-116 cells; **H.** The impact of SOD1 acetylation on cytosolic superoxide levels under CPT stimulation. HCT116 cells stably knock-down SOD1 (shSOD1) were transfected with Flag-tagged SOD1 (WT), SOD1K71R or SOD1K71Q, and cells were treated with CPT. The cytosolic superoxide level was measured by DHE staining and FACS analysis. Data represent mean ± SEM (*n* = 3), \*\**P* \< 0.01. **I.** SOD1 Acetylation promotes cells sensitive to CPT treatment. The sensitivity of cells HCT116 shSOD1 cells expressing the wildtype, K71R or K71Q SOD1 were treated with CPT (1 nM) for 14 days and measured by a cologeneic assay.](oncotarget-06-20578-g004){#F4} We then used CPT as a representative to follow up the impact of DNA damage on SOD1 acetylation. CPT caused DNA damage, as reflected by p53 upregulation, induced SOD1 acetylation in a dose- and time-dependent manner. The induced increase of SOD1 acetylation was closely associated with a decline of SOD1 activity (Figure [4C](#F4){ref-type="fig"} and [4D](#F4){ref-type="fig"}), and the interaction with CCS ([Supplemental Figure S5A](#SD1){ref-type="supplementary-material"} and [S5B](#SD1){ref-type="supplementary-material"}). Consistently, CPT treatment disrupted the interaction between SOD1 and CCS but failed to affect the CCS binding to K71R mutant, suggesting an acetylation-dependent impact on CCS binding (Figure [4E](#F4){ref-type="fig"}). In line with these findings, CPT treatment suppressed dimerization of wildtype SOD1 but did not affect either K71R or K71Q mutant (Figure [4F](#F4){ref-type="fig"}, [Supplemental Figure S6A](#SD1){ref-type="supplementary-material"} and [S6AB](#SD1){ref-type="supplementary-material"}). Importantly, we also observed that treatment with CPT largely disrupted the interaction between SOD1 and SIRT1 (Figure [4G](#F4){ref-type="fig"}), which may explain the enhanced SOD1 acetylation upon DNA damage. Together, these data provided the first evidence showing that chemotherapies induce SOD1 acetylation and impair its enzymatic activity in cancer cells, which may result from disrupted interaction with SIRT1. SOD1 is essential for maintaining the redox homeostasis of cancer cells. CPT treatment increased ROS generation in cancer cells ([Supplemental Figure S4B](#SD1){ref-type="supplementary-material"}) whilst suppressed SOD1 activity (Figure [4C](#F4){ref-type="fig"} and [4D](#F4){ref-type="fig"}). This may suggest the further enhanced oxidative stress caused by SOD1 inactivation. We hence examined the impact of acetylation of SOD1 on cytosolic ROS level in HCT-116 cells, where the endogenous SOD1 was depleted using shRNA. shRNA-resistant SOD1 wild-type, K71R and K71Q SOD1 were transfected into HCT-116 cells respectively followed by the treatment of CPT. As expected, SOD1 depletion led to a dramatic increase of intracellular ROS level. Ectopic expression of wildtype SOD1 was able to rescue the ROS generation resulted from SOD1 depletion. In contrast, the acetylation-mimetic K71Q mutant, with impaired enzymatic activity, failed to reverse the intracellular ROS level (Figure [4H](#F4){ref-type="fig"}). However, it appeared puzzling that the nonacetylatable K71R mutant behaved similar to the wildtype SOD1 in ROS clearance upon CPT treatment. We speculated that this was likely due to the small proportion of acetylated form induced by CPT among total amount of exogenous SOD1. Indeed, cell sensitivity to CPT treatment measured by a cologeneic assay was consistent with results of the ROS clearance. Cells expressing acetylation-mimetic SOD1 K71Q mutant were more sensitive than either wild-type SOD1 and K71R mutant to CPT treatment (Figure [4I](#F4){ref-type="fig"}). SOD1 acetylation sensitizes cancer cells to DNA damaging agents {#s2_6} --------------------------------------------------------------- The substantial impact of K71Q mutant shown above suggests a possibility that the abundance of SOD1 acetylation may be a determinant of the sensitivity to the CPT-based chemotherapies, which are used in the clinical therapy of various types of human cancers including the first line treatment for colon cancer. We then probed the status of SOD1 acetylation at K71 in a panel of colon cancer cells, and found that the basal level of SOD1 acetylation varied largely across the cells (Figure [5A](#F5){ref-type="fig"}). Some cells lines, like HCT-8 and HCT-16, displayed massive abundance of intrinsic SOD1 acetylation. These data suggest that SOD1 acetylation status may confer a distinct antioxidant capacity across cancer cells, and those with low capacity may be more susceptible to CPT-induced oxidant stress. Indeed, we found that cells with high SOD1 acetylation were relatively more sensitive to CPT treatment. We also tested whether Ac-SOD1 level alteration in responses to CPT was correlated with the CPT sensitivity of these cells as well. Ac-SOD1 level was examined after 12 hr exposure to CPT treatment in the colon cancer cell lines ([Supplemental Figure S7](#SD1){ref-type="supplementary-material"}). It was found that cell with higher basal level of Ac-SOD1 showed more significant increase of Ac-SOD1 level upon CPT treatment, suggesting a correlation between Ac-SOD1 level change and the response to CPT treatment. ![SOD1 acetylation is associated with the response to CPT treatment\ **A.** SOD1 acetylation and cell sensitivity to CPT-11 in colon cancer cell lines. **B.**, **C.** Intratumoral SOD1 acetylation level and response to CPT-11 treatment of patient-derived xenograft models. **B.** Growth curves of tumors from indicated PDX models treated with vehicle control or 10 mg/kg CPT-11 daily for 21 consecutive days. Data represent mean tumor volume ± SEM (*n* = 6), \*\*\**P* \< 0.001; \**P* \< 0.05; n.s., not significant. **C.** Correlation between SOD1 K71 acetylation and sensitivity to CPT-11. T/C%, relative tumor volume versus vehicle control on day 21.](oncotarget-06-20578-g005){#F5} Apart from colon cancer, we also tested whether basal Ac-SOD1 levels were correlated with the sensitivity to CPT treatment in lung cancer cells. The sensitivity of 13 lung cancer cells towards topotecan, a CPT analogue, was extracted from Cancer Cell Line Encyclopedia (CCLE) database. Immunoblotting detection of Ac-SOD1 level from those cells revealed that the basal level of Ac-SOD1 was correlated with the sensitivity to CPT treatment in lung cancer cells ([Supplemental Figure S8](#SD1){ref-type="supplementary-material"}). This data suggested that correlation of Ac-SOD1 and camptothecin-sensitivity could be a general mechanism beyond colon caner For further confirmation, we proceeded to validate this finding in patient-derived xenograft (PDX) models, which are believed to faithfully resemble the characteristics of human tumors in many aspects including heterogeneity, histology and genetic alterations \[[@R30]--[@R33]\]. CPT-11 efficacy was screened in a small panel of PDX models across different cancer types including lung cancer (LU0299, LU0743, LU0375, LU0350, LU0377), liver cancer (LI0941) and esophageal cancer (ES0204). We observed the diverse response of these models to CPT-11 treatment and subgrouped the models into CPT sensitive and resistant subset (Figure [5B](#F5){ref-type="fig"}). Meanwhile, we also measured the basal level of SOD1 acetylation on K71. The models with higher level of SOD1 acetylation were more responsive to the treatment (Figure [5C](#F5){ref-type="fig"}). These results suggest a potential value of SOD1 K71 acetylation in stratifying the responsive subset to CPT-11 based chemotherapy. DISCUSSION {#s3} ========== The increased generation of ROS and altered redox status in cancer cells offers an interesting therapeutic window that cancer cells are more sensitive than normal cells to agents causing further accumulation of ROS \[[@R4]\]. In fact, direct or indirect affects on ROS amount have been widely believed to contribute to the anticancer efficacy of cytotoxic anticancer agents, in particular genotoxic agents. Generation of high levels of ROS has been observed in patients receiving various chemotherapy treatment \[[@R24]--[@R29]\], though the mechanism of ROS generation may vary among the agents \[[@R34]\]. Apart from the widely studied ROS generation, the molecular insights into the ROS homeostasis changes by genotoxic agents have been very limited. In this study, we have provided the first evidence showing that genotoxic agents caused ROS accumulation was able to impair the antioxidant capacity of cancer cells via diminishing the activity of antioxidant enzyme SOD1. Our findings suggest the existence of a positive feedback mechanism in which ROS per se mediates the impairment of the antioxidative enzyme (defence) system of cancer cells (Figure [6](#F6){ref-type="fig"}). The feedback inhibition of SOD1 further raises the cytosolic ROS level, reinforces oxidative stress, and promotes the effectiveness of the anticancer agents. ![A schematic model showing the positive feedback of ROS induction resulted from SOD1 acetylation](oncotarget-06-20578-g006){#F6} It has long been noticed that the increase of ROS level and DNA damage, can be found one being caused by the other one; ROS induces DNA damage while DNA damage agents could also increase ROS generation. Cytotoxic anticancer agents, including cisplatin, mitomycin C, doxorubicin, CPT and ultraviolet radiation induced ROS are important for the induction of cell apoptosis and anticancer efficacy of these agents \[[@R24]--[@R29]\]. While in certain cancer cells, chemotherapeutic agents induced persistent ROS stress may induce adaptive stress responses including activation of redox-sensitive transcription factors, leading to an increase in the expression of ROS-scavenging enzymes, such as SOD and glutathione, to counteract with ROS stress. All these events enable cells to survive with the high level of ROS and render cancer cells more resistant to chemotherapeutic agents \[[@R6], [@R35]\]. Accordingly, modulating ROS-scavenging enzymes activity could enhance the anti-tumor activity of genotoxic agents via ROS mediated apoptosis induction. Intriguingly, our findings provided new insights by showing an apposing mechanism, in which the genotoxic agents, in parallel to ROS induction, are able to paralyze the antioxidant defence of cancer cells to facilitate their anticancer efficacy. Our findings are particularly interesting given the fact that cancer cells often maintain a high antioxidant capacity to cope with the massive ROS resulted from fast growth. This finding highlighted the role of antioxidant defence system in determining the efficacy the genotoxic anticancer agents, and may lead to a better understanding of the anticancer mechanism of genotoxic agents. The key molecular mechanism behind involves the acetylation of SOD1 on the lysine 71 residue. We have shown that acetylation decreases SOD1 activity by impairing the interaction between SOD1 and CCS, and hence decreasing the output of enzymatically active SOD1 homodimers in the maturation process of SOD1. We also noticed that the mutation of lysine 71 to arginine, which abolished the acetylation of SOD1, did not obviously promote the activity of SOD1. One possible explanation is that most cancer cells feature a high antioxidant capacity, in which SOD1, as a critical anti-oxidant enzyme, may maintain a high basal activity and can hardly to be further enhanced. Likewise, we have observed that the acetylation mimetic K71Q mutant exhibits striking impact on CPT-induced ROS generation and cell survival whereas the K71R mutant did not obviously promote scavenging of ROS or cell viability. We speculated that this might also partially result from the small proportion of acetylated form induced by CPT among total amount of exogenous SOD1. We also exploited how SOD1 K71 acetylation decreases CCS interaction with SOD1. As the human SOD1-CCS complex structure has not been solved yet, we took advantage of the solved crystal structure of yeast SOD1-CCS complex (ySOD1-yCCS) (PDBID: 1JK9) \[[@R36]\] and aligned the human SOD1 to yeast SOD1 in the ySOD1-yCCS crystal structure (PDBID:1JK9) using PyMol software. According to the structural alignment ([Supplemental Figure S9](#SD1){ref-type="supplementary-material"}), hSOD1 shares high structural similarity with ySOD1, and K71 (T71 in yeast SOD1) does not lie in the protein-protein interaction surface. Also, a previous crystallography study has suggested that the binding between SOD1 and CCS may involve conformational changes. We speculated that K71 acetylation may interfere with this conformational change through allosteric regulation and hinder its binding to CCS1. Previous studies have suggested two possible mechanisms by which CPT treatment may influence SOD1 acetylation. Genotoxic stress was reported to inactive SIRT1 via the ATM/ATR-DBC1 signaling \[[@R37], [@R38]\]. To test this possibility, we examined the SOD1 acetylation in ATM knockdown cells but did not observe the apparent reduction in either DBC1 phosphorylation or SOD1 acetylation, which largely ruled out the possible involvement of this pathway ([Supplemental Figure S10](#SD1){ref-type="supplementary-material"}). Alternatively, ROS induced by CPT is known to inhibit SIRT1 activity by evoking oxidative modifications on its cysteine residues \[[@R39]\]. In this case, we did not observe the alteration of SIRT1 activity upon CPT treatment ([Supplemental Figure S11](#SD1){ref-type="supplementary-material"}). Instead, our data suggested that CPT disrupted the interaction between SIRT1 and SOD1, which may possibly result from oxidative modification of SIRT1 by ROS. SIRT1 is known to play a critical role in coordinating the cellular response to stress. SIRT1 expression and activity are regulated by cellular stressors, including metabolic, genotoxic, oxidative and phototoxic stress. SIRT1 impacts cell survival by deacetylating substrate proteins to drive the cell towards a cytoprotective pathway \[[@R40]\]. Moderate overexpression of SIRT1 provides cells protection against oxidative stress by increasing the activity of catalase \[[@R41]\]. SIRT1 may also protect against oxidative stress through the modulation of family of forkhead transcription factors (FOXO) \[[@R42], [@R43]\] and the induction of manganese SOD (MnSOD) expression \[[@R44]\]. Our result supported and complemented the previous studies by showing that SIRT1 also regulates the oxidative metabolism via directly deacetylating and modulating the activity of SOD1. All these evidence suggests an important role of SIRT1 in determining the antioxidant capacity of cancer cells and hence the outcome of chemotherapy. Indeed, a previous study has observed the correlation between SIRT1 expression level and the CPT sensitivity in prostate cancer cells, although in very limited number of cell lines \[[@R45]\]. An apparent caveat is that SIRT1 expression level may not be necessarily associated with its activity. Indeed, we have observed that in colon cancer cell lines and the PDX tumors, SIRT1 protein level was not correlated with the CPT sensitivity (Figure [5A](#F5){ref-type="fig"} and [5C](#F5){ref-type="fig"}). According to our results, the basal level of SOD1 acetylation varies largely among either the cancer cell lines or patients tumor tissues; high level SOD1 acetylation is closely correlated with the increased response to CPT treatment. We speculate that while cancer cells often feature an increased antioxidant capacity, high level of SOD1 acetylation represents an intrinsic silencing of SOD1, and is also an indicator of low activity of SIRT1. Thus abundant basal level of SOD1 acetylation is able to stratify the subset with low capacity to copy with oxidative stress of cancer cells. The clinical value of SOD1 acetylation may deserve further investigation in clinical practice to increase the response rate of CPT-based chemotherapy regimen. MATERIALS AND METHODS {#s4} ===================== Plasmids {#s4_1} -------- The FLAG/HA-tagged form of SOD1 was generated by subcloning Xho I-Hind III an cassette of SOD1 into the Flag/HA-pcDNA3.1 mammalian expression vector. The plasmid pECE encoding SIRT1/SIRT1-H363Y with a FLAG tag was purchased from Addgene. The RNAi Consortium (TRC) Lentiviral shRNAs against SOD1 (Clone ID: TRCN0000039808, targeting the 3′UTR region of SOD1) and against SIRT1 (Clone ID: TRCN0000039808) were purchased from Thermo. Mutations in pcDNA3.1-SOD1-FLAG were introduced by the change site directed mutagenesis kit (Saibaisheng Gene Technolog, Shanghai, China). Sequences were verified by automated sequence analysis (Sangon Biotech, Shanghai, China). siRNA transfection {#s4_2} ------------------ For siRNA transfection, HCT116 cells were plated at 3×10^5^ cells/ml in OPTI-MEM serum-free medium and transfected with siRNA duplex using Lipofectamine RNAiMAX Reagent Agent (Life Technologies) according to the manufacturer\'s instructions. siRNAs were ordered from Sigma-Aldrich. The sequences were as follows: siSOD1 \#1: TTC GAG CAG AAG GAA AGT AAT GGA CCA dTdT; siSOD1 \#2: GGC CUG CAU GGA UUC CAU G dTdT. Cell culture {#s4_3} ------------ Human colon cancer HCT-116 cells purchased from American Type Culture Collection (ATCC) were cultured in McCOY\'s 5A medium (Life Technology) supplemented with 10% FBS. HCT116 cell lines stably transfected with short hairpin RNA targeting SOD1 or SIRT1 (ThermoFisher) (shSOD1/shSIRT1) were constructed according to manufacturer\'s instructions and maintained in McCOY\'s 5A medium supplemented with 1 μg/μl puromycin dihydrochloride (Sigma). Immunoprecipitation assay {#s4_4} ------------------------- Immunoprecipitation of Flag-tagged SOD1 was carried out using anti-FLAG M2 beads. Equal amounts of proteins in lysis Buffer were used for precipitation. Input samples represent \~1% of protein amounts used for immunoprecipitation. The following antibodies were used for immunoprecipitation and followup immunoblotting: monoclonal rabbit anti-SOD1 (Epitomics); monoclonal rabbit anti-Sir2/SIRT1 (Epitomics); monoclonal mouse anti-acetylated-Lysine(Cell Signaling Technology); monoclonal mouse anti-P53 (Santa Cruz Biotechnology); monoclonal mouse anti-CCS (Santa Cruz Biotechnology); polyclonal mouse anti-FLAG M2 affinity Gel (Sigma); monoclonal mouse anti-DYKDDDDK-Tag (Abmart, Shanghai, China); monoclonal mouse anti-HA-tag (Abmart); monoclonal rabbit anti-GAPDH (Epitomics). Antibodies specifically recognizing acetylation at lysine 71 were prepared by PTM BioLab, Inc. (Hangzhou, China). In-gel SOD1 activity assay {#s4_5} -------------------------- Cells were harvested and lysed in lysis buffer (25 mM HEPES; pH 7.9, 100 mM NaCl, 1 mM EDTA, 1% Triton X-100, 10% Glycerol, 1 mM Na~3~VO~4~, 1 mM PMSF and protease inhibitors). The supernatant was subjected to non-denaturing gel electrophoresis and nitro blue tetrazolium (NBT, Beyotime Institute of Biotechnology, Nanjing, China) staining to examine SOD1 activity as described previously with slight modifications \[[@R9]\]. Cell lysate was loaded directly without boiling to a nondenaturing 12% polyacrylamide gel. After electrophoresis, the gel was incubated for 45 min in the dark in 50 mM potassium phosphate buffer (pH 7.8) supplemented with 0.34 mM NBT and 14 mM riboflavin (Beyotime Institute of Biotechnology). The gel was subsequently exposed to light for 15 min after 1 μl/ml TEMED was added to the reaction mixture. Assessment of intracellular superoxide anions {#s4_6} --------------------------------------------- Cells were incubated with 3 μM dihydroethidium (hydroethidine, DHE; Invitrogen) in PBS for 30 min at 37°C following the treatment with CPT at 100 nM for 12 hr. Cells were washed once with PBS, trypsinized, and immediately analyzed using flow cytometry on a FACSCalibur flow cytometer (BD biosciences; software CellQuest Pro). Visualization of SOD1 monomers and dimers {#s4_7} ----------------------------------------- Cells were rinsed once in PBS prior to lysis in lysis buffer (25 mM HEPES (PH 7.5), 150 mM NaCl, 1% NP-40, 100 mM MgCl~2~ 6H~2~O, 1 mM Na~3~VO~4~, 1 mM EDTA, 1 mM PMSF, and protease inhibitor cocktail). 5 mg/ml protein lysis was incubated for 30 min at room temperature with or without 20 mM Dimethyl pimelate. The samples were incubated with 0.2 M Tris-HCl (PH 7.5) for 15 min at room temperature and analyzed using immunoblotting. Clonogenic assay {#s4_8} ---------------- Cells were plated in triplicate (500 cells/well in 6 well plates), treated with indicated doses of CPT and allowed to grow for 2 weeks. Plates were fixed in methanol and stained with 0.5% crystal violet solution. Colonies were counted and results were expressed as percentages of the untreated control. Animal studies {#s4_9} -------------- Animal studies using patient derived xenograft models were conducted by Crown Bioscience, Inc (Taicang, China) in strict accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. Prior to the initiation of treatment, mice were randomized into control and treated groups (*n* = 6 per group). Mice were treated intravenously with CPT-11 daily at 10 mg/kg or vehicle control for 21 consecutive days. Tumor growth was monitored by the measurement of tumor size using caliper every other day. Mice were sacrificed and tumor tissues were collected 2 hr after the last dosing. Statistical analysis {#s4_10} -------------------- A two-tailed independent Student\'s t test was used to compare the continuous variables between the two groups. The *p* value \< 0.05 was considered to be statistically signiﬁcant. SUPPLEMENTARY FIGURES AND TABLES {#s5} ================================ We thank Dr. Zhongyi Cheng (PTM BioLab, Inc, Hangzhou, China) for the assistance to generate Ac-SOD1-K71 antibody. This work was supported by grants from the National Science and Technology Major Project of the Ministry of Science and Technology of China (No. 2012ZX09301001-007, 2012ZX09301001-004, 2015ZX09101009), the National Marine "863" Project (No. 2013AA092902) and National Natural Science Foundation (No. 81102461). M.H. is a young scholar funded by the National Natural Science Foundation of China (NO. 81202549) and the Pujiang Scholar Program Grant by the Shanghai Metropolitan Government (No. 12PJ1410400). **CONFLICTS OF INTEREST** The authors declared no conflict of interest.

{ "pile_set_name": "PubMed Central" }

Background {#Sec1} ========== Betel quid chewing is one of the main causes leading to oral cancer in Taiwan \[[@CR1]\]. Arecoline, one of main effective components in betel quid, was reported to lead to DNA damage and apoptosis through the formation of reactive oxygen species (ROS) and contribute to oral carcinogenesis \[[@CR2]-[@CR5]\]. Therefore, the modulation of ROS level may be helpful for oral cancer prevention and therapy. Grape seed extract (GSE) is a common dietary health supplement due to its natural ROS modulating ability \[[@CR6]\]. Commercial preparations of GSE are marketed in the world as a dietary health supplements due to their natural free radical scavenging ability \[[@CR6]\]. The cancer chemoprevention and anticancer potential of GSE has been well reviewed previously \[[@CR7]\] including skin, colorectal, prostate, breast, lung, and gastric cancers. However, the GSE effects with respect to oral cancer cells are less studied as yet. ROS modulation effect has been well reviewed \[[@CR8],[@CR9]\]. For example, cellular ROS may regulate apoptosis through the mitochondrial pathway \[[@CR10]-[@CR13]\]. Pro-oxidants induce ROS specifically targeting cancer cells, thereby activating signal transduction pathways that are responsible for cell cycle arrest and/or apoptosis \[[@CR14]\]. Similarly, GSE was reported to generate a strong superoxide radical-associated oxidative stress and result in the apoptosis of non-small-cell lung cancer cells \[[@CR15]\] as well as in the induction of DNA damage \[[@CR16]\]. Different concentrations of GSE were reported to generate diverse biological effects in several cancer studies \[[@CR17]-[@CR20]\]. For example, high concentrations (25--100 μg/ml) of GSE showed cytotoxicity or antiproliferation of human bladder \[[@CR17]\], colorectal \[[@CR21]\], and breast \[[@CR18]\] cancer cell lines. In contrast, a low concentration (2.5 μg/ml) of GSE was reported to inhibit the micronuclei frequency and ROS generation in a lymphocyte culture, demonstrating that its antioxidant property has a protective effect during oxidative stress \[[@CR19]\]. However, more detailed mechanisms between cancer chemoprevention and anticancer effects of GSE in terms of concentration effects remain unclear. Since GSE is a natural ROS scavenger, we hypothesize that GSE modulates ROS to further regulate proliferation, apoptosis, mitochondrial function, and DNA damage. Since concentration responses of GSE for these regulations may be relevant, in this study we aim to define the critical concentrations that may or may not be able to induce apoptosis in oral cancer cells. Methods {#Sec2} ======= GSE source {#Sec3} ---------- The IH636 premium grade proanthocyanidin grape seed (*Vitis vinifera*) extract (GSE, commercially known as ActiVin®) was purchased from InterHealth Nutraceuticals Inc. (Benicia, CA, USA), which included 75--80% oligomeric proanthocyanidins and 3--5% monomeric proanthocyanidins as described previously \[[@CR22]\]. Cell cultures {#Sec4} ------------- Cell lines of human oral gingival cancer Ca9-22 \[[@CR23]\] and gingival fibroblast HGF-1 \[[@CR24]\] were routinely maintained in DMEM/F12 medium (Gibco, Grand Island, NY) containing 10% fetal bovine serum, 0.03% glutamine, 1 mM sodium pyruvate, and penicillin/streptomycin mixtures. Cells were kept at 37°C in a humidified incubator containing 5% CO~2~. Determination of cell viability {#Sec5} ------------------------------- Viability analysis was performed using Cell Titer 96™ Aqueous One solution cell proliferation (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4- sulfophenyl)-2H-tetrazolium) MTS) assay kit (Promega Madison, WI, USA) as described previously \[[@CR25]\] with minor modification. In brief, cells were treated with various concentrations of GSE in fresh media in triplicates. The non-toxic concentration of DMSO (less than 1% v/v) was used to prepare test solutions in all assays. The plates were then incubated for 24 h under standard growth conditions. Subsequently, MTS reagent was loaded to each well (5 mg/ml in PBS) and cells were again incubated for another 2 h. Then, absorbance of each well was recorded directly at 490 nm by ELISA multi-Plate Reader (MTX Lab Systems, Inc., Vienna, VA, USA). Determination of sub-G1 population {#Sec6} ---------------------------------- Measurement of DNA content for cell cycle analysis were carried out by flow cytometry, based on a previously described protocol \[[@CR26]\]. In brief, Ca9-22 cells were treated with either DMSO only or different GSE concentrations for 24 h. After incubation, cells were harvested for washing and fixing in 70% ethanol overnight. After harvest, cells were resuspended in 1 ml PBS containing 10 μg/ml PI (Sigma, St Louis, MO, USA) in the dark. Subsequently, cells were analyzed using a flow cytometer (FACScan; Becton-Dickinson, Mansfield, MA) at excitation and emission settings of 480 and 525 nm, respectively, and Win-MDI software (<http://facs.scripps.edu/software.html>). Determination of apoptosis by annexin V/PI {#Sec7} ------------------------------------------ The induction of apoptosis by GSE-treated Ca9-22 cells was analyzed by annexin V staining as previously described \[[@CR27]\]. Briefly, cells were treated with either vehicle or various GSE concentrations for 24 h. Subsequently, the cells were trypsinized, washed twice with PBS and stained with fluorescein isothiocyanate (FITC)-labelled annexin V. Then, the samples were measured with a flow cytometer (FACSCalibur; Becton-Dickinson) for the quantification of apoptotic cells at excitation and emission settings of 480 and 525 nm, respectively, and Win-MDI software. Determination of apoptosis by pan-caspase activity {#Sec8} -------------------------------------------------- The induction of apoptosis by GSE-treated Ca9-22 cells was analyzed by activation of caspases (caspase-1, 3, 4, 5, 6, 7, 8, 9) by the generic caspase activity assay kit (Abcam, Cambridge, UK) as previously described \[[@CR27]\]. Briefly, cells were treated with either vehicle or various GSE concentrations for 24 h. After harvest, the cells were suspended and stained with 1 X fluorescent TF2-Val-Ala-Asp (VAD)-FMK at the cell incubator for 1 h. Then, the samples were measured with a flow cytometer (BD Accuri C6; Becton-Dickinson, Mansfield, MA, USA) and a BD Accuri C6 Software (version 1.0.264) for the quantification of pan-caspase positive populations at excitation and emission settings of 480 and 525 nm, respectively. Determination of intracellular ROS {#Sec9} ---------------------------------- Intracellular redox state were determined by the ROS-sensitive dye 2′,7′-dichlorodihydrofluorescein diacetate (DCFH-DA) (Sigma Chemical Co., St. Louis, MO, USA) as previously described \[[@CR25],[@CR28]\]. Ca9-22 cells were treated with various concentrations of GSE for 24 h. Subsequently, cells were harvested, thoroughly washed, resuspended in 10 μM DCFH-DA in PBS and then incubated at 37°C for 30 min in darkness. After incubation, cells were washed, resuspended in PBS, and analyzed with a FACSCalibur flow cytometer at excitation and emission settings of 480 and 525 nm, respectively, and Win-MDI software. Determination of mitochondrial membrane potential {#Sec10} ------------------------------------------------- Mitochondrial membrane potential (MitoMP) was determined by flow cytometry using MitoProbe™ DiOC2(3) assay kit (Invitrogen, San Diego, CA, USA) as described previously \[[@CR25]\]. In brief, cells were incubated with various GSE concentrations at 37°C for 24 h. Subsequently, cells were incubated in culture medium (containing 50 μM of DiOC~2~(3)) at 37°C for 20 min in an incubator. After washing and resuspension in PBS, cells were subjected to flow cytometric analysis. The fluorescence intensity was measured using 488 and 525 nm filter settings for the excitation and emission wavelengths, respectively. The data were analyzed with Win-MDI software. Determination of DNA double strand breaks (DSBs) by γH2AX/PI double staining {#Sec11} ---------------------------------------------------------------------------- DSBs were measured by flow cytometry as described previously \[[@CR25]\]. Ca9-22 cells were incubated with various GSE concentrations for 24 h, followed by fixation with 70% ethanol overnight. After washing twice with BSA-T-PBS (1% bovine serum albumin and 0.2% Triton X-100 in PBS), cells were treated with 100 μl of BSA-T-PBS solution containing 0.2 μg monoclonal antibody against p-Histone H2A.X (Ser 139) (Santa Cruz Biotechnology, Santa Cruz, CA, USA) for overnight at 4°C. After washing, cells were resuspended in Alexa Fluor 488-tagged secondary antibody (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA) at a 1:100 dilution for 1 h at 4°C. After washing, cells were resuspended in 1 ml PBS containing 5 μg/ml PI and analyzed by a FACSCalibur flow cytometer and Win-MDI software. Statistical analysis {#Sec12} -------------------- Statistical analysis was performed with JMP 9 software. One-way ANOVA with Tukey's HSD Post Hoc Test was used to analyze significant differences between treatments. Unless otherwise indicated, all experiments were repeated in triplication. Results {#Sec13} ======= Cell viability {#Sec14} -------------- To access the potential harmful effects of GSE on Ca9-22 cells, cell viability was determined by MTS reagent. At low concentrations of GSE treatment (1--10 μg/ml), the cell viabilities maintained around 91% (Figure [1](#Fig1){ref-type="fig"}). In contrast, a significantly concentration-dependent decrease in cell viability (*P* \< 0.005--0.0001) was observed at high GSE concentrations (50, 100, 200 and 400 μg/ml). The IC~50~ of GSE for Ca9-22 cells was 150 μg/ml at 24 h incubation. However, both the low and high concentrations of GSE were not harmful to normal oral HGF-1 cells.Figure 1**Cytotoxicity of GSE treatments on oral cancer Ca9-22 cells and normal oral HGF-1 cells.** Cells were treated with either vehicle control (DMSO) or with 2, 4, 8, 10, 50, 100, 200 and 400 μg/ml of GSE for 24 h. Cell viability was detected by the MTS assay. The percent cell viability in the experimental groups was adjusted to the DMSO-treated group representing 100% viability. Data, mean ± SD (n = 10 and 5 for Ca9-22 and HGF-1 cells, respectively). Treatments with the same capital letter are nonsignificant. Cell cycle distribution by GSE treatments {#Sec15} ----------------------------------------- To investigate if GSE treatments cause change in cell cycle distribution in Ca9-22 cells, a standard PI-staining protocol was applied to GSE-treated Ca9-22 cells. In Figure [2](#Fig2){ref-type="fig"}A, the cell cycle distributions were stable at low concentrations (1--10 μg/ml) of GSE but the subG1 populations were gradually accumulated at high concentrations of GSE. In Figure [2](#Fig2){ref-type="fig"}B, the change in the sub-G1 populations (%) of Ca9-22 cells was not significant at low concentrations of GSE. However, the changes in the sub-G1 populations (%) significantly increased to 5.61, 16.73, 25.69 and 26.80 in a concentration-dependent manner (*P* \< 0.0001) when GSE concentrations were increased at 50, 100, 200 and 400 μg/ml, respectively. Additionally, the percentage changes in other cell cycle phases (i.e. the G1, S and G2/M phases) did not exhibit significant changes in all treatment groups compared to untreated samples.Figure 2**Analysis on distribution of cell cycle in GSE-treated Ca9-22 cells.** Ca9-22 cells were treated with indicated GSE concentrations (0--400 μg/ml) for 24 h before being harvested, fixed and stained with PI for cell cycle analysis. **(A)** Representative histograms for cell cycle phases in GSE-treated Ca9-22 cells. **(B)** Quantitative analysis on distribution of cell cycle phases. Data, mean ± SD (n = 3). Treatments with the same capital letter are nonsignificant. Apoptotic cell death: annexin V/PI {#Sec16} ---------------------------------- To determine the degree of apoptosis of GSE-induced cell death in Ca9-22 cells, the annexin V-FITC staining was determined by flow cytometry. In Figure [3](#Fig3){ref-type="fig"}A, the apoptosis signals were similar at low concentrations (1--10 μg/ml) of GSE but they gradually increased at high concentrations of GSE. In Figure [3](#Fig3){ref-type="fig"}B, the annexin V intensity of GSE-treated Ca9-22 cells was weak at low concentrations of GSE. However, the percentage changes in the annexin V intensity significantly increased to 14.19%, 20.03%, 44.53, and 72.86% in a concentration-dependent manner at high concentrations (50, 100, 200 and 400 μg/ml) of GSE (*P* \< 0.05--0.0001).Figure 3**GSE induced annexin V/PI-based apoptosis in Ca9-22 cells.** Cells were treated with indicated GSE concentrations (0--400 μg/ml) for 24 h. **(A)** Histograms of representative annexin V-FITC profile in GSE-treated Ca9-22 cells. **(B)** Quantitative analysis for the apoptotic cells (%). Apoptosis was counted at the intensity of right gated region in **(A)**. Data, mean ± SD (n = 3). Treatments with the same capital letter are nonsignificant. Apoptotic cell death: caspase activity {#Sec17} -------------------------------------- To further determine the degree of apoptosis of GSE-induced cell death in Ca9-22 cells, the multiple caspase (pan-caspase) activity staining was determined by flow cytometry. In Figure [4](#Fig4){ref-type="fig"}A, the apoptosis signals based on pan-caspase intensities were similar at low concentrations (1--10 μg/ml) of GSE but they gradually increased at high concentrations of GSE. In Figure [4](#Fig4){ref-type="fig"}B, the pan-caspase positive intensity of GSE-treated Ca9-22 cells was weak at low concentrations of GSE. However, the percentage changes in the pan-caspase positive intensity concentration-responsively increased at high concentrations (50, 100, 200 and 400 μg/ml) of GSE, respectively (*P* \< 0.01--0.001).Figure 4**GSE induced pan-caspase-based apoptosis in Ca9-22 cells.** Cells were treated with indicated GSE concentrations (0--400 μg/ml) for 24 h. **(A)** Histograms of representative pan-caspase activity profile in GSE-treated Ca9-22 cells. Caspase positive % was indicated in the right gated region of each panel. **(B)** Quantitative analysis for the pan-caspase positive (%). Data, mean ± SD (n = 3). Treatments with the same capital letter are nonsignificant. Intracellular ROS {#Sec18} ----------------- ROS plays a pivotal role in regulating cellular apoptosis. To determine whether Ca9-22 cells exposed to GSE bore higher levels of ROS, we monitored the intracellular ROS levels using the fluorescent dye DCFH-DA as a specific ROS scavenger. In Figure [5](#Fig5){ref-type="fig"}A, the ROS signals were similar at low concentrations (1--10 μg/ml) of GSE but they were gradually increased at high concentrations of GSE. In Figure [5](#Fig5){ref-type="fig"}B, no significant elevation of ROS levels became apparent in Ca9-22 cells treated with GSE concentrations lower than 10 μg/ml. However, after exposing Ca9-22 cells to GSE at 50, 100, 200 and 400 μg/ml for 24 h, the ROS levels increased significantly in a concentration-dependent manner to 2.10%, 5.10%, 27.77%, and 63.17%, respectively (*P* \< 0.005--0.0001).Figure 5**Concentration-dependent ROS generation in GSE-treated oral cancer Ca9-22 cells.** Ca9-22 cells were incubated with indicated concentrations of GSE (0--400 μg/ml) for 24 h. **(A)** Representative histograms of flow cytometric analysis on ROS levels from GSE-treated cells. **(B)** Quantitative analysis of ROS intensity by means of DCF positivity percentage. ROS is counted at the intensity of right gated region in **(A)**. Data, mean ± SD (n = 3). Treatments with the same capital letter are nonsignificant. MitoMP {#Sec19} ------ To examine the involvement of GSE-induced mitochondrial dysfunction in Ca9-22 cells, the flow cytometry-based Rh123 staining was performed. In Figure [6](#Fig6){ref-type="fig"}A, the mitoMP signals were similar in low concentrations (1--10 μg/ml) of GSE but they were gradually decreased in high concentrations of GSE after 100 μg/ml of GSE. In Figure [6](#Fig6){ref-type="fig"}B, there was not a significant elevation of mitoMP levels in Ca9-22 cells incubated with GSE concentrations lower than 10 μg/ml. In contrast, after exposing Ca9-22 cells to GSE at 50, 100, 200 and 400 μg/ml for 24 h, the mitoMP levels were significantly decreased in a concentration-dependent manner to 101.14%, 91.69%, 66.97%, and 15.01%, respectively (*P* \< 0.005--0.0001).Figure 6**Reduction of MitoMP in GSE-treated Ca9-22 cells.** Ca9-22 cells were treated with indicated GSE concentrations (0--400 μg/ml) for 24 h before the addition of fluorescent dye Rh123 (0.5 μg/ml) for 30 min. **(A)** Representative histograms on MitoMP levels for GSE-treated Ca9-22 cells. **(B)** Quantitative analysis on reduction of MitoMP in vehicle controls and GSE-treated cells. MitoMP is counted at the intensity of the left gated region in **(A)**. Data, mean ± SD (n = 3). Treatments with the same capital letter are nonsignificant. DNA damages caused by GSE treatment {#Sec20} ----------------------------------- To detect whether GSE treatments cause DNA double strand break (DSB) in Ca9-22 cells, samples were analyzed using flow cytometry to quantify levels of the phosphorylated γH2AX protein. In Figure [7](#Fig7){ref-type="fig"}A, the γH2AX signals were similar in low concentrations (1--10 μg/ml) of GSE but they were gradually increased in high concentrations of GSE. In Figure [7](#Fig7){ref-type="fig"}B, there was not a significant elevation of γH2AX levels in GSE-treated Ca9-22 cells under low concentrations (lower than 10 μg/ml). In contrast, after exposing Ca9-22 cells to GSE at 50, 100, 200 and 400 μg/ml for 24 h, the γH2AX levels were significantly increased in a concentration-dependent manner to 3.38%, 5.88%, 19.02%, and 35.53%, respectively (*P* \< 0.001).Figure 7**GSE treatments caused γH2AX-based DNA damage in Ca9-22 cells.** Cells were treated with different GSE concentrations for 24 h before subjected to anti-phospho-γH2AX/PI double staining for measuring DNA double strand breaks by flow cytometry.**(A)** A representative γH2AX/PI profile for GSE-treated cells.**(B)** Quantitative analysis of γH2AX-based DNA damage in GSE-treated cells. The mean fluorescence intensity of phospho-γH2AX for gated regions in **(A)** was calculated. Data, mean ± SD (n = 3). Treatments with the same capital letter are non-significant. Discussion {#Sec21} ========== Accumulating evidence of the antiproliferative effect of GSE had been reported in several oral cancer cell lines. For example, high concentrations (50--600 μg/ml) of GSE of *Vitis vinifera* were found to inhibit cell proliferation and induce apoptosis of the KB cells but less harmful to non-cancerous human umbilical vein endothelial cells (HUVEC) by trypan blue assay at 24 h GSE treatment \[[@CR29]\]. Similarly, we found that the low and high concentrations of GSE to normal oral HGF-1 cells based on MTS analysis. The KB cells was used to be regarded as the oral cancer cell line, however, it was recently confirmed to be the contaminant cervical cancer HeLa cells \[[@CR30]\]. Moreover, the low concentrations of GSE were not investigated in this study. Recently, the differential concentration effect of GSE to differentially inhibit proliferation of oral cancer cells has been demonstrated. For example, low concentrations of GSE (10--20 μg/ml) did not displayed the antiproliferation of oral cancer CAL 27 cells but high concentrations of GSE (30--80 μg/ml) were able to inhibit its proliferation \[[@CR31]\]. Similarly, we found that low (1--10 μg/ml) and high (50--400 μg/ml) concentrations of GSE displayed the differential cytotoxic effects to cell viability of oral cancer Ca9-22 cells. Similar results also reported in other cancer cells. In the example of skin cancer HaCaT cells, high concentrations of GSE (IC~50~ = 76.44 μg GAE/ml) displayed the growth inhibitory effect, but low concentrations of GAE (10--20 μg GAE/ml) protected against UVB irradiation (50--100 mJ/cm^2^)-induced skin cancer \[[@CR20]\]. These findings suggested that different cancer cell lines may require different but high concentrations of GSE for antiproliferation purpose. ROS induction by GSE was reported in non-small-cell lung cancer H1299 and A549 cells but it only tested at high concentrations (20--100 μg/ml) without detecting the mitochondrial function \[[@CR15]\]. ROS generation of high GSE (40 μg/ml) also reported to induce apoptosis in head and neck cancer Detroit 562 and FaDu cells \[[@CR32]\]. In oral cancer CAL 27 cells, GSE also reported to induce mRNA overexpression of apoptosis-associated signaling such as caspase-2 and caspase-8 \[[@CR31]\]. In head and neck cancer cells, GSE also reported to induce DNA damage \[[@CR32]\]. Our results further validated that GSE at high concentrations (50--400 μg/ml) have high oxidative stress and apoptosis in terms of ROS generation, mitochondrial depolarization, annexin V/PI staining, and caspase activation but not for low concentrations (\<10 μg/ml) of GSE in oral cancer Ca9-22 cells. Moreover, this differential concentration effect of GSE was also found in cancer cell migration. For example, GSE was reported to inhibit migration and invasion of breast cancer MDA-MB231 cell \[[@CR18]\]. High concentrations (50--100 μg/ml) of GSE inhibited cell proliferation and induced apoptosis. Conversely, low GSE (25 μg/ml) concentrations decreased cell migration and invasion. Therefore, the differential concentration effect of GSE in oral cancer cell migration is warranted for further investigation. Conclusion {#Sec22} ========== We demonstrated that GSE shows differential concentration effects in the antiproliferation of oral cancer cells through differential expressions of apoptosis, oxidative stress, and DNA damage. We showed that the antiproliferative effect of high GSE concentrations is associated with an overproduction of ROS causing DNA damage and apoptosis of cancer cells. **Competing interests** The authors declare that they have no competing interests. **Authors' contributions** C-YY, M-FH, B-HC, and H-WC participated in the writing and the study design. J-YT and H-WC performed statistical analysis. Z-WY, K-TL, C-YH, and Y-HH performed survival assay and flow cytometer analysis. H-WH, S-YL, T-FF, B-HC, and H-WC analyzed and discussed the GSE data. B-HC and H-WC coordinated and oversaw the study. All authors read and approved the final manuscript. This work was partly supported by funds of the Ministry of Science and Technology (MOST 103-2320-B-037-008), the ChiMei-KMU Joint Project (103CM-KMU-09), the Kaohsiung Medical University "Aim for the Top Universities Grant, grant No. KMU-TP103A33", the Kaohsiung Municipal Ta-Tung Hospital (kmtth-102-011), the National Sun Yat-sen University-KMU Joint Research Project (\#NSYSU-KMU 104-p036), and the Health and welfare surcharge of tobacco products, the Ministry of Health and Welfare, Taiwan, Republic of China (MOHW104-TDU-B-212-124-003). We also thank for the help in English editing by Dr. Hans-Uwe Dahms and technical support with the flow cytometer by Mr. Yi-An Chung.

{ "pile_set_name": "PubMed Central" }

Introduction ============ *Staphylococcus aureus*is the most common cause of septic arthritis (SA) \[[@B1],[@B2]\]. SA has shown no change in incidence in spite of advances in antimicrobial therapy and is responsible for residual functional impairment and for a high mortality rate among debilitated patients. Risk factors include older age, diabetes mellitus, rheumatoid arthritis (RA), immunodeficiency, and a pre-existing joint disease. In SA, *S. aureus*contributes to more than two thirds of identified organisms \[[@B3],[@B4]\]. In an epidemiological study of SA in an adult population of 116 patients by Abid and colleagues \[[@B5]\] between 1999 and 2004, *S. aureus*was the most common organism isolated from blood as well as synovial fluid (18.8%). Cleeman and colleagues \[[@B6]\] studied 23 culture-positive cases of SA of the glenohumeral joint between 1986 and 2000, and 52% had a different primary site of infection identified, 70% of which were *S. aureus*-positive and 17% of which were methicillin-resistant. In a retrospective analysis by Moumile and colleagues \[[@B7]\] of the bacterial etiology of acute osteoarticular infections in 406 children with clinically suspected osteoarticular infections, 74 (18%) had a positive bacterial culture: 38 cases of SA and 36 cases of bone infections (osteitis and osteomyelitis), the most commonly recovered pathogen being *S. aureus*(44%). Goergens and colleagues \[[@B8]\] reviewed the clinical presentation, management, and organisms responsible for acute hematogenous osteomyelitis (AHO) and SA in Australia between 1998 to 2002, and *S. aureus*was the most common identifiable causative organism, accounting for 76% of isolated organisms in AHO and 39% of isolated organisms in SA. *S. aureus*remains the most common organism causing AHO and SA, and multidrug-resistant *S. aureus*(MRSA) is on the increase as well. Progressive joint destruction despite appropriate antibiotic therapy and synovial fluid aspiration may indicate a potential role for host-derived proteases. Several matrix metalloproteinases (MMPs) are induced in host cells in response to infectious stimuli. Normally, MMPs assist in clearing infections, initiating immune responses, and in tissue remodeling \[[@B9]\]. Excessive MMPs cause matrix degradation and joint destruction as in various forms of arthritis \[[@B10]\]. Cytokines interleukin (IL)-1, IL-6, tumor necrosis factor (TNF)-α, and interferons (IFN-α and IFN-γ) are released from host cells in response to *S. aureus*infection and these are potent inducers of MMPs \[[@B11]-[@B15]\]. Staphylococcal capsule polysaccharides, toxins, cell wall-attached adhesions, and possibly also the chromosomal DNA are virulence determinants in *S. aureus*arthritis. These bacterial components might affect the innate immune response and inflammation \[[@B1]\]. Alternatively, the bacterial products, secreted or intracellular, could directly affect the transcriptional machinery or signal transduction pathways related to MMP expression. Previous studies have shown the induction of proteolytic enzymes in chondrocytes in response to bacteria-free culture supernatants from *S. aureus*\[[@B16]\]. Also, peptidoglycan (PGN) from *S. aureus*has been shown to be capable of inducing arthritis \[[@B17]\]. A recent study showed that *S. aureus*PGN induces MMP-1, -3, and -13 in human synovial fibroblasts \[[@B18]\]. Purified PGN is chemically modified and may not really represent the native PGN. Also, there is a wide variety of bacterial components, including the superantigens, cell wall components, and extracellular toxins, which could stimulate the host cells. The full potential of synovial fibroblasts in terms of multiple MMP expression in response to *S. aureus*components has not yet been addressed. To determine the global impact of *S. aureus*components on primary human fibroblasts with respect to MMP expression, we exposed de-identified normal human dermal fibroblasts and synovial fibroblasts derived from de-identified patients with RA and osteoarthritis (OA) to whole cell lysate and culture supernatants (whole and fractionated) derived from *S. aureus*wild-type and mutant strains that induce less severe SA in murine models. Materials and methods ===================== Bacterial strains ----------------- *S. aureus*strain isolated from a patient with SA was obtained from American Type Culture Collection (ATCC) (Manassas, VA, USA). A de-identified clinical isolate (U1) and mutants lacking staphylococcal accessory gene regulator (U155: *Sar*^-/-^) and accessory gene regulator (U929: *Agr*^-/-^) and a strain lacking both *Sar A*and *Agr*(*Sar-Agr*^-/-^) derived from that clinical isolate were obtained from M. Smeltzer and were used in this study. The U155 strain was grown in the presence of tetracycline (5 μg/ml); U929 was grown in presence of kanamycin (50 μg/ml) and neomycin (50 μg/ml); and U930 was grown in the presence of tetracycline (5 μg/ml), 50 μg/ml kanamycin, and 50 μg/ml neomycin/ml for selection of the respective mutants. Strains grown in the presence of antibiotics were centrifuged, washed, and resuspended in Dulbecco\'s modified Eagle\'s medium (DMEM)/F-12 medium for inoculation in order to remove the antibiotics. Preparation of whole and fractionated bacterial culture supernatants and bacterial cell lysates ----------------------------------------------------------------------------------------------- To collect supernatants and bacterial cell pellets for experiments, bacterial strains (isogenic parent strain and mutants strains) were grown in DMEM/F-12 containing 2% fetal bovine serum (FBS) without any antibiotics. Supernatants and cell pellets were obtained by centrifugation from 12-hour bacterial cultures. The supernatants were sterilized through 0.22-μm filters to ensure that they were free of any bacteria. The cell pellets were treated with lysostaphin (20 U/ml) for 20 minutes at 37°C followed by repeated freezing and thawing. The lysates were clarified by centrifugation at 12,000 *g*for 20 minutes and were filtered through 0.22-μm filters. The ATCC strain was also grown in the presence of 5 and 15 ng/ml recombinant human rhIL-1β (R&D Systems, Inc., Minneapolis, MN, USA). The cell lysates were prepared as described above. Total protein concentrations were measured by the calorimetric method (Bio-Rad, Hercules, CA, USA) in accordance with the manufacturer\'s instructions. The culture supernatants from the ATCC strain were fractionated into \<30, 30 to 50, and \>50 kDa molecular-weight fractions using respective Centricon filter centrifugation. Fibroblast cultures ------------------- Dermal fibroblasts from de-identified normal volunteers and synovial fibroblasts from de-identified RA patients and OA patients were maintained in DMEM/F-12 containing 10% FBS, 100 U/ml penicillin, and 100 μg of streptomycin. All the fibroblast cell lines were from a cell culture bank established by A. Postlethwaite in accordance with the full approval of the institutional review board of the University Of Tennessee Health Science Center (Memphis, TN, USA). Treatment of fibroblasts with *S. aureus*supernatants, lysates, and rhIL-1β/rhTNF-α ----------------------------------------------------------------------------------- For studies measuring MMP production, 10^5^fibroblasts harvested by trypsinization were added to each well of 24-well tissue-culture plates (Falcon; Falcon Plastics, Inc., Washington, PA, USA). Three days later, confluent monolayers of fibroblasts were treated with phosphate-buffered saline, 25 μg of total proteins from bacterial cell lysates, 25 μg of total proteins from culture supernatants, and 15 μg of protein from each fraction of culture supernatant per well (only from ATCC strain). Fibroblasts were cultured in an incubator containing a humidified atmosphere containing 5% CO~2~at 37°C. Fibroblasts were cultured for 8 hours for RNA analysis and 48 hours for protein analysis. Fibroblasts were also treated with a combination of 10 μg each of rhIL-1β/TNF-α for 8 hours and 48 hours. For mRNA analysis, cells were harvested after 8 hours of respective treatments, and total RNA was isolated using TRI-Reagent (Sigma-Aldrich, St. Louis, MO, USA) followed by isopropanol precipitation. The fibroblast culture supernatants were collected 48 hours after respective treatments and centrifuged to remove any cell debris. All samples were stored at -80°C until analyzed. Fractionation of *S. aureus*culture supernatants ------------------------------------------------ Culture supernatants from *S. aureus*were purified using the Amicon Centricon filter device from Millipore Corporation (Billerica, MA, USA). Using this device, an approximately 2.0-ml volume was concentrated into an approximately 30-μl volume. Using the 10,000, 30,000, and 50,000 kDa cutoff filter devices, we fractionated the whole culture supernatants to \<30, 30 to 50, and \>50 kDa fractions. Assay for MMP and tissue inhibitor of metalloproteinases mRNAs -------------------------------------------------------------- Total RNA was reverse-transcribed into cDNAs by using avian myeloblastoid virus (AMV) RTase and Oligo dT primers. Qualitative profiling of multiple MMP mRNAs was performed using a Multi-MMP-mRNA kit from SuperArray Bioscience Corporation (Frederick, MD, USA) in accordance with the manufacturer\'s protocol. Relative quantification of MMP mRNAs was performed using SYBR green real-time polymerase chain reaction (PCR) on the cDNAs obtained. Assays for MMP-1, -2, -3, -7, -8, -9, -10, -11, -12, -13, and -14 were performed. Gene-specific oligonucleotide primers for tissue inhibitor of metalloproteinases (TIMP)-1, TIMP-2, and TIMP-3 were synthesized at Integrated DNA Technologies (Coralville, IA, USA). The messages of TIMP mRNAs were quantified using SYBR green real-time reverse transcription-PCR (RT-PCR). The message levels were expressed as ratios of the threshold cycle values of respective MMP or TIMP messages to those of the housekeeping gene glyceraldehyde phosphate dehydrogenase (*GAPDH*). Assays for MMP protein ---------------------- Multiple MMP proteins were determined using an MMP protein array kit from RayBiotech, Inc. (Norcross, GA, USA) strictly following the manufacturer\'s instructions. Briefly, the 100 μl of culture supernatants was applied to the membrane with arrayed antibodies. After blocking the free spaces on the membrane, a cocktail of biotin-labeled antibodies was added and the membranes were incubated for 2 hours at ambient temperature. After repeated washing, horseradish peroxidase-conjugated streptavidin was added onto the membrane and incubated for 2 hours at ambient temperature. After extensive washing, the detection buffer was added and the signals were detected by capturing the enhanced chemiluminiscence onto a Kodak x-omat AR film (Eastman Kodak, Rochester, NY, USA). The film was photographed and scanned for documentation. Semiquantitative profiling of mRNAs of MAPK family -------------------------------------------------- A human MAPK gene family multigene-12 RT-PCR profiling kit (Superarray Bioscience Corporation) was used for the qualitative assessment of extracellular signal regulated kinase (ERK) 2/MAPK2, ERK1, MAPK4, ERK3, ERK5, c-jun N-terminal kinase (JNK) 1, JNK2, JNK3, p38b MAPK, p38g MAPK, and p38delta MAPK mRNA in fibroblasts in response to *S. aureus*culture supernatant and cell lysate. Total RNA was reverse-transcribed into cDNAs using AMV RTase and Oligo dT primers, and the messages were amplified using the primer sets supplied by the manufacturer. The expression level of the housekeeping gene *GAPDH*in each sample was used to assess the qualitative differences in respective message levels between samples. The experiments were repeated three times and each time the assays were set up in duplicate. The PCR products were analyzed on a 2% agarose gel and were stained with SYBR green. The intensities of the bands were estimated by densitometric scanning software from Alpha Innotech Corporation (San Leandro, CA, USA). The values expressed are ratios of the densities of the MAPK genes to those of the housekeeping gene *GAPDH*. Cell-based enzyme-linked immunosorbent assay for tyrosine phosphorylation ------------------------------------------------------------------------- A cell-based phosphotyrosine enzyme-linked immunosorbent assay (ELISA) kit from RayBiotech, Inc., was used to quantitate tyrosine phosphorylation in human dermal fibroblasts in response to *S. aureus*components and IL-1/TNF. Approximately 30,000 cells were seeded into each well in a 96-well plate. Cells were incubated at 37°C, 5% CO~2~overnight. The cells were then exposed to *S. aureus*cell lysate (25 μg/ml), *S. aureus*(ATCC) culture supernatant (25 μg/ml), or 10 ng/ml each of rhIL-1β and rhTNF-α for 30 minutes. The medium was removed from the wells, and the cells were treated with the fixing solution followed by quenching solution. The fixed, quenched cells were treated with blocking solution for 3 hours at ambient temperature, and after washing the cells were exposed to anti-phosphotyrosine-horseradish peroxidase for 1 hour followed by washing and the addition of one-step substrate solution. The plates were incubated in the substrate solution for 30 minutes, the color reaction was stopped, and the optical densities were read at 450 nm. The experiments were repeated three times and each time the experiments were run in triplicates. Statistical analysis -------------------- Each treated sample was compared with the untreated sample using Student\'s test. Sigma Stat (Systat Software, Inc., San Jose, CA, USA) program was used for statistical computation, and Sigma Plot (Systat Software, Inc.) was used to create graphs. A *p*value of less than 0.05 was considered significant. Results ======= Induction of multiple MMP proteins by *S. aureus*in human dermal fibroblasts ---------------------------------------------------------------------------- Culture supernatant and cell lysate from *S. aureus*(ATCC) induced the expression of immunoreactive proteins of MMP-1, MMP-2, MMP-10, and MMP-13 in dermal fibroblasts (Figure [1](#F1){ref-type="fig"}). Upregulation of TIMP-1 and TIMP-2 was also noted in *S. aureus*culture supernatant and cell lysate-treated fibroblasts. There were no notable changes in the expression levels of other MMP proteins (MMP-8 and MMP-9) in the cells in response to treatment. The expression pattern and level of expression were similar in *S. aureus*components and IL-1β/TNF-α-treated fibroblasts. Induction of multiple MMP mRNAs by *S. aureus*in human dermal and synovial fibroblasts -------------------------------------------------------------------------------------- Multiple MMP mRNA profile in dermal and synovial fibroblasts in response to *S. aureus*components was determined by SYBR green real-time PCR. Culture supernatants and cell lysate from *S. aureus*significantly enhanced the expression of multiple MMP mRNAs (namely, MMP-1, -2, -3, -7, -10, and -11; *p*\< 0.05). As in the case of MMP protein expression pattern, the response of the fibroblasts in terms of MMP mRNA expression was similar in *S. aureus*component-treated and rhTNF-α- and rhIL-1β-treated fibroblasts (Figure [2](#F2){ref-type="fig"}). Unlike untreated dermal fibroblasts (Figure [2](#F2){ref-type="fig"}), untreated synovial fibroblasts from patients with RA (Figure [3a](#F3){ref-type="fig"}) and OA (Figure [3b](#F3){ref-type="fig"}) had higher basal multiple MMP mRNA expression, indicating an activated status of the synovial fibroblasts from a pathological site. Similar to the dermal fibroblasts, most MMP mRNAs tested were elevated in RA and OA fibroblasts in response to *S. aureus*components. Levels of significantly elevated MMPs (MMP-1, -2, -3, -7, -10, and -11; *p*\< 0.05) are shown in Figure [3a,b](#F3){ref-type="fig"}. The MMP mRNA expression pattern in response to IL-1β/TNF-α and *S. aureus*lysate was similar in RA and OA fibroblasts (Figure [3a,b](#F3){ref-type="fig"}) as well. Interestingly, no significant differences were noted in MMP-13 mRNA levels between the treated and untreated fibroblasts. All other MMPs tested were expressed at very low levels, could not be quantified, and hence were not included in the graph. Two more dermal fibroblast lines, RASF and OASF cell lines, were tested for multiple MMP mRNA expression profile upon exposure to *S. aureus*culture supernatants and bacterial cell lysates. Essentially the same profile as described above was obtained from the additional cell lines (data not shown). Because fibroblasts are heterogeneous in terms of their origin and some of their features, it is likely that fibroblasts from different sources may respond slightly differently in terms of MMP expression. Potentiation of MMP protein expression in human fibroblasts by *S. aureus*grown in presence of rhIL-1β ------------------------------------------------------------------------------------------------------ We have observed significant changes in gene expression in *S. aureus*grown in the presence of rhIL-1β (manuscript submitted). To test whether *S. aureus*grown in the presence of rhIL-1β would have any impact on MMP expression, dermal fibroblasts were exposed to 25 μg/ml per well bacterial cell lysate obtained from *S. aureus*grown in the presence of 5 or 15 ng/ml rhIL-1β. The supernatants were collected and expression of multiple MMP protein was assessed by multi-MMP-Array kit from RayBiotech, Inc., as described previously. The data presented in Figure [4](#F4){ref-type="fig"} show that production of MMP-2, -3, and -8 is greater in fibroblasts treated with cell lysate obtained from the *S. aureus*strain grown in the presence of rhIL-1β. TIMP-4 expression was also slightly enhanced in fibroblasts treated with lysate obtained from the *S. aureus*grown in the presence of 15 ng/ml IL-1β. Induction of MMP mRNA in human dermal fibroblasts by fractionated culture supernatants from *S. aureus* ------------------------------------------------------------------------------------------------------- The MMP-inducing active components in the culture supernatants were mostly in the 30 to 50 and \>50 kDa molecular-weight range as evidenced by significantly elevated expression of MMP-1 and MMP-3 by Centricon fractions 30 to 50 and \>50 kDa in dermal fibroblasts (Figure [5](#F5){ref-type="fig"}; *p*\< 0.05). Although the fractions are not identified beyond their molecular weight, this does rule out some of the already-characterized low-molecular-weight extracellular (secreted) products of *S. aureus*. MMP mRNA induction by *Sar*, *Agr*, and *Sar/Agr*mutants of *S. aureus* ----------------------------------------------------------------------- Blevins and colleagues \[[@B19]\] have shown that *S. aureus*strains lacking the regulatory loci *Sar*or *Agr*result in less severe SA and osteomyelitis in murine models of these diseases. We therefore tested the ability of cell lysates and culture supernatants obtained from these mutants and their isogenic parent strain to induce MMP-1 and MMP-3 mRNAs in human dermal fibroblasts. The mutants and isogenic strains enhanced MMP-1 and MMP-3 production by fibroblasts to a similar degree (Figure [6](#F6){ref-type="fig"}). Induction of TIMP mRNA expression in human fibroblasts by *S. aureus*wild-type and *Sar/Agr*mutants --------------------------------------------------------------------------------------------------- TIMPs are members of the MMP gene family and play an important role in the overall availability of active MMPs. Hence, it is important to determine the TIMP expression profile of fibroblasts in response to *S. aureus*and *S. aureus*components. In our current study, we used culture supernatants obtained from an *S. aureus*strain isolated from synovial fluid of a patient with SA (ATCC), a clinical isolate (U1), and its *Agr/Sar A*double-loci-deleted mutant U930 (strains obtained from M. Smeltzer). The results presented in Figure [7a,b](#F7){ref-type="fig"} indicate a notably increased induction of TIMP-1, -2, and -3 mRNA by the *Agr/Sar A*deletion mutant (U930) of the isogenic parent wild-type strain (U1) and the ATCC strain isolated from the synovium of a patient with arthritis. It may be speculated that the effective MMP available upon infection with *Agr/Sar*deletion mutant is likely to be less compared with the parent isogenic strain. However, further studies to examine expression of other MMPs as well as analysis to estimate enzymatically active MMPs by zymogram will be required to determine whether genes under the control of *Sar*or *Agr*have any effect on the expression of functional MMPs. MAPK gene expression in synovial fibroblasts from patients with RA and OA ------------------------------------------------------------------------- Members of the MAPK gene family (such as ERK1/2 and p38MAPK) are involved in the induction of MMPs through activation protein (AP-1) transcription factors. We therefore analyzed the mRNA expression levels of 12 members of the MAPK family using the MultiGene-12 RT-PCR profiling kit from Superarray Bioscience Corporation. Synovial fibroblasts obtained from patients with RA and OA were exposed to 25 μg of total proteins from bacterial culture supernatant or cell lysate, and total RNA was isolated 6 hours later, reverse-transcribed, and assayed for mRNA of 12 MAPK genes. Several of the MAPK family members were upregulated. The ratio between the intensities of each MAPK gene to that of *GAPDH*is depicted in Figure [7](#F7){ref-type="fig"}. Significant increases in ERK2, ERK1, MAPK4, JNK1, JNK2, p38b, and p38g were observed in dermal fibroblasts treated with *S. aureus*culture supernatant and cell lysate-treated compared with untreated fibroblasts (Figure [8a](#F8){ref-type="fig"}; *p*\< 0.05) and in synovial fibroblast-treated compared with untreated fibroblasts (Figure [8b](#F8){ref-type="fig"}; *p*\< 0.05). Similar increases in these MAPK gene family members were noted in IL-1β/TNF-α-treated fibroblasts (Figure [8a,b](#F8){ref-type="fig"}; *p*\< 0.05). Tyrosine phosphorylation in human dermal fibroblasts exposed to *S. aureus*culture supernatant ---------------------------------------------------------------------------------------------- Using a cell-based ELISA system (RayBiotech, Inc.), tyrosine phosphorylation was assessed in human dermal fibroblasts after 30-minute exposure to 25 μg of total protein from filtered culture supernatant of *S. aureus*and in fibroblasts treated with 10 ng/ml each of rhIL-1β and rhTNF-α. There was a significant increase in phosphotyrosine in *S. aureus*culture supernatant-treated cells, similar to that observed in IL-1β/TNF-α-treated cells (Figure [9](#F9){ref-type="fig"}; *p*\< 0.05). Overall, our data indicate that *S. aureus*components induce multiple MMP expression in human dermal and synovial fibroblasts and that the response is similar to that induced by IL-1β/TNF-α. The expression pattern of MAPK gene expression also indicates the possibility of a signal transduction pathway akin to that induced by the inflammatory cytokine pathway. Our data also indicate that the virulence gene loci (namely, *Sar*and *Agr*) are not determinants of *S. aureus*-induced MMP mRNA expression. Discussion ========== We have shown that the culture supernatants and whole bacterial lysate from *S. aureus*induce multiple MMPs from human dermal and synovial fibroblasts. Several genes of the MAPK pathways were upregulated in treated fibroblasts, and phosphotyrosine proteins were significantly elevated. Using fractionated *S. aureus*culture supernatants, we have shown that the best MMP induction was by components that fall within the molecular-weight range of 30 to 50 kDa. Interestingly, culture supernatants and bacterial cell lysates obtained from *S. aureus*grown in the presence of rhIL-1β induced notably higher levels of MMPs compared with *S. aureus*grown in the absence of rhIL-1β. The overall spectrum of MMP induction by *S. aureus*components was similar to that elicited by a combination of IL-1β and TNF-α. Our *in vitro*MMP mRNA expression analysis showed that mutants lacking *Sar A*and *Agr*loci and their parent isogenic strain induced comparable levels of MMP mRNAs; however, the mutant strains induced notably higher levels of TIMP-1, -2, and -3 mRNAs in human fibroblasts. To our knowledge, this is the first report on multiple MMP/TIMP induction by fractionated *S. aureus*culture supernatants and whole bacterial cell lysates in human dermal and synovial fibroblasts. SA is the most commonly reported bacterial complication of RA. The risk is highest in severe, longstanding, seropositive disease. The clinical presentation of joint infection is frequently atypical, and in 25% of cases, the infection is polyarticular. *S. aureus*is the most common causative organism \[[@B20]\]. Staphylococcal infections can be hard to eradicate from RA joints and often surgery is required \[[@B21]\]. TNF-α plays an important role in the host defense against infection. Inhibition of its activity could therefore be anticipated to augment the risk of infection in patients with RA. The reasons for the predominance of *S. aureus*in SA and the mechanisms of pathogenecity are not yet fully understood. The synovium of patients with RA is rich in IL-1β. We have previously shown that *S. aureus*can bind to IL-1β and use it as a growth factor \[[@B22]\]. A recent report by McLaughlin and Hoogewerf \[[@B23]\] showed that the growth and replication of *S. aureus*in a biofilm are significantly increased by the addition of rhIL-1β. We have also observed that rhIL-1 can modulate the gene expression in *S. aureus*including the bicomponent leukotoxins and some of the surface adhesion molecules collectively called MSCRAMMs (microbial surface components recognizing adhesive matrix molecules) in addition to some of the genes in the pathogenecity island of *S. aureus*(Kanangat and colleagues, manuscript submitted). We speculate that the IL-1-rich synovial milieu might potentially contribute to the increased frequency of *S. aureus*in patients with RA-SA and that the host-derived MMPs induced by *S. aureus*might accelerate the pathogenesis of SA. Our data on the induction of MMPs by *S. aureus*culture supernatants and cell lysates compares well with the previous report by Williams and colleagues \[[@B16]\], who demonstrated MMP-1 and -3 expression by articular cartilage upon exposure to purified culture supernatant from *S. aureus*. We have extended this observation by showing expression of a wide range of MMPs, including MMP-7 (with a proven *in vivo*association with *S. aureus*-induced SA), by human synovial as well as dermal fibroblasts in response to *S. aureus*components. The profile was similar to that induced by a combination of IL-1/TNF, which might indicate the involvement of an inflammatory cytokine-mediated pathway in the observed induction of MMPs by *S. aureus*. *S. aureus*culture supernatants and cell lysates have a wide variety of proteins, and identification of the components that are actually responsible for inducing the MMP induction is essential to determine the mechanisms of induction as well as to rationally design intervening agents against bacterial products. Toward this we end, we have narrowed down the possible candidates to molecular-weight groups of the range of \>30 to 50 based on our experiments using Centricon filtration of the culture supernatants. Because the molecular weight of the chemically purified PGN used in previous studies is not known, we are not in a position to determine whether PGN is included in the stated molecular-weight range. At this time, we have not identified the components beyond the molecular level; nevertheless, this rules out the possibility of some of the recently described low-molecular-weight proteins such as the 19-kDa extracellular fibrinogen-binding protein that inhibits complement activation \[[@B24]\]. Certain complement components have been reported to activate MMPs \[[@B25]\]. The results of the fractionated supernatants also tentatively rule out the possibility of the exotoxin akin to the toxic shock syndrome protein described by Ren and colleagues \[[@B26]\] and the enterotoxin H described by Su and Wong \[[@B27]\]. Although speculative at this juncture, it is possible that the active components in the 30 to 50 kDa could potentially be the novel 38.5-kDa protein named extracellular matrix-binding protein described by Hussain and colleagues \[[@B28]\]. Joint destruction by *S. aureus*is very rapid if not treated appropriately. Although direct erosion of the joint architecture by *S. aureus*proteases/toxins cannot be completely ruled out, continued degradation of extracellular matrix component and the joint architecture even after clearing the infection and debris from the joint cavity indicates the possibility of host-derived proteases in causing joint pathology. Previous studies have shown the release of active MMP-1 and MMP-3 by human articular cartilage upon exposure to sterile purified *S. aureus*culture medium \[[@B16]\]. The enzymatic profile was similar to that induced by IL-1. The authors concluded that the collagenase and stromelysin released by articular cartilage could contribute to extensive destruction of human cartilage in SA. The exoproteases of *S. aureus*have been proposed as virulence factors during *S. aureus*infections. Calander and colleagues \[[@B29]\], using wild-type *S. aureus*strain 8325-4 and its mutants lacking aureolysin, serine protease, and cysteine protease, demonstrated in a murine SA model that inactivation of the exoprotease genes did not affect the frequency or the severity of joint pathology. Intra-articular injection of PGN into murine joints triggered arthritis in a dose-dependent manner \[[@B17]\]. A single injection of this compound caused massive infiltration of macrophages and polymorphonuclear cells with signs of cartilage and/or bone destruction, lasting for at least 14 days, indicating that PGN exerts a central role in joint inflammation triggered by *S. aureus*\[[@B17]\]. The significance of MMP-7 expression in SA was examined by Gjertsson and colleagues \[[@B30]\] using MMP-7-deficient mice and congeneic controls. These mice were inoculated (intravenously) with an arthritogenic dose of *S. aureus*LS-1, and the mice deficient for MMP-7 developed significantly less-severe arthritis both clinically and histologically despite significantly increased numbers of live bacteria within the internal organs. Interestingly, *in vitro*responses to staphylococcal antigens and superantigens were not different between MMP-7^+/+^and MMP-7^-/-^mice in terms of cytokine production. MMP-7 facilitates migration of both macrophages and neutrophils, and the authors therefore conclude that modulation of SA by MMP-7 might be due to changes in peripheral leukocyte distribution. Also, studies by Wang and colleagues \[[@B31]\] have shown that addition of PGN to whole human blood resulted in enhanced levels of MMP-9 within 1 hour and significant enhancement of MMP-9 secretion from the neutrophils was obvious within 30 minutes of incubation with *S. aureus*PGN. Host-cell MMPs are necessary for efficient clearance of infection by accelerating the recruitment of effector cells to kill pathogen and to resolve inflammation and for subsequent tissue remodeling \[[@B9]\]. However, excessive MMPs after infection and inflammation may cause tissue damage leading to immunopathology. MMPs are secreted by both inflammatory and connective tissue cells such as fibroblasts, fibroblast-related cells, chondrocytes, neutrophils, and monocytes/macrophages in response to both infectious assaults and inflammatory cytokines such as TNF-α and IL-1β \[[@B32],[@B33]\]. The regulation of MMP secretion is dependent on the cell type and stimulus. Signal transduction pathways that involve the MAPK and prostaglandin E2/cyclic AMP pathways are considered to be crucial, although the involvement of other pathways cannot be ruled out \[[@B34],[@B35]\]. The transcription factors c-jun (cellular homologue of viral oncogene jun) and c-fos (cellular homologue of FBJ murine sarcoma virus oncogene), members of the AP-1 family, are involved in the transcription of MMP-1 and these AP-1 factors are induced through MAPK signal transduction. We determined the mRNA levels of some of the MAPK family members in synovial fibroblasts from patients with RA or OA treated with *S. aureus*and lysate culture supernatant or IL-1β/TNFα. mRNAs for several of the MAPK family members were upregulated by *S. aureus*lysates and culture supernatants similar to those treated with IL-1β/TNF-α. Also, the overall phosphotyrosine expression was enhanced in fibroblasts treated with *S. aureus*components. The increase in phosphotyrosine in fibroblasts treated with *S. aureus*components or IL-1β/TNFα was comparable. The importance of the proinflammatory cytokines TNF-α and lymphotoxin (LT)-α in an experimental model of *S. aureus*sepsis and arthritis was examined by Hultgren and colleagues \[[@B36]\]. Using TNF-α/LT-α-double-deficient mice, they showed that in mice inoculated intravenously with a toxic shock syndrome toxin-1-producing *S. aureus*strain LS-1, mortality in TNF-α/LT-α-deficient mice was 67%, with no mortality in the controls. Those results correlated with a significantly decreased phagocytosis *in vitro*and inefficient bacterial clearance *in vivo*in mice lacking the capacity to produce TNF-α/LT-α. Infection of mice with a lower dose of staphylococci resulted in an overall low mortality, but the frequency of arthritis was significantly higher in the wild-type group compared with the TNF-α/LT-α-deficient mice (40% versus 13%). Synovitis and erosivity were lower in TNF-α/LT-α-deficient mice compared with wild-type controls. This study demonstrates the detrimental role of TNF-α/LT-α as mediators of the inflammatory response in *S. aureus*arthritis. TNF-α is a potent inducer of several types of MMPs. Although IL-1 or TNF induced by *S. aureus*could very well be contributing to the joint destruction (either through induction of MMPs or through other degradative pathways), studies by Kimura and colleagues \[[@B37]\] showed that blocking TNF and IL-1 does not significantly prevent the late-stage destruction of joint architecture in arthritis induced by *S. aureus*. In the murine heat-killed *S. aureus*-induced arthritis model, TNF-α and IL-1β peaked at 2 and 24 hours after the injection of heat-killed *S. aureus*, respectively. Simultaneous administration of anti-TNF-α monoclonal antibody (mAb) and IL-1 receptor antagonist (IL-1ra) with *S. aureus*resulted in significant inhibition (80%) of 12-hour leukocyte infiltration. However, leukocyte infiltration at 24 hours and beyond and the loss of proteoglycan in *S. aureus-*induced arthritis were not affected by anti-TNF-α mAb, IL-1ra, or their combination. These results suggest that TNF-α and IL-1β involvement in the pathogenesis of *S. aureus*-induced arthritis may be limited to the initial phases of inflammation. The authors suggested that suppressing TNF-α and IL-1 may not be effective in the clinical treatment of Gram-positive bacteria-induced arthritis. With respect to the molecular pathways involved in *S. aureus*-induced MMP expression in fibroblasts, our results suggest that *S. aureus*components could use a pathway(s) similar to that of IL-1β/TNF-α given that the MMP expression pattern, MAPK gene expression, and phosphotyrosine levels were similar in fibroblasts treated with *S. aureus*components or IL1β/TNF-α. It is also important to note that *S. aureus*is capable of inducing synthesis of inflammatory cytokines such as IL-1β and TNF-α from host cells \[[@B13]\]. Whether the MMP induction in fibroblasts by *S. aureus*component(s) is due to the cytokine/chemokine induced by *S. aureus*is not known at present. Previous studies by Wang and colleagues \[[@B31]\] have shown that inhibitors of p38 MAPK (SB202190) and ERK1/2 (PD98059) and inhibitors of Src Tyrosine kinase (PP2) and PI3-K (LY294002) effectively blocked PGN-mediated MMP-9 upregulation in neutrophils. The potential involvement of the Toll-like receptor (TLR)-2 in *S. aureus*PGN-induced joint inflammation and destruction was postulated in a study by Kyburz and colleagues \[[@B18]\]. Cultured synovial fibroblasts obtained from patients with RA or OA were stimulated with PGN. The expression of various integrins was determined by fluorescence-activated cell sorting. TLR-2 and MMP mRNAs as measured by real-time PCR were upregulated in fibroblasts treated with staphylococcal PGN. The levels of IL-6 and IL-8 in the culture supernatants were also increased by treatment with PGN. We demonstrated that cultured synovial fibroblasts express low levels of TLR-2 and TLR-9 mRNA. Anti-TLR-2 mAbs significantly inhibited production of IL-6 and IL-8 induced by stimulation with PGN. The authors concluded that bacterial PGNs activate synovial fibroblasts, partially via TLR-2, to express integrins, MMPs, and proinflammatory cytokines. There was no mention regarding MMP expression after TLR blockade, and it remains unclear whether TLR is involved in MMP expression in a more direct way. Our preliminary results have shown that *S. aureus*culture supernatant and whole cell lysate induce the mRNA expression of several members of the TLR family, including TLR-2 (data not shown). To elucidate the MMP induction by *S. aureus*, we turned to two well-characterized mutant strains of *S. aureus*lacking *Sar A*and *Agr*. *Agr*and *Sar*are the two best-characterized loci responsible for modulating the expression of *S. aureus*virulence factors \[[@B38]\]. *S. aureus*strains lacking either locus have been shown to result in attenuation of *S. aureus*in several models of staphylococcal diseases \[[@B39]-[@B41]\]. Recent investigations by Blevins and colleagues \[[@B19]\] have also shown that mutation of S*ar A*and/or *Agr*caused reduced capacity to induce both SA and osteomyelitis. The exact mechanisms of reduced effectiveness of *Sar/Agr*mutants to cause SA or osteomyelitis are not known. Studies by Nilsson and colleagues \[[@B41]\] showed that mice inoculated with the *Sar A*^+^staphylococcal strain exhibited a more pronounced T- and B-lymphocyte activation and higher levels of serum IL-6 and IFN-γ, compared with a *Sar A*^-/-^mutant, and infection with *Sar A*^+^staphylococci induced pronounced weight loss as well. These studies suggested that S*ar A*locus might control molecules that are important virulence determinants in the induction and progression of SA. We therefore tested the MMP-1, -3, and -13 expression patterns in response to *Sar*, *Agr*, or *Sar/Agr*mutants in human dermal fibroblasts. The three MMPs were selected due to their known involvement in various models of arthritis and their respective degrading actions on collagen type I, II, and III and proteoglycan, which are important constituents of connective tissues and cartilage of the joints. Our results did not show any significant differences in MMP-1 and MMP-3 mRNA levels, and 13 mRNA levels were minimal and could not be quantified with reasonable accuracy in dermal fibroblasts upon exposure to culture supernatants or cell lysates obtained from the mutants and isogenic parent strain (Figure [6](#F6){ref-type="fig"}). However, interestingly, the expression of TIMPs was notably increased in fibroblasts treated with *Sar/Agr*mutants compared with isogenic parent strain (Figure [7a,b](#F7){ref-type="fig"}). This could mean that the effective biologically active MMPs are less abundant in cells treated with the *Sar/Agr*mutants compared with cells treated with isogenic parent strain. It will be necessary to estimate the levels of biologically active MMPs (by zymograms) to determine the net effect of *Sar/Agr*mutants on MMP expression. Temporal estimation of biologically active MMPs in the joints after infection with isogenic parent and mutants will help to clarify the issue of MMPs as a factor in the observed differences in severity of diseases caused by wild-type and mutant strains. Bacteria may secrete proteolytic enzymes such as the thermolysin family secreted by *Pseudomonas aeruginosa*and *Vibrio cholera*which activate pro-MMP-1, -8, and -9 \[[@B42]\]. Also, proteases from the oral pathogen *Porphyromonas gingivalis*activate MMP-1, -3, and -9 \[[@B43]\]. If the bacterial-derived proteases are required for virulence, such proteases can be attractive therapeutic targets because their inhibition can be achieved without affecting the normal expression and function of MMPs. There are reports of other staphylococcal virulence factors associated with the pathogenesis and severity of SA \[[@B43]-[@B45]\]. Whether these virulence factors are associated with MMP/TIMP expression remains to be seen. In addition to the bone and joint infections, *S. aureus*is also the prime causative agent in many skin and soft tissue infections (SSTIs), which can be manifested as superficial to deep-seated and at times become life-threatening \[[@B46]-[@B48]\]. Due to lack of validated clinical evidence, it is often difficult to recommend general treatment options \[[@B49]\]. The pathogenesis of SSTI is not understood well, and the treatment is guided mostly by epidemiological pattern and microbiological information \[[@B50]\]. Due to the emergence of MRSA, it is important to understand the mechanisms of tissue destruction in soft tissue infections which could lead on the identification of novel therapeutic targets. Our current *in vitro*data and the *in vivo*data reported previously by others implicate that host-derived metalloproteinases could be involved, at least in part, in tissue destruction. Excessive expression of these metalloproteinases induced by *S. aureus*could lead to the destruction of the soft tissue connective tissue architecture. Conclusion ========== We have shown that *S. aureus*is a potent inducer of multiple MMPs in human dermal and synovial fibroblasts. Our studies also indicate that MAPK-mediated signal transduction pathway involving proteins that are phosphorylated at tyrosine residues might play a role in *S. aureus*-induced MMP expression. Enhanced expression of immunoreactive MMPs by cell lysate obtained from *S. aureus*grown in the presence of rhIL-1 indicates that an inflamed milieu such as RA synovium might augment the MMP induction potential of *S. aureus*. More specific identification of the component(s) of *S. aureus*involved in the upregulation of MMP and associated signal transduction pathways may help in identifying novel targets for intervention. Based on our results, we propose that biologically active MMPs (reflected as a ratio of MMPs to TIMPs) induced by *S. aureus*could potentially accelerate the joint destruction in SA. Abbreviations ============= Agr = accessory gene regulator; AHO = acute hematogenous osteomyelitis; AMV = avian myeloblastoid virus; AP-1 = activation protein; ATCC = American Type Culture Collection; DMEM = Dulbecco\'s modified Eagle\'s medium; ELISA = enzyme-linked immunosorbent assay; ERK = extracellular signal regulated kinase; FBS = fetal bovine serum; GAPDH = glyceraldehyde phosphate dehydrogenase; IFN = interferon; IL = interleukin; IL-1ra = interleukin-1 receptor antagonist; JNK = c-jun N-terminal kinase; LT = lymphotoxin; mAb = monoclonal antibody; MAPK = mitogen-activated protein kinase; MMP = matrix metalloproteinase; MRSA = multidrug-resistant *Staphylococcus aureus*; OA = osteoarthritis; PCR = polymerase chain reaction; PGN = peptidoglycan; RA = rheumatoid arthritis; RT-PCR = reverse transcription-polymerase chain reaction; SA = septic arthritis; Sar = staphylococcal accessory regulator; SSTI = skin and soft tissue infection; TIMP = tissue inhibitor of metalloproteinases; TLR = Toll-like receptor; TNF = tumor necrosis factor. Competing interests =================== The authors wish to state that they have no commercial or other association (such as pharmaceutical stock ownership, consultancy, advisory board membership, patents, or research funding) that might contribute to competing interests. Authors\' contributions ======================= SK conceived the idea, designed the study, and prepared the manuscript draft. AP, AK, KH, and DS all helped in the experimental design and data analysis and provided creative and critical suggestions after reviewing the manuscript. MS provided insight into the murine SA model, provided the mutant and isogenic strains of *S. aureus*, and critically reviewed the manuscript. WA assisted with the RT-PCR work, cell culture maintenance, MMP protein arrays, as well as in review of literature. All authors read and approved the final manuscript. Acknowledgements ================ This work was supported by a Public Health Service Grant (R03 AR47936-01) from the National Institute of Arthritis and Musculoskeletal and Skin Diseases (National Institutes of Heath); a University of Tennessee Health Science Center (UT-HSC) Connective Tissue Diseases\' Center of Excellence grant, a special fund from the Department of Medicine, UT-HSC, Memphis, TN, USA; and a pilot study grant from the Department of Veterans Affairs, Memphis, TN, USA. Figures and Tables ================== ![Induction of multiple matrix metalloproteinase (MMP) protein by *Staphylococcus aureus*in human dermal fibroblasts. Confluent monolayers of human primary dermal fibroblasts were exposed to **(I)**phosphate-buffered saline, **(II)**25 μg/well of total protein from clarified whole cell lysates of *S. aureus*cells, **(III)**25 μg of total protein from the filtered culture supernatant of *S. aureus*, and **(IV)**treated with 10 ng/m each of interleukin-1β/tumor necrosis factor-α (IL-1β/TNF-α) for 48 hours. The fibroblast culture supernatants were harvested and the MMPs and tissue inhibitors of metalloproteinases (TIMPs) were detected using the MMP-Array kit from RayBiotech, Inc. (Norcross, GA, USA). Culture supernatant and cell lysate from *S. aureus*induced the expression of MMP-1, MMP-2, MMP-10, and MMP-13. A similar pattern was observed in IL-1β/TNF-α-treated fibroblasts, which were slightly more intense. There was upregulation of both TIMP-1 and TIMP-2 as well.](ar2086-1){#F1} ![Induction of multiple matrix metalloproteinase (MMP) mRNAs in human dermal fibroblasts (HDF) by *Staphylococcus aureus*. Confluent monolayers of fibroblasts were exposed to 25 μg/well of total protein from culture supernatants (SUP), 25 μg/well of total protein from bacterial cell lysate, or 10 ng/ml each of rhTNF-α and rhIL-1β for 8 hours. Total RNA was isolated and reverse-transcribed. The cDNA was amplified and quantified using SYBR green real-time polymerase chain reaction. All the MMPs tested were expressed in dermal fibroblasts exposed to *S. aureus*components. Both culture supernatants and cell lysate from *S. aureus*significantly enhanced the expression of multiple MMPs (namely, MMP-1, -2, -3, -7, -10, and -11; *p*\< 0.05). The dermal fibroblasts responded similarly to a combination of rhTNF-α and rhIL-1β in terms of multiple MMP mRNA expression. IL, interleukin; TNF, tumor necrosis factor.](ar2086-2){#F2} ![Induction of multiple matrix metalloproteinase (MMP) mRNAs in human synovial fibroblasts (SFs) by *Staphylococcus aureus*. MMP mRNA expression profile in confluent monolayers of SFs obtained from de-identified **(a)**rheumatoid arthritis (RA) and **(b)**osteoarthritis (OA) patients treated with culture supernatants (SUP) (25 μg of total protein/well) and cell lysates (25 μg of total protein/well) were determined using SYBR green real-time reverse transcription-polymerase chain reaction. Unlike untreated dermal fibroblasts, untreated SFs from RA and OA patients had higher basal multiple MMP mRNA expression. Similar to the dermal fibroblasts, most MMP mRNAs tested were elevated in RA and OA fibroblasts in response to *S. aureus*cell lysate and culture supernatants. Levels of significantly elevated MMPs (MMP-1, -2, -3, -7, -10, and -11; *p*\< 0.05) are shown in **(a)**and **(b)**. The MMP mRNA expression pattern in response to IL-1β/TNF-α and *S. aureus*lysate was similar in RA and OA fibroblasts. IL, interleukin; TNF, tumor necrosis factor.](ar2086-3){#F3} ![Potentiation of matrix metalloproteinase (MMP) protein expression in human fibroblasts by *Staphylococcus aureus*grown in the presence of recombinant human interleukin (rhIL)-1β. The cell pellets from *S. aureus*grown in the presence or absence of rhIL-1β were washed repeatedly to prepare cell lysate. Enzyme-linked immunosorbent assay (R&D Systems, Inc., Minneapolis, MN, USA) was carried out to make sure that the bacterial cell lysate was devoid of any traces of rhIL-1β. Confluent monolayers of human dermal fibroblasts were exposed to 25 μg/well of IL-1β-free cell lysate obtained from *S. aureus*for 48 hours. The supernatants were collected and expression of multiple MMP protein was assessed using multi-MMP-Array kit from RayBiotech, Inc. (Norcross, GA, USA) as described in the text. The data presented show that MMP-2, -3, and 8 are relatively more expressed in fibroblasts treated with *S. aureus*cell lysate obtained from strain grown in the presence of 5 and 15 ng/ml rhIL-1β. Tissue inhibitor of metalloproteinases (TIMP)-4 expression was also slightly enhanced in fibroblasts treated with *S. aureus*lysate obtained from strain treated with 15 ng/ml IL-1β.](ar2086-4){#F4} ![Induction of matrix metalloproteinase (MMP)-1 and MMP-3 mRNA in response to *Staphylococcus aureus*components separated by molecular-weight difference. Fractionated *S. aureus*culture supernatant (15 μg/ml per well) obtained by centrifuging through Centricon filters with molecular-weight cutoff points \<30 kDa, 30 to 50 kDa, and \>50 kDa was added to confluent monolayers of dermal fibroblasts. Total RNA was isolated after 8 hours. The mRNA levels of MMP-1 and MMP-3 were measured by SYBR green real-time reverse transcription-polymerase chain reaction. The active components were mostly in the 30 to 50 and \>50 kDa molecular-weight range as evidenced by significantly elevated expression of MMP-1 and MMP-3 by Centricon fractions 30 to 50 and \>50 kDa (*p*\< 0.05; 30 to 50 kDa and \>50 kDa fractions were compared with \<30 kDa fractions).](ar2086-5){#F5} ![Matrix metalloproteinase (MMP) mRNA induction by staphylococcal accessory regulator (*Sar*), accessory gene regulator (*Agr*), and *Sar/Agr*mutants of *Staphylococcus aureus*. MMP-1 and MMP-3 mRNAs in confluent monolayers of human dermal fibroblasts exposed to 25 μg of total protein/well per milliliter from cell lysate and 25 μg/well per milliliter of total protein from culture supernatants (SUP) of parent (U1) mutant strains (U155 *Sar*^-/-^; U929 *Agr*^-/-^; U930 *Sar/Agr*^-/-^) were measured by SYBR green real-time reverse transcription-polymerase chain reaction 8 hours after exposure. No significant differences in mRNA levels of MMP-1 and MMP-3 induced by the mutants and their isogenic parent strain were observed.](ar2086-6){#F6} ![Enhanced expression of tissue inhibitor of metalloproteinases (TIMP) mRNAs in human dermal fibroblasts treated with culture supernatants **(a)**and whole bacterial cell lysates **(b)**obtained from *Sar A*^-/-^and *Sar A*^-/-^Agr^-/-^mutants of *Staphylococcus aureus*and isogenic parent strain. Culture supernatants and cell lysates obtained from an *S. aureus*strain isolated from synovial fluid of a septic arthritis patient (obtained from American Type Culture Collection, Manassas, VA, USA), a clinical isolate (U1), its *Sar A*mutant (U929), and *Agr/Sar A*double-loci-deleted mutant (U930) (strains obtained from M. Smeltzer) were used. Equal amounts of total protein from all samples were added to confluent monolayers of fibroblast cultures. Total RNA was harvested from the cells 12 hours after exposure to the supernatants. The levels of TIMP-1, -2, and -3 mRNAs were estimated by real-time SYBR green reverse transcription-polymerase chain reaction. The values plotted represent the ratios of the threshold cycle values of TIMPs to that of the housekeeping gene *GAPDH*(glyceraldehyde phosphate dehydrogenase). Agr, accessory gene regulator; Sar, staphylococcal accessory regulator; SUP, culture supernatant.](ar2086-7){#F7} ![Mitogen-activated protein kinase (MAPK) family mRNA expression profile in human dermal fibroblasts **(a)**and human synovial fibroblasts **(b)**exposed to *Staphylococcus aureus*whole cell lysate and filtered culture supernatant. Confluent monolayers of de-identified human dermal fibroblasts from normal volunteer and synovial fibroblasts from a patient with rheumatoid arthritis were treated with 25 μg/ml per well of *S. aureus*whole lysate, filtered culture supernatant, and a combination of rhIL-1β and TNF-α (10 ng/ml each) for 8 hours. Cells were harvested and total cellular RNA was isolated and reverse-transcribed. The MAPK family gene expression was analyzed using the Human MAPK Gene Family I Multigene-12 reverse transcription-polymerase chain reaction (PCR) profiling kit. The semiquantitative values were generated by determining the ratios of the band intensities of respective PCR products to that of housekeeping gene *GAPDH*(glyceraldehyde phosphate dehydrogenase) in each sample. The experiments were repeated three times. The band densities were determined using three-dimensional densitometric scanning software from Alpha Innotech Corporation (San Leandro, CA, USA). Of the 11 tested Human MAPK Family I genes (ERK1, ERK2, MAPK2/4, ERK3, ERK5, JNK1, JNK2, JNK3, p38b MAPK, p38g MAPK, and p38delta), ERK1, ERK2, JNK1, JNK2, MAPK4, and p38b were elevated significantly in both dermal and synovial fibroblasts upon exposure to *S. aureus*components (*p*\< 0.05 for individual MAPK gene family member when values from untreated fibroblasts and treated fibroblasts were compared). The response was similar in both cell lines in response to IL-1/TNF. ERK, extracellular signal regulated kinase; IL, interleukin; JNK, c-jun N-terminal kinase; SUP, filtered culture supernatant; TNF, tumor necrosis factor.](ar2086-8){#F8} ![Tyrosine phosphorylation in human dermal fibroblasts exposed to *Staphylococcus aureus*culture supernatant. Using a cell-based enzyme-linked immunosorbent assay system (RayBiotech, Inc., Norcross, GA, USA), tyrosine phosphorylation was assessed in human dermal fibroblasts exposed to 25 μg of total protein/ml from filtered culture supernatant of *S. aureus*, 25 μg/ml total protein from *S. aureus*cell pellet lysate, and in fibroblasts treated with 10 ng/ml each of rhIL-1β and rhTNF-α. These assays were carried out in 96-well plates supplied by the manufacturer. The cells were exposed to various agents for 30 minutes before assaying for tyrosine phosphorylation. The assays were carried out in triplicate. There was significant increase in tyrosine phosphorylation in *S. aureus*culture supernatant-treated cells similar to IL-1β/TNF-α-treated cells compared with untreated fibroblasts (*p*\< 0.05). IL, interleukin; OD, optical density; SUP, culture supernatant; TNF, tumor necrosis factor.](ar2086-9){#F9}

{ "pile_set_name": "PubMed Central" }

We thank Fox et al. ([@B2]) for their interest in our research and welcome this opportunity to respond to their commentary. They argue that participants in our studies enjoyed "just thinking" more than we claimed (Wilson et al., [@B5]). We found some irony in their position, because we began this line of research with a similar hypothesis. As the data came in we were surprised that participants did not enjoy deliberative thought very much, even when we went to some lengths to give them time to prepare and choose their topics (i.e., in our "prompted fantasy" conditions). We are thus in the rather amusing position of explaining why both Fox et al.\'s interpretation and our initial hypotheses are wrong. Our view can be summarized with a simple thought experiment. Imagine you are a venture capitalist, and a filmmaker comes to you asking for money to make a full-length feature film. You decide it would be wise to show the film maker\'s 15-min trailer to some focus groups. You find that (a) on average, the focus groups rate the trailer as much less enjoyable than a variety of everyday activities, such as listening to music and surfing the web; (b) between a third and a half of the people in the focus groups disobey the instructions to watch the trailer and instead prefer to do something else, such as playing with their cell phones; and (c) the majority of the men (and a quarter of the women) in the focus groups choose to give themselves an unpleasant electric shock while watching the trailer, even though they had said earlier that they would pay to avoid getting the shock. Would you invest? Probably not. These are, of course, precisely the results we found in our studies when we left people alone with their thoughts. As such, we do not think it is unreasonable to conclude that the participants found deliberative thought to be neither engaging nor enjoyable. Although our space is limited, we will respond to a few of Fox et al.\'s specific arguments. First, the results they report in Figures 1A--C are entirely consistent with the data we reported in our Table 1. Instead of quibbling over what it means for participants to rate something as somewhat enjoyable and somewhat boring, we believe it is more informative to compare the ratings of participants under different experimental conditions and to observe their actual behavior. In Study 8, participants rated everyday external activities as much more enjoyable than deliberate thinking (the distribution of responses in the two conditions was almost non-overlapping). Fox et al. suggest that we stacked the deck in favor of the external activities by letting people choose among activities that "were not banal or boring." But note that participants in the "thinking" condition were completely free to think about anything they wished, to choose topics that were "tailored to their personal interests," and to switch between topics. They could choose to think about past accomplishments, current interests, or future plans, or even, to build on Fox et al.\'s analogy, chocolate and sex. The fact that participants much preferred the everyday external activities is quite telling, we think, and does not speak well for the appeal of "just thinking." We also find it revealing that 32--54% of participants (depending on the study) disobeyed instructions to "just think" and performed external activities they were specifically asked to avoid, such as using their cell phones or doing school work. (Note that these figures include only those who admitted to doing the banned activities.) If deliberative thinking is as enjoyable as Fox et al. suggest, why did so many participants disobey explicit instructions to do it for relatively short periods of time? Fox et al. are unimpressed by the results of our study in which 67% of men and 25% of women administered at least one shock to themselves. Their argument that participants were simply curious is at odds with the fact that all participants had already experienced the shock and that we specifically selected only those participants who reported that they would pay to avoid another shock. Fox et al. argue that 14 of the 18 participants who shocked themselves reported that they did so out of "curiosity about (or interest in) the quality of the shock or its effects as their motivation." But note that what many of these participants were interested in was how unpleasant the shock would be, such as the participant who wrote, "I wanted to see if the shock felt as painful the second time as it did the first time." If thinking is so enjoyable, why would participants be willing to re-experience an unpleasant stimulus just to see if it is still unpleasant? (Not to mention the fact that participants\' reports about the reasons for their behavior are, in large part, *post-hoc* rationalizations that may not reflect their actual reasons; Nisbett and Wilson, [@B4]; Johansson et al., [@B3]). Lastly, we note an error in one of Fox et al.\'s statements. Participants did not rate their "enjoyment of 'just thinking,"' but rather their enjoyment of the thinking *period*, including any shocks they administered. Thus, if people shocked themselves because they were bored, and the shocks helped alleviate that boredom, it is not surprising that the shockers rated the overall experience similarly to the non-shockers. More fundamentally, Fox et al. seem to have misunderstood what our studies were about. They described our research as having investigated "spontaneous thought," which one of the authors defined as "the unintended, non-working, non-instrumental mental content that comes to mind unbidden and effortlessly" (Christoff, [@B1], p. 52). But in fact we examined directed conscious thought, because in virtually all of our studies we specifically instructed participants to think in enjoyable ways. As Christoff ([@B1]) noted, "It is impossible, after all, to instruct participants to have a spontaneous thought" (p. 57). No other studies, to our knowledge, have examined how successfully people can engage in deliberative thinking when the goal is enjoyment or positive affect. Fox et al.\'s comparison of our research to previous studies of spontaneous thought is thus moot. Clearly some kinds of thinking are enjoyable. As we said at the end of our article, "There is no doubt that people are sometimes absorbed by interesting ideas, exciting fantasies, and pleasant daydreams." But we also noted that "it may be particularly hard to steer our thoughts in pleasant directions and keep them there" (p. 77). The preponderance of the evidence supports this conclusion, we believe, and is contrary to both Fox et al.\'s and our own initial hypotheses. Conflict of interest statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The writing of this article was aided by National Science Foundation Grant BCS-1423747. [^1]: This article was submitted to Cognition, a section of the journal Frontiers in Psychology. [^2]: Edited by: Daniel Lakens, Eindhoven University of Technology, Netherlands [^3]: Reviewed by: Bernhard Hommel, Leiden University, Netherlands

{ "pile_set_name": "PubMed Central" }

Background {#Sec1} ========== End-stage renal disease (ESRD) is increasing worldwide; at the end of 2009 there were nearly 38,000 people living with end-stage renal disease in Canada \[[@CR1]\]. Most patients with ESRD do not receive a kidney transplant and require dialysis to survive. Quality of life for patients receiving dialysis has been rated similarly to patients with metastatic cancer \[[@CR2]\], in part because of the intrusiveness of ongoing dialysis (whether hemodialysis or peritoneal dialysis), and in part because of symptoms related to the disease, including depression which occurs in up to 30% of patients \[[@CR3], [@CR4]\]. Qualitative research provides insight into the views of respondents, including their attitudes, needs, beliefs and feelings. This provides some insight into patients' perspectives on their illness. Understanding patients' views can help health care providers tailor the type of treatments offered to patients, as well as understand when, how and what information to provide to patients about the treatment options. This is particularly relevant in patients with chronic diseases, such as kidney failure, since they deal with their illness and the consequences of its treatment on a daily basis. We build on prior work to determine the research priorities of patients on or nearing dialysis, their caregivers and the health care professionals who look after them. To develop the list of top research priorities, we initially surveyed patients, their caregivers and health care professionals across Canada, collating, and ranking their responses. A top ten list was subsequently developed at an in-person meeting \[[@CR5]\]. During this priority setting exercise, we noted that many of the items expressed by respondents represented not only unanswered research questions, but also expressed the needs, beliefs or feelings that were specifically related to their illness or its treatment. In this report, we use survey responses from patients and caregivers to conduct a thematic analysis of their views. This consideration of the views of patients and caregivers with kidney disease is meant to inform health care service providers about ways to support treatment decision-making, enhance communication, address psychosocial wellbeing and improve patient satisfaction. Methods {#Sec2} ======= We undertook a qualitative, descriptive study, based on a secondary analysis of survey data, using thematic content analysis to synthesize the views of patients on or nearing dialysis and their caregivers. Participants {#Sec3} ------------ Patients, caregivers and health care professionals who care for patients on or nearing dialysis were invited to complete a survey distributed online through partner organizations (Kidney Foundation of Canada), social media (Twitter, Facebook), or a paper-based version available in 10 Canadian hemodialysis centers. The survey was open for 3 months (October 2012 -- January 2013). For this analysis, we only analyzed the responses of patients and their caregivers. Survey {#Sec4} ------ The survey consisted of seven open-ended questions on topics relevant to patients with ESRD (overall management of kidney failure, treatment options, dialysis access options, prognosis, diet, symptoms and lifestyle). These open-ended questions were intended to elicit questions that were answerable by research, for example: "Are there questions about decisions regarding the way in which kidney failure can be treated that you would like to see answered by research?" In previous work, these research questions were prioritized to develop the top ten most important unanswered research questions from the perspective of patients and their caregivers (<http://www.cann-net.ca/patient-information/dialysis-research-priorities-survey#results>). While the goal of this previous work was to consider respondent statements in the context of developing a list of unanswered research questions, in the current work, we were interested in how the statements reflected respondents, needs, beliefs and feelings. In the survey responses, some patient statements only reflected unanswered research questions, and did not inform views. To eliminate these statements, which were less informative in the context of this work, two reviewers classified all responses independently, and separated these statements out. The remainder of the statements formed the dataset for the content analysis. Qualitative content analysis {#Sec5} ---------------------------- Statements were analyzed by thematic content analysis, a method that enables systematic analysis of the content in communication, such as answers to a survey, and reflects the content of the data set. Thematic content analysis is a method for identifying, analyzing, and reporting themes within data, where concepts or categories are derived from the data in an inductive or deductive manner \[[@CR6]\]. These concepts or categories serve to represent the data by providing knowledge and new insights, with the outcome a condensed and broad description of the phenomenon under study \[[@CR7]\]. Content analysis goes beyond the summary of the data and involves a qualitative interpretation by the researcher, providing a "vicarious experience" for the audience \[[@CR8]\]. A descriptive theme was defined as a subject that captured important data related to end-stage renal disease\[[@CR9]\]. Themes were not defined a priori but emerged after the initial readings of the data. Themes were identified as having captured an important element of the views of patients with chronic kidney disease and their caregivers and were not necessarily the most prevalent subjects within the data. Themes were not created to be mutually exclusive, but in order to represent the best conception of the data. Synthesis of findings {#Sec6} --------------------- We followed the process outlined by Braun et al. \[[@CR6]\] to synthesize the results. The data were first read by one author (LB) several times, to become familiar with the content and generate an initial idea of themes. The thematic content analysis involved three phases: the development of descriptive themes along with the creation of mutually exclusive rules for inclusion; statement by statement coding done independently by two authors (LB & LM); and the organization of the statements into meaningful categories by the same two authors (LB & LM). Any disagreements between the two authors were resolved by consensus, with the input of a third reviewer (BM) if necessary. We defined 4 descriptive themes, which were re-assessed to ensure that they worked with and represented the data. Lastly, a selection of powerful and compelling examples were selected, with context, to represent each theme. Results {#Sec7} ======= The survey elicited a total of 1820 statements from all respondents (patients, health care providers and caregivers) (Table [1](#Tab1){ref-type="table"}). Of these, 245 were removed because they were out of scope (not relevant to patients on or nearing dialysis) or were unclear. Of the remaining 1575, 1054 were from patients or caregivers. Statements that only expressed a research question, for example "can stem cell research improve my kidneys" or "what research is being done with regard to a mechanical replacement kidney, much like the artificial heart", were removed. There were a total of 544 remaining statements, from 189 individual respondents, included for the thematic content analysis.Table 1**Patient and caregiver characteristics**Type of respondentn (%)Total n= 189Patient receiving hemodialysis in a clinic91 (48.1)Patient on home hemodialysis32 (16.9)Patient on peritoneal dialysis22 (11.6)Patient on dialysis, no detail2 (1.1)Patient, within a year of starting dialysis6 (3.2)Care provider36 (19.0)**Age**18 -- 295 (2.6)30 -- 3915 (7.9)40 -- 4926 (13.8)50 -- 5935 (18.5)60 -- 6949 (25.9)70 -- 7923 (12.2)80 and over10 (5.3)Prefer not to answer26 (13.8)**Gender**Male73 (38.6)Female89 (47.1)Prefer not to answer27 (14.3)**Ethnicity**Aboriginal3 (1.6)Asian10 (5.3)Black9 (4.8)Mixed3 (1.6)Other7 (3.7)White123 (65.1)Prefer not to say34 (18.0)**Province**Atlantic26 (13.8)British Columbia7 (3.7)Ontario59 (31.2)Prairies65 (34.4)Quebec4 (2.1)Territories1 (0.5)Prefer not to say27 (14.3) Four major descriptive themes were identified as being central to the views of patients with chronic kidney disease and their caregivers: gaining knowledge, maintaining quality of life, sustaining psychosocial wellbeing, and ensuring appropriate care. A thematic schema of the analytical framework is represented in Figure [1](#Fig1){ref-type="fig"}.Figure 1**Components of each theme as an overview of the thematic analysis.** Gaining knowledge {#Sec8} ================= The theme gaining knowledge encompassed seeking knowledge/lack of information, communication, timing of information, and included how patients and their caregivers learned about topics related to their chronic kidney disease. Seeking knowledge/lack of information {#Sec9} ------------------------------------- Respondents sought knowledge not only about the different treatment options available (either dialysis modality or access), but also wanted to determine which was the "best", and why, for both their overall health and quality of life. They specifically expressed the desire for a thorough explanation of all options. A caregiver stated: "*Yes, I would like to see all of these options thoroughly explained at the very beginning. This is a whole new world for patients and care givers, making it very difficult to make good decisions*." Respondents verbalized that they felt uninformed about many aspects of their treatment and care. A hemodialysis patient expressed: "*Again left in dark till things happened, policy seems to be on a need to know. When you start dialysis why don\'t they warn you about the symptoms of crashing? Not nice not knowing what is happening".* There was a recurring concern regarding symptom management, and respondents frequently voiced that they were unsure of why certain complications occurred (lack of energy, itchiness). They also felt that the lack of information interfered with their ability to cope with and manage these symptoms, as stated by a hemodialysis patient: "*Why do we get restless legs, itchy skin and have difficulty sleeping? No one ever explained why nor what can be done*". This lack of information extended to other areas of their disease, including fluid management and diet restrictions. A peritoneal dialysis patient wrote: *"More information on why some foods are okay and others aren\'t. I had to find detailed lists of phosphate containing foods on the internet*". Respondents also stated that they were unsure about their prognosis on dialysis, wanting to know how long they would need to stay on dialysis, the progression of their disease, as well as their life expectancy. A patient receiving hemodialysis in a clinic expressed: "*Generally a summation of what will happen physically as the years pass on dialysis would be beneficial to me as a patient. What should I expect in terms of medical problems and deterioration due to the end stage renal disease*". Lack of information extended to those seeking a transplant as well. A patient receiving hemodialysis in a clinic wanted to know: "*At what age, and with what present medical issues does a person with kidney failure become ineligible for a transplant? Persons with kidney failure want to know all about the process of selection regarding a transplant*". One dialysis respondent expressed that they wanted a holistic take on information, stating *"We need to be better informed about the side effects of dialysis on your mind as well as your body, such as focus, cognitive ability and lack thereof. Also, how your whole life will change: how you have to live and interact with others and how totally frustrating it can be".* Communication {#Sec10} ------------- In addition to not receiving the information they felt they needed, participants raised concerns about the communication between patient and healthcare provider, specifically in how the information was delivered. "*An explanation in layman\'s terms would be useful*". A hemodialysis patient felt information was not always communicated appropriately. "*I started dialysis two weeks ago. There are a number of acronyms used, which I do not understand. How can we improve communication with patients?*" This same respondent expressed that he found it "*very frightening*" when first discussing kidney disease. Respondents also expressed a desire for improved communication for managing their disease, and being informed. A home hemodialysis patient wondered: "*Why can't the nurses share what's going on with our blood work and inform us as patients*". In the particular case of diet restrictions, a long term dialysis patient stated: "*Just because the patient was told this once at the beginning does not mean they remember. There should be some kind of annual review of the do's and don'ts as well as product updates*". A respondent receiving hemodialysis in a clinic stated it as follows: "*They need to go one step further than just give us a piece of paper saying what you can and cannot eat or drink and explain the consequences of not following a diet and watching what you drink*". Many respondents verbalized their desire for a transplant, but felt that the selection process and position on the waiting list could be communicated in a clearer manner. Some felt this process was not transparent, perhaps due to a lack of communication on how the process works. "*Why do I have to wait so long? Why haven\'t they made an effort to get me on the transplant list? Will I ever get my transplant?*". Overall, some respondents voiced that they would like to feel part of the team and more involved in the decision making around their care. This was expressed by a patient receiving hemodialysis in a clinic through the following statement: "*I've been through being diagnosed and I still see it happening: the nephrologist tells you what you have, the treatment and the diet, but never acts like you're a part of the team. Why is it that up until the moment you are told the above, you think you have a say in your treatment, but the specialist doesn't see it that way*". Timing of information {#Sec11} --------------------- Respondents also questioned whether information could have been given earlier in the disease process. They felt that had they known the severity of their disease earlier on, they could have taken more action to prevent or delay the progression. Respondents also wanted to know early on about their treatment options for dialysis, given the impact of the decision. One patient wrote, "*I feel that the patient should be told earlier about their choices. Also important knowing about how dialysis can affect your life and that of your family, friends, employer, etc. It is very traumatic and life changing*". For those who had already made a decision about dialysis options, a respondent receiving hemodialysis in a clinic recounted feeling pressure to make a decision and needed more time. "*Can you make the whole orientation process much clearer? Also, a strong focus was given to home hemodialysis, which I wasn\'t comfortable doing. Pushing us, especially those on the senior side, to use home dialysis with that machine is terrifying, even with training. Can you build layers of consulting to ease those of us into our new lifestyle? It may be repetitive but we\'d make better decisions in the long run*". Another patient receiving hemodialysis in a clinic recounted his experience as follows: "*When I first started dialysis, it would have been very helpful to know about the side effects of dialysis. I believe it would have helped prepare myself and my spouse for the onset of all this stuff. I received very little information when I first started*". Source of information {#Sec12} --------------------- Peers appeared to be an important influence for both patients and caregivers on deciding their treatment options. Respondents valued the experience of their peers in the process of choosing their modality and were interested to know how they came to their decision. One caregiver stated: "*Indeed it was a very difficult decision to make regarding the topic of which dialysis treatment would be best suited for us. We could have used more info on real people who have done both and what their experiences were and what the pros and cons are. Also the success rate of each and is there any difference or research on who does better on the different treatments".* Respondents also looked to their peers in terms of their care, benchmarking their treatment against others, and wanting to know why there were differences. "*How come some people have shorter times \[on dialysis\] than others, and more frequently during the week?*". Maintaining quality of life {#Sec13} =========================== The theme maintaining quality of life encompassed the patient's general lifestyle, freedom to travel and sexual vitality. These were all identified as being important in maintaining a quality of life similar to pre-dialysis. General lifestyle {#Sec14} ----------------- Respondents most often expressed a strong desire to maintain their lifestyle, prior to the commencement of dialysis, what they commonly referred to as a "*normal life*". One respondent receiving hemodialysis in a clinic expressed: "*What are the best options for a normal life*?". Respondents also wanted to know how they could plan a life and maintain work while undergoing dialysis, as one respondent phrased it: "*How to plan a life and work style that is compatible with staying fit while undergoing dialysis"*. One dialysis patient expressed a desire for normalcy and control in her ability to prepare the appropriate foods: *"I see cookbooks for people of all kinds of conditions - heart problems, diabetes, gluten intolerance etc. I have never come across a cookbook for kidney patients. Can\'t some dietitian make one for us?".* Respondents who experienced side effects or complications from dialysis or their illness often described the impact this had on their quality of life. "*I\'ve lost my eyesight and leg because of diabetes-how can I have a normal life? My wife is fed up, tired of my problems, how can I make this better? I need her to look after me*". Freedom to travel {#Sec15} ----------------- A large proportion of respondents verbalized their desire for freedom to travel and stated that traveling was extremely important in maintaining their quality of life. Respondents identified several barriers related to travel. Limited access to other dialysis centers was one barrier identified by a respondent on hemodialysis: "*Can more access to hemodialysis machines be set up so that people who like to travel can get appointments in different parts of the country? It is very difficult to get time to dialyze where we boat and like to camp*". Another respondent identified not knowing how to plan travel while on dialysis and resources available for travel (both personal resources required and those made available by the health system) as barriers. "*Can persons on dialysis be provided with a comprehensive and detailed directory regarding travel anywhere throughout Canada, including any special perks or considerations available by public or private services in terms of costs, access, assistance, etc.".* Respondents were also concerned with whether it was safe to travel while dialyzing and any possible consequences traveling had on their health. Sexual vitality {#Sec16} --------------- Respondents expressed the desire for more forthright communication around sexual vitality, emphasizing that sexuality was part of maintaining a normal life for them. A young woman with kidney disease expressed it as follows: "*Sexuality is never spoken about. Young patients often have many questions about fertility, sex and how to deal with having a line or catheter and a sexual life*". Respondents, and family members, were especially concerned about how they could maintain their sexuality, either by addressing erectile dysfunction, lack of sexual desire or other challenges to sexuality related to dialysis itself (e.g. femoral dialysis lines). "*People are complaining about groin lines and trying to understand how they can still have sex. Sex is very important to any couple. Open communication regarding this topic would be much appreciated. All we need is to be on dialysis and divorced!".* Sustaining psychosocial wellbeing {#Sec17} ================================= The theme sustaining psychosocial wellbeing included psychosocial support for both patients on or nearing dialysis and their caregivers, in addition to their overall wellbeing. Support for themselves & caregivers {#Sec18} ----------------------------------- Respondents were concerned about their access to support for themselves and their caregivers, whether at the start of treatment or after years on dialysis. "*How about having a counseling session of encouragement before a patient goes on dialysis because I was terribly afraid and I wanted to run away*". One caregiver expressed a need for psychosocial care. "*What do you do about the mental care of a dialysis patient that suffers from depression due to or caused by lengthy treatment of dialysis?*". Respondents were also concerned about their caregivers and the toll the disease took on them, illustrated by the following two quotes from patients on hemodialysis. "*What types of things do family members have concerns about but are not voicing? What effect does CKD have on them*?". "*My wife left me because I was no longer that strong man she married, she wasn\'t getting the emotional love she needed and she thought I was going to die. Can more support/documents be made available to the spouses?".* Some described the need for psychosocial support that extended beyond the dialysis unit. "*Someone outside of the dialysis unit that is familiar with personal issues of chronic kidney disease other than my family doctor, especially dealing with sleeping and sexuality*". Overall wellbeing {#Sec19} ----------------- Respondents emphasized that the lack of quality of life they experienced while on dialysis contributed to feelings of depression or inadequacy. "*I cannot make myself breakfast/get dressed/put on socks. My quality of life is very bad but I feel all these symptoms are a part of my life. All I can do is smile and keep doing what I can, but it is very hard*". Many comments reflected the overwhelming and deep-seated impact of dialysis and kidney disease on everyday life. "*Without work you are living under the poverty line so are not only dealing with loss of health but now can barely afford to live. Family wants to distance themselves so they won\'t be reminded that death is imminent and won\'t feel the pain. Canada has made travel impossible. Most dialysis places don\'t want dialysis patients from another place, for various reasons; might infect their patients, too much trouble to accommodate. We cannot leave the country unless we are willing to pay for our own dialysis care. We are already destitute financially so that is impossible unless you have money. These kinds of restrictions, along with diet, and overall health add to feelings of depression".* Ensuring appropriate care {#Sec20} ========================= The theme ensuring appropriate care included quality of care, care delivery and patient resources. Quality of care encompassed the satisfaction and standard of care patients received, while care delivery centered around the health system and its resources that impacted care of the patient. Patient resources focused on any perceived or real limitations to accessing health care due to constraints of personal financial resources of patients. Quality of care {#Sec21} --------------- Few respondents communicated concerns about their quality of care, with many expressing satisfaction with their nephrologists and other members of their health care team. "*No questions but very satisfied with all treatment and information by all dialysis staff I have had contact with*". Respondents did question whether their disease could have been diagnosed earlier, and suggested that education of their primary care physician may have led them to being treated earlier. "*How do we educate family doctors to test for and refer patients at an early stage, so it is not such a shock when they are told they have kidney failure*". Care delivery {#Sec22} ------------- Many of the views on care delivery expressed by respondents centered around perceived or real constraints on resources that affected their access to care, due to resources at the level of the health system. For example, some respondents perceived that home hemodialysis was a better treatment option, and that this should be offered more. "*Why are more people not trained and guided towards self care, peritoneal dialysis or home hemodialysis*". Another example was the lack of availability of all treatment options at all health centres, as stated by one patient receiving hemodialysis in a clinic. "*I would like to try the button hole method but because of time and money and the clinic's part, I have not been able to try it. I'm trying to find a way to preserve the life of my vein*". Patient resources {#Sec23} ----------------- As respondents are often limited in their ability to work and earn income, they were concerned about personal costs incurred because of their illness and the need for dialysis, as well as whether one dialysis therapy would incur more expenses to them: "*What personal costs could I expect if I chose one \[dialysis therapy\] over the other?*" Several dialysis patients expressed concerns about the impact that medication costs have on their finances, as highlighted by the following two comments. "*We need certain medications to deal with kidney failure and dialysis, so why aren\'t they free?*" "*Why are some medications covered and others not, for example sensipar? Last year I spent over \$20,000 on medications. How does a 18-year dialysis patient survive his co-pay while maintaining their dignity and live with these costs?"* Another hemodialysis patient expressed frustration about the lack of financial resources available to them. "*Since dialysis patients can\'t work full-time because it takes time out of your life, why aren\'t we compensated financially? We need dialysis to survive*!" Discussion {#Sec24} ========== We identified four themes that emerged from views of patients' on or nearing dialysis and their caregivers: gaining knowledge, maintaining quality of life, sustaining psychosocial wellbeing, and ensuring appropriate care. Respondents identified a lack of information and communication between patient and provider as a significant concern, particularly with respect to information about dialysis modality and access. Respondents also expressed significant concerns about their quality of life on dialysis, and expressed a desire to have as much of a "normal" life as possible. They identified a variety of issues that could be addressed by both the health system and health care providers to improve their quality of life. With respect to the health system, there is an urgent need for more psychosocial support, and resources to facilitate travel, particularly for patients on hemodialysis. For health care providers, better communication is key, along with more information on why symptoms happen, how they can best be managed, and information about their future prognosis. The majority of respondent's expressed a need for gaining knowledge, which was thought of either as seeking knowledge or lack of information. This need for information was expressed by both dialysis patients and caregivers, echoing previous studies that found that the primary need of families of chronic dialysis patients was information \[[@CR10], [@CR11]\]. The timing of when information was presented to patients and caregivers was also a common theme, as was the importance of re-enforcing and updating material previously communicated. The information needs of people as they pass through the various stages of their disease changes over time. A careful consideration of the type of information, including how much information, and at what stage of their illness it is provided, could enable a more efficient transfer of knowledge from health provider to patient. Further, education may alleviate concerns and stress, and stress has been found to affect dialysis modality selection \[[@CR12]\]. Providing predialysis education can not only enable patients to choose the modality best suited to them, but helped in their understanding of their disease\[[@CR13]\], and may extend their time prior to dialysis initiation \[[@CR14]\]. Indeed, previous research has noted that patients need information spread over an extended period of time, with increasing amount of detail and specificity as renal replacement therapy nears \[[@CR15]\]. Clinical practice guidelines in the UK specify that information should also be tailored to the stage of disease \[[@CR16]\]. Our results also highlighted that transferring knowledge is dependent on the quality of communication among people, echoing the results of a study that found that perceived knowledge of kidney disease was related to the quality of communication \[[@CR17]\]. Many respondent comments reflected the need for improved communication between patients/caregivers and health care providers, though increased communication is not merely repetition of the same information. Improving health literacy of patients with chronic kidney disease may be one step in not only ameliorating communication, but may also addressing the information needs of these patients, and how knowledge is retained \[[@CR18]\]. Health literacy encompasses communication between patients, their social networks, and providers. Limited literacy has been linked with reduced knowledge, less adherence, hospitalization and death \[[@CR19], [@CR20]\]. Finally, many respondents identified the importance of peer influence as a source of information, which may be particularly important since previous studies identified peer support as a means of providing practical information about kidney disease \[[@CR21]\]. Peer support may also give purpose to patients already on dialysis by valuing their experiences. As peers have "been there, done that", they have first hand comprehension of some of the difficulties of treatment, and may be a more influential source of information than clinicians for some important treatment decisions. We identified the need for more resources for patients and their caregivers, including improved access to travel, more information on how to cook meals within the limitations of the renal diet, more financial support, and more psychosocial care for themselves and their caregivers. Patients on hemodialysis consistently requested more information on how to arrange hemodialysis in other cities, and were concerned with the lack of hemodialysis spots available to enable travel. With respect to meal planning, although there is high-quality information available online to assist dialysis patients and caregivers (<http://www.kidneycommunitykitchen.ca/kkcookbook>), this information may not be sufficiently broad to address all cultures, tastes or financial means. Further, the number of comments we received requesting additional meal planning resources suggests patients are not aware of existing resources and are still facing barriers with respect to meal preparation. The difficulty of preparing meals to meet the dietary restrictions of those on dialysis is a need that has been met by pre-packaged frozen meals in the United States \[[@CR22]\]. Finally, with respect to the need for psychosocial care, the availability of trained professionals for psychosocial care in and out of the dialysis unit, along with support for caregivers, may help relieve the burden of living with this disease. Moreover, since depression is very common \[[@CR23]\], resources to help patients prevent and manage depression appears to be a critical need for patients. In addition to professional support, social support has also been found to improve outcomes, including compliance, in those with end-stage renal disease \[[@CR24], [@CR25]\] and resources should be made available to help foster important social relationships. In light of this information, kidney care and dialysis programs, in collaboration with patients and caregivers, should take inventory of what information they provide, the resources available, and how this information is communicated to patients. A careful examination of where a program's current resources are being directed and whether this aligns with the needs of patients and their caregivers is necessary. Further, detailed information on how to access resources should be made clear and reiterated often. Although the challenges that patients face may vary slightly across the country, there are likely more similarities than differences in many of the views of patients. As such, the development of certain educational resources, including materials and access to online resources, could be a collaborative process across programs, reducing the need for duplication of effort. While these educational materials will not replace face-to-face visits, they may provide helpful information for patients needing further information in between visits. Programs may also explore the idea of shared decision making as an innovative way of communicating and involving patients in critical decisions about their health \[[@CR26]\]. Our paper has limitations. We did a qualitative analysis of survey responses, which did not enable us to ask follow-up questions to further explore a theme. Analyzing written responses, without further clarification from respondents, could result in misclassification of certain statements. Though we sought to have a representative national response, our study was limited by having a low response rate from British Columbia and Quebec, and from some patient types (First Nations and the elderly). While we think that many of the views expressed in this report will be common across kidney care programs, the extent of the issue may vary based on geographic regions, health care systems and the availability of local resources. The information provided by this study could be supplemented by local questionnaires to best assess local needs, and inform local program changes. Conclusions {#Sec25} =========== We identified four themes from this analysis of the views of patients' on or nearing dialysis and their caregivers: gaining knowledge, maintaining quality of life, sustaining psychosocial wellbeing, and ensuring appropriate care. Importantly, respondents identified a variety of issues that could be addressed by both health systems and health care providers to improve their quality of life. These include the development of patient materials and resources, or sharing of existing resources across Canadian renal programs, along with adopting better communication strategies. Other concerns, such as the need for increased psychosocial and financial support, require consideration by health care funders. **Competing interests** The author(s) declare that they have no competing interests. **Authors' contributions** KK, BH, AL & BM have made substantial contributions to conception and design, or acquisition of data; LB, LM & BM have made substantial contributions to analysis and interpretation of data; all authors have been involved in drafting the manuscript or revising it critically for important intellectual content; all authors have given final approval of the version to be published; and all authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. Drs. Hemmelgarn and Manns are supported by Alberta Innovates - Health Solutions (AI-HS) (formerly Alberta Heritage Foundation for Medical Research (AHFMR) salary awards). Drs. Manns, and Hemmelgarn are supported by an alternative funding plan from the Government of Alberta and the Universities of Alberta and Calgary. This research was supported by two grants from CIHR (operating grant number 284262 team grant number 251048), an interdisciplinary team grant from Alberta Innovates -- Health Solutions, the Interdisciplinary Chronic Disease Collaboration, and from the Kidney Foundation of Canada. The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The authors are indebted to Erin Lillie for coordinating the primary study, which provided us with the responses for this study.

{ "pile_set_name": "PubMed Central" }

INTRODUCTION {#s1} ============ Nasopharyngeal carcinoma (NPC), derived from the epithelial lining of the nasopharrynx, is a commonly occurring cancer with the highest incidence of metastasis among head and neck cancers in southern China and Southeast Asia \[[@R1]--[@R4]\]. Because of its high radio-sensitivity, radiotherapy has been the main strategy for treatment of NPC. However, distant relapse remains the major cause of treatment failure in NPC \[[@R5]\], and the molecular mechanisms underlying NPC metastasis are poorly understood. In order to provide a basis for the development of novel therapeutics for NPC, it is crucial to obtain a better understanding of the molecular mechanisms used by cancer cells to facilitate their survival during metastasis. While there are many abnormal features of metastatic cancer cells, resistance to anoikis is particularly interesting as it enables cell survival with loss of attachment to the extracellular matrix (ECM) \[[@R6], [@R7]\]. Metastatic cancer cells develop anoikis resistance by triggering several signaling pathways \[[@R6], [@R8], [@R9]\], including the production of reactive oxygen species (ROS) and activation of mitochondrial metabolism \[[@R10], [@R11]\]. Under physiological conditions, ROS are constantly generated as by-products of aerobic metabolism in the mitochondria \[[@R12], [@R13]\]. Consistent with the pro-anoikis abilities of ROS, treatment with antioxidants suppresses anoikis in breast cancer \[[@R10]\]. Moreover, energy deficiency (diminished ATP levels) caused by ECM detachment can reduce the viability of breast cancer cells \[[@R11]\]. ROS is a collective term for the chemical species that are formed as a result of incomplete reduction of oxygen. ROS include superoxide anion radical (O~2~^·−^), peroxyradical (ROO−), hydrogen peroxide (H~2~O~2~), singlet oxygen (^1^O~2~), perhydroxyl radical (HO~2~.), and the extremely reactive hydroxyl radical (.OH). The mitochondrial enzyme manganese superoxide dismutase (MnSOD) efficiently converts O~2~^·−^ to H~2~O~2~ and thereby critically changes mitochondrial destructive effects. Because superoxide arises primarily from the mitochondria, MnSOD plays a pivotal role in its detoxification \[[@R14], [@R15]\]. However, the role of MnSOD as a critical feature of highly metastatic NPC has not been elucidated. In this study, we assess the role of MnSOD in mediating the survival of ECM-detached NPC cells. We provide evidence, both *in vitro* and *in vivo*, that up-regulation of MnSOD in a highly metastatic NPC cell line increases anoikis resistance after ECM detachment by decreasing mitochondrial O~2~ ^·−^ and accelerating hydrogen peroxide diminished. In addition, we show that the β-catenin pathway is robustly activated following a loss of ECM attachment. Activation of this pathway acts as a pro-survival signal, inducing a MnSOD-dependent cytoprotective response that is characterized by reduced mitochondrial O~2~ ^·−^ production and resistance to anoikis. Taken together, our findings suggest that β-catenin/MnSOD play central roles in mediating an antioxidant effect that enables cancer cells to survive and migrate to distant sites during metastasis. RESULTS {#s2} ======= Highly metastatic NPC cells have an advantage in anchorage-independent cell growth {#s2_1} ---------------------------------------------------------------------------------- Malignant cells can often survive and grow without adhesion to the ECM making this growth a hallmark of the malignant phenotype \[[@R16], [@R17]\]. There is a well-established model to evaluate the relationship between metastatic potential and anoikis resistance \[[@R17], [@R18]\]. In this model, a highly metastatic cellular clone, S18, is compared with the poorly metastatic clone, S26. We tested clonogenesis of the S18 and S26 cells using a soft agar assay, and assessed cell viability and apoptosis of cells by CCK8 and AnnexinV/PI staining, respectively. Even though there was no obvious difference in cell viability between S18 and S26 cells under attached culture conditions (Figure [1A](#F1){ref-type="fig"}), S18 cells survived better in anchorage-independent cell growth conditions (Figure [1B--1C](#F1){ref-type="fig"}). When the NPC cells were cultured in a three-dimensional model to form clones in Matrigel, the clones generated from S18 cells formed more fully-filled structures when compared with S26 cells (Figure [1D](#F1){ref-type="fig"}). Moreover, anoikis was induced in S26 cells after 12, 24, and 48 hours of suspension conditions (Figure [1E](#F1){ref-type="fig"}). These observations demonstrate that cell viability is elevated with the aggressiveness of human NPC cells in response to matrix detachment. ![S18 cells exhibit enhanced anoikis resistance\ The viability of S18 and S26 cells under adherence (**A**) and suspended (**B**) conditions. (**C**) Colony numbers of S18 and S26 cells in soft agar. (**D**) Acini were generated using the indicated cell lines cultured in Matrigel for 12 days, and cells were stained with DAPI. Representative images are shown (× 200). \**p* \< 0.05. (**E**) Flow cytometric analyses of S18 and S26 cell apoptosis under suspended conditions for 12 hours, 24 hours, and 48 hours. Values were plotted as mean ± S.D. \**p* \< 0.05 compared with control group.](oncotarget-07-32408-g001){#F1} ROS induced by matrix detachment plays a critical role in NPC anoikis {#s2_2} --------------------------------------------------------------------- Inadequate matrix attachment generates ROS and causes anoikis \[[@R10], [@R11]\]; therefore, we investigated the level of ROS in ECM-detached NPC cells. Since ROS are located in different subcellular compartments \[[@R19]\], we examined several different sub-cellular locations. There were no differences in DCF-DA fluorescence (for H~2~O~2~), DHE fluorescence (for Cytoplasmic O~2~^·−^) or Mito-SOX fluorescence (for mitochondrial O~2~^·−^) between S18 and S26 cells cultured on adherent plates. However, S18 cells had decreased H~2~O~2~ and mitochondrial O~2~^·−^ levels under detached conditions when compared with S26 cells (Figure [2A--2C](#F2){ref-type="fig"}). Since MnSOD and catalase are key antioxidant enzymes involved in scavenging ROS, we examined the protein levels of MnSOD and catalase in NPC cells grown in suspension. The MnSOD protein levels was up-regulated in S18 cells as compared to S26 cells. However, the protein levels of catalase was decreased in S26 cells when detached (Figure [2D](#F2){ref-type="fig"}). These results help understand why S18 cells have less O~2~^·−^ and H~2~O~2~. In addition, soft agar assays (Figure [2E](#F2){ref-type="fig"}) and Annexin/PI staining (Figure [2F](#F2){ref-type="fig"}) showed that treatment with 1 mM of an antioxidant compound N-acetyl-L-cysteine (NAC) stimulated the anchorage-independent growth in S26 cells, while treatment with 30 μM H~2~O~2~ decreased the anchorage-independent growth of S18 cells. These observations indicate that the level of ROS is inversely correlated with malign ant cell viability during the detachment process. ![Enhanced ROS is critical for anoikis in NPC cells\ (**A**--**C**) S18 and S26 cells were incubated for 30 min in the presence of DCH-DA (10 μM), DHE (10 μM) or Mito-SOX (5 μM). Representative flow cytometry plots show the separate analysis of H~2~O~2~ (A), cytoplasmic O~2~^·−^ (B) and mitochondrial O~2~^·−^ (C) content in S18 and S26 cells under attachment and detachment cell growth condition after 24 hours. (**D**) The S18 and S26 cells were cultured in suspension conditions for 24 h, total cell lysates were subjected to western blot analyses performed to detect MnSOD and Catalase levels.(**E**) Quantification of soft agar colony numbers of S18 and S26 cells treated with H~2~O~2~ (30 μM) or NAC (1 mM). (**F**) Flow cytometric analyses of S18 cells and S26 cells by AnnexinV/PI under suspended conditions in the present of H~2~O~2~ (30 μM) or NAC (1 mM) as described. Values were plotted as mean ± S.D. \**p* \< 0.05 compared with control cells.](oncotarget-07-32408-g002){#F2} MnSOD expression is elevated in detached highly metastatic NPC cells {#s2_3} -------------------------------------------------------------------- We next investigated the molecular changes necessary for anoikis resistance and balancing the ROS concentrations. We examined Cu/ZnSOD activity as well as MnSOD activity in detached NPC cells. There was a difference in total SOD activity and MnSOD activity, but not Cu/ZnSOD activity, between S18 and S26 cells under detached conditions (Figure [3A--3C](#F3){ref-type="fig"}). When cells were cultured on adherent plates, there were no differences in Cu/Zn-SOD or MnSOD mRNA levels between S18 and S26 cells (Figure [3D](#F3){ref-type="fig"}). However, when cells were grown in suspension, both the mRNA and protein levels of MnSOD were upregulated in S18 cells as compared to S26 cells (Figure [3D--3E](#F3){ref-type="fig"}). These results suggest that MnSOD is necessary for the antioxidant defense in highly metastatic S18 cells and is critical for resistance to anoikis. ![Matrix detachment induces MnSOD in high-metastasis S18 cells\ (**A**--**C**) S18 and S26 cells were cultured under detachment condition for 12 hours. Then, total SOD (A), Cu/Zn-SOD (B) and MnSOD (C) enzyme activity was measured using an SOD Assay Kit-WST. (**D**) Transcript levels of Cu/Zn-SOD and MnSOD in S18 and S26 cells under attachment and detachment. (**E**) MnSOD levels in S18 and S26 cells by western blot analyses. Values were plotted as mean ± S.D. \**p* \< 0.05 compared with control.](oncotarget-07-32408-g003){#F3} MnSOD reduces ROS levels and anoikis-mediated cell death {#s2_4} -------------------------------------------------------- To further confirm the role of MnSOD in providing resistance to anoikis, we utilized MnSOD siRNA and a MnSOD overexpression plasmid. siRNA knockdown of MnSOD in S18 cells (Figure [4A](#F4){ref-type="fig"}) resulted in higher levels of Mito-SOX fluorescence (for mitochondria O~2~^·−^) (Figure [4B](#F4){ref-type="fig"}) and apoptosis (Figure [4E](#F4){ref-type="fig"}), corresponding to over 20% reduction of survival after 24 hours in suspended conditions and rendering these cells sensitive to NAC (Figure [4E](#F4){ref-type="fig"}). Therefore, MnSOD-mediated elimination of mitochondrial O~2~^·−^ in response to matrix detachment protects against anoikis. Conversely, overexpression of MnSOD in S26 cells (Figure [4F](#F4){ref-type="fig"}) reduced Mito-SOX fluorescence (Figure [4G](#F4){ref-type="fig"}) and decreased the percentage of apoptotic cells (Figure [4J](#F4){ref-type="fig"}), further supporting an essential role of MnSOD in anoikis resistance. However, SOD converts superoxide radical into hydrogen peroxide thus may establish an elevation of H~2~O~2.~ Interestingly, subsequent analysis of the data demonstrated that MnSOD can promote protein expression level of catalase (Figure [4C, 4H](#F4){ref-type="fig"}) and H~2~O~2~ diminished (Figure [4D, 4I](#F4){ref-type="fig"}), which suggest that more MnSOD would be unlikely to yield more H~2~O~2~ in NPC cells. ![MnSOD confers resistance to anoikis\ S18 cells (**A**--**E**) were transfected with sicon or siMnSOD, S26 cells (**F**--**J**) were transfected with pcDNA3.1-vector or pcDNA3.1-MnSOD for 24 hours, after which they were cultured in poly-HEMA-coated plates under anchorage-independent conditions for 24 hours following which flow cytometry analysis for mitochondrial O~2~^·−^ (B, G), western blot analyses for catalase levels(C, H) were performed. H~2~O~2~ concentration (D, I) and apoptosis (E, J) were detected with Amplex^®^ Red Hydrogen Peroxide Kit and AnnexinV/PI, respectively. The data are presented as mean ± S.D. of triplicate determinations. For differences between two groups, the Students *t*-test was employed. \**p* \< 0.05.](oncotarget-07-32408-g004){#F4} β-catenin signaling upregulated MnSOD and critical for anoikis resistance {#s2_5} ------------------------------------------------------------------------- Recently, several studies have suggested involvement of the Wnt/β-catenin pathway in ROS scavenging, and Wnt signaling is also known to function in tumor metastasis \[[@R20]\]. β-catenin transcriptionally regulates several antioxidant genes in response to oxidative stress \[[@R21]\]. Hence, we studied the putative connection between MnSOD and the canonical Wnt/β-catenin pathway. Interestingly, loss of β-catenin expression disrupted cell-cell adhesion in S26 cells and increased the nuclear accumulation of β-catenin in S18 cells after 6 hours of suspension (Figure [5A--5B](#F5){ref-type="fig"}). Knockdown of β-catenin in S18 cells reduced MnSOD protein levels in suspension (Figure [5C](#F5){ref-type="fig"}). Flow cytometric analysis confirmed that mitochondrial O~2~ ^·−^ levels were elevated in siβ-catenin cells compared with control cells following 12 hours of suspension (Figure [5D](#F5){ref-type="fig"}) and accompanied by increased anoikis (Figure [5E](#F5){ref-type="fig"}). Moreover, S26 cells overexpressing β-catenin had increased MnSOD protein levels (Figure [5F](#F5){ref-type="fig"}) and suppressed anoikis when in suspension (Figure [5H](#F5){ref-type="fig"}). On the contrary, Mito-SOX fluorescence intensity was reduced in S26 cells overexpressing β-catenin (Figure [5G](#F5){ref-type="fig"}). Together, these findings provide evidence for the involvement of β-catenin working upstream of MnSOD to promote anoikis resistance in metastatic NPC cells. ![β-catenin signaling promotes anoikis resistance via MnSOD up-regulation\ (**A**) S18 cells were under adherent or suspended condition for 6 hours and 12 hours as indicated, total cell lysates, MnSOD, and β-catenin were subjected to western blotting. (**B**) Immunofluorescence staining of β-catenin. β-catenin in the cellular nucleus of S26 had a low intensity compared with that in the S18 after grown in suspension for 24 hours while there had no obvious difference of that when attachment cultured. (**C**) S18 cells were transfected with siRNA against β-catenin (siβ-catenin) or sicon. Immunoblots for indicated proteins were measured. (**D**) Flow cytometric analysis of mitochondrial O~2~ ^·−^ in S18 cells after cultured in suspension condition for 24 hours. (**E**) Flow cytometric analysis of cells by AnnexinV/PI under attachment or suspended conditions for 6 h,12 h, 24 h and 48 h as indicated. (**F**--**H**) S26 cells were transfected with empty vector or β-catenin for 24 hours and grown in attached or suspension condition for another 24 hours. Immunoblots of β-catenin (F), Flow cytometric analysis of mitochondrial O~2~^·−^ (G) and apoptosis rate (H) were shown. Values are plotted as mean ± S.D. \**p* \< 0.05 compared with control.](oncotarget-07-32408-g005){#F5} Knockdown of MnSOD inhibits metastasis by promoting anoikis {#s2_6} ----------------------------------------------------------- Additionally, we utilized the shRNA (shMnSOD) to develop a stably transfected S18-shMnSOD cell line (Figure [6A](#F6){ref-type="fig"}). Under detached conditions, there was a difference in H~2~O~2~ and mitochondrial O~2~^·−^ (but not cytoplasmic O~2~^·−^) between S18-shcon and S18-shMnSOD cells (Figure [6B](#F6){ref-type="fig"}). S18-shMnSOD cells also displayed increased sensitivity to anoikis and decreased cell viability following 24 hours of suspension, as compared to S18-shcon cells (Figure [6C--6E](#F6){ref-type="fig"}). Activation of caspase3, and 9 were involved in mitochondrial ROS-induced apoptosis (Figure [6F](#F6){ref-type="fig"}). To test the consequences of MnSOD loss *in vivo*, we used a nude mouse tumor model administering S18-shcon or S18-shMnSOD cells via intravenous injection to force them to grow in suspension. After six weeks, mice injected with S18-shMnSOD cells had amassed less tumor burden in the lungs compared with the S18-shcon group (Figure [6G](#F6){ref-type="fig"}). In addition, there were discernible differences in tumor histology when comparing lung samples from S18-shcon to S18-shMnSOD injected animals (Figure [6H](#F6){ref-type="fig"}). These data suggest that MnSOD is critical for the NPC colonization of lung tissue *in vivo.* ![Diminished MnSOD expression promotes anoikis *in vitro* and reduces tumor formation *in vivo*\ (**A**) Using lentiviral transduction of shRNA, MnSOD expression was reduced in S18 cells. Western blot analysis for MnSOD and β-actin confirms the success of the established stable cell lines. (**B**) Cells were incubated for 30 min in the present of DCH-DA (10 μM), DHE (10 μM) or Mito-SOX (5 μM). Representative flow cytometry plots show the separate analysis of H~2~O~2~, cytoplasmic O~2~^·−^ and mitochondrial O~2~^·−^ content in S18-shcon and S18-shMnSOD cells under detachment cell growth condition for 24 hours. (**C**) Cell viability by CCK8 assays under suspension conditions. (**D**) Colony numbers of S18-shcon and S18-shMnSOD cells grown in soft agar. (**E**) The percent of apoptotic cells shown in response to cell culture under suspension as indicated. (**F**) Western blot analysis for markers of apoptosis (caspase 3 and caspase 9) in S18-shcon and S18-shMnSOD cells grown in suspension cultures for 24 hours was performed. (**G**) S18-shcon and S18-shMnSOD cells (2 × 10^6^ cells per injection) were injected into nude mice via tail vein (*n* = 8). Six weeks after injection, mice were killed. Representative images of shcon and shMnSOD tumors on the lungs fixed in picric acid and the average numbers of tumor nodules were shown. (**H**) Immunohistochemistry staining with hematoxylin and eosin was performed on shcon or shMnSOD lung tumor tissues, lack of tumor nodules in shMnSOD lung tissues. Scale bars represent 500 μm or 200 μm. The data are presented as mean ± S.D. For differences between two groups, the Students *t*-test was employed. \**p* \< 0.05.](oncotarget-07-32408-g006){#F6} MnSOD expression is elevated with progressing NPC tumor stage {#s2_7} ------------------------------------------------------------- We next used immunohistochemistry to analyze MnSOD expression in tumors from NPC cancer patients (tissue microarrays, Shanghai Outdo Biotech Co., Ltd.). In NPC samples stratified by stage, MnSOD expression increased with histologic tumor grade, being highest at histologic grade IV and lowest in grade I (Figure [7A--7B](#F7){ref-type="fig"}). We also examined MnSOD expression in NPC samples with known clinical outcomes. In a Kaplan-Meier model, MnSOD protein expression was a strong predictor of poor survival rate, *P* = 0.0015 (Figure [7C](#F7){ref-type="fig"}). Correlations between MnSOD expression and various clinic pathological features are summarized in Table [1](#T1){ref-type="table"}. MnSOD expression did not correlate with sex; however, it was closely associated with tumor metastasis. These results show that MnSOD expression is elevated with progressing tumor stage and correlated with poor survival, suggesting an association between high MnSOD expression and tumor aggressiveness in human NPC. ![Upregulation of MnSOD in nasopharyngeal carcinoma\ (**A**) Representative immunostaining of MnSOD in tissue microarrays across different pathologic stages. (**B**) Quantification of MnSOD in tissue microarrays (Shanghai Outdo Biotech Co., Ltd., *n* = 118). The IHC intensity of MnSOD was analyzed using ImageJ and statistical analysis was performed using one-way analysis of variance. Error bars represent SD, \**p* \< 0.05. (**C**) Kaplan-Meier survival curves based on MnSOD staining for patients with stage I--IV (log-rank test).](oncotarget-07-32408-g007){#F7} ###### Correlation between MnSOD expression and various clinical pathological features Variables No.of cases (*n* = 118) Expression of MnSOD *P* ----------------------- ------------------------- --------------------- ----- ---- ------- Gender Male 87 8 32 47 0.974 Female 31 3 12 16 Pathology grading I 5 3 1 1 0.000 II 42 6 29 7 III 40 1 9 30 IV 31 1 5 25 Primary tumor T1 9 3 3 3 0.004 T2 43 4 24 15 T3 30 2 9 19 T4 36 2 8 26 Lymph node metastasis N0 40 5 17 18 0.002 N1 45 2 23 20 N2 28 2 3 23 N3 5 2 1 2 Distant metastasis M0 105 10 43 52 0.046 M1 13 1 1 11 Correlations between MnSOD staining index scores and other categorical factors were analyzed using the Pearson chi-square test of independence. DISCUSSION {#s3} ========== Anoikis is apoptosis induced by cell detachment from the extracellular matrix is a phenomenon \[[@R22]\]. Tumor cells need anoikis resistance to survive before metastasis, making it an essential trait of malignancy \[[@R8]\]. However, the mechanisms underlying anoikis resistance remain poorly understood in nasopharyngeal carcinoma (NPC) cells, which have a high rate of metastasis. In the present study, we found that the highly metastatic NPC cells, S18, have an advantage in anchorage-independent cell growth and exhibit strong anoikis resistance when compared with the poorly metastatic S26 NPC cell line (Figure [1](#F1){ref-type="fig"}). These results suggest that anoikis resistance is a critical feature of highly aggressive NPC cancer cells. When adherent cells detach from the extracellular matrix (ECM) it causes metabolic defects, including marked increases in the amount of reactive oxygen species (ROS) \[[@R10]\]. Although low levels of ROS regulate cellular signaling and play a role in normal cell proliferation, recent studies have shown that excessive amounts or persistent elevation of ROS might lead to increased anoikis. Neutralization of ROS by antioxidant enzymes or NAC enhances survival of breast cancer cells when deprived of ECM \[[@R10], [@R11]\]. Considering the importance of ROS and their implications in anoikis, as anticipated, we found that the highly metastatic S18 cells have lower intracellular ROS production. Specifically, S18 cells have low mitochondrial O~2~^·−^and H~2~O~2~ but not cytoplasmic O~2~^·−^, when compared with S26 cells (Figure [2](#F2){ref-type="fig"}). These results suggest that the level of ROS is inversely correlated with malignant cell viability during the detachment process. MnSOD is an antioxidant mitochondrial matrix protein that catalyzes O~2~^·−^ into H~2~O~2~ \[[@R23], [@R24]\]. Our data show that the highly metastatic S18 cell line display higher MnSOD protein and activity levels compared to the S26 cell line, when cells are grown in suspended conditions. These results are consistent with other reports, which found that induction of MnSOD by detachment allows cells to survive longer in suspension and depletion of MnSOD sensitizes cells to anoikis \[[@R15]\]. However, SOD converts superoxide radical into hydrogen peroxide may establish a steady flow of H~2~O~2~ originating from mitochondria. Interestingly, MnSOD can promote protein expression level of catalase thus would be unlikely to yield more H~2~O~2~ (Figure [4](#F4){ref-type="fig"}). Futhermore, though we observed Cu/Zn-SOD (a cytosolic superoxide dismutase \[[@R25]\]) was highly expressed in both S18 and S26 cells, it was not induced by detachment from the ECM in the first 24 hours (Figure [3](#F3){ref-type="fig"}). These results indicate that MnSOD, but not Cu/Zn-SOD contributes to maintaining mitochondrial bioenergetics and promoting anoikis resistance in NPC. To further confirm the role of MnSOD in anoikis resistance, we found that the anchorage-independent growth ability of S18 cells is inhibited upon MnSOD knockdown, while growth is enhanced with MnSOD overexpression in S26 cells (Figure [4](#F4){ref-type="fig"}). These findings suggest that increased MnSOD activity contributes to anoikis resistance, which is a critical feature of highly aggressive NPC cancer cells. These results are strongly supported by our *in vivo* studies, which showed that the lung tumor burden in mice injected with MnSOD-deficient cells is reduced (Figure [6G](#F6){ref-type="fig"}). Clinically significant elevations in MnSOD expression are associated with increased tumor invasion and metastasis in certain cancer types including gastric and esophageal cancer, breast cancer, lung carcinoma \[[@R26]--[@R29]\]. However, the role of MnSOD as a primary participant in the malignant transformation process in NPC cancer cells has been not investigated. Our clinical studies demonstrated that MnSOD was expressed at increased levels with progressing tumor stage, indicating an association between high MnSOD expression and tumor aggressiveness. Moreover, high expression of MnSOD correlates with extremely poor survival rates (Figure [7](#F7){ref-type="fig"} and Table [1](#T1){ref-type="table"}). Another implication of the data presented here is that activation of β-catenin plays a role in the anoikis resistant of NPC. We found that the β-catenin levels were substantially decreased following matrix detachment in poorly metastatic NPC cells. Upon cell detachment, β-catenin accumulates and translocates to the nucleus, as shown with immunofluorescent staining of highly metastatic S18 cells (Figure [5B](#F5){ref-type="fig"}). Knockdown of β-catenin blocks MnSOD expression and induces S18 cells to become sensitive to apoptosis in suspension culture. Furthermore, overexpression of β-catenin increases MnSOD expression and represses anoikis in S26 cells. These results support an emerging role for β-catenin in regulating MnSOD in NPC, thereby making it an attractive therapeutic target to prevent tumor metastasis. Even though we document the properties of β-catenin/MnSOD in anoikis resistance, the precise mechanism by which β-catenin is activated in suspended cells and how β-catenin is involved in the regulation of the MnSOD warrants further investigation. In summary, we established the critical role of MnSOD in conferring mitochondrial O~2~^·−^ and H~2~O~2~ degradation during anoikis resistance. Activation of β-catenin up-regulates MnSOD and might be the underlying molecular mechanism that defines the aggressive form of NPC. Thus, β-catenin could potentially be exploited as a therapeutic target to induce anoikis in circulating tumor cells and prevent metastatic growth. MATERIALS AND METHODS {#s4} ===================== Cells culture and generation of stable cell lines {#s4_1} ------------------------------------------------- S18 and S26 were graciously provided by Professor Qian (Department of Nasopharyngeal Carcinoma, Sun Yat-sen University Cancer Center, Guangzhou, China). Cells were maintained in RPMI 1640 supplemented with 10% FBS. Penicillin and streptomycin were added to all cultures. Puromycin (0.5 ug/mL) was used for the generation of stable cell lines. Cell viability assays {#s4_2} --------------------- Cells were plated at a density of 20,000 cells per well in 48-well or 150,000 cells per well in 6-well poly-HEMA-coated (or normal) plates. Cell viability was measured using the CCK8 assay (Dojindo, Japan) according to the manufacturer\'s protocol. For cell counting, detached cells were stained with trypan blue and placed in a hemocytometer. Soft agar colony assays {#s4_3} ----------------------- A total of 20,000 cells were suspended in RPMI 1640 (10% FBS) containing 0.7% agarose and layered on solidified medium containing 1.2% agarose in six-well plates. After solidification, the top layer was covered with the medium. When necessary, 1 mM NAC and 30 μM H~2~O~2~ was added to each layer and the cover medium. The medium was replaced every week. Images were taken after 3--4 weeks and analyzed with ImageJ software. Acini formation assays {#s4_4} ---------------------- 12-well plates were initially coated with 30 ul 100% Matrigel and allowed to solidify, forming a gelled bed of basement membrane measuring approximately 1 mm in thickness. Cells (1 × 10^4^/ml) were seeded onto this bed as a single-cell suspension in RPMI 1640 (10% FBS) containing 2% Matrigel. The medium was replaced every 4 days. Images were obtained using a confocal laser scanning microscope after 2 weeks. Anoikis assays {#s4_5} -------------- Apoptosis was assessed by the AnnexinV/PI detection. Cells (4 × 10^5^) were plated onto poly-HEMA-coated six-well plates in growth medium to prohibit attachment. After 12, 24, and 48 h in suspension, cells were harvested, clumps were separated by trypsin (0.05%) and a mesh screen process utilized to prepare monoplast suspensions which were stained per the manufacturer\'s recommendation. Experiments were repeated three times. ROS assays {#s4_6} ---------- Cells were plated at a density of 200,000 cells per well in 6-well poly-HEMA-coated (or normal) plates. After 12 h and 24 h, H~2~DCF-DA (sigma), dihydroethidium (Life Technologies) or MitoSOX (Invitrogen) was added to each well at a concentration of 10 μM, 10 μM or 5 μM, respectively. After 30 min, the samples were run on the flow cytometer to detect fluorescence. Each sample was collected using 20,000 events. Amplex^®^ Red Hydrogen Peroxide/Peroxidase Assay Kit (Invitrogen, A22188) was also used to detect H~2~O~2~ according to the manual operation. Superoxide dismutase (SOD) activity {#s4_7} ----------------------------------- SOD activity was measured using a Cu-Zn/Mn-SOD assay kit (WST) (Beyotime Institute of Biotechnology, Jiangsu, China). Briefly, total SOD activity was measured by reduction rate inhibitions of 2-(4-iodophenyl)-3-(4-nitrophenyl)-5-(2,4-disulfophenyl)-2H-tetrazolium and monosodium salt (WST-1). MnSOD activity was measured by adding 10 mM potassium cyanide to inactivate Cu-Zn/SOD activity. The difference between total SOD and MnSOD activity was considered as the Cu-Zn/SOD activity. SOD activity was expressed as units/mg of protein (one unit was defined as the amount of enzyme that inhibited WST-1 reduction by 50%). RNA isolation and quantitative RT-PCR {#s4_8} ------------------------------------- Total RNA isolation from cells grown under attached and suspended conditions was carried out with TRIzol (Invitrogen, Grand Island, NY, USA), following manufacturer\'s protocol. 500 ng total RNA was used for reverse transcription and quantitative real-time PCR analysis (qRT-PCR). Relative mRNA quantities were determined using the comparative cycle threshold (ΔΔCt) method. β-actin was used for normalization. The primers used for *MnSOD*, *Cu, ZN-SOD*, and *ACTB* are as follows: *MnSOD* forward, 5′-CTG TTG GTG TCC AAG GCT CA-3′, *MnSOD* reverse, 5′-GTA GTA AGC GTG CTC CCA CA-3′. *Cu, Zn-SOD* forward, 5′-TGG TTT GCG TCG TAG TCT CC-3′, *Cu, Z-SOD* reverse, 5′-CTT CGT CGC CAT AAC TCG CT-3′, *ACTB* forward, 5′-GCA CTC TTC CAG CTT CCT T-3′, *ACTB* reverse, 5′-GTT GGC GTA CAG GTC TTT GC-3′. The experiments were performed in triplicate. siRNA transfection and short hairpin RNA {#s4_9} ---------------------------------------- To establish stable cells expressing either MnSOD short hairpin RNA(shRNA), S18 cells were transfected with Lentiviral RNAi system (The lentivirus skeleton plasmid were gifts from Prof. Peng Xiang (Sun Yat-sen University, China) and selected by 0.5 ug/mL puromycin (Invitrogen, Carlsbad, CA, USA) for 15 days. For siRNA knockdown, MnSOD siRNA and a control siRNA were purchased from RiboBio. According to the manufacturer\'s instructions, transfections were performed at approximately 60% confluency using HiPerFect Transfection Reagent (Qiagen). Western blotting {#s4_10} ---------------- Cells were harvested and lysed for total protein extraction. Protein concentration was determined using Bio-Rad DC protein assay kit (Bio-Rad Laboratories) according to manufacturer\'s protocol. Aliquots of equal amounts of protein from the cell lysates were subjected to Western blot analysis. Antibodies used include those specific for MnSOD(BD Biosciences), catalase (proteintech), caspase3 (Cell Signaling Technology), caspase9 (Cell Signaling Technology), β-catenin (Cell Signaling Technology) and β-actin (Sigma, MO, USA). Animal experiments {#s4_11} ------------------ Male athymic mice between 5 and 6 weeks of age were obtained from Shanghai Institutes for Biological Sciences (Shanghai, China). 2 × 10^6^ cells were injected into the lateral tail vein of 7-week-old nude mice. Six weeks after injection, the mice were sacrificed, and lungs were removed. Lungs were fixed in Bouin\'s solution for 24 hours and stored in 70% ethanol before analysis. All the animal procedures were outlined in the guidelines of Institutional Animal Care and Use Committee at SYSU. Immunofluorescence {#s4_12} ------------------ S18 and S26 cells plated on culture slides or smear after the suspension culture were fixed in ice-cold 4% paraformaldehyde. Then, samples were permeabilized for 20 min with 0.01% Triton X-100 in PBS and blocked with normal goat serum at 37°C for 1 hour. After washing 3 times with PBS, they were incubated with β-catenin (Cell Signaling Technology) overnight at 4°C, then were followed by fluorescent secondary antibodies (1:200 dilution in blocking buffer, 1 h, room temperature). After immunolabeling, cells were washed, stained with 4, 6-di-amino-2-phenylindole (DAPI) (Sigma, St. Louis, MO, USA) for 5 min at room temperature. Cells were visualized under a confocal microscope (Axio Observer Z1, ZEISS, Jena, Germany). Immunohistochemistry {#s4_13} -------------------- Tissue microarray (Shanghai Outdo Biotech Co., Ltd., *n* = 118) were deparaffinized and dehydrated with graded alcohol. The samples were pretreated with 0.01 M citrate buffer (pH 6.0) for 2 min at 100°C in an autoclave; then the slides were allowed to cool to room temperature. Endogenous peroxidase activity was quenched by incubation in methanol containing 3% H~2~O~2~ for 10 min at room temperature. After several washes in Phosphate-Buffered Saline Tween-20 (PBST) (pH 7.2), the sections were blocked with goat serum for 60 min at room temperature and then incubated with MnSOD antibody (BD Biosciences) overnight at 4°C in a humidified chamber. After a brief rinse in PBST, sections were incubated for 40 min at 37°C with a biotin-conjugated secondary antibody (mouse) followed by incubation with DAB for 15 s. After rinsing with distilled water, sections were counterstained with hematoxylin. As a negative control, slides were incubated in PBS in place of primary antibody. The IHC intensity of MnSOD was analyzed using ImageJ. Statistical analyses {#s4_14} -------------------- Data are presented as means ± SD. The difference between experimental groups was assessed by ANOVA or two-tailed paired Student\'s *t*-test. The log-rank test was used for survival analysis. A value of *p* \< 0.05 was considered statistically significant. We thank the members of the laboratory for their helpful comments on the manuscript. **CONFLICTS OF INTEREST** No potential conflicts of interest was disclosed. **GRANT SUPPORT** This study was supported by National Nature Science Foundation of China, Grant Number: 81172163, 81272338, 81272515, 81200706, 8137 0945, 81471033, 81572342, 81570871, 81570764; National Key Sci-Tech Special Project of China, Grant Number: 2013ZX09102--053, 2015GKS-355; Program for Doctoral Station in University, Grant Number: 20120171110 053,20130171110053 Key Project of Nature Science Foundation of Guangdong Province, China, Grant Number: 2015A030311043; Guandong Natural Science Fund, Grant Number: 2014A020212023, 2014A0303 13073, 2015A030313103, 2015A030313029; Key Sci-tech Research Project of Guangzhou Municipality, China, Grant Number: 2014J4100162, 201508020033; Chang jiang Scholars and Innovative Research Team in University, number: 985 project PCSIRT 0947; Fundamental Research Funds for the Central Universities of China (Youth Program 13ykpy06, 31610046). The funding agencies had no role in study design, data collection, and analysis, decision to publish, or preparation of the manuscript.

{ "pile_set_name": "PubMed Central" }

CASE REPORT {#s1} =========== A 16-year-old girl developed multiple painful and tender soft tissue swellings over her neck, trunk and extremities since 2 years of age. These swellings appeared spontaneously or after trivial injury and either used to resolve spontaneously or they ossified into mature bone. These flares of heterotopic ossification resulted in gradual restriction of movements of neck, trunk, shoulder, elbow, hip and knee joints. There was associated congenital hallux valgus bilaterally (Fig. [1](#OMV022F1){ref-type="fig"}a). There was no history of similar illness in her family. The radiographs of the girl showed certain characteristic features of fibrodysplasia ossificans progressiva (FOP). The cervical skiagram showed characteristic anomalies of the cervical spine in form of large posterior elements and fusion of the facet joints between C2 and C7 **(**Fig. [1](#OMV022F1){ref-type="fig"}b**)** \[[@OMV022C1]\]. A CT scan of dorso-lumbar spine showed a peculiar 'Y-shaped' soft tissue ossification extending from the inferior angle of bilateral scapulae to the midline soft tissue of posterior back (lattisimus dorsi muscles and thoracolumbar fascia). It also demonstrated dorso-lumbar scoliosis and fusion of posterior elements of thoracic and lumbar vertebrae (Fig. [1](#OMV022F1){ref-type="fig"}c). Figure 1:(**a**) Characteristic malformed great toe and hallux valgus. (**b**) The cervical skiagram shows large posterior elements and fusion of the facet joints between C2 and C7. (**c**) A CT scan of dorso-lumbar spine shows Y-shaped soft tissue ossification extending from the inferior angle of bilateral scapula to the midline soft tissue of posterior back (lattisimus dorsi muscles and thoracolumbar fascia), dorso-lumbar scoliosis with convexity to left, squaring of thoracic and vertebral bodies and fusion of posterior elements of thoracic and lumbar vertebrae. DISCUSSION {#s2} ========== FOP is a rare genetic disorder of connective tissue characterized by congenital malformed great toes and progressive heterotopic ossification leading to severe disability. Classic FOP is caused by a recurrent activating mutation (617G\>A; R206H) in the gene ACVR1/ALK2 encoding Activin A receptor type I/Activin-like kinase 2, a bone morphogenetic protein type I receptor \[[@OMV022C2]\]. This genetic defect which leads to the formation of a heterotopic skeleton involves normal skeletal morphogenesis at heterotopic sites \[[@OMV022C3]\]. During the first decade of life, sporadic episodes of painful soft tissue swellings (flare-ups) occur which are often precipitated by trivial injury. These flare-ups transform skeletal muscles, tendons, ligaments, fascia and aponeuroses into heterotopic bone causing restriction of movements of various joints. The flares may partially respond to a short course of steroids, but overall prognosis is reduced with median lifespan of 40 years. Most patients die of complications of thoracic insufficiency syndrome \[[@OMV022C4]\]. CONFLICT OF INTEREST STATEMENT {#s3} ============================== None declared.

{ "pile_set_name": "PubMed Central" }

All relevant data are within the paper. Data can also be found online: doi:[10.5061/dryad.mb622](https://doi.org/10.5061/dryad.mb622) Introduction {#sec001} ============ Temperature influences the rates of virtually all developmental processes in bird embryos, including metabolism, development, and growth \[[@pone.0184670.ref001]--[@pone.0184670.ref003]\]. Thus, it is not surprising that, within a zone of thermal tolerance, higher mean temperatures result in higher avian embryonic growth rates than lower mean temperatures \[[@pone.0184670.ref004]--[@pone.0184670.ref013]\], albeit with some potentially negative effects on hatchability. Many birds---especially domesticated species---maintain the incubation temperature of their eggs relatively constant and close to optimal temperature (typically 35--38°C) through behavioral and physiological manipulations \[[@pone.0184670.ref014]\]. As a result, most studies evaluating temperature effects on avian development use a constant incubation temperature equal to mean daily-temperatures of a particular species' natural environment \[[@pone.0184670.ref014]\]. For economically important species such as the domestic chicken (*Gallus gallus*) or turkey (*Meleagris gallopavo*), for example, constant temperatures during incubation represent the natural incubation conditions and certainly the commercial industry standard. Consequently, relatively few studies have used cyclic thermal protocols to evaluate hatchability or post-natal growth---e.g., \[[@pone.0184670.ref015]\]. However, constant (non-cyclic) incubation temperatures do not fully represent the naturally encountered daily thermal fluctuations that eggs of many precocial, non-domesticated birds. Some species may lay multiple eggs in a clutch (e.g., ducks, gallinaceous birds, and ratites), depositing eggs in their nest at a rate of one egg per day. In such species, true incubation by one or more parents does not begin until the penultimate or ultimate egg is laid (see \[[@pone.0184670.ref002]\] for review, \[[@pone.0184670.ref016]\]). Thus, for many precocial birds there is an extended *pre-incubation* period---the focal period of this study---where eggs that are laid first are unattended for the entire pre-incubation period and thus lack the thermal protection and stabilization provided by an incubating parent. Pre-incubation egg temperatures closely follow what can be large diurnal thermal fluctuations of the environment \[[@pone.0184670.ref017]\]. The magnitude of the diurnal thermal fluctuations varies locally, seasonally, and from year to year, being for example typically greater during drought years than non-drought years. Pre-incubation egg temperatures can paradoxically reach actual incubation temperatures, which could begin development of the fertilized egg. In fact, potentially lethal temperatures have been measured in pre-incubation eggs found in ground-nests of precocial birds. For example, internal egg temperatures of pre-incubated northern bobwhite eggs (*Colinus virginianus*) in their nest but not yet being tended, followed ambient thermal fluctuations in non-drought years and often exceeded 40°C during peak thermal intensity \[[@pone.0184670.ref012], [@pone.0184670.ref017], [@pone.0184670.ref018]\]. In fact, bobwhite nest temperatures in drought conditions regularly peaked at 45°C and were recorded as high as 60°C \[[@pone.0184670.ref019]\]. Bobwhite eggs in ground nests could be severely impacted by further increases in temperature variability and magnitude \[[@pone.0184670.ref007]\]. Indeed, the impacts of climate change and local habitat warming, especially on the thermal biology of species including birds and their development, only continue to grow---for an entry into the voluminous and expanding literature see \[[@pone.0184670.ref020]--[@pone.0184670.ref024]\]. The objective of this study was thus to determine if pre-incubation is actually an unrecognized critical period for subsequent avian morphological and physiological development that can be revealed by experimental thermal protocols (e.g., \[[@pone.0184670.ref023]\]). Specifically, we investigated if exposure to diurnally oscillating temperatures or a constant temperature of equal heating degree-hours (the product of temperature (°C) X time (hours)) during pre-incubation affects subsequent morphological and physiological development of northern bobwhite embryos during their actual incubation period, the duration of incubation, the hatchability of eggs, and the hatching synchrony of eggs. We hypothesized that bobwhite embryos exposed to oscillating temperature treatments would exhibit a differential rate of development, metabolism, and hatching compared to those exposed to a constant temperature of the same mean value, and that an increased heat load would negatively affect development, metabolism and hatching success. Materials and methods {#sec002} ===================== Fertilized northern bobwhite eggs were collected from captive-reared breeding pairs of flight ready birds at Lake Cumberland Game Bird Farm (Mill Springs, KY, USA). Eggs were packaged and shipped to the University of North Texas (Denton, USA) on the day of collection. Eggs arrived on-site within 2 business days with a written record of the date and time of egg collection. Upon arrival, bobwhite eggs were randomly divided into 5 groups of 15 eggs, placed on plastic egg trays blunt end up, and given a unique identifying number with an indelible marker. Each egg was then weighed to the nearest 0.01 g with a digital scale (Ohaus Explorer Pro, Pinebrook, NJ, USA) and placed into the assigned pre-incubation treatment. Lake Cumberland Game Bird Farm was approved for egg production by the United States Department of Agriculture (USDA) and certified by the USDA National Poultry Improvement Plan. This research was approved by the University of North Texas Institutional Animal Use Care Committee, protocol \# 0808. Pre-incubation protocol {#sec003} ----------------------- Each of the 5 egg groups (N = 15 eggs per group) were assigned to a thermal chamber (G.Q.F. 1583 Hova-Bator with circulated air) with a unique experimental thermal protocol of oscillating or constant temperatures for a 12-d pre-incubation period. The groups consisted of a low and high oscillating temperature regime (LT~Osc~, HT~Osc~), a low and high constant temperature regime (LT~Const~, HT~Const~), and a commercial regime (T~Comm~) that was standard in the commercial poultry production industry \[[@pone.0184670.ref025]\]. The specifics of these incubation regimes are given in [Table 1](#pone.0184670.t001){ref-type="table"}. 10.1371/journal.pone.0184670.t001 ###### Diel thermal regimes for oscillating and constant treatments for eggs of northern bobwhites during a 12-d pre-incubation period. ![](pone.0184670.t001){#pone.0184670.t001g} Time of Day Oscillating Thermal Regime Constant Thermal Regime ------------- ---------------------------- ------------------------- --------- --------- --------- 0000--0759 25.00°C 30.00°C 28.85°C 33.85°C 20.00°C 0800--1059 30.00°C 35.00°C 28.85°C 33.85°C 20.00°C 1100--1359 35.00°C 40.00°C 28.85°C 33.85°C 20.00°C 1400--1659 40.00°C 45.00°C 28.85°C 33.85°C 20.00°C 1700--2359 25.00°C 30.00°C 28.85°C 33.85°C 20.00°C Peak temperatures of the low and high groups were selected based on thermal studies which described temperatures peaking at ≥40°C in non-drought years and ≥45°C in simulated drought years \[[@pone.0184670.ref017], [@pone.0184670.ref019]\], in Texas. The experimental temperature regime for each group (low and high) ensured that the temperature of the constant treatment was equivalent to the mean of the oscillating treatment, and heating degree-hours were equivalent within the low and high groups. A heating degree-hour was defined as 1°C above physiological zero (25°C for *Colinus virginianus*) \[[@pone.0184670.ref026]\] for 1 hour, such that 26°C for 1 h = 1 heating degree-hour, and 27°C for 1 h = 2 heating degree-hours. Thus, the mean temperature for the LT~Osc~ group was equal to that of the LT~Const~ group (28.85°C) and mean temperature for the HT~Osc~ group was equal to that of the H~Const~ group (33.85°C). Further, each low group (LT~Osc~ and LT~Const~) received 92.4 heating degree-hours per day, and each high group (HT~Osc~ and H~Const~) received 212.4 heating degree-hours per day, for the entire pre-incubation. The commercial regime (T~Comm~; 20°C) received no heating degree-hours per day, since pre-incubation temperatures did not reach physiological zero. Oscillating chambers were kept at base temperatures of 25°C (LT~Osc~) and 30°C (HT~Osc~) from midnight (0000h) to 0800h and were increased by 5°C at 0800h, 1100h, and 1400h. At 1700h, chamber temperatures were returned to their base temperature. These diurnal fluctuations were carried out each day of the 12-d pre-incubation period. The constant temperature chambers were kept at the intended temperatures throughout the entire 12-day pre-incubation period ([Table 1](#pone.0184670.t001){ref-type="table"}). All thermal chambers maintained a relative humidity (RH) of 60% to minimize compounding factors with temperature treatments. Eggs were not turned during this pre-incubation period. Assessment of development during pre-incubation {#sec004} ----------------------------------------------- To establish a baseline of pre-incubation development for egg groups subjected to each pre-incubation thermal regime, and to investigate whether various pre-incubation thermal groups affected subsequent embryonic development, eggs from each group were randomly selected for analysis (23 eggs per thermal regime group were selected over the duration of the study). Following the pre-incubation protocol, on the morning of the 13^th^ day of incubation, selected eggs were weighed to the nearest 0.01 g to determine the amount of water loss during pre-incubation. Further, eggs were opened and if embryos were present, they were separated from the egg (yolk-free), weighed to the nearest 0.01 g, aged, and staged according to morphological indicators of development \[[@pone.0184670.ref027], [@pone.0184670.ref028]\]. Incubation protocols {#sec005} -------------------- After the pre-incubation period, on the morning of the 13^th^ day, eggs from each group were removed from the thermal chamber, weighed to the nearest 0.01 g, and immediately placed into the incubator (G.Q.F. 1502 Sportsman incubator, G.Q.F. Manufacturing Co., Savannah, GA, USA). Incubator temperature was maintained at a constant 37.5±0.5°C with a RH of 60%. Eggs were turned automatically every 3 h for the first 19 days of the 23 day incubation \[[@pone.0184670.ref001]\]. On day 20 (D20) of incubation, eggs were weighed to the nearest 0.01 g and placed in the hatching chamber of the same incubator and egg turning was stopped \[[@pone.0184670.ref001]\]. Hatching was determined when eggs were star-pipped, defined as an externally pipped egg where the embryo created a small hole in the shell (approximately 3 mm^2^) to initiate hatching \[[@pone.0184670.ref024]\]. This definition of hatching was used to compensate for any artificial hatching difficulties. Upon star-pipping, the duration of incubation, degree of hatching synchrony, and percentage of eggs hatched was recorded. The stage at the time of death was recorded for eggs that did not hatch. A subset of incubating eggs was allowed to go to hatching, and the hatchlings were subsequently maintained for one day at 37.5±0.5°C with a RH of 60%, for oxygen consumption measurements (see below). Oxygen consumption {#sec006} ------------------ During incubation, on incubation days 10, 12, 14, 16, 18, 20, 22, and on 1-d post hatch, VO~2~ of the embryos *in ovo* was recorded via flow-through respirometry as an indicator of development and timing of physiological processes (e.g., internal pipping). On these selected days, 6 eggs from an individual group were removed and immediately placed into individual metabolic chambers located in a modified incubator (37.5°±0.5°C) as part of the flow-through respirometry system. The respirometry system was operated by pumping compressed normoxic air (159.22 mmHg) into an airflow manifold (MF-8 Sable Systems International, Las Vegas, NV, USA) which split the airflow into 7 channels and set the rate air flow through the system at 100±1 ml·min^-1^ \[[@pone.0184670.ref028]\]. The seven Nalgene air tubes then went into the incubator in a sealed opening and connected to copper coils of the same diameter, to warm the air to incubation temperature. Air then flowed through Nalgene tubes connected the copper coils, then to the 7 metabolic chambers (modified 2.5 oz sterilized Gerber baby food jar) simultaneously. Chamber 1 was kept empty (the "blank") and used as a reference for normoxic air, while chambers 2--7 contained eggs. From each chamber, air flowed to a multiplexor (Sable Systems International, Las Vegas, NV, USA) that regulated the sequence of airflow from the chambers to a column of Drierite and soda lime, which removed water vapor and carbon dioxide, respectively. A sub-sample pump then pulled a low-pressure air sample of each chamber, sequentially through the oxygen analyzer (FC-1B, Sable Systems International, Las Vegas, NV, USA). Each chamber was sampled at 5 sec consecutive intervals for 15 min. Oxygen measurements from the analyzer were sent to the computer via a Sable Systems Universal Interface II and processed by system software (Data acquisition system 2.0 and Datacan V, Sable Systems International, Las Vegas, NV, USA). VO~2~ was measured as the difference in oxygen concentration between the normoxic air flowing into the blank chamber (chamber 1) and the expired air flowing out of each treatment chamber (chambers 2--7). Statistical analyses {#sec007} -------------------- All data was tested with a Shapiro--Wilks normality test \[[@pone.0184670.ref029]\] and Hartley's Fmax test \[[@pone.0184670.ref029]\] before specific statistical analyses were performed. An independent t-test \[[@pone.0184670.ref029]\] was used to compare oscillating to constant temperature egg groups. An ANOVA \[[@pone.0184670.ref029]\] was used to identify the relationship between groups for all experiments. A Kaplan-Meier survival analysis \[[@pone.0184670.ref030]\] was used to compare survival rates and time to mortality among groups. Significance between groups was determined with a Student--Newman--Keuls (SNK) multiple range *post hoc* test \[[@pone.0184670.ref029]\]. All statistical tests were conducted using SigmaStat 3.5 software (Systat Software Inc. San Jose, CA), or manually \[[@pone.0184670.ref029], [@pone.0184670.ref031]\]. Statistical decisions were made with a 0.05 level of probability. All data are reported as mean±S.E. unless otherwise indicated. Results {#sec008} ======= Pre-incubation development {#sec009} -------------------------- For the subsample of embryos extracted directly after the 12-d pre-incubation period, mean stages of development were different among pre-incubation treatments (ANOVA, P\<0.001). Oscillating and constant groups significantly differed from each other ([Fig 1](#pone.0184670.g001){ref-type="fig"}) and developmental stage of all groups were statically unique (P\<0.05). During pre-incubation, eggs of the high-constant temperature (H~Const~) group were the most developed, advancing to a mean developmental stage of 21.4±0.2, equivalent to 5.3 incubation days at 37.5°C \[[@pone.0184670.ref027], [@pone.0184670.ref028]\]. Eggs exposed to high-oscillating temperatures (HT~Osc~) developed to a mean stage of 13.7±0.2, equivalent to 2.2 incubation days at 37.5°C. Low-oscillating (LT~Osc~) embryos developed to a mean stage of only 2.5±0.1, equivalent to 0.5 incubation days at 37.5°C, while pre-incubation development was lowest in low-constant (LT~Const~) eggs, which progressed to a mean stage of just 1.1±0.1, or \<0.5 incubation days at 37.5°C ([Fig 1](#pone.0184670.g001){ref-type="fig"}). Commercial pre-incubation eggs (T~Comm~) had no pre-incubation development. ![Mean developmental stages of northern bobwhite embryos after exposure to specific thermal groups during a 12-d pre-incubation period.\ Numbers above bars show days of incubation equivalent to the indicated developmental stages. Letters indicate statistical groupings. n = 23 eggs per group. Commercially incubated eggs did not have any pre-incubation development.](pone.0184670.g001){#pone.0184670.g001} Survival curves hatching success {#sec010} -------------------------------- Mortality occurred prior to hatching in all groups including the commercial group, as anticipated. The survival rate and stage of development at the time of death was significantly different between groups (Log rank survival analysis; P\<0.001). A Holm-Sidak Pairwise MCP statistically grouped the treatments receiving the lowest amounts of heating degree-hours (LT~Osc~, LT~Const~, and T~Comm~) into one group. These groups had a statistically higher survival rate compared to the high heating degree-hour groups (HT~Osc~ and H~Const~; P\<0.001, [Fig 2](#pone.0184670.g002){ref-type="fig"}). Mean survival time for LT~Const~ eggs was 22.2±1.4 days. The low oscillating group had a mean survival time of 22.1±0.7 days, and this was not statistically different from the LT~Const~ group (P = 0.25). Mean survival for HT~Osc~ eggs was 12.2±1.8 days which was significantly different from H~Const~ eggs (4.47±0.3 days), none of which hatched during any trials ([Fig 2](#pone.0184670.g002){ref-type="fig"}). ![Kaplan-Meier survival curves showing survival rate and time to mortality of northern bobwhite embryos exposed to different thermal protocols during a 12-day pre-incubation period.\ Stair-step declines show time of mortality. n = 45 eggs per treatment group. Letters indicate statistical groupings with a P\<0.05 level of significance.](pone.0184670.g002){#pone.0184670.g002} Commercial eggs hatched at a significantly higher rate (80.8±3.6%) than all treatment groups ([Fig 3](#pone.0184670.g003){ref-type="fig"}; ANOVA, P\<0.001). Hatching success of the LT~Const~ treatment (63.3±6.7%) was not significantly different from the LT~Osc~ treatment (53.5±6.4%; P = 0.24). HT~Osc~ hatch rates (6.0±2.1%) did not statistically differ (P = 0.53) from H~Const~ hatch rates (0.0%). Hatch rates for LT~Osc~ was significantly higher than for the HT~Osc~ groups (P\<0.001). ![Hatching success of northern bobwhite embryos exposed to different thermal protocols during a 12 day pre-incubation period.\ n = 45 eggs per group. Letters indicate statistical groupings.](pone.0184670.g003){#pone.0184670.g003} Pre-incubation thermal treatment not only affected survival and hatching success, but also the timing of specific hatching events---internal pipping, external pipping, and hatching. All of these developmental landmarks were significantly different between commercial and treatment groups ([Fig 4](#pone.0184670.g004){ref-type="fig"}; ANOVA, P\<0.001). HT~Osc~ eggs internally pipped earlier (D18.6±0.2) than LT~Const~ eggs (D19.7±0.2), LT~Osc~ eggs (D20.5±0.2), and T~Comm~ eggs (D20.6±0.2). Similarly, HT~Osc~ eggs externally pipped earlier (D20.0±0.0) than LT~Const~ eggs (D21.5±0.3), LT~Osc~ eggs (D21.5±0.2), and T~Comm~ eggs (D21.7±0.2). HT~Osc~ eggs also hatched earlier (D21.0±0.0) than T~Comm~ eggs (D23.0±0.0) and LT~Osc~ eggs (D23.5±0.2), while LT~Const~ eggs took the longest to hatch of treatment groups (D23.8±0.1). ![Time to internal and external pipping and hatching for northern bobwhite eggs exposed to T~Comm~, LT~Const~, LT~Osc~, and HT~Osc~ thermal treatments during a 12 day pre-incubation period.\ Sample size in parentheses. Events are statistically grouped (0.05 level of significance) with lower-case, italicized, and capital lettering.](pone.0184670.g004){#pone.0184670.g004} Additionally, the duration from internal pip to hatching was significantly different between treatment groups (P\<0.001) ([Fig 4](#pone.0184670.g004){ref-type="fig"}). The hatching duration was not significantly different between LT~Const~ (3.7±0.4 days) and LT~Osc~ (3.2±0.1 days) treatments (P = 0.16), but HT~Osc~ had a shorter duration of hatching events than LT~Osc~ (P\<0.001). The duration between internal pip and hatch was the shortest for HT~Osc~ (2.4±0.2 days) and T~Comm~ eggs (2.4±0.2 days. All eggs in a given group hatched essentially synchronously--i.e., within 24 h of each other. Oxygen consumption {#sec011} ------------------ Oxygen consumption (VO~2~) increased significantly as a function of development ([Fig 5](#pone.0184670.g005){ref-type="fig"}). VO~2~ in LT~Const~ and LT~Osc~ groups was similar in overall pattern throughout incubation, but absolute values differed significantly (P\<0.05) on D10, D15, D18, D20, and D21 of incubation ([Fig 5A](#pone.0184670.g005){ref-type="fig"}). HT~Osc~ embryos had a significantly higher VO~2~ than H~Const~ embryos on D10 but not on D15, however H~Const~ embryos did not survive past day 15 of incubation, so no further comparisons were possible ([Fig 5B](#pone.0184670.g005){ref-type="fig"}). Hatchling VO~2~ was higher than at any point in embryonic development in all groups, but there were no significant differences between treatments in hatchlings. ![Oxygen consumption of northern bobwhite embryos as a function of development and pre-incubation thermal groups.\ Statistically similar means are grouped within boxes. \* No H~Const~ eggs survived past day 15.](pone.0184670.g005){#pone.0184670.g005} To further aid in VO~2~ comparison, semi-log plots were created ([Fig 6](#pone.0184670.g006){ref-type="fig"}). Slopes relating VO~2~ to development were not significantly different between LT~Const~ and LT~Osc~ pre-incubation treatments ([Fig 6A](#pone.0184670.g006){ref-type="fig"}; t test for slopes, P\>0.50). Slopes relating VO~2~ to development were significantly different between LT~Osc~ and HT~Osc~ pre-incubation treatments ([Fig 6B](#pone.0184670.g006){ref-type="fig"}; t-test for slopes, 0.02\>P\>0.001). As noted above, H~Const~ eggs did not survive past day 15 of incubation, so no oxygen consumption comparisons were made with this treatment group. ![Regressions relating oxygen consumption to development.\ A) No significant differences existed between the slope of the lines for LT~Const~ and LT~Osc~ (ANOVA; P\>0.50). B) The slope of the line for relating LT~Osc~ was significantly higher than for HT~Osc~ (0.02\>P\>0.001). Coefficient of determination (r^2^) and n values are shown.](pone.0184670.g006){#pone.0184670.g006} Egg mass and water loss {#sec012} ----------------------- Mean egg weight loss among thermal groups during the pre-incubation period was higher in LT~Osc~ and HT~Osc~ egg groups than in all other pre-incubation groups (P\<0.001), with a percentage loss of 4.9±0.2% and 3.9±0.5% respectively. There was no difference between HT~Const~ eggs (2.8±0.5%) and LT~Const~ eggs (2.6±0.1%) and between LT~Const~ eggs and T~Comm~ eggs (2.14±0.04%; P\>0.05) ([Fig 7](#pone.0184670.g007){ref-type="fig"}). ![Mean percentage of fresh egg weight loss for northern bobwhite eggs measured after exposure to oscillating and constant thermal treatment for a 12 day pre-incubation period.\ n = 45 eggs per group. Letters (lower-case, italicized, and capital) indicate statistical groupings within each treatment.](pone.0184670.g007){#pone.0184670.g007} Unlike pre-incubation, no significant difference (P = 0.88) in the percentage of fresh egg weight loss during incubation was detected between HT~Osc~ and H~Const~ (P = 0.05) or LT~Osc~ and LT~Const~ (P = 0.88) groups. However, there was a significant (P\<0.05).difference in egg weight loss between T~Comm~ eggs (11.5±0.3%) and H~Const~ eggs (14.9±1.5%) during incubation ([Fig 7](#pone.0184670.g007){ref-type="fig"}). Combining the pre-incubation and incubation periods, all treatment groups lost a greater percentage of fresh egg mass than T~Comm~ eggs (ANOVA, P\<0.05). The mean percentage of fresh egg mass lost by T~Comm~ eggs eggs was 11.5±0.3%, lower than H~Const~ (17.7±1.9%), HT~Osc~ (14.9±1.5%), LT~Const~ (15.9±0.6%), and LT~Osc~ eggs (18.3±0.7%). No difference was detected in percentage of fresh egg mass lost between HT~Osc~ and H~Const~ (P = 0.28), or LT~Osc~ and LT~Const~ (P = 0.12) ([Fig 7](#pone.0184670.g007){ref-type="fig"}). Discussion {#sec013} ========== Pre-incubation modification of subsequent embryonic development {#sec014} --------------------------------------------------------------- This study shows that quail embryos actually begin to develop during pre-incubation when eggs are exposed to temperatures above "physiological zero". As expected, more development was obtained with more heating degree-hours and less development occurred with fewer heating degree-hours, even when slightly above physiological zero. Eggs exposed to high heat loads (HT~Osc~ and H~Const~) and low heat loads (LT~Osc~ and LT~Const~) had equal quantities of heating degree-hours within groups, but they all exhibited differential pre-incubation development and growth rates. Groups exposed to a constant temperature had the largest variation between high and low groups, with a difference of \~5 days of development between H~Const~ and LT~Const~, even though each group received the same amount of heating degree-hours ([Fig 1](#pone.0184670.g001){ref-type="fig"}). The difference in groups with oscillating temperature was much less, with about 1.7 days of development difference between HT~Osc~ and LT~Osc~. Thus, at least for quail development, the quantity of heat (e.g., number of heating degree-hours) is not the only driver of development and that the nature of the heat (oscillating or constant) is also a major factor. Recently, intragenerational epigenetic modification of embryonic development has been induced by the length of pre-incubation storage in chicken embryos \[[@pone.0184670.ref032]\]. These changes were not the result of high temperatures, since the eggs were stored well below physiological zero. Whether an oscillating temperature at cool temperatures with no likelihood of beginning embryonic development would actually alter embryonic development remains to be investigated. Survival and hatching assessment {#sec015} -------------------------------- The thermal protocol used during the pre-incubation period in *Colinus virginianus* had a profound effect on subsequent embryonic survival. While high pre-incubation temperatures severely affected subsequent embryonic survival, it is interesting to note that an equivalently high oscillating temperature, while leading to only poor survival, nonetheless was more successful in producing viable embryos than a constant high temperature. Observations of the effect of pre-incubation temperature and other conditions on subsequent embryonic survival have mostly derived from interest in poultry production or basic experimentation \[[@pone.0184670.ref032]--[@pone.0184670.ref038]\]. Such studies on the pre-incubation period have almost invariably considered steady-state storage conditions, rather than oscillations in environmental parameters, such as the oscillations in temperature employed in the current study. Few studies on oscillating temperatures exist for any aspect of avian egg storage---the few such studies have been directed towards inducing temperature oscillations once formal incubation has begun \[[@pone.0184670.ref015]\]. Embryonic survival and hatching success was not statistically different between constant and oscillating groups. However, H~Const~ embryos did not hatch, dying at \~4 to 15 days of incubation ([Fig 2](#pone.0184670.g002){ref-type="fig"}). The death of H~Const~ embryos is noteworthy, since eggs were kept for 12 days at 33.85°C, well within the natural incubating range of 30--35°C measured in nature for this species \[[@pone.0184670.ref017]\]. While the objective of this study was not to determine the cause of death in eggs at constant high temperature, we speculate that constant temperatures near incubation temperature require eggs to be turned daily for survival. Indeed, studies have shown that, while turning of quail eggs is not required when eggs are stored at 12.8°C, turning is required during incubation at temperatures between 35--38°C \[[@pone.0184670.ref001], [@pone.0184670.ref039], [@pone.0184670.ref040]\]. High mortality was not experienced in LT~Const~ groups (28.85°C) presumably because the mass of the embryo was that at \<0.5 day of age (i.e. a much smaller embryo), and thus was not affected by the absence of turning. Bobwhite quail eggs exposed to high oscillating temperatures internally pipped, externally pipped, and hatched two days earlier than all other groups, ultimately showing a reduced incubation period ([Fig 4](#pone.0184670.g004){ref-type="fig"}). This should not be surprising since the high oscillating temperature protocol (HT~Osc~) during pre-incubation enabled approximately 2 additional days of development than in LT~Osc~ and LT~Const~ groups. Both LT~Osc~ and LT~Const~ eggs, which developed approximately 0.5 days during the pre-incubation period, exhibited an extended incubation time of 0.5 d ([Fig 4](#pone.0184670.g004){ref-type="fig"}). HT~Osc~ eggs not only started development earlier, but also showed overall reduced incubation duration of HT~Osc~ eggs. This may be an adaptation to the thermally stressful environments inhabited by northern bobwhite quail, where both high temperature and temperature fluctuation are common and would affect the eggs during their pre-incubation period. Collectively, our results suggest that oscillating temperatures act differently on developing organisms than constant temperatures of equal heating degree-hours. The impact of these differences is increased as heat loads are increased, i.e., higher heating degree-hours result in more variation in development in embryos exposed to constant and oscillating temperatures. Further, within oscillating groups, higher oscillating heat loads caused an increase in the rate of development during pre-incubation, a decrease in the rate of development during incubation, and had more of a negative effect on hatching. Hatching synchrony {#sec016} ------------------ While the current study did not focus on hatching synchronization, it is interesting to note that, irrespective of thermal regime, all quail eggs hatched within 24h of each other--i.e. "synchronously" \[[@pone.0184670.ref041]\]. Northern bobwhites do not begin incubation until the last egg is laid \[[@pone.0184670.ref042]\]. Consequently, all eggs receive equal amounts of incubation, resulting in synchronous hatching. This hatching adaptation seems beneficial by allowing the incubating parent to quickly transition behavior from incubating sessile eggs to brooding motile hatchlings \[[@pone.0184670.ref041], [@pone.0184670.ref043]\]. Importantly, limited studies have investigated hatching synchrony in warmer climates where nest temperatures fluctuate during pre-incubation and can often exceed physiological zero (the temperature above which embryonic development begins). In warmer climates, those eggs laid early could actually begin to develop prior to incubation, resulting in asynchronous development at the onset of incubation and asynchronous hatching at the other end of embryonic development. In this case, eggs might hatch asynchronously unless a synchronizing mechanism takes place \[[@pone.0184670.ref006], [@pone.0184670.ref044]--[@pone.0184670.ref046]\]. Hatching synchrony has been observed in the laboratory among quail eggs that differed in developmental stages by 2 days \[[@pone.0184670.ref046]\], but not by up to 5 days, which is the probable result of northern bobwhites in warm climates given our experimental findings on stage of development in high temperature groups at the onset of incubation. Interestingly, eggs must remain in physical contact with each other for synchronous hatching to occur, especially in the last 20% of incubation when pipping occurs \[[@pone.0184670.ref046], [@pone.0184670.ref047]\]. Regardless of relatedness or order of laying, viable bobwhite eggs of the same age hatch synchronously if touching \[[@pone.0184670.ref048]\]. When isolated, their time of hatch is spread over a longer period \[[@pone.0184670.ref045]\]. Bobwhite eggs accelerated hatching when placed in contact with more-advanced eggs (24 h advanced in development) and delayed hatching when placed in contact with more-retarded eggs (24 h retarded in development) \[[@pone.0184670.ref045]\], suggesting that both an acceleration and delay of hatching time occur between earlier laid and later laid eggs in the wild. Oxygen consumption and development {#sec017} ---------------------------------- Oxygen consumption during avian embryonic development has been very well documented, and the reader is directed to recent discussions of that literature \[[@pone.0184670.ref016], [@pone.0184670.ref049]--[@pone.0184670.ref051]\]. Many factors influence avian embryonic VO~2~, most notably body mass, oxygen availability, and environmental temperature. Yet, in precocial birds, there is a representative pattern of VO~2~ change during development in which VO~2~ typically increases until approximately 80% of incubation time. Thereafter, rate of VO~2~ change plateaus until the embryo internally pips into the air cell. VO~2~ increases once more, followed by final increase at the time the chick pips the shell and hatches \[[@pone.0184670.ref047]\]. The oxygen consumption pattern of change in the current study ([Fig 5](#pone.0184670.g005){ref-type="fig"}) was generally that described for precocial birds \[[@pone.0184670.ref047]\], and the pattern is consistent with reports of developing northern bobwhite quail \[[@pone.0184670.ref052]\] as well as the Coturnix quail, *Coturnix coturnix* \[[@pone.0184670.ref041]\], a precocial bird with similar sized eggs. Importantly, in our study the thermal protocol during pre-incubation significantly affected VO~2~, which was different between LT~Const~ and LT~Osc~ thermal groups. The slopes of the normalized VO~2~ data were not different between LT~Const~ and LT~Osc~ treatments ([Fig 6A](#pone.0184670.g006){ref-type="fig"}), indicating no overall differences in rate of development to match the elevation of development induced by higher temperature. There were significant differences in VO~2~ between HT~Osc~ and LT~Osc~ groups logically due to HT~Osc~ embryos having developed more during pre-incubation ([Fig 1](#pone.0184670.g001){ref-type="fig"}), thus requiring more oxygen on any given day during incubation. However, the lower slope of VO~2~ for HT~Osc~ eggs during incubation showed that HT~Osc~ eggs developed at a slower rate during incubation than LT~Osc~ eggs after a higher rate of development during pre-incubation ([Fig 6B](#pone.0184670.g006){ref-type="fig"}). Additionally, HT~Osc~ eggs did not exhibit the same oscillating VO~2~ patterns in the last 20% of incubation as LT~Osc~ and LT~Const~ eggs. Rather, VO~2~ of HT~Osc~ eggs stayed relatively constant during the pipping process ([Fig 5B](#pone.0184670.g005){ref-type="fig"}). Why and how VO~2~ during incubation is affected by pre-incubation thermal protocols awaits further investigation. Implications for natural bobwhite quail populations {#sec018} --------------------------------------------------- The intention of comparing LT~Osc~ (simulated non-drought temperatures) and HT~Osc~ (simulated drought temperatures) groups was to better understand the differences between drought and non-drought conditions as well as the potential impacts of global warming on avian development generally, and quail populations specifically. In this context, one of the most interesting findings is the significant difference in hatch rates between high-oscillating (HT~Osc~, 6.0±2.1%) and low-oscillating (LT~Osc~, 53.5±6.4%) groups ([Fig 2](#pone.0184670.g002){ref-type="fig"}). The HT~Osc~ thermal group was pre-incubated at a temperature 5.0°C higher than the LT~Osc~ group, a temperature increase coincident with the predicted increase in global average temperature of 5±1°C in the United States by the period 2071--2100 relative to the period 1961--1990 \[[@pone.0184670.ref053]\]. Our results show that a 5°C increase in pre-incubation temperature of just 12 days reduces hatching rate of northern bobwhites by approximately one half. Reductions in the percentage of juvenile bobwhite quail at different lattitudes \[[@pone.0184670.ref054]\] or during drought \[[@pone.0184670.ref055], [@pone.0184670.ref056]\] could, according to our findings, be caused by increased heat loads during the 12 day pre-incubation period. Eggs in ground nests could be severely impacted by further increases in temperature variability and magnitude as global warming continues. Thus, our study indicates that the dynamic nature and magnitude of diel temperature should be considered when evaluating the biological response to predicted temperatures of global warming. We thank D. Reyna, G. Bowling, and H. Valls for providing logistical support with egg incubation, and E. Dzialowski for providing training on measuring oxygen consumption. [^1]: **Competing Interests:**The authors have declared that no competing interests exist.

{ "pile_set_name": "PubMed Central" }

1. Introduction {#sec1} =============== Cancer is a complex disease that is very heterogenic and variable at cellular level and also differs from one patient to the other in its behaviour, development, and outcome \[[@B1]\]. Physical, metabolic, and behavioural variations of cancer cells from normal ones arise through the accumulation of genetic modifications and help them to proliferate rapidly, escape from host immune surveillance, and ultimately invade distant tissues \[[@B2]\]. Histopathological, genetic, and epigenetic and clinical outcome variations between and within different types of cancers have been the greatest challenge to understand the disease and develop novel therapies \[[@B3]\]. Surgery and radiation therapy were the most preferred means of treatment to control cancer before 1950 and after 1960, respectively \[[@B4]\]. Chemotherapy can be done before surgery to shrink the tumor or after surgery to kill the remaining cancer cells \[[@B5]\]. However, most of the chemotherapeutic drugs lack specificity and tend to rapidly damage normal dividing tissues, causing side effects such as immunosuppression, neurotoxicity, and hair loss \[[@B6]\]. Moreover, resistance has also reduced therapeutic efficacy of some anticancer chemotherapeutic drugs \[[@B7]\]. In order to address these limitations, tapping nature as a major source of chemically diverse novel anticancer compounds is a consistently proven track \[[@B8]\]. Screening natural products yield more hit with more "drug-like" characteristics (absorption and metabolism) as compared to screening of rationally designed compounds \[[@B9]\]. Furthermore, screening medicinal plants based on traditional use provides a higher chance of finding active plants relative to the random approach \[[@B10]\]. Ethiopia has a rich and diverse heritage of traditional medical practices, known for using plants to prepare more than 90% of the remedies \[[@B11]\]. In addition, the country has more than 6,500 higher plant species of which, around 12% are endemic \[[@B12]\]. Reports indicate that up to 80% of the population relies on traditional remedies as a primary source of health care \[[@B13]\]. Only few ethnobotanical reports from different agroecological zones of Ethiopia are available in the literature regarding medicinal plants used for cancer treatment. These include *Bersama abyssinica*, *Buddleja polystachya*, *Clerodendrum myricoides*, *Dovyalis abyssinica*, *Ekebergia capensis*, *Myrsine melanophloeos*, *Olea capensis*, *Pentas lanceolata*, *Sideroxylon oxyacanthum*, and *Zingiber officinale* \[[@B14]\]; *Bidens macroptera*, *Clematis simensis*, *Ferula communis*, and *Punica granatum* \[[@B15]\]; *Rumex abyssinicus* \[[@B16]\]; *Zanthoxylum chalybeum* \[[@B17]\]; *Phytolacca dodecandra* and *Vinca rosea* \[[@B18]\]; *Kalanchoe lanceolata*, *Stephania abyssinica*, and *Vernonia hymenolepis* \[[@B19]\]; *Plumbago zeylanica* \[[@B20]--[@B22]\]; *Acalypha acrogyna*, *Carissa spinarum*, *Maytenus ovatus*, and *Salvia nilotica* \[[@B23]\]; *Croton macrostachyus* \[[@B24]\]; and *Dorstenia barnimiana* \[[@B25], [@B26]\]. In view of this fact and considering the weak traditional recording and knowledge transfer system and an alarming rate of environmental degradation, finding anticancer plants and documenting their ethnobotanical information constitute an urgent and indispensable task. Therefore, the main aim of this study was to establish an inventory of medicinal plants traditionally used to treat cancer in eleven districts of Ethiopia. 2. Materials and Methods {#sec2} ======================== 2.1. Description of the Study Areas {#sec2.1} ----------------------------------- This ethnobotanical study was conducted in four national regional states of Ethiopia: Oromia, Amhara, Afar, and Southern Nations, Nationalities, and People. The survey included different districts from each region, namely, Bale Robe and Goba from Oromia, Bahir Dar Zuria and Filiklik from Amhara, Gewane from Afar, and Wondo Genet, Sodo Zuria, Doyo Gena, North Bench, Mizan Aman, and Shako from Southern Nations, Nationalities, and People Regional State ([Figure 1](#fig1){ref-type="fig"}). These geographically, culturally, and agroecologically different study areas ([Table 1](#tab1){ref-type="table"}) were selected mainly based on the availability of traditional healers and recommendations from health workers. 2.2. Data Collection {#sec2.2} -------------------- A team comprising a botanist and researchers from Addis Ababa University was set up, and health authorities were contacted for permission and identification of traditional herbalists living in each study area. Altogether, 117 traditional healers were approached using the snowball technique and 74 traditional healers who used herbs to manage cancer-like symptoms were selected. Ethnobotanical data were collected between January and August 2016, mainly through individual interviews with the selected traditional herbalists using a semistructured interview questionnaire. The questionnaire was prepared in Amharic language and translated to different local languages for traditional healers who do not speak Amharic. This questionnaire was designed to obtain information in the following areas: (i) general data on the informant, (ii) school attendance, (iii) use of plants for cancer treatment, (iv) source of the plant material, (v) part of the plant used, (vi) method of medicinal preparation, (vii) route of administration, and (viii) side effects. A traditional healer for the purpose of this study is "a person who is recognized by the community in which s/he lives as competent to provide healthcare by using plants and plant products." Each traditional healer was approached, briefed about the purpose of the research, and asked for his/her verbal consent in talking about cancer and its treatment. They were assured of the confidentiality of the information they provided. If plants were mentioned for their anticancer purposes, a botanical sample was collected. These specimens were pressed and preserved for later identification at the National Herbarium, Addis Ababa University, Addis Ababa, and a voucher specimen of each plant was deposited in the institute. All botanical names have been transcribed according to the nomenclature system used by the Plant List ([http://www.theplantlist.org](http://www.theplantlist.org/)). 2.3. Data Analysis {#sec2.3} ------------------ The relative importance of medicinal plants used in the management of cancer-like symptoms in study areas was assessed using the relative frequency of citation (RFC), use value (UV), informants consensus factor (ICF), and cultural importance index (CI). ### 2.3.1. Relative Frequency of Citation (RFC) {#sec2.3.1} The RFC was calculated by dividing the number of informants that cite a particular plant species (FC) by the total number of informants in the survey (*N*) \[[@B29]\]:$$\begin{matrix} {\text{RFC} = \frac{\text{FC}}{N}.} \\ \end{matrix}$$ ### 2.3.2. Use Value (UV) {#sec2.3.2} The use value demonstrates the relative importance of plant species to treat particular ailment, and it is determined by the following formula \[[@B30]\]:$$\begin{matrix} {\text{UV} = {\sum\frac{U_{i}}{N_{i}}},} \\ \end{matrix}$$where "UV" stands for the use value of a species, "*U*~*i*~" stands for the number of use reports cited by informants for that plant species, and "*N*~*i*~" is the total number of informers who reported the particular plant species *i*. ### 2.3.3. Informant Consensus Factor (ICF) {#sec2.3.3} Informant consensus factor (ICF) was calculated to determine the homogeneity of the information collected about particular plant species to treat specific ailment. It was estimated using the following formula \[[@B31]\]:$$\begin{matrix} {\text{ICF} = \frac{\text{Nur} - \text{Nt}}{\text{Nur} - 1},} \\ \end{matrix}$$where Nur is the number of use reports of informants for particular ailment category and Nt refers to the number of species used for the ailment category by all informants. ### 2.3.4. Cultural Importance Index (CI) {#sec2.3.4} Cultural importance index (CI) is calculated by the sum of the use reports (UR) of informants mentioning each species use (from *i*~1~ to *i*~*N*~) in each use category and adding all the UR of each category (from *u*~1~ to *u*~NC~) divided by the total number of informants *N*. This index is determined by the following formula \[[@B29]\]:$$\begin{matrix} {\text{CI}_{i}{\sum\limits_{u = u_{1}}^{u_{\text{NC}}}{\sum\limits_{i = i_{1}}^{i_{N}}\frac{\text{UR}u_{i}}{N}}},} \\ \end{matrix}$$where CI is an ethnobotanical index that indicates the spread of the use along with the diversity of uses of each species. 3. Results {#sec3} ========== The informants consisted of 66 male and 8 female traditional healers and they were divided into three age groups: 20--40, 41--60, and ≥ 61 years. Out of 74 interviewed traditional healers, most of them (*N*%) were adults aged between 41 and 60 years. Majority of the respondents (70.2%) gained their knowledge from family members and 82% of all interviewed respondents practiced ethnomedicine for more than 25 years. More than 70% of the respondents were either only at their primary level of education or did not have a formal education at all ([Figure 2](#fig2){ref-type="fig"}). Traditional healers usually used their intuition and relied on the chronicity and growth of external mass, as a means to diagnose cancer. Lumpy growth was the most commonly cited criteria used to diagnose cancer, followed by ulcerative wounds and bleeding ([Table 2](#tab2){ref-type="table"}). However, there were instances where some of the healers claimed to have treated patients already diagnosed with cancer at modern health institutions. Traditional healers identified cancer as "*Nekersa*" in Bahir Dar Zuria and Filiklik, "*Naqarsa*" in Bale Robe and Goba, "*Sissac*" in Gewane, "*Xoka* or *Toka*" in Doyo Gena, "*Balamo*" in Wondo Genet, "*Kums* or *niamt*" in North Bench, and "*Kanser*" in Sheko and Sodo Zuria district. Out of the 6 specific cancer types (skin, breast, lung, cervical, throat, and intestinal) claimed to be treated by the respondents, skin cancer was a dominant one followed by breast cancer. A total of 53 plant species belonging to 30 families were reported for their anticancer use ([Table 3](#tab3){ref-type="table"}). The result of this study showed that shrubs (49.1%), herbs (33.9%), trees (13.2%), and climbers (3.8%) were the main sources of potential anticancer medicinal plants. This study also indicated that leaves (56.7%) were the most commonly used plant parts followed by roots (21.7%), bark (6.7%), stem (1.7%), seeds (1.7%), whole plant (1.7%), leaves and roots (5%), leaves or stem (1.7%), and leaves or seeds (1.7%) ([Figure 3](#fig3){ref-type="fig"}). Most of the reported plants occurred naturally in wild (96.2%); however, cultivation was also a source (3.8%). Reported medicinal plants have been traditionally claimed to be used to treat different types of ailments including cancer. However, only few have been scientifically investigated for their antiproliferative or cytotoxic activity ([Table 4](#tab4){ref-type="table"}). While comparing the amount and distribution of anticancer plants in the past ten years, regardless of the study areas, all respondents believed that the amount and distribution of these plants are reduced. In the current study, the highest UVs were recorded for *Aloe* spp. (6), *Albizia schimperiana* (4), *Sida schimperiana* (4), *Achyranthes aspera* (4), *Brucea antidysenterica* (4), *Cleome brachycarpa* (3), *Leonotis ocymifolia* (3), and *Prunus africana* (3). The lowest UVs were obtained for *Acokanthera schimperi*, *Acmella caulirhiza*, *Cineraria abyssinica*, and *Gnidia involucrata* ([Table 3](#tab3){ref-type="table"}). A total of 228 use reports have been documented and categorized into seven categories ([Table 5](#tab5){ref-type="table"}). Among these, other ailments (46.3%) and skin cancer (26.5%) had the highest use reports. Furthermore, ICF values were also calculated and ranged from 0 to 0.42. The highest ICF values were recorded in other ailments (0.42) and breast cancer (0.32) followed by skin cancer (0.23) category ([Table 5](#tab5){ref-type="table"}). The other ailments category comprises of diseases such as stomach ache, malaria, wart, swelling, wounds, evil eye, toothache, bleeding, gastrointestinal disorder, headache, bone fracture, cough, snake bite, herpes simplex, tonsillitis, hypertension, dandruff, fever, and hemorrhoid. The ICF value of the remaining four categories (lung cancer, colon cancer, cervical cancer, and throat cancer) was zero. Quantitative ethnobotanical indexes such as RFC and CI were calculated in this study to analyze the ethnobotanical information. According to RFC values, *Croton macrostachyus* (0.1), *Vernonia auriculifera* (0.04), *Clematis simensis* (0.04), and *Acmella caulirhiza* (0.04) are the most frequently cited among all reported plants. *Croton macrostachyus* (0.16), *Dorstenia barnimiana* (0.12), and *Aloe* spp. (0.08) rank 1^st^, 2^nd^, and 3^rd^ in position, respectively, according to the CI reference. Our result also shows that the Pearson correlation coefficient of RFC was positively and negatively correlated to CI and UV, respectively ([Table 6](#tab6){ref-type="table"}). Most of the reported remedies, prepared from these plants, were either applied topically (50%) or taken orally (41.7%). The remaining remedies were prepared to be administered either topically or orally (3.3%), both topically and orally (1.7%), and intranasally (1.7%). Usually, fresh plants were finely chopped, dried, and pounded to powder form. Then, the powder of either one or the combination of more than one plant was either mixed with drinking water or pasted and applied topically. In other cases, fresh plant parts were decocted and taken orally or crushed and applied topically. Water was the main medium in preparation of most remedies and additives like honey, milk, and butter were also used. To determine the amount of plant parts used to prepare remedies, traditional healers used spoon, fingertip, and number (in case of fresh leaves). Adverse effects reported by respondents include vomiting, diarrhea, and skin ulcers. 4. Discussion {#sec4} ============= Despite the rich biodiversity of the study areas, broad acceptability, and centuries-old tradition of using traditional medicines, the number of anticancer plants reported in this study is far less than expected. As it was reported by different ethnobotanical studies conducted in different parts of Ethiopia, this could be attributed to the attitude of many traditional healers to guard their indigenous medical knowledge as a family secret and hence hesitant to share with the researchers \[[@B13], [@B32], [@B73]\]. Justifying the lower number of female traditional healers (8, 11%) participated in this study, these studies also inferred that traditional healers usually pass their knowledge to the first son of the family. In this study, in agreement with the studies conducted in Fiche district \[[@B35]\], Ghimbi district \[[@B20]\], and Hawassa city \[[@B17]\] of Ethiopia, the predominant botanical families recorded, listing over 5 plant species each, were Asteraceae, Fabaceae, and Lamiaceae. This could be due to the fact that these families are the largest in the flora of Ethiopia and Eritrea \[[@B15], [@B21], [@B143]\]. Moreover, cytotoxicity studies conducted on different Mexican plants reported that the highest number of plant species with both *in vitro* and *in vivo* antineoplasic activities was from these families \[[@B20]\]. The highest UVs recorded in this study include *Aloe* spp. (6), *Achyranthes aspera* L. (4), *Albizia schimperiana* (4), *Sida schimperiana* (4), and *Brucea antidysenterica* (4). The highest ICF value (0.42) recorded for "other ailments" category, in this study, suggests that informants are in agreement with the use of particular plant species to treat ailments in this category. The lowest ICF value (0) obtained was for lung, colon, cervical, and throat cancer categories. This might be due to the cultural and ecological differences of the study sites and the difficulty to pinpoint the physical symptoms of lung, colon, cervical, and throat cancer as compared to the breast and skin cancer. The present study also revealed that RFC and CI values of some reported species are similar. However, there is a distinct difference in species ranking using each index. *C. macrostachyus* is placed in the first position according to both RFC and CI index. This could be due to the fact that this species is mentioned by many informants and is the most recognized plant in most study areas. Furthermore, CI value of *C. macrostachyus* is also high, suggesting the diversified use of the plant. *V. auriculifera* and *C. simensis* ranked next to *C. macrostachyus*, according to RFC index. On the other hand, *D. barnimiana* and *Aloe* spp. ranked 2^nd^ and 3^rd^ by CI index. It has been suggested that UV value is a good measure of use diversity, than the number of citations \[[@B144]\]. In agreement with this, UV value in our study is driven by species with greatest number of use rather than those cited by more informants. The Pearson correlation coefficient of −0.36, between RFC and UV, shows significant negative association between the local importance of each medicinal plant and relative importance of use of plants. This result is in contrast to previous studies that reported a significant positive correlation between RFC and UV \[[@B145], [@B146]\]. On the other hand, there is a significant positive correlation between RFC and CI (*r*^2^ = 0.74, *p* \< 0.001) implying that their pattern matches across species. The species with larger RFC value usually have higher CI, such as *Croton macrostachyus* and *Vernonia auriculifera*. Leaves and roots are the most commonly used plant parts in the preparation of remedies in the study districts. Similarly, other ethnobotanical studies conducted in different parts of Ethiopia also reported that leaves are the dominant plant part followed by root \[[@B16]--[@B18], [@B20]\]. The preference towards leaves may be because leaves are the main photosynthetic organs in plants and the primary reservoirs for secondary metabolites with medicinal values \[[@B36]\]. In contrary to other ethnobotanical studies \[[@B17], [@B18]\], where the common use of concoctions and oral route were reported, in the current study majority of the reported remedies are prepared from a single plant species and applied topically. Comparative analysis of this study with other ethnobotanical surveys of plants used traditionally in treating and managing cancer in Ethiopia \[[@B18]\], Kenya \[[@B147]\], Cameroon \[[@B37]\], Nigeria \[[@B19], [@B38]\], South Africa \[[@B39]\], and Bangladesh \[[@B148]\] revealed some similarities in the plants cited in these surveys. Of the 30 plant species cited to be used in Ethiopia \[[@B18]\], 7 species are identified in our study: *Bersama abyssinica* Fresen., *Brucea antidysenterica* JF. Mill., *Calpurnia aurea* (Ait.) Benth. *Dodonaea angustifolia* L.f., Dorstenia barnimiana Schweinf, *Kalanchoe petitiana* A. Rich., and *Prunus africana* (Hook. f) Kalkm. Although herbal remedies are believed by the general public to be safe \[[@B46]\], some research findings suggested otherwise. For instance, traditionally used Thai anticancer plants *Ganoderma lucidum* (Fr.) Karst., *Houttuynia cordata* Thunb., and *Saussurea involucrata* Matsum. & Koidz. were reported to cause side effects such as headache, insomnia, constipation, and diarrhea \[[@B62]\]. Similarly, side effects such as vomiting, diarrhea, and skin necrosis, associated with the use of traditional herbal remedies, were reported in this and other ethnobotanical studies conducted in Ethiopia \[[@B149], [@B150]\]. Few side effects reported in this study, as compared to other ethnobotanical studies conducted in Ethiopia, could be attributed to the frequent use of the topical route of administration. Nevertheless, considering the probability of underreporting adverse effects, extensive toxicological investigations should be conducted to protect the public. *In vitro* cytotoxicity and antioxidant properties of some of the plants reported in our study have also been studied. Among these plants, potent cytotoxic activity was reported for knipholone anthrone isolated from *Kniphofia foliosa*, with IC~50~ value that ranges between 0.9 ± 0.1 and 3.3 ± 0.4 *μ*g/mL \[[@B89]\]. Similarly, Nibret and Wink reported the cytotoxic activity of the crude extract of *Acokanthera schimperi* with IC~50~ value of 7.1 *μ*g/mL \[[@B73]\]. Studies conducted on the leaves of *Cineraria abyssinica* \[[@B100]\], bark of *Senna singueana* \[[@B116]\], and bark and leaves of *Rumex nepalensis* \[[@B79]\] also revealed potent radical scavenging activity of these plants. 5. Conclusion {#sec5} ============= The present study showed that traditional healers in eleven districts of Ethiopia use different medicinal plants to manage cancer-like symptoms. Frequency of citation value ranked *Croton macrostachyus* Del., *Clematis simensis* Fresen., *Dorstenia barnimiana* Schweinf, *Vernonia auriculifera* Hiern, and *Acmella caulirhiza* Del. as most cited plant species in study areas. Hence, based on these findings, we are currently evaluating the *in vitro* antiproliferative activities of reported medicinal plant species with a higher frequency of citation against human breast adenocarcinoma (MCF-7), human uterine cervical adenocarcinoma (SiSo), human lung carcinoma (A-427), and human bladder cancer (RT-4) cell lines using crystal violate assay. However, considering the rapid disappearance of the traditional knowledge of medicinal plants and an urgent need for new anticancer agents, additional studies have to be conducted to document and scientifically validate traditionally used Ethiopian anticancer plants. The authors wish to thank all traditional healers, individuals, data collectors, and local administrative authorities in all study districts that made this survey possible. This research was supported by the thematic research grant from the Addis Ababa University (grant number: TR/35/2015). Data Availability ================= The authors declare that all data supporting the finding of this study are included in this article and its supplementary information files. Ethical Approval ================ Ethical approval was obtained from Addis Ababa University, College of Health Sciences Ethics Review Board (Ref no. ERB/SOP/126/12/2015). Consent ======= Each participant consented before the interview. Conflicts of Interest ===================== The authors declare that they have no conflicts of interest. Authors\' Contributions ======================= A.B., E.E., and K.A. jointly conceived the study. S.T. conducted the ethnobotanical study and taxonomical identification. S.T., A.B., E.E., T.G., and K.A. enriched the draft manuscript for its intellectual content. All authors read and approved the final manuscript. ![Map of Ethiopia showing the location of study districts.](ECAM2020-7683450.001){#fig1} ![Demographic details of the interviewed informants.](ECAM2020-7683450.002){#fig2} ![Frequency of plant parts used for the preparation of medicinal remedy.](ECAM2020-7683450.003){#fig3} ###### Vegetation type, climatic condition, and demographic data of the study areas \[[@B27], [@B28]\] (source: National Meteorological Service Agency of Ethiopia). District Distance from capital city (km) Approximate population (2015) Number of interviewed healers Area size (km^2^) Geographical location Average elevation above sea level (m.a.s.l) Vegetation type Climatic condition (2014) ----------------- --------------------------------- ------------------------------- ------------------------------- ------------------- ------------------------------- --------------------------------------------- ----------------------- --------------------------- ------------ Bale Robe 432 65,284 2 8.87 7°07′11.65″ N 40°00′24.82″ E 2480 DAF 745.6 9.2--23.2 Goba 444 47,135 7 20.15 7°00′41.66″ N 39°58′33.96″ E 2614 DAF 736.3 9.5--23.8 Bahir Dar Zuria 578 206,708 16 1443.37 11°34′27.15″ N 37°21′40.87″ E 1800 CTW, DAF, and FLV/MFS 1547.1 12.7--27.6 Filiklik 188 142,722 7 806.98 10°02′12.74″ N 38°14′27.65″ E 1853 CTW and DAF 880.2 12.9--22.0 Gewane 344 39,186 6 967.85 10°29′59.99″ N 40°44′59.99″ E 568 ACB 586.7 19.5--36.7 Wondo Genet 270 196,277 12 226.45 7°05′3.01″ N 38°37′8.02″ E 1742 DAF 928.7 15--29.6 Sodo Zuria 383 145,092 2 25.62 6°51′10.11″ N 37°45′39.49″ E 1854 CTW and DAF 1569.2 14.8--25.2 Doyo Gena 258 95,393 14 130.57 7°21′20.22″ N 37°47′07.15″ E 2300 DAF 1334.5 11--22.8 North Bench 587 126,308 4 392.65 6°37′53.43″ N 35°33′56.83″ E 2367 CTW 1671.8 16--33.3 Mizan Aman 565 64,996 3 24.45 6°59′37.13″ N 35°34′55.92″ E 1441 CTW and MAF 1963.7 14.8--28.8 Shako 617 51,195 1 48,089.63 ha 7°33′42.37″ N 35°39′11.83″ E 1800 CTW and MAF 1906.9 11.4--22.4 *Note*. Vegetation type: DAF: dry evergreen Afromontane forest and grassland complex; CTW: Combretum-Terminalia woodland and wooded grassland; FLV/MFS: freshwater marshes and swamps, floodplains, and lake shore vegetation; ACB: Acacia-Commiphora woodland and bushland proper; MAF: moist evergreen Afromontane forest. m.a.s.l: meter above sea level; mm: millimeter; °C: degree Celsius; km^2^: kilometer square. ###### Symptoms that are used by traditional healers to diagnose cancer. Cancer types Reported symptoms Number of traditional healers ------------------------------------------------------------------------------ ----------------------------------------------------------------------------------------- ------------------------------- Skin Lumpy growth 32 Spreading pea-sized growth 1 Ulcerative growth and oozing blood 1 Breast Lumpy growth 17 Lumpy growth on one breast and progressive weight loss 1 Ulcerative wounds on breast 5 Ulcerative wounds on breast and swelling on armpit and neck 1 The patient was receiving anticancer treatment for breast cancer in hospital 12 Cervical Foul-smelling bloody vaginal discharge, pain during sexual intercourse, and weight loss 1 Colon Chronic rectal bleeding and weight loss 1 Lung Coughing up blood 1 Throat Coughing and swelling on the neck 1 ###### List of candidate medicinal plants traditionally used for cancer treatment in the study areas. Voucher number Botanical name (family) Vernacular name Districts Growth form Habitat Parts used Preparation Type of cancer treated Application UV RFC CI ---------------- ---------------------------------------------------------- ------------------- ----------------- ------------- ------------ ------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------ ------------------ ------ ------- ------- Acanthaceae Bele-047 *Justicia schimperiana* (Hochst. ex Nees) T. Anderson *Kitkit* North Bench Shrub Wild Roots Fresh roots are crashed and boiled, and the cool decoction is drunk before meal Lung Oral 1 0.027 0.067 Bele-057 *Justicia schimperiana* (Hochst. ex Nees) T. Anderson *Gulbana* Doyo Gena Shrub Wild Leaves Fresh leaves are pounded, and the juice is applied on the affected area Skin Topical Aloaceae Bele-060 *Aloe* sp. *Gurta waqota* Doyo Gena Shrub Wild Leaves Fresh roots are crashed, and the sap is applied on the affected area Skin Topical 6 0.014 0.081 Amaranthaceae Bele-044 *Achyranthes aspera* L. *Koch ashite* Mizan Aman Herb Wild Leaves Leaves are roasted on metal plate, pounded into powder, mixed with animal butter, and smeared on the affected part Skin Topical 4 0.014 0.054 Apiaceae Bel-046 *Centella asiatica* (L.) Urb. *Gorongoch* Sheko Herb Wild Leaves Young leaves are crashed, and the sap sniffed Throat Intranasal 2 0.014 0.027 Bel-002 *Hydrocotyle mannii* Hook.f *Ye\'ti medhanit* North Bench Herb Wild Leaves Young leaves are crashed and applied on the affected area Skin Topical 1 0.014 0.014 Apocynaceae Bel-003 *Acokanthera schimperi* (A.DC.) Schweinf. *Merenz* Bahir Dar Zuria Shrub Wild Leaves Young leaves are crashed and applied Skin Topical 0.5 0.027 0.027 Bel-009 *Carissa spinarum* L. *Agam* Bahir Dar Zuria Shrub Wild Leaves Leaves are crashed and infused in cold water overnight and drunk before meal and applied on the affected area Skin Oral 1 0.027 0.027 Asclepiadaceae Bel-040 *Calotropis procera* (Aiton) Dryand. *Qumbo* Gewane Shrub Wild Roots Fresh roots are crashed, and the sap is applied on the affected area Breast Topical 3 0.014 0.027 Bel-036 *Pentarrhinum insipidum* E. Mey. *Barohula* Gewane Shrub Wild Roots Fresh roots are crashed, and the sap is applied on the affected area Breast and skin Topical 1 0.014 0.014 Bel-037 *Echidnopsis dammanniana* Sprenger *Mureli* Gewane Herb Wild Stem Stems are cut, and the sap is applied Skin Topical 2 0.014 0.027 Asphodelaceae Bel-020 *Kniphofia foliosa* Hochst. *Shushube* Bale Goba Shrub Wild Roots Dry roots are pounded, and the powder is mixed with honey Cervical and breast Oral 1 0.027 0.027 Asteraceae Bel-045 *Acmella caulirhiza* Delile *Kust asht* Mizan Aman Shrub Wild Leaves Young leaves are chewed by the healer and spit on Breast Topical 0.67 0.04 0.054 Bel-049 *Acmella caulirhiza* Delile *Bitisa* Wondo Genet Shrub Wild Leaves Fresh leaves are crashed and infused in cold water Breast Oral Bel-030 *Artemisia absinthium* L. *Natrara* Sodo Zuria Herb Wild Leaves Dried leaves are ground and macerated in coffee or tea Breast Oral 2 0.014 0.027 Bel-029 *Artemisia afra* Jacq. ex Willd. *Agufa* Doyo Gena Herb Wild Leaves Dried leaves are ground and macerated in coffee or tea Breast Oral 1 0.014 0.014 Bel-031 *Artemisia annua* L. *Artemisia* Sodo Zuria Tree Cultivated Leaves Dried leaves will be ground and decocted in hot water Breast Oral 1 0.014 0.014 Bel-021 *Cineraria abyssinica* Sch.Bip. ex A.Rich. Unknown Bale Robe Herb Wild Leaves Fresh leaves are pounded, and the sap is applied on the affected area Skin Topical 1.5 0.027 0.054 Bel-058 *Guizotia scabra* (Vis.) Chiov. *Sheshota* Doyo Gena Shrub Wild Leaves Fresh leaves are pounded, and the sap is applied on the affected area Skin Topical 1 0.014 0.014 Bel-034 *Solanecio gigas* (Vatke) C. Jeffrey *Arbaba* Doyo Gena Shrub Wild Leaves Fresh leaves are pounded and the sap is applied on the affected area Skin Topical 2 0.014 0.027 Bel-025 *Vernonia auriculifera* Hiern *Barawa* Doyo Gena Shrub Wild Leaves Fresh leaves are pounded, and the sap is applied on the affected area Skin Topical 1.33 0.041 0.081 Bel-056 *Vernonia auriculifera* Hiern *Reji* Wondo Genet Shrub Wild Leaves Fresh leaves are chewed by the healer and spit on Skin Topical Capparidaceae Bel-039 *Cleome brachycarpa* (Forssk.) Vahl ex DC. *Berbere* Gewane Herb Wild Leaves Fresh leaves are pounded, and the sap is applied on the affected area Breast and skin Topical 3 0.014 0.014 Commelinaceae Bel-026 *Commelina benghalensis* L. *Laluncha* Doyo Gena Herb Cultivated Roots Fresh roots are pounded, and the sap is applied on the affected area Skin Topical 2 0.014 0.027 Crassulaceae Bel-019 *Kalanchoe petitiana* A. Rich. *Anchura* Bale Goba Shrub Wild Leaves Fresh leaves are roasted for 2 minutes and applied on the affected area Breast and skin Topical 1.5 0.027 0.041 Euphorbiaceae Bel-012 *Croton macrostachyus* Hochst. ex Delile *Bisana* Filiklik Tree Wild Leaves or stem Fresh leaves or succulent stems are crashed, and the sap is applied on the affected area Breast and skin Topical 0.75 0.1 0.16 Bel-035 *Croton macrostachyus* Hochst. ex Delile *Besena* Doyo Gena Tree Wild Bark Dry bark is pounded, and the powder is applied on the affected area Skin Topical Bel-048 *Croton macrostachyus* Hochst. ex Delile *Masichoo* Wondo Genet Tree Wild Leaves Fresh leaves are crashed, macerated in cold water, and drunk Breast and skin Oral Bel-032 *Euphorbia schimperiana* Scheele *Gendalelata* Doyo Gena Shrub Wild Roots Fresh roots are pounded, and the sap is applied on the affected area Skin Topical 1 0.014 0.014 Fabaceae Bel-014 *Albizia schimperiana* Oliv. *Sessa* Filiklik Tree Wild Leaves The mixture of fresh leaves of *Albizia schimperiana* and *Carissa spinarum* is macerated in cold water for 2 days, and the macerated liquid is drunk Breast, intestinal, and skin Oral 4 0.014 0.014 Bel-004 *Calpurnia aurea* (Aiton) Benth. *Digita* Bahir Dar Zuria Shrub Wild Leaves or seeds Dry leaves or seeds are ground, macerated in cold water, and drunk Breast Oral 2 0.014 0.027 Bel-023 *Crotalaria agatiflora* Schweinf. Unknown Bale Goba Shrub Wild Seeds Dry seeds are ground, mixed with honey, and applied Skin Topical 1 0.014 0.014 Bel-028 *Crotalaria incana* L. *Chelke* Doyo Gena Shrub Wild Leaves Fresh leaves are crashed, and the sap is applied on the affected area Skin Topical 1 0.014 0.014 Bel-007 *Senna singueana* (Delile) Lock *Gefa* Bahir Dar Zuria Shrub Wild Leaves Fresh leaves are crashed, macerated, and drunk Skin Oral 2 0.014 0.027 Lamiaceae Bel-043 *Ajuga leucantha* Lukhoba *Tiks asht* North Bench Herb Wild Leaves Fresh leaves are crushed, and the sap is applied on the affected area Breast Topical 1 0.014 0.014 Bel-024 *Leonotis ocymifolia* (Burm.f.) Iwarsson *Armagusa* Bale Goba Herb Wild Leaves Fresh leaves are crashed, macerated overnight, and drunk Breast and skin Oral 3 0.014 0.014 Bel-054 *Ocimum gratissimum* L. *Mekedesisa* Wondo Genet Herb Wild Roots Fresh roots are crushed, boiled, and drunk Skin Oral 2 0.014 0.027 Bel-059 *Pycnostachys abyssinica* Fresen. *Tontona* Doyo Gena Herb Wild Leaves Fresh leaves are crushed, and the sap is applied on the affected area Skin Topical 2 0.014 0.027 Bel-042 *Salvia nilotica* Juss. ex Jacq. *Barnbanch* North Bench Shrub Wild Whole plant Dry plant parts are ground, mixed with honey, and applied Breast Topical 2 0.014 0.027 Bel-022 *Thymus schimperi* Ronniger *Tosigne* Bale Goba Herb Wild Leaves Dry leaves are decocted and drunk Breast Oral 2 0.014 0.027 Malvaceae Bel-051 *Sida schimperiana* Hochst. ex A. Rich. *Kotijebessa* Wondo Genet Shrub Wild Roots and leaves Fresh leaves and roots are crashed, macerated, and drunk Breast and skin Oral 4 0.014 0.027 Melianthaceae Bel-001 *Bersama abyssinica* Fresen. *Azamir* Bahir Dar Zuria Shrub Wild Bark Dry bark is ground, macerated, and drunk before meal Breast Oral 1 0.014 0.014 Moraceae Bel-008 *Dorstenia barnimiana* Schweinf. *Work Bemeda* Bahir Dar Zuria Herb Wild Roots Dry roots are ground, mixed with water and honey, and drunk, or dry roots are ground, mixed with honey, and applied on the affected area Breast Oral or topical 0.6 0.068 0.12 Myrtaceae Bel-006 *Syzygium guineense* (Willd.) DC. *Dokima* Bahir Dar Zuria Tree Wild Leaves and roots Dry leaves and roots of *Syzygium guineense* and dry leaves of *Osyris quadripartita* are ground, mixed, decocted, and drunk Skin Oral 2 0.014 0.027 Oxalidaceae Bel-052 *Oxalis corniculata* L. *Qinta* Wondo Genet Herb Wild Leaves and roots Fresh leaves and roots are crashed and applied with a bandage Breast Topical 2 0.014 0.027 Polygonaceae Bel-018 *Rumex nervosus* Vahl *Emboacho* Filiklik Shrub Wild Roots Dry roots are ground, macerated, and drunk Skin Oral 3 0.014 0.041 Bel-033 *Rumex nepalensis* Spreng. *Goecho* Doyo Gena Herb Wild Roots Dry roots are ground and taken with food Colon Oral 1.5 0.027 0.041 Bel-053 *Rumex nepalensis* Spreng. *Sharibicho* Wondo Genet Herb Wild Bark Fresh bark is crashed and squeezed, and the sap is applied Skin Topical Ranunculaceae Bel-010 *Clematis simensis* Fresen. *Yeazo Hareg* Bahir Dar Zuria Climber Wild Leaves Fresh roots of *Dorstenia barnimiana* mixed with fresh leaves of *Clematis simensis*, pounded, and applied Breast Topical 0.67 0.041 0.054 Rosaceae Bel-011 *Prunus africana* (Hook.f.) Kalkman *Tikur enchet* Bahir Dar Zuria Tree Wild Bark Dry bark is ground, decocted, and drunk Breast and skin Oral 3 0.014 0.014 Rutaceae Bel-016 *Clausena anisata* (Willd.) Hook.f. ex Benth. *Limich* Filiklik Shrub Wild Leaves Dry leaves are ground, mixed with honey, and eaten Breast Oral 2 0.014 0.027 Santalaceae Bel-013 *Osyris quadripartita* Salzm. ex Decne. *Keret* Filiklik Shrub Wild Leaves Dry leaves are ground, decocted, and drunk Breast Oral 2 0.027 0.027 Sapindaceae Bel-005 *Dodonaea viscosa* subsp. *angustifolia* (L.f.) J.G.West *Kitkita* Bahir Dar Zuria Tree Wild Roots Dry roots are ground, mixed with honey, and applied or dry roots are ground, decocted, and drunk Breast, skin and cervical Topical or oral 1 0.014 0.041 Simaroubaceae Bel-017 *Brucea antidysenterica* J.F.Mill. *Abalo* Filiklik Tree Wild Leaves Dry leaves are ground, pasted with cold water, and applied Skin Topical 4 0.014 0.054 Solanaceae Bel-027 *Discopodium penninervium* Hochst. *Chechanga* Doyo Gena Shrub Wild Leaves Fresh leaves are crashed and applied Skin Topical 1 0.014 0.014 Thymelaeaceae Bel-055 *Gnidia involucrata* Steud. ex A.Rich. *Bito* Bahir Dar Zuria Herb Wild Roots Dry roots are ground, mixed with honey, and eaten Breast Oral 0.5 0.027 0.027 Verbenaceae Bel-050 *Lantana trifolia* L. *Hanshebello* Wondo Genet Shrub Wild Leaves Fresh leaves are ground, macerated in cold spring water, and drunk Breast and skin Oral 2 0.014 0.014 Bel-015 *Lippia adoensis* Hochst. *Kessie* Filiklik Shrub Wild Leaves Dry leaves are ground, macerated in cold water, and drunk Skin Oral 2 0.014 0.027 Vitaceae Bel-038 *Cyphostemma serpens* (Hochst. ex A.Rich.) Desc. *Eiriti* Gewane Climber Wild Roots Dry roots are ground, pasted with honey and eaten, and applied Skin Oral and topical 1 0.014 0.014 UV = use value; RFC = relative frequency of citation; CI = cultural importance index. ###### Cross-reference of cancer treatment candidate plant species collected from the study areas with the published literature. ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Botanical name (family) Biological activity/chemical constituents Illnesses/symptoms claimed to be treated traditionally ------------------------------------------------------------------------ ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- *Justicia schimperiana* (Hochst. ex Nees) T. Anderson (Acanthaceae) Saponins, alkaloids, terpenoids and flavonoids \[[@B32]\]\ Wound \[[@B15], [@B21]\]; rabies \[[@B15], [@B18]--[@B20], [@B36]--[@B39]\]; jaundice \[[@B15], [@B16], [@B21], [@B23], [@B38], [@B40]\]; gonorrhea \[[@B17], [@B36], [@B39]\]; liver cirrhosis \[[@B18], [@B26]\]; seizure \[[@B19], [@B41]\]; stomach ache \[[@B15], [@B25], [@B38]\]; helminths \[[@B15], [@B42], [@B43]\]; skin burn/lesion \[[@B23], [@B44]\]; arthritis \[[@B21], [@B23]\]; hepatitis \[[@B45], [@B46]\]; evil eye \[[@B15], [@B46]\]; dysentery \[[@B15], [@B21]\]; malaria \[[@B36], [@B39]\]; common cold, asthma, and headache \[[@B36], [@B39], [@B47]\]; otitis \[[@B48]\]; toothache \[[@B49]\]; and rheumatism \[[@B50]\] *In vitro* cytotoxicity \[[@B33]\]; *in vitro* antioxidant activity on DPPH assay \[[@B34]\]; *in vivo* suppression of parasitaemia on *Plasmodium berghei*-infected mice in the 4-day suppressive test \[[@B32]\]; and *in vivo* hepatoprotective activity in mice intoxicated with CCL~4~ \[[@B35]\] *Aloe* sp. (Aloaceae) Anthrones and chromones \[[@B51]\], pyrones, coumarins, alkaloids, glycoproteins, naphthalenes, and flavonoids \[[@B52]\]\ Wound \[[@B21], [@B55]\]; eye disease \[[@B21], [@B46], [@B48], [@B56]\]; snake bite \[[@B21], [@B48], [@B56]\]; malaria \[[@B20], [@B21], [@B44], [@B48], [@B54]\]; easing labour \[[@B44]\]; tropical ulcer, colon cleaner, and gallstone \[[@B48]\]; amoeba, abdominal pain, impotence, and urine retention \[[@B21]\]; dandruff \[[@B46], [@B56]\], hemorrhoids and hepatitis B \[[@B46]\]; ascariasis \[[@B21]\]; diabetes \[[@B54]\]; asthma \[[@B55]\]; foot strain \[[@B57], [@B58]\]; wart and anthrax \[[@B20]\]; external injury \[[@B59]\]; and liver swelling, splenomegaly, and skin inflammation \[[@B56]\] 7‐O‐methylaloeresin showed *in vitro* antioxidant activity in DPPH assay \[[@B51]\], and methanol and ethanol extract showed *in vivo* parasitaemia suppression on *Plasmodium berghei*-infected mice in the 4-day suppressive test \[[@B53], [@B54]\] *Achyranthes aspera* L. (Amaranthaceae) Phytosteroids, polyphenols, and saponins \[[@B60]\]\ Bleeding \[[@B21], [@B24], [@B26], [@B62]--[@B64]\]; retained placenta \[[@B21], [@B62]\]; stomach ache and external swelling \[[@B17]\]; rhesus factor incompatibility in pregnancy \[[@B40], [@B55]\]; epistaxis \[[@B19]\]; hepatitis and evil eye \[[@B24]\]; tonsillitis \[[@B21], [@B57]\]; snake bite and paralysis \[[@B21]\]; dysentery \[[@B59]\]; herpes zoster \[[@B26]\]; anthrax \[[@B21], [@B49]\]; nasal infection and ophthalmic infection \[[@B64]\]; excessive menstruation and tape worm infection \[[@B15]\]; and gonorrhea \[[@B65]\] Methanol extracts have showed *in vivo* wound healing activity \[[@B61]\] *Centella asiatica* (L.) Urb. (Apiaceae) Terpenoids (triterpenes, asiaticoside, centelloside, madecassoside, brahmoside, brahminoside (saponin glycosides), asiaticentoic acid, centellic acid, centoic acid, madecassic acid, terminolic acid, betulic acid, *β*-caryophyllene, trans-*β*-farnesene and germacrene D (sesquiterpenes), *α*-pinene, and *β*-pinene \[[@B66], [@B67]\]\ Genital infection and lymphadenitis \[[@B63]\]; topical swelling \[[@B26], [@B70]\]; gastritis, headache, and evil eye \[[@B70]\]; bleeding \[[@B40]\]; wound \[[@B24]\]; abdominal ache \[[@B71]\]; meningitis \[[@B72]\]; and tinea corporis \[[@B47]\] Methanol extract inhibited the proliferation of human gastric adenocarcinoma (MK-1), human uterine carcinoma (HeLa), and murine melanoma (B16F10) cells *in vitro* \[[@B68]\]; aqueous extracts induced apoptosis in colonic crypts and exerted chemopreventive effect on colon tumorigenesis in male F344 rats \[[@B69]\] *Hydrocotyle mannii* Hook.f (Apiaceae) No previous reports Eye infection \[[@B63]\] and cataract \[[@B72]\] *Acokanthera schimperi* (A.DC.) Schweinf. (Apocynaceae) *In vitro* cytotoxicity \[[@B73]\]; *in vitro* antiviral activity against coxsackie B3, influenza A, and herpes simplex type1 virus \[[@B74]\]; *in vitro* antimicrobial activity against *Staphylococcus aureus*, *Pseudomonas aeruginosa, Trichophyton mentagrophytes* \[[@B75]\]; and *in vivo* parasitaemia suppression in *Plasmodium berghei*-infected mice \[[@B76]\] Wound \[[@B16], [@B44], [@B77], [@B78]\]; hepatitis \[[@B15], [@B16], [@B22], [@B44]\]; gonorrhea \[[@B19], [@B25]\]; evil eye \[[@B62]\]; bone fracture \[[@B24]\]; hemorrhoids \[[@B44]\]; scabies \[[@B21]\]; malaria and tonsillitis \[[@B48], [@B56]\]; psychiatric disease \[[@B55]\]; and skin diseases \[[@B65]\] *Carissa spinarum* L. (Apocynaceae) *In vitro* antioxidant activity on DPPH assay and *in* antiproliferative activity \[[@B79]\] Throat cancer \[[@B23], [@B80]\]; evil eye \[[@B16], [@B21], [@B24], [@B49], [@B62], [@B70], [@B72], [@B81]\]; snake bite \[[@B23], [@B80]\]; gonorrhea \[[@B20], [@B65]\]; stomach ache \[[@B20], [@B70]\]; impotence and headache \[[@B20]\]; tonsillitis \[[@B17], [@B56], [@B70]\]; wound and febrile illness \[[@B16]\]; bleeding after delivery \[[@B44]\]; muscle cramps \[[@B49]\]; toothache \[[@B47]\]; and premature ejaculation \[[@B56]\] *Calotropis procera* (Aiton) Dryand. (Asclepiadaceae) Latex contains phytochemicals such as alkaloids, sterols, fatty acids, starches, sugars, oils, tannins, resins, and gums, and enzymatic proteins such as proteases, chitenases, lipases, peptidases, esterase, peroxidases, papain, hevein, and lectins \[[@B82]\]\ Wound \[[@B16], [@B21], [@B81]\]; hemorrhoids \[[@B16], [@B19], [@B44]\]; wart \[[@B16], [@B57]\]; snake bite \[[@B23], [@B87]\]; kidney stone, tuberculosis, and scabies \[[@B16]\]; swelling \[[@B58]\]; skin rash \[[@B21], [@B49]\]; tinea capitis \[[@B21]\] *In vivo* hepatoprotective \[[@B83]\]; hypoglycemic effect \[[@B84]\]; strong anti-implantation (antifertility) \[[@B85]\]; crude latex showed antioxidant and antiapoptotic activities against the toxicity of 4-nonylphenol \[[@B86]\] *Pentarrhinum insipidum* E. Mey. (Asclepiadaceae) No previous reports *Echidnopsis dammanniana* Sprenger (Asclepiadaceae) No previous reports Snake bite \[[@B56]\] *Kniphofia foliosa* Hochst. (Asphodelaceae) 2-Acetyl-1-hydroxy-8-methoxy-3-methylnaphthalene, 10-(chrysophanol-7′-yl)-10-(ξ)-hydroxychrysophanol-9-anthrone, chryslandicin, knipholone, and chrysophanol \[[@B88]\]\ No previous reports 10-(Chrysophanol-7′-yl)-10-(ξ)- hydroxychrysophanol-9-anthrone showed *in vitro* antiplasmodial activity against chloroquine-sensitive 3D7 strain of *Plasmodium falciparum* and knipholone selectively inhibited leukotriene metabolism in *in vitro* a human blood assay \[[@B88]\]; knipholone anthrone showed *in vitro* cytotoxicity \[[@B89]\] and antioxidant activity on DPPH assay \[[@B90]\] *Acmella caulirhiza* Delile (Asteraceae) Unsaturated alkylamides like spilanthol and N-isobutylnona-2E,4E-dien-8ynamide \[[@B91]\]\ Swelling \[[@B15]\]; tonsillitis \[[@B20], [@B63]\]; and toothache \[[@B40], [@B87]\] *In vitro* antiplasmodial activity \[[@B92]\] *Artemisia absinthium* L. (Asteraceae) Camphor, davanone, ethyl (E)-cinnamate, (E)-nerolidol, and chamazulene \[[@B93]\]\ Hypertension, stomach ache, severe abdominal cramp \[[@B18]\] and sour throat \[[@B40]\] Essential oils showed *in vitro* antiparasitic effects against promastigote and axenic amastigote forms of *Leishmania donovani* and *Leishmania aethiopica* and *in vitro* cytotoxicity on THP-1 (human leukaemia) cell lines \[[@B93]\]; and *in vitro* cytotoxicity on human leukaemia cell lines \[[@B94]\] *Artemisia afra* Jacq. ex Willd. (Asteraceae) Epoxylinalol and dihydrocostunolide \[[@B94]\]; camphor, davanone, bornyl acetate, 4-terpineol, and chamazulene \[[@B95]\]\ Stomach ache \[[@B18], [@B42]\]; evil eye \[[@B16], [@B17], [@B62]\]; headache \[[@B42], [@B77]\]; eye disease, tinea capitis infection, hematuria, and stabbing pain \[[@B77]\]; antifertility agent \[[@B33]\]; malaria \[[@B42], [@B62]\]; ascariasis \[[@B18]\]; epilepsy and febrile illness \[[@B46], [@B65]\] *In vitro* cytotoxicity on human leukaemia cell lines \[[@B73]\]; and *in vitro* antioxidant effect on DPPH assay \[[@B95]\] *Artemisia annua* L. (Asteraceae) *In vitro* inhibition of immune mediators of angiogenesis \[[@B96]\]; the sesquiterpene (Z)-7-acetoxy-methyl-11-methyl-3-methylene-dodeca-1,6,10-triene showed moderate cytotoxic activities against the human tumor cell lines of HO8910 (ovary), 95-D (lung), QGY (liver), and HeLa (cervix) by MTT assay and induced apoptosis on 95-D tumor cells \[[@B97]\]; artemisinin and quercetagetin 6,7,3′,4′-tetramethyl ether showed significant cytotoxicity against P-388, A-549, HT-29, MCF-7, and KB tumor cells \[[@B98]\] No previous reports *Cineraria abyssinica* Sch.Bip. ex A.Rich. (Asteraceae) *In vitro* radical scavenging activity on DPPH assay \[[@B99]\]; flavonoidal glycoside (rutin) showed *in vitro* antibacterial activity \[[@B100]\] No previous reports *Guizotia scabra* (Vis.) Chiov. (Asteraceae) *In vitro* cytotoxicity on human leukaemia cell lines \[[@B73]\], and *in vitro* antiviral activity \[[@B101]\] Wound \[[@B20]\]; epilepsy \[[@B40]\]; and ectoparasite infestation \[[@B47]\] *Solanecio gigas* (Vatke) C. Jeffrey (Asteraceae) *In vitro* antiviral activity against human immunodeficiency virus type 1 and type 2 cytotoxicity on human T-lymphocytic MT-4 cell lines \[[@B102]\] Skin diseases \[[@B62]\]; retained placenta \[[@B40]\]; hepatitis \[[@B64]\]; evil eye \[[@B15]\] *Vernonia auriculifera* Hiern (Asteraceae) Tannins, flavonoids, terpenoids, and saponins \[[@B103]\] Toothache \[[@B72]\]; snake bite \[[@B42]\]; skin cut \[[@B47]\] *Cleome brachycarpa* (Forssk.) Vahl ex DC. (Capparidaceae) No previous reports *Commelina benghalensis* L. (Commelinaceae) Phlobatannins, carbohydrates, tannins, glycosides, volatile oils, resins, balsams, flavonoids, and saponins \[[@B104]\]\ Helminths \[[@B65]\]; skin infection \[[@B72]\] Ethanol extract showed *in vivo* sedative and anxiolytic activity \[[@B105]\] *Kalanchoe petitiana* A. Rich. (Crassulaceae) Polyphenols, alkaloids, flavonoids, tannins, saponins, and steroids \[[@B106]\]\ Breast and skin cancer \[[@B107]\]; swelling \[[@B40], [@B77]\]; tapeworm infection, trachoma, and syphilis \[[@B77]\]; lymphadenopathy and evil eye \[[@B22]\]; sore muscles \[[@B108]\]; itching skin \[[@B63]\]; and bone fracture \[[@B23]\] *In vitro* antimicrobial activity against *Escherichia coli*, *Pseudomonas aeruginosa*, and *Staphylococcus aureus* \[[@B75]\]; and *in vivo* wound healing activity \[[@B106]\] *Croton macrostachyus* Hochst. ex Delile (Euphorbiaceae) Ethanol extract showed *in vitro* antioxidant activity on DPPH assay \[[@B79]\] Tumor, rabies, and wart \[[@B24]\]; skin cancer and wound \[[@B17]\]; gonorrhea \[[@B20], [@B23], [@B62]\]; headache \[[@B18], [@B109]\]; snake bite \[[@B18], [@B72]\]; malaria \[[@B16], [@B18]--[@B20], [@B110]\]; helminths \[[@B18], [@B111]\]; tinea nigra \[[@B40]\]; ringworm \[[@B17], [@B62]\]; tinea versicolor \[[@B16], [@B25]\]; heart failure \[[@B62]\]; bleeding \[[@B18], [@B24]\]; hepatitis \[[@B16], [@B18], [@B24]\]; stomach ache \[[@B16], [@B18], [@B23]\]; diarrhea \[[@B16], [@B18]\]; lymph adenitis and rheumatism \[[@B18]\]; bloat, scabies, and urine retention \[[@B16]\]; retained placenta and leprosy \[[@B19]\] *Euphorbia schimperiana* Scheele (Euphorbiaceae) *In vitro* cytotoxic effect against breast cancer (MCF7), hepatocellular carcinoma (HEPG2), and cervix cancer (HELA) cells \[[@B112]\] Syphilis \[[@B108]\] *Albizia schimperiana* Oliv. (Fabaceae) *In vitro* cytotoxicity on human leukaemia cells \[[@B73]\] Evil eye \[[@B20]\]; kidney infection and liver cirrhosis \[[@B18]\] *Calpurnia aurea* (Aiton) Benth. (Fabaceae) 3*β*,4*α*,13*α*-Trihydroxylupanine, calpaurine, lupinine, and epilupinine calpurmenine and calpurmenine pyrrolecarboxylic acid ester, 13-hydroxylupanine, its tiglate and pyrrolecarboxylic acid esters (calpumine), virgiline and virgiline pyrrolecarboxylic acid ester \[[@B113]\]; 4*β*-hydroxy-13*α*-*O*-(2′-pyrrolylcarbonyl)-lupanine (digittine) and 4*β*,13*α*-dihydroxylupanine \[[@B114]\]; alkaloids, tannins, flavonoids, and saponins \[[@B35]\]\ Tumor \[[@B22], [@B26], [@B80]\]; stomach ache \[[@B21], [@B62], [@B70], [@B81]\]; wound and skin infection \[[@B62]\]; Gonorrhoea and syphilis \[[@B16]\], amoebiasis \[[@B16], [@B80]\]; ascariasis and gastric ulcer \[[@B23]\]; diarrhea \[[@B21], [@B38], [@B70]\]; scabies and pubic hair louse \[[@B40]\]; diabetes mellitus and hypertension \[[@B19]\]; herpes zoster, hemorrhoids and tinea capitis \[[@B21]\]; and swelling and tuberculosis \[[@B58]\] Methanol extract showed *in vitro* antimicrobial activity against *Staphylococcus aureus*, *Escherichia coli*, and *Pseudomonas aeruginosa* \[[@B75]\] and type 1 and type 2 human immunodeficiency virus and showed cytotoxicity on human T-lymphocytic MT-4 cell lines \[[@B102]\]; methanol and dichloromethane crude extracts showed *in vitro* cytotoxicity on human leukaemia cells \[[@B73]\]; and ethanol extracts showed *in vitro* antioxidant activity on DPPH assay \[[@B79]\] *Crotalaria agatiflora* Schweinf. (Fabaceae) Methanol and dichloromethane crude extracts showed *in vitro* cytotoxicity on human leukaemia cells \[[@B73]\] No previous reports *Crotalaria incana* L. (Fabaceae) Dihydrosenecionine isomer, nemorensine isomer, integerrimine and anacrotine \[[@B115]\]\ Methanol and dichloromethane crude extracts showed *in vitro* cytotoxicity on human leukaemia cell lines \[[@B73]\] *Senna singueana* (Delile) Lock (Fabaceae) Methanol extracts showed *in vitro* antioxidant activity on DPPH assay \[[@B116]\] Stomach ache \[[@B58], [@B62], [@B70]\]; wound and swellings \[[@B62]\]; teeth infection and sprain \[[@B58]\] *Ajuga leucantha* Lukhoba (Lamiaceae) No previous reports Diarrhea \[[@B70]\] *Leonotis ocymifolia* (Burm.f.) Iwarsson (Lamiaceae) Methanol and dichloromethane crude extracts showed *in vitro* cytotoxicity on human leukaemia cells \[[@B73]\] Ascariasis \[[@B62]\], febrile illness \[[@B16], [@B62]\]; eye disease \[[@B16]\]; headache and neck ulcer \[[@B55]\]; and snake bite \[[@B15]\] *Ocimum gratissimum* L. (Lamiaceae) Essential oil contains constitutes *γ*-terpinene, *β*-phellandrene, limonene, and thymol and showed *in vivo* antiplasmodial activity against *Plasmodium berghei* infection \[[@B117]\] Allergy reaction \[[@B18], [@B20]\]; rheumatism, headache and eye disease \[[@B18]\]; febrile illness and general malaise \[[@B40]\]; sun stroke \[[@B24]\]; malaria \[[@B44]\] *Pycnostachys abyssinica* Fresen. (Lamiaceae) No previous reports Eye disease \[[@B18], [@B47]\]; ascariasis and wound \[[@B18]\]; diarrhea, stomach ache, amoebiasis, stomach bloating, and food poisoning \[[@B70]\]; headache \[[@B63]\] *Salvia nilotica* Juss. ex Jacq. (Lamiaceae) Essential oil contains germacrene D, guaiol, and *trans*-caryophyllene as major constituents and showed activity against both Gram-positive and Gram-negative pathogenic bacteria; the oil also showed *in vitro* antioxidant activity on DPPH assay \[[@B118]\] Tonsillitis and constipation \[[@B62]\]; herpes simplex \[[@B18], [@B38]\]; wound \[[@B40]\]; lymphadenitis \[[@B63]\]; and hemorrhoids and diarrhea \[[@B65]\] *Thymus schimperi* Ronniger (Lamiaceae) Phenol and flavonoid compounds, and aqueous methanol extract showed *in vitro* radical scavenging ability, iron reducing power, and total antioxidant capacity \[[@B119]\] Diabetes \[[@B62]\]; hypertension \[[@B18], [@B40]\]; tonsillitis \[[@B18]\]; toothache \[[@B18], [@B21]\]; abdominal pain \[[@B21]\]; and cough \[[@B38], [@B55]\] *Sida schimperiana* Hochst. ex A. Rich. (Malvaceae) No previous reports "Shotelaye" (hydrops fetalis) \[[@B21], [@B22]\]; cough and fever \[[@B62]\]; diarrhea \[[@B18]\]; wound \[[@B25], [@B62]\]; bleeding and evil eye \[[@B24]\]; glandular disease and rabies \[[@B40]\]; amoebic dysentery, and liver disease \[[@B65]\]; paralysis \[[@B21]\]; epilepsy \[[@B43]\] *Bersama abyssinica* Fresen. (Melianthaceae) Flavonol glycosides isoquercetrin, hyperoside, quercetin-3-*O*-arabinopyranoside, kaempferol-3-O-arabinopyranoside, xanthone glycoside, mangiferin \[[@B115]\]\ Tumor, dysentery and roundworms \[[@B107], [@B109]\]; ascariasis \[[@B15], [@B38], [@B81], [@B109]\]; wound \[[@B20]\]; stomach ache \[[@B17]\]; snake bite and liver diseases \[[@B70]\]; tonsillitis \[[@B72]\]; bronchitis and febrile illness \[[@B42], [@B43]\] Ethanol water extracts showed *in vitro* antioxidant activity on DPPH assay and antiproliferative activity on human liver carcinoma cell line and normal human fetal lung cells \[[@B79]\]; methanol extract showed *in vitro* antioxidant activity on DPPH assay \[[@B115]\], and antiviral activity against type 1 human immunodeficiency virus \[[@B102]\] *Dorstenia barnimiana* Schweinf. (Moraceae) Phytochemical screening showed the presence of coumarins \[[@B34]\] Cancer \[[@B26]\]; hepatitis, syphilis and rabies \[[@B25], [@B26]\]; skin cancer, dysentery, wart and fever \[[@B25]\]; pulmonary tuberculosis, leprosy, and stomach illness \[[@B22]\] *Syzygium guineense* (Willd.) DC. (Myrtaceae) Methanol and dichloromethane crude extracts showed *in vitro* cytotoxicity on human leukaemia cells \[[@B73]\] and antimicrobial activity \[[@B120]\] Stomach ache \[[@B17]--[@B19], [@B23]\]; diarrhea \[[@B15], [@B18], [@B19], [@B24]\], kidney infection, liver cirrhosis, and tonsillitis \[[@B18]\]; syphilis \[[@B23], [@B80]\]; malaria, hemorrhoid, internal worms, snake bite, and gonorrhea \[[@B65]\] *Oxalis corniculata* L. (Oxalidaceae) *In vivo* antitumor activity against Ehrlich ascites carcinoma on mice \[[@B121]\] Wound \[[@B17]\]; arthritis \[[@B63]\]; tape worm infection \[[@B21]\] *Rumex nervosus* Vahl (Polygonaceae) Alkaloids, flavonoids, terpenoids, tannins, glycosides, and volatile oils \[[@B122]\] Breast cancer, gastritis, and snake bite \[[@B16]\]; wart \[[@B15], [@B22]\]; hepatitis \[[@B49], [@B55]\]; skin rash \[[@B16], [@B21]\]; bleeding \[[@B15], [@B40], [@B81], [@B109]\]; wound \[[@B40], [@B49], [@B55], [@B62], [@B109], [@B110]\]; scabies and acne vulgaris \[[@B62]\]; ascariasis and herpes simplex \[[@B21]\]; stomach ache and dysentery \[[@B22]\]; diarrhea \[[@B49]\]; eye problems and round worm \[[@B55]\] *Rumex nepalensis* Spreng. (Polygonaceae) Anthraquinones, naphthalenes, tannins, stilbenoids \[[@B123]\]\ Wound, ascariasis, abdominal bleeding, gastric ulcer, and hemorrhage \[[@B23], [@B80]\]; gastritis \[[@B18]\]; stomach problems \[[@B108]\]; leishmaniasis \[[@B25]\]; abdominal cramp and ear infection \[[@B63]\]; tonsillitis \[[@B18], [@B25]\] Ethanol water extracts showed *in vitro* antiproliferative activity on human liver carcinoma cell line and on normal human fetal lung cells and antioxidant activity on DPPH assay \[[@B79]\], and methanol and dichloromethane crude extracts showed *in vitro* cytotoxicity on human leukaemia cells \[[@B73]\] *Clematis simensis* Fresen. (Ranunculaceae) Triterpenoids, saponins, alkaloids, polyphenols, and unsaturated sterols \[[@B120]\]\ Cancer and hemorrhoid \[[@B15]\]; wart and evil eye \[[@B24], [@B40]\]; wound \[[@B15], [@B24], [@B40], [@B63], [@B81]\]; tonsillitis \[[@B62]\]; eye infection \[[@B63]\]; leg swelling, malaria, and mental illness \[[@B49]\]; stomach ache \[[@B47]\] *In vivo* anti-inflammatory and antinociceptive activities \[[@B124]\] *Prunus africana* (Hook.f.) Kalkman (Rosaceae) No previous reports Benign prostatic hyperplasia and prostate gland hypertrophy \[[@B20]\]; cancer, respiratory disorders, bad breathe, diarrhea, gonorrhea, tuberculosis, and ear problems \[[@B22]\]; swelling \[[@B40]\]; wounds \[[@B19], [@B22]\]; tonsillitis \[[@B23], [@B80]\] *Clausena anisata* (Willd.) Hook.f. ex Benth. (Rutaceae) Carbazole alkaloids, peptide derivatives, sitosterol, and stigmasterol \[[@B125]\]\ Skin irritation \[[@B20]\]; toothache \[[@B40]\]; ascariasis \[[@B19]\]; evil eye \[[@B24], [@B25], [@B63]\] Methanol and dichloromethane crude extracts showed *in vitro* cytotoxicity on human leukaemia cells \[[@B73]\] *Osyris quadripartita* Salzm. ex Decne. (Santalaceae) Alkaloids, phenols, terpenoids, tannins, saponins, and flavonoids \[[@B126]\]\ Cancer \[[@B62]\]; anaphylactic shock, evil eye, and epilepsy \[[@B18]\]; eczema \[[@B40]\]; toothache \[[@B46]\] Methanol extracts showed *in vitro* antimicrobial activity against *Escherichia coli*, *Pseudomonas aeruginosa, Staphylococcus aureus*, *Candida albicans*, and *Trichophyton mentagrophytes* \[[@B11]\]; *in vitro* inhibition of NO production and cytotoxicity against MCF-7 and NCI-H460 cell lines \[[@B127]\] *Dodonaea viscosa* subsp. *angustifolia* (L.f.) J.G.West (Sapindaceae) Alkaloids, terpenoids, saponins, tannins, sugars, phenolics, and flavonoids \[[@B128]\]\ Malaria \[[@B57]\] Methanol extracts showed *in vivo* nonsensitizer effect in mice using the mouse ear swelling test method \[[@B129]\], *in vitro* antiviral effect against type 1 human immunodeficiency virus \[[@B102]\], and *in vitro* free radical scavenging activity on DPPH assay \[[@B128]\] *Brucea antidysenterica* J.F.Mill. (Simaroubaceae) Flavonoids, amino acids, and vitamin C \[[@B130]\]\ Cancer/tumor \[[@B107]\]; wart \[[@B24]\]; rabies \[[@B18], [@B62]\]; leprosy \[[@B62]\] *In vitro* antiplasmodial activity against *Plasmodium berghei* infection \[[@B131]\] *Discopodium penninervium* Hochst. (Solanaceae) 5*α*,17*β*-Dihydroxy-6*α*,7*α*-epoxy-1-oxowitha-2,24-dienolide, withanone, and withanolide A \[[@B132]\], 5,6-epoxy-16-oxygenated withanolides, jaborosalactone-L, and 17-epiacnistin-A \[[@B133], [@B134]\]; 6*α*,7*α*-epoxy-1-oxo-5*α*,12*α*,17*α*-trihydroxywitha-2,24-dienolide and a coloratane sesquiterpene, 7*α*,11*α*-dihydroxy-4(13),8-coloratadien-12,11-olide, withanone, 5*α*,17*β*-dihydroxy-6*α*,7*α*-epoxy-1-oxowitha-2,24-dienolide, 7*α*,11*α*-dihydroxy-8-drimen-12,11-olide, withasomnine, and (E,Z)-9-hydroxyoctadeca-10,12-dienoic acid \[[@B135]\]\ Skin detoxification \[[@B62]\]; and liver disease \[[@B70]\] Jaborosalactone-L showed cytotoxicity only to the murine macrophage cell line, RAW 264.7, but the 16*α*-oxygenated withanolides exhibited cytotoxicity to both human (COR-L23 and ECV 304) and murine (L929 and RAW 264.7) carcinoma cell lines with IC~50~ values ranging from 1.2 to 150 *μ*M \[[@B136]\]. 6*α*,7*α*-Epoxy-1-oxo-5*α*,12*α*,17*α*-trihydroxy-witha-2,24-dienolide inhibited COX-2 and LTB4 formation; 7*α*,11*α*-dihydroxy-4(13),8-coloratadien-12,11-olide and withasomnine inhibited LTB~4~ biosynthesis but showed minor inhibition of COX-1 and COX-2 \[[@B135]\] *Gnidia involucrata* Steud. ex A.Rich. (Thymelaeaceae) Flavonoids and glycosides \[[@B137]\] Ascariasis, evil eye, anthrax, intestinal helminths, and gland swelling \[[@B18]\] *Lantana trifolia* L. (Verbenaceae) Flavone glycosides (scutellarein-*7-O*-*β*-D-apiofuranoside and apigenin-*7-O*-*β*-D-apiofuranosyl-(1⟶2)-*β*-D-apiofuranoside) and the flavone celtidifoline (5,6,40,50-tetrahydroxy-7,30-dimethoxyflavone) \[[@B138], [@B139]\] Headache \[[@B70]\]; malaria \[[@B71]\] *Lippia adoensis* Hochst. (Verbenaceae) Limonene, perillaldehyde, piperitenone, and 2-methyl-6-methylene-2,7-octadien-4-one \[[@B140]\], sesquiterpene hydrocarbon (germacrene D) \[[@B141]\]\ Eczema, fungal infections, common cold, and cough \[[@B62]\]; intestine swelling \[[@B18]\]; gastrointestinal disorder \[[@B40]\]; abdominal irritation and acute stomach illness \[[@B46]\] Methanol extract showed *in vitro* cytotoxicity on human leukaemia cell lines \[[@B73]\], and antimicrobial activity against *Staphylococcus aureus*, *Escherichia coli*, and *Pseudomonas aeruginosa* \[[@B75]\]; water extracts showed *in vivo* protection/relieve against acetic acid induced writhing in mice model \[[@B142]\] *Cyphostemma serpens* (Hochst. ex A.Rich.) Desc. (Vitaceae) No previous reports ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ###### Informants consensus factor for different ailment categories. No. Category No. of species \% of all species No. of use reports \% of all use reports ICF ----- --------------- ---------------- ------------------- -------------------- ----------------------- ------ 1 Skin 25 30.5 32 26.5 0.23 2 Breast 20 24.4 29 23.9 0.32 3 Cervical 1 1.22 1 0.83 0 4 Colon 1 1.22 1 0.83 0 5 Lung 1 1.22 1 0.83 0 6 Throat 1 1.22 1 0.83 0 7 Other disease 33 40.2 56 46.3 0.42 *Total* *82* ^*∗*^ *121* ^*∗*^Each plant species may be listed in several categories. ###### Summary of stats for relative frequency of citation (RFC) and cultural importance index (CI). -------------------------------------------------------------------- ------------ -------------------- --------- --------- Mean Standard deviation Minimum Maximum UV 1.8 1.1 0.5 6 RFC 0.02 0.015 0.014 0.1 CI 0.034 0.027 0.014 0.16 Association between RFC and CI by using Pearson correlation method UV RFC CI UV 1 RFC −0.36^*∗*^ 1 CI 0.003 0.858^*∗∗*^ 1 -------------------------------------------------------------------- ------------ -------------------- --------- --------- ^*∗*^Correlation is significant at 0.05 level. ^*∗∗*^Correlation is significant at 0.001 level. [^1]: Academic Editor: Sebastian Granica

{ "pile_set_name": "PubMed Central" }

Introduction {#s1} ============ Insomnia is both a symptom of depression and a separate disorder that may precede depression [@pone.0094674-Baglioni1]. Women sleep poorly during pregnancy and sleep even less in the postpartum period [@pone.0094674-Dorheim1], [@pone.0094674-Ko1]. At the same time, the perinatal period is a time of increased risk for depressive disorders [@pone.0094674-MunkOlsen1]. Insomnia during pregnancy may, therefore, influence the risk of postpartum depression. Insomnia is defined as repeated difficulty with sleep initiation, duration, consolidation, or quality that occurs despite adequate time and opportunity for sleep, and results in some form of daytime impairment [@pone.0094674-AASM1]. Among postpartum women, there may not be adequate time and opportunity for sleep; the newborn baby requires attention and will disturb maternal sleep in the first months after delivery [@pone.0094674-Insana1]. However, postnatal women may be exposed to similar practical challenges with sleep and infant care postpartum, but some may develop larger sleep problems than others or be more vulnerable to these changes. A review of 21 longitudinal studies found that individuals with insomnia had a twofold risk of developing depression and that residual insomnia after recovery from depression was associated with an increased risk of relapse [@pone.0094674-Baglioni2]. A large, population-based study from Norway found that not only did depression precede insomnia but also that insomnia preceded depression by many years [@pone.0094674-Sivertsen1]. Similar results have been found in Finland [@pone.0094674-Salo1]. Furthermore, treating insomnia in depressed patients leads to a better outcome of depression than treating depression on its own [@pone.0094674-Manber1]. Several studies suggest that insomnia may be co-morbid with depression, with a shared genetic component and a common final pathway [@pone.0094674-Zhang1]. Cross-sectional studies have found that insomnia and depressive symptoms are conditions associated with each other both before [@pone.0094674-Dorheim1], [@pone.0094674-Kizilirmak1] and after delivery [@pone.0094674-Dorheim2], [@pone.0094674-Marques1]. Depression during pregnancy is a risk factor for preterm birth [@pone.0094674-Dayan1] and may affect the physiology of the newborn\'s brain [@pone.0094674-Diego1]. Furthermore, maternal depression both during pregnancy and after delivery may affect the infant\'s cognitive and emotional development [@pone.0094674-Evans1], [@pone.0094674-Gerardin1], and these effects may persist through the childhood and teenage years [@pone.0094674-Murray1], [@pone.0094674-Murray2]. Prenatal depression is associated with sleep disturbances in the newborn (less deep sleep and more disorganized sleep) [@pone.0094674-Field1], and both prenatal and postnatal depression is associated with infant sleep problems one year after delivery [@pone.0094674-Gerardin1], [@pone.0094674-Pinheiro1]. Poor maternal sleep has been linked to perceived sadness of the infant (perceived by the mother) [@pone.0094674-Tikotzky1]. Poor sleep may also influence the relationship between the parents of the baby [@pone.0094674-Insana1], [@pone.0094674-Donnellan1], and a poor relationship with the partner is a risk factor for postpartum depression [@pone.0094674-OHara1]. Longitudinal studies of change in sleep and depression across childbirth are few. One study of 44 low-risk women suggested that change in subjective sleep, more than objective sleep, predicts postpartum mood in the first week postpartum [@pone.0094674-Bei1]. Marques et al., in a study of 382 women, found that insomnia in pregnancy was a predictor for postpartum depressive symptoms along with negative affect, but insomnia did not remain a risk factor when controlling for lifetime depression [@pone.0094674-Marques1]. Conversely, Okun et al., in a study of 51 perinatal women with a history of postpartum depression, found that women with *fewer* sleep complaints in late pregnancy had a more rapid recurrence (less than 4 weeks) of postpartum depression, whereas women with sleep problems during pregnancy developed postpartum depression later on [@pone.0094674-Okun1]. Change in sleep in the postpartum period may predict the development of depression more than changes in hormones do [@pone.0094674-Okun2]. More information is needed regarding the longitudinal relationship between insomnia and depressive symptoms in the perinatal period. Aims of the study {#s1a} ----------------- The aims of the study were to describe changes in sleep patterns, insomnia, and depressive symptoms before and after delivery, and to evaluate whether insomnia during pregnancy may predict postpartum depression. The study was done in a large, population-based cohort of pregnant women followed through childbirth until eight weeks after delivery, controlling for a range of known risk factors for perinatal depression. Material and Methods {#s2} ==================== Study population and design {#s2a} --------------------------- The Akershus Birth Cohort is a longitudinal questionnaire study targeted at all women giving birth at Akershus University Hospital, Norway. The hospital is located near the capital city of Oslo and serves a population of 350,000 from urban and rural areas. All women scheduled to give birth at the hospital were approached in gestational week 17, when they underwent routine fetal ultrasound. Women were included if they gave consent to participate and were able to complete a questionnaire in Norwegian. Recruitment took place from November 2008 until April 2010. Consenting women were handed a questionnaire at gestational week 17 and, thereafter, received a questionnaire by mail at week 32 of pregnancy and eight weeks after delivery. Information was also retrieved from the birth records at the maternity ward. [Figure 1](#pone-0094674-g001){ref-type="fig"} displays a flowchart of the recruitment and retention of study participants. The current paper examined insomnia and depressive symptoms as measured in week 32 of pregnancy and week eight postpartum. In total, 2,386 women returned the third questionnaire, making the participation rate 51.2% (n = 2386) of the 4,662 women who originally consented to participate, and 81.1% of the women who returned the second questionnaire. The final sample consisted of 2,088 women, as 298 women had missing data for the main outcome variables in one or both questionnaires (insomnia or depressive symptoms). ![Study flow chart.\ Legend: EPDS -- Edinburgh Postnatal Depression Scale. BIS -- Bergen Insomnia Scale.](pone.0094674.g001){#pone-0094674-g001} Variables {#s2b} --------- ### Measure of sleep {#s2b1} The Bergen Insomnia Scale (BIS) was used to assess for insomnia (Appendix A) [@pone.0094674-Pallesen1]. This questionnaire consists of six items, of which the first four pertain to night factors such as sleep onset delayed more than 30 minutes, waking up for more than 30 minutes during the night, waking up more than 30 minutes earlier than planned, and not feeling adequately rested after sleep. These items correspond to the DSM-IV-TR criterion A for insomnia [@pone.0094674-APA1]. The last two items assess level of daytime impairment (affecting work/studies or private life) due to sleepiness and dissatisfaction with sleep, corresponding to the DSM-IV-TR criterion B [@pone.0094674-APA1]. Each item is rated on average occurrence from 0 to 7 days per week, giving a possible total sum score from 0 to 42. Women were scored as having insomnia if they had experienced at least one A criterion and one B criterion for three days or more per week during the last month. The BIS has been validated against other self-report scales, as well as polysomnographic data [@pone.0094674-Pallesen1]. Cronbach alpha for the BIS was 0.85 in week 32 of pregnancy and 0.74 in week eight postpartum. Three questions from the Pittsburgh Sleep Quality Index (PSQI) [@pone.0094674-Buysse1], [@pone.0094674-Pallesen2] were included concerning the average time at which participants went to bed, average wake-up time in the morning, and average total sleep duration per night during the previous month. From these data, habitual sleep efficiency (time asleep divided by total time spent in bed) was calculated. When calculations resulted in sleep efficiencies exceeding 100% (due to women\'s self-reports), sleep efficiency was recorded as 100%. The women also reported the estimated time their baby was awake during the night. The use of sleep medication was reported for the final 10 weeks of pregnancy. ### Measure of depressive symptoms {#s2b2} The Edinburgh Postnatal Depression Scale (EPDS) [@pone.0094674-Cox1], [@pone.0094674-EberhardGran1] was used to measure depressive symptoms. The EPDS is a 10-item, self-rating questionnaire developed to screen for depression in the postpartum period; it addresses symptoms present during the last seven days. The scale also has good psychometric properties for use during pregnancy [@pone.0094674-Bunevicius1]. Each question has four possible responses, related to scores from 0 to 3, for a maximum score of 30. Cronbach alpha for the EPDS was 0.85 in late pregnancy and 0.86 in week eight postpartum. A cutoff of 10 or above was found to have good psychometric properties for a diagnosis of depression among Norwegian postpartum women, and this cutoff has also been used in previous studies of pregnant women [@pone.0094674-EberhardGran2]. Information on previous depression was measured by the Lifetime Major Depression Scale [@pone.0094674-Kendler1]. This scale consists of five questions (concerning sadness, appetite changes, lack of energy, self-blame, and concentration) constructed to measure lifetime history of major depression based on the DSM-IV criteria. Prior depression was defined as having had at least three symptoms at the same time with duration of at least two weeks. ### Other study factors {#s2b3} The demographic information collected included maternal age, marital status (married or cohabitating versus single/widowed/divorced), number of previous children, and level of education (elementary school, completed high school, or higher education). In pregnancy week 32, we asked about the experience of 10 specific, stressful life events during the last year, as used by previous studies in Norway [@pone.0094674-Dorheim2], [@pone.0094674-EberhardGran3]. Fear of childbirth was measured by the Wijma Delivery Expectancy/Experience Questionnaire version A (W-DEQ) [@pone.0094674-Wijma1], [@pone.0094674-GarthusNiegel1]. This is a 33-item, self-assessment scale where each item ranges from 0 to 5, with a total sum ranging from 0 to 165. Serious fear of childbirth was defined as a W-DEQ sum score ≥85. This cutoff is commonly used to distinguish women with a fear of childbirth from women without a fear of childbirth [@pone.0094674-Ryding1]. Symptoms of anxiety were measured by the first 10 items in the Hopkins Symptom Checklist (SCL-25) that comprise the anxiety score (HSCL-A) [@pone.0094674-Derogatis1]. Each item ranges from "not at all" (score 1) to "extremely" (score 4). The Norwegian version of SCL-25 has been validated against the ICD-10 criteria for anxiety and depression [@pone.0094674-Sandanger1]. Statistical methods {#s2c} ------------------- The data distributions were checked for normality using p-p charts. For numerical data, means and standard deviations were calculated. Differences in means between groups were tested by student\'s independent t-test, and differences between mean scores before and after delivery for the same women were tested by paired samples t-tests. For categorical data, statistical differences in proportions were examined with chi square. Differences in prevalence of insomnia before and after delivery were tested by McNemar\'s test. Pearson\'s correlation was used to test the association between BIS and EPDS. Multivariate linear regressions were used for testing associations with postpartum BIS and EPDS values simultaneously. Factors individually associated with either of these two variables were entered into the model. Parameters with non-normal distribution (EPDS and HSCL) were log transformed before the analyses, and validity of the model was tested by Box\'s test of equality of covariance matrices and Levene\'s test for equality of variances. Wilks\' Lambda was calculated to test the significance of the multivariate tests, and non-significant variables, except age, were excluded. All analyses were performed in SPSS 18.0 for Windows. The significance level was set to a probability (p) of less than 0.05. Ethical considerations {#s2d} ---------------------- All women asked to participate were given written information explaining the purpose of the study and informed that participation was voluntary. Written informed consent was obtained from all participants, by asking them to sign with full name, address and their date of birth at a separate form. These forms were sent to the project leader, and kept separate from the other data. At the field for date of birth, the following information was given "If you are less than 18 years old, you cannot participate in the study". In spite of this, three girls aged 17 returned questionnaires, and one had returned all three questionnaires and was included in the present study. This was not discovered until a later phase of the study. The age of majority for health in Norway is 16 years, therefore guardians were not asked for consent, as this would break confidentiality. Since the girl had taken time to complete the questionnaires, and the study did not involve any active interventions, we decided to keep her data in the study. The Regional Committee for Ethics in Medical Research in Norway approved the inclusion of the women aged 17 in the study, provided she now, as adult, gave her consent. Authors did not have any access to identifying information. Anonymization was done before transferring the data from the birth records to the researchers, by personnel at the birth department having this as their job. The forms were sent out to the women\'s home addresses containing only an anonymous form number, by people not having access to the completed forms returned, and who did not participate in the analyses. Therefore, the data was anonymized upon collection. The study was approved by the Regional Committee for Ethics in Medical Research in Norway, approval number S-08013a. Results {#s3} ======= The mean age of the postpartum women was 31.5 years (SD 4.7; range 17.4--45.7), and the number of children each woman had (including the new-born) ranged from 1 to 6, median one child, interquartile range 1--2; 11.8% had three or more children. Primipara comprised 50.8% (n = 1,058) of the sample. Two-thirds of the women (n = 1,345; 67.5%) had an education above 12 years of school. Nearly all the women were in stable relationships; only 48 (2.3%) were single. A majority of the women, 70.6% (n = 1,464), were exclusively breastfeeding; 17.8% (n = 370) combined breastfeeding with supplemental formulas, and 241 women (11.6%) did not breastfeed. The response rate among women scoring high for depressive symptoms (EPDS ≥10) in week 32 was 68.0%, significantly lower than among low scorers (74.1%; p = 0.008). [Table 1](#pone-0094674-t001){ref-type="table"} displays changes in sleep from late pregnancy to postpartum week eight. Mean sleep duration was significantly reduced after delivery (mean change 49 minutes; p\<0.001). The amounts of time women spent in bed after delivery were similar to the time spent in bed in week 32 of pregnancy, resulting in the mean sleep efficiency being reduced from 84.0% to 74.9%, p\<0.001. The babies were awake, on average, for one hour and 53 minutes during the night, 25 minutes less than their mothers. Women reported significantly more frequent night-time awakenings in the postpartum period compared to late pregnancy, but the total sum score of insomnia (BIS) improved from 17.2 to 15.4 (p\<0.001). The prevalence of insomnia according to the DSM-IV criteria also decreased from 61.5% in late pregnancy to 53.9% after delivery (p\<0.001). We did not ask about the present use of sleep medication, but only 12 women (0.6%) reported having used sleep medication during the final months of pregnancy. 10.1371/journal.pone.0094674.t001 ###### Sleep before and after delivery among 2088 women. ![](pone.0094674.t001){#pone-0094674-t001-1} Sleep variable Week 32 of pregnancy Week 8 postpartum Mean change 95% CI of change p value ------------------------- ---------------------- ------------------- ------------- ------------------ --------------- Sleep duration (h:mm) 7:19 (1:28) 6:30 (1:18) −0:49 −0:53; −0:44 \<.001 Sleep efficiency (%) 84.0 (14.0) 74.9 (13.9) −9.19 −9.94; −8.42 \<.001 Time in bed (h:mm) 8:44 (1:16) 8:47 (1:26) 0:03 −0:01; 0:07 .21 Time baby awake (h:mm) \- 1:53 (1:10) \- **BIS. sum score** **17.19 (10.3)** **15.47 (8.9)** **−1.72** **−2.17; −1.27** **\<.001** Sleep initiation 2.73 (2.4) 1.49 (2.0) −1.24 −1.34; −1.13 \<.001 Sleep maintenance 2.89 (2.3) 4.24 (2.8) 1.35 1.21; 1.49 \<.001 Early awakening 2.30 (2.3) 1.29 (2.0) −1.01 −1.12; −0.88 \<.001 Non-restorative sleep 3.54 (2.2) 3.69 (2.3) .15 −.26: −.03 .013 Daytime impairment 2.28 (2.2) 1.79 (2.0) −.49 .38; .58 \<.001 Satisfaction with sleep 3.46 (2.3) 2.98 (2.3) −.48 .37; .60 \<.001 n (%) n (%) McNemars test **Insomnia DSM-IV** **1284 (61.6)** **1126 (53.8)** **−7.8%** **−10.4; −5.2** **\<.001** Criterion A Nighttime 1687 (80.8) 1786 (85.4) 4.6% 2.6; 6.6 \<.001 Criterion B Daytime 1350 (64.8) 1150 (55.0) −9.8% −12.3; −7.2 \<.001 BIS - Bergen Insomnia Scale - sub scores represent number of days per week with the insomnia symptom. h:mm - hours:minutes. Primiparous women reported significantly lower sleep efficiencies (3.8% difference; p\<.001), and reported that their babies were awake significantly longer (26 minutes more; p\<.001) during the night compared to babies of multiparous women ([Table 2](#pone-0094674-t002){ref-type="table"}). They had significantly higher BIS scores (15.9 versus 14.9; p\<0.001), mainly due to more problems with sleep-onset latency and night-time awakenings. This did not, however, result in a higher proportion of insomnia. Women who breastfed did not have significantly different BIS scores or prevalence of insomnia compared to women who were not breastfeeding. However, they reported fewer nights with sleep onset insomnia (BIS 1 sub score 1.4 versus 2.1; p\<0.001) and more nights with wakefulness after sleep onset insomnia (BIS 2 sub score 4.3 versus 3.7; p = 0.003). 10.1371/journal.pone.0094674.t002 ###### Sleep after delivery, according to parity among 2088 women. ![](pone.0094674.t002){#pone-0094674-t002-2} Sleep variable Primipara Multipara Mean difference 95% CI p-value ------------------------- ----------------- ----------------- ----------------- ----------------- ------------ mean (SD) mean (SD) t-test Sleep duration (h:mm) 6:33 (1:24) 6:26 (01.11) −0:07 −0:14; 0:00 .031 Sleep efficiency (%) 73.1 (14.6) 76.8 (12.8) 3.77 2.57; 4.96 \<.001 Time in bed (h:mm) 9:07 (1:33) 8:26 (1:13) −0:40 −0:47; −0:33 \<.001 Time baby awake (h:mm) 2:07 (1:12) 1:41 (1:05) −0:26 −0:32; −0:20 \<.001 mean (SD) mean (SD) **BIS, sum** **15.97 (9.1)** **14.94 (8.8)** **−1.03** **−1.79; −.26** **\<.001** Sleep initiation 1.74 (2.1) 1.23 (1.9) −.51 −.67; −.32 \<.001 Sleep maintenance 4.55 (2.7) 3.92 (2.8) −.63 −.87; −.40 \<.001 Early awakening 1.35 (2.1) 1.22 (2.0) −.13 −.31; .04 .13 Non-restorative sleep 3.64 (2.3) 3.74 (2.3) .10 −.10; .29 .34 Daytime impairment 1.74 (2.0) 1.84 (2.0) .10 −.08; .27 .27 Satisfaction with sleep 2.95 (2.3) 3.00 (2.4) .05 −.15; .25 .78 n (%) n (%) Chi X^2^ **Insomnia DSM-IV** **567 (53.3%)** **556 (53.9%)** **0.6%** **−3.4; 4.7** **.79** Criterion A Nighttime 916 (86.1%) 866 (84.6%) −1.5% −4.4; 1.3 .17 Criterion B Daytime 579 (54.6%) 568 (55.1%) 0.5% −3.5; 4.6 .79 BIS - Bergen Insomnia Scale - sub scores represent number of days per week with the insomnia symptom. h:mm - hours:minutes. Women who recovered from being depressed during pregnancy (EPDS ≥10 in pregnancy only) still had significantly higher insomnia (BIS) scores postpartum compared to women who scored low on the EPDS at both times ([Table 3](#pone-0094674-t003){ref-type="table"}). Furthermore, [table 3](#pone-0094674-t003){ref-type="table"} shows that women who developed new depression (EPDS ≥10 postpartum only) had higher insomnia (BIS) scores in late pregnancy compared to women who remained well. Estimated sleep duration and sleep efficiency did not predict recovery from depression or later development of depression. Women who developed new depression postpartum had the largest negative change in mean sleep duration (−90 minutes) and sleep efficiency (−13.6%). The longer periods of wakefulness at night for postpartum women scoring 10 or more on the EPDS were mainly accounted for by the baby being awake more, as the differences between maternal and infant wake times were similar across the groups. 10.1371/journal.pone.0094674.t003 ###### Differences in insomnia scores and other sleep parameters according to depressive status[+](#nt105){ref-type="table-fn"} before and after delivery among 2088 women. ![](pone.0094674.t003){#pone-0094674-t003-3} Not depressed[+](#nt105){ref-type="table-fn"} pre or postpartum New onset postpartum Recovered postpartum Depressed[+](#nt105){ref-type="table-fn"} both pre and post --------------------------- ----------------------------------------------------------------- -------------------------------------------------------------------------------------- --------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------- **BIS, sum score** 3rd trimester 15.8 (9.9) 18.9 (9.9)[\*\*](#nt108){ref-type="table-fn"} [a](#nt110){ref-type="table-fn"} 22.7 (10.1)[\*\*](#nt108){ref-type="table-fn"} [a](#nt110){ref-type="table-fn"} 26.2 (10.0)[\*\*](#nt108){ref-type="table-fn"} ^b,^ [\*](#nt109){ref-type="table-fn"} ^c^ Postpartum 14.0 (8.3) 22.1 (9.2)[\*\*](#nt108){ref-type="table-fn"} [a](#nt110){ref-type="table-fn"} 17.8 (8.7)[\*\*](#nt108){ref-type="table-fn"} [a](#nt110){ref-type="table-fn"} 23.8 (8.2)[\*\*](#nt108){ref-type="table-fn"} ^c^ **Sleep duration (h:mm)** 3rd trimester 7:22 (1:24) 7:22 (1:18) 7:11 (1:54) 6:50 (1:36)[\*](#nt109){ref-type="table-fn"} ^b^ Postpartum 6:36 (1:18) 5:52 (1:18)[\*\*](#nt108){ref-type="table-fn"} ^a^ 6:32 (1:30) 5:56 (1:30)[\*](#nt109){ref-type="table-fn"} ^c^ Baby wake time 1:50 (1:08) 2:23 (1:15)[\*\*](#nt108){ref-type="table-fn"} ^a,^ [c](#nt112){ref-type="table-fn"} 1:50 (1:04) 2:15 (1:23)[\*](#nt109){ref-type="table-fn"} ^a,^ [c](#nt112){ref-type="table-fn"} Difference mum/baby 0:26 0:19 0:24 0:25 **Sleep efficiency** 3rd trimester 84.8% (13.3) 84.2% (13.6) 79.3% (17.5)[\*\*](#nt108){ref-type="table-fn"} ^a^ 78.6% (15.9)[\*](#nt109){ref-type="table-fn"} ^b^ Postpartum 75.5% (13.3) 70.6% (15.5)[\*\*](#nt108){ref-type="table-fn"} ^a^ 75.3% (14.1) 70.8% (17.4)[\*](#nt109){ref-type="table-fn"} ^c^ Depression was defined by EPDS (Edinburgh Postnatal Depression Scale) score ≥10. BIS - Bergen insomnia Scale SD: Standard Deviation. h:mm - hour: minutes. \*\*p\<.001. \*p\<.01. a- difference from women with no depression before or after delivery. b- difference from women with new onset depression postpartum. c- difference from women who recovered from depression in pregnancy. [Table 4](#pone-0094674-t004){ref-type="table"} shows the results of a multivariate linear regression analysis of factors associated with higher insomnia (BIS) scores after delivery and factors associated with postpartum depression (EPDS scores). Insomnia scores during pregnancy were not a risk factor for depressive symptoms after delivery when adjusted for previous depression. 10.1371/journal.pone.0094674.t004 ###### Risk factors for insomnia and for depressive symptoms eight weeks after delivery, results from a multivariate linear regression analysis among 1914 women. ![](pone.0094674.t004){#pone-0094674-t004-4} Variable B Adj. B P value 95% CI p value ------------------------------------------------- ------------------------- ------- ----------- ---------- ------------ ---------- ----------------------------------- **Insomnia** [a](#nt113){ref-type="table-fn"} **Intercept** **−4.30** **.213** **−11.08** **2.48** **.037** (BIS score) Primipara .98 1.51 \<.001 .74 2.29 \<.001 after delivery Previous depression 3.89 1.00 .018 .17 1.82 \<.001 Fear of delivery 5.72 2.23 .002 .82 3.63 .008 Stressfactors last year .67 .19 .11 −.04 .42 .001 BIS week 32 .35 .30 \<.001 .26 .33 \<.001 Log EPDS week 32 7.27 1.91 .008 .51 3.30 \<.001 Log HSCL-A week 32 12.27 3.09 .012 .67 5.51 \<.001 Education level .33 .81 .023 .11 1.50 .020 Maternal age .01 .08 .058 −.003 .17 .16 **Depression** [b](#nt114){ref-type="table-fn"} **Intercept** **−.39** **.001** **−.63** **−.15** [\*](#nt115){ref-type="table-fn"} (log EPDS) Primipara .03 .04 .002 .02 .07 after delivery Previous depression .22 .09 \<.001 .06 .11 Fear of delivery .23 .02 .35 −.03 .07 Stressfactors last year .05 .02 \<.001 .01 .02 BIS week 32 .008 .00 .66 −.001 .002 Log EPDS week 32 .58 .43 \<.001 .38 .48 Log HSCL-A week 32 .80 .23 \<.001 .14 .31 Education level −.01 .03 .025 .004 .05 Maternal age −.004 .00 .76 −.003 .004 R^2^ = 0.209. R^2^ = 0.356. \*p values for Wilk\'s Lambda for the significance of the multivariate tests (the p values are common for insomnia and depression). CI - Confidence Interval. BIS - Bergen Insomnia Scale. EPDS - Edinburgh Postnatal Depression Scale. HSCL A - Hopkins Symptoms Checklist, Anxiety module. Discussion {#s4} ========== We found that women slept fewer hours at night after delivery compared to during late pregnancy, and that they reported more nights with night-time awakenings. Hence, sleep efficiency decreased from 84% to 75%. In spite of this, the self-reported insomnia scores improved, and the prevalence of insomnia according to the DSM-IV criteria decreased. Primiparous women reported lower sleep efficiency and higher insomnia scores than multiparous. Primiparous women also had increased risk of depressive symptoms after delivery. Insomnia scores in pregnancy, but not sleep duration or sleep efficiency, were associated with depressive symptoms postpartum. However, when adjusted for previous depression, insomnia in pregnancy did not predict depressive symptoms after delivery. Change in sleep from pregnancy to postpartum {#s4a} -------------------------------------------- We found that women slept fewer hours and less efficiently after delivery when compared to sleep during pregnancy. Montgomery-Downs et al. found longer sleep duration after delivery than in our study (7.2 hours compared to 6.5 hours) but similar low sleep efficiency [@pone.0094674-MontgomeryDowns1]. In spite of reduced sleep time and sleep efficiency, insomnia scores and prevalence of insomnia were reduced in our study. Although they reported more nights of being awake for longer than 30 minutes, the postpartum women had fewer evenings with delayed sleep onset and fewer mornings with awakenings more than 30 minutes earlier than desired. Women reported less daytime impairment and were more satisfied with their sleep after delivery compared to during pregnancy. New mothers may have expected sleep disturbances after delivery to be more severe than they experienced, and therefore they were more satisfied with the (shorter amount of) sleep they got. In addition, waking up because of an infant may be more tolerable and may produce less worry compared to being unable to sleep for other reasons. Worry about one\'s sleep may contribute to insomnia, as individuals who tend to focus cognitively on their insomnia and ruminate about their poor sleep are less likely to get a good night\'s sleep [@pone.0094674-Baglioni1]. This worry may consequently lead to hyper-arousal, an important factor in maintaining insomnia [@pone.0094674-Riemann1]. None of the women were working outside the home at the time of the postpartum follow-up, due to the right of paid maternity leave in Norway for the first 45 postnatal weeks. This may explain some of the lower insomnia rate, as sleep problems are more common during periods of active employment [@pone.0094674-Lampio1]. However, the majority of the women in our study, 63%, were on sick leave in week 32, and these women had a higher rate of insomnia compared to working pregnant women [@pone.0094674-Dorheim3]. Taking care of a new-born baby the first weeks after delivery may be hard work and not considered leisure time. However, compared to work outside the home, there may be more flexibility in terms of completing tasks and more opportunities for daytime napping. Our study did not measure daytime napping, as this is not part of the Bergen Insomnia Scale or the Pittsburgh Sleep Quality Index, but a study of pregnant women in Taiwan found decreased sleep among women with longer work hours and women who had longer daytime naps [@pone.0094674-Tsai1]. Future studies should include questions regarding daytime naps. Primiparous women reported more days with delayed sleep-onset time and were awake during more nights compared to multiparous women, possibly due to less experience and confidence in child care and more worry and hyper-arousal. Coo Calcagni et al. found that first-time mothers experienced more stress and had poorer subjective sleep than multiparous women [@pone.0094674-CooCalcagni1]. Other studies have also found poorer sleep among first-time mothers [@pone.0094674-Dorheim2], [@pone.0094674-Signal1]. Salo et al. found that difficulties initiating or maintaining sleep increased the long-term risk of depression in the general population [@pone.0094674-Salo1]. These sleep factors, more prevalent among primipara, may be mediators that also increase the risk of postpartum depression. However, primiparous women also had an increased risk of depression when adjusting for level of insomnia, so there may be factors other than insomnia contributing to this increased risk for depression among first-time mothers. Previous studies from Norway have found conflicting results regarding parity and postpartum depression [@pone.0094674-Dorheim2], [@pone.0094674-EberhardGran3], [@pone.0094674-Berle1]. Parity does not seem to be among the strongest risk factors for postpartum depression [@pone.0094674-OHara1]. We found no associations between breastfeeding and total insomnia scores, which is in line with other studies [@pone.0094674-Dorheim2], [@pone.0094674-MontgomeryDowns2]. From our data, it seems that although breastfeeding may increase the number of nights with awakenings, it also decreases nights with prolonged sleep-onset latency. Insomnia and depression {#s4b} ----------------------- We found that insomnia, but not sleep efficiency or sleep duration, was a risk factor for postnatal depression in the univariate analyses. This is similar to findings by Park et al., using actigraphy and subjective sleep reports. They found that subjective assessments of sleep may be more accurate predictors of postpartum depression than total amount of sleep measured by actigraphy [@pone.0094674-Park1]. In our study, women who recovered from antepartum depression still had residual insomnia symptoms compared to women who scored low for depression at both time points. Similarly, women who developed depression after delivery but were not depressed in pregnancy had higher insomnia scores in pregnancy compared to women who remained well. This may reflect the co-morbid and bidirectional relationship between insomnia and depression [@pone.0094674-Manber1], [@pone.0094674-JanssonFrojmark1]. It may also reflect a pre-clinical state where women have insomnia as a precursor for or an early symptom of depression. The residual insomnia found among women who recovered from depression in pregnancy has previously been reported in non-pregnant populations and is associated with an increased risk of relapse of depression [@pone.0094674-Manber1], [@pone.0094674-Li1]. This may explain why insomnia in pregnancy did not predict postpartum depression when adjusted for previous depression. Antenatal depression and a prior history of depression are among the strongest risk factors for postnatal depression [@pone.0094674-Milgrom1]. Marques et al. found similar results to our study: Insomnia in pregnancy seemed to predict depression after delivery; however, when adjusted for lifetime history of depression and negative affect in pregnancy, the association disappeared [@pone.0094674-Marques1]. In a study of women with previous postpartum depression, Okun et al. found that changes in postpartum sleep during the first 17 weeks predicted recurrence of depression in these at-risk women [@pone.0094674-Okun2]. Insomnia in pregnancy may thus be seen as a mediator between previous depression and a new depressive episode postpartum for women at risk. As insomnia is an easy complaint to look for during pregnancy, insomnia may thus warn the clinician of an increased risk of postpartum relapse of depression. This may in turn provide opportunities for early interventions to prevent clinical depression. Treating insomnia during pregnancy and the early postpartum period for women with previous depression may be a way of preventing postpartum depression. One pilot study of insomnia treatment among 12 women with postpartum depression and insomnia showed promising results on sleep efficiency, total wake time, subjective mood, insomnia severity, sleep quality, and fatigue [@pone.0094674-Swanson1]; however, more studies on treatment of postpartum women are needed. An encouraging finding of our study is that women with insomnia in pregnancy who have never been depressed before may be reassured that insomnia in itself does not seem to increase the risk for postpartum depression. For many, insomnia is a chronic condition. The present study found that a woman who slept poorly during pregnancy was more likely to sleep poorly after delivery as well, regardless of any depression. Morin and colleagues found that 74% of people with insomnia had the complaint for more than one year, and more than 25% of those who recovered experienced relapse [@pone.0094674-Morin1]. The perinatal period may therefore be no exception to this. Hyper-arousal may be one factor explaining that both anxiety, as measured by the HCSL-Anxiety scale, and fear of delivery were factors associated with postpartum insomnia [@pone.0094674-Riemann1]. There may be an interaction between cognitive and autonomic hyper-arousal that contributes to the maintenance of both insomnia and continued emotional disturbance [@pone.0094674-Baglioni3], [@pone.0094674-LeBlanc1]. Strengths and limitations {#s4c} ------------------------- This is a large study from a general population of pregnant women. However, as women who did not understand the Norwegian language were excluded, these results may not be representative for the newer immigrant population. The longitudinal design makes it possible to examine the direction of effects and to evaluate whether depressive symptoms precede insomnia or whether insomnia is a precursor for depression in this period with decreased sleep among all women. For longitudinal population-based studies, a retain rate of more than 50% after three questionnaires is acceptable, and a large number of participants remained in the study. According to Galea and Tracey, declines in participation rates are not likely to have substantial influence on exposure-disease associations [@pone.0094674-Galea1]. Women with depressive symptoms during pregnancy were less likely to remain in the study, and this may have led to selection bias. We found, however, a strong correlation between depression and insomnia in our sample with possibly fewer depressed women than in the population, and the real relationship may be stronger than the results presented here. The questionnaires included several known risk factors for postpartum depression, making it possible to adjust for confounders. We controlled for stressful life events the previous year, but did not assess for past life events, such as childhood adversities, which may also be of importance for emotional dysfunction during pregnancy [@pone.0094674-Dayan2]. The tools used are all validated and have been used in several other studies in Norway. Depression and insomnia were measured by self-report and not verified by clinical interviews or sleep diaries. As the diagnosis of insomnia relies on the subjective experiences of the individual experiencing sleep problems [@pone.0094674-AASM1], self-report may be an adequate way of measuring insomnia. The BIS is based on recall of insomnia symptoms during the past month, and the prospective use of sleep diaries might have given a more accurate estimate of current insomnia. However, BIS scores correspond well with polysomnographic sleep registration, except for item 4 (feeling adequately rested after sleep) and item 5 (daytime functioning) [@pone.0094674-Pallesen1]. The EPDS is not diagnostic of depression, and a cut-off at 9/10 at the Norwegian version is relatively low compared to validations in other languages [@pone.0094674-EberhardGran4]. There may therefore be a risk of some false positives in [Table 3](#pone-0094674-t003){ref-type="table"}. However, the continuous scale of the EPDS has been recommended for use in population research to find factors associated with depression [@pone.0094674-Green1]; hence the cut-off on the EPDS did not influence the risk factors identified in [Table 4](#pone-0094674-t004){ref-type="table"}. Suggestions for clinical intervention and further research {#s4d} ---------------------------------------------------------- Insomnia during pregnancy may be a valid clinical marker for recurrence of depression postpartum among women at risk. Further studies should evaluate whether treatment of insomnia during pregnancy may prevent postpartum depression among women with other risk factors for the condition. Cognitive behavioral therapy for insomnia may have a positive effect on anxiety [@pone.0094674-Belleville1], and treatment of depression with mindfulness meditation improves subjective sleep quality with corresponding changes in polysomnography [@pone.0094674-Britton1]. Lee and Gay found that treatment of sleep problems among new parents using a modified sleep-hygiene intervention improved sleep, especially among less socioeconomically advantaged women [@pone.0094674-Lee1]. Low socioeconomic status is a risk factor for postpartum depression [@pone.0094674-OHara1], and improving sleep for these mothers may therefore be one way of preventing or relieving depressive symptoms in this group. The authors thank the women who volunteered their time to participate in this study. We also thank Tone Breines Simonsen, Wenche Leithe, and Ishtiaq Khushi for assistance in the data collection, and Ingvild Dalen, Stavanger University Hospital, and Geir Egil Eide, Western Norway Regional Health Authority, for assistance in the statistical analyses. [^1]: **Competing Interests:**Signe Dørheim has had paid speaking engagements for AstraZeneca and Pfizer. Bjørn Bjorvatn has had paid speaking engagements for Glaxo and Confex. Malin Eberhard-Gran has indicated no financial conflicts of interest. This does not alter the authors\' adherence to PLOS ONE policies on sharing data and materials. [^2]: Conceived and designed the experiments: SD BB MEG. Performed the experiments: MEG. Analyzed the data: SD BB MEG. Contributed reagents/materials/analysis tools: SD. Wrote the paper: SD BB MEG.

{ "pile_set_name": "PubMed Central" }

Today, most patients who require renal replacement therapy are treated using dialysis membranes. To improve the outcomes of chronic dialysis patients, unique developments have been made in the past several decades. Cellulose membranes have been mostly replaced by synthetic polymeric membranes with improved biocompatibility. The development of high-flux membranes and more efficient treatment modes, such as hemodiafiltration, have resulted in improved removal rates of uremic toxins, particularly those in the middle molecular-weight range (\>500 Da). Beyond the membranes used for conventional hemodialysis, new membranes with increased pore size have been developed for specific treatments. These specialized membranes allow for the removal of higher molecular-weight molecules, such as mediators of sepsis/inflammation, or the removal of nephrotoxic light chains of immunoglobulins. However, these membranes allow the passage of plasma proteins, such as albumin ([@bibr1-ijao.5000603]), whose loss is undesirable. The newest generation of highly selective and permeable medium cut-off (MCO) membranes enables the removal of large molecules, as do high cut-off (HCO) membranes, while simultaneously maintaining low passage of albumin. For uremic solutes in the 15,000 to 45,000 Da- size range, the MCO membranes offer improved clearance in comparison with that of high-flux membranes used in HD mode and equivalent clearance to that of high-flux membranes used in high-volume hemodiafiltration (HDF) mode ([@bibr2-ijao.5000603]). Therefore, the use of MCO membranes simplifies the delivery of high-removal treatments for end-stage renal disease (ESRD) patients, with a removal spectrum that extends the current possibilities of the best available therapies. This improvement should allow clinicians to surpass the benefits of HDF while using regular HD equipment, i.e., not requiring large amounts of high-quality fluid and a more complex setup. These special MCO membranes should aid in raising the standard of treatment available for all chronic HD patients and may potentially decrease inflammation and improve general patient outcomes. Future studies are being planned to demonstrate the clinical benefits of this new membrane type. Development of Dialysis Membranes with Different Permeabilities {#section1-ijao.5000603} =============================================================== Renal failure (RF) is characterized by the loss of the ability of the kidneys to excrete wastes, concentrate urine, conserve electrolytes, and maintain fluid balance. Acute kidney injury can be a life-threatening illness, with a mortality of between 50% and 80% ([@bibr3-ijao.5000603]). ESRD occurs in the late stages of chronic kidney disease and is associated with an irreversible loss of kidney function. Renal replacement therapy (RRT) replaces kidney function in patients with both types of renal failure. There are 2 possible alternatives to treat chronic RF: organ transplantation or, more commonly, dialysis. However, the natural secretion of kidney hormones, which influences blood pressure, cannot be achieved with dialysis treatment modalities. Over 2 million people worldwide currently receive treatment with dialysis or a kidney transplant to stay alive, yet this number may represent only 10% of people who actually need treatment to live ([@bibr4-ijao.5000603]). Of the 2 million people who receive treatment for kidney failure, the majority are treated in only 5 countries: the United States, Japan, Germany, Brazil, and Italy. These 5 countries represent only 12% of the world population. Only 20% are treated in approximately 100 developing countries that make up over 50% of the world population ([@bibr4-ijao.5000603]). In medicine, hemodialysis (HD) is the process of removing blood from a patient, purifying the blood through an artificial kidney (dialyzer), and then returning it to the patient\'s bloodstream. On the basis of the developments of Willem Kolff and Nils Alwall in the 1940s, the dialyzer membrane systems have undergone multiple development cycles and are now the basis of an effective, reliable, and cost-effective treatment approach. In the early days of dialysis, large and unwieldy plate dialyzers made of cellulose membranes were used. Until the 1970s, cellulosic membranes were used exclusively and represented the majority of membranes used in hemodialysis worldwide ([@bibr5-ijao.5000603]). However, cellulosic membranes may activate the complement system ([@bibr6-ijao.5000603]) and induce other adverse biological reactions, e.g., leucopenia, owing to the accumulation of granulocytes in lung capillaries ([@bibr7-ijao.5000603]), inhibition of granulocyte metabolism, and release of enzymes from granulocytes and monocytes ([@bibr8-ijao.5000603]). The drawbacks associated with the decreased hemocompatibility of cellulosic membranes can be mitigated by the partial substitution of the hydroxyl groups by (i) acetylation or the (ii) introduction of diethylamino groups or (iii) benzyl groups ([@bibr9-ijao.5000603]). Cellulose acetate is a cellulose-derived material that has been modified by esterification. Depending on the degree of substitution, such a membrane can be categorized as cellulose acetate, cellulose diacetate, or cellulose triacetate. Another biocompatibility improvement of cellulosic membranes is surface-coating them with polyethylene glycol (PEG), polyacrylonitrile (PAN-RC®) ([@bibr10-ijao.5000603]), or vitamin E (Excebrane®). The use of vitamin-E-coated membranes, compared with uncoated cellulose membranes, appears to significantly decrease the activation and migration of monocytes and granulocytes ([@bibr11-ijao.5000603]). Cellulosic membranes belong to the class of "low-flux" membranes, which have a low permeability: substances of up to 5,000 Da can be eliminated, whereas nearly 100% of the (ß2-microglobulin is rejected. The area of synthetic polymeric membrane development began with the discovery by Geyjo in 1985 that (ß2-microglobulin is the amyloidogenic precursor of AB-amyloidosis ([@bibr12-ijao.5000603]), together with the introduction of the hemofiltration treatment mode. Because the mass transfer of low-flux membranes is mainly diffusive and is not suitable for use in therapies that require high convective transport, such as hemofiltration (HF) and hemodiafiltration (HDF), membranes with higher permeability were required. Synthetic polymeric membranes close this gap, showing higher permeability and better biocompatibility. The hydrophobic base material is mainly polysulfone or polyethersulfone (polyarylethersulfone). These high-flux membranes allow for the removal of so-called middle molecular-weight molecules. Membranes prepared from hydrophobic/hydrophilic polymer blends are the predominant type of synthetic polymeric membrane ([@bibr9-ijao.5000603]). High-flux membranes are highly permeable to β2-microglobulin. The pore structure of high-flux membranes allows for the passage of so-called middle molecules of up to 20,000 Da. In addition, significant protein loss, for example, in albumin (68,000 Da), is prevented. Beyond the immunoglobulins, albumin is the most important plasma protein that must be retained in chronic dialysis patients. High-flux membranes are suitable for hemodialysis, hemofiltration, and hemodiafiltration. High-flux synthetic membranes show good transport properties and exhibit only slight complement activation ([@bibr13-ijao.5000603]), a minimal decrease in leukocytes ([@bibr14-ijao.5000603]), and a low release of leukocyte elastase ([@bibr15-ijao.5000603]). In a recent study, no significant survival benefit with either high-flux or low-flux membranes has been found in the overall population, but the use of high-flux membranes confers a significant survival benefit in patients with serum albumin ≤4 g/dL ([@bibr16-ijao.5000603]). Beyond membranes used for conventional hemodialysis, high cut-off (HCO) membranes with increased pore size have been developed for specific treatments. In high cut-off membranes, the pore sizes are shifted toward a larger pore diameter than that of conventional high-flux membranes showing a higher permeability for substances above 15,000 Da ([@bibr17-ijao.5000603]). Examples of these substances include inflammatory cytokines, which are over-expressed in acute inflammatory diseases and contribute to the pathogenesis of septic acute kidney injury or free light-chain proteins. Monoclonal light chains of immunoglobulin are excessively produced by malignant plasma cells in patients with multiple myeloma (MM) and can cause renal lesions such as cast nephropathy. In this condition, serum concentrations may be elevated by more than 100-fold compared with normal serum concentrations. The increase in the pore sizes of HCO membranes translates into an increase in both the convective and diffusive permeability, thus allowing for the removal of molecules, such as mediators of sepsis/inflammation ([@bibr18-ijao.5000603]), or the removal of nephro-toxic light chains of immunoglobulins ([@bibr19-ijao.5000603]). However, the albumin loss with this membrane is substantial, from approximately 9 to 23 g/treatment ([@bibr20-ijao.5000603], [@bibr21-ijao.5000603]), and it is not indicated for use in chronic dialysis but instead is recommended mainly for acute applications ([@bibr22-ijao.5000603]). Over the past decade, various membranes, such as the cellulose triacetate-based FH 70/150 (Sureflu; Nipro ([@bibr23-ijao.5000603])), the polysulfone-based APS-1050 (Asahi ([@bibr24-ijao.5000603])), the PMMA-based BK-F/BG 2.1(Toray ([@bibr25-ijao.5000603])), and the Helixone polysulfone based FX-E (Fresenius Medical Care ([@bibr26-ijao.5000603])), have been manufactured with the purpose of increasing membrane permeability, given the need for the increased clearance of low molecular weight proteins and protein-bound solutes. Because the loss of albumin is associated with the removal of protein-bound uremic toxins, protein-leaking membranes have been developed for hemodialysis. Dialysis treatments with such membranes, compared with conventional high-flux dialysis membranes, provide greater clearance of those molecules, albeit at the cost of some albumin loss into the dialysate ([@bibr27-ijao.5000603]). The newest generation of highly selective and permeable medium cut-off (MCO) membranes meets both requirements for high-quality and effective dialysis treatment: the removal of large middle-molecules up to a molecular weight of 45,000 Da, similar to a high cut-off membrane, and a low loss of albumin (high-flux membrane). For uremic solutes in the 15,000 to 45,000-Da size range, the MCO membranes offer improved clearance in comparison with those of high-flux membranes used in HD mode and equivalent clearance to high-flux membranes in HDF mode ([@bibr2-ijao.5000603]). Uremic Toxins and Removal Mechanisms {#section2-ijao.5000603} ==================================== Because 'artificial kidneys' remove uremic toxins primarily via size exclusion, the classification of these solutes on the basis of their molecular weight and removal behavior is key for membrane development. In maintenance hemodialysis patients, the insufficient removal of mid-sized and protein-bound uremic toxins is associated with endothelial injury and chronic inflammation, thus contributing to subsequent cardiovascular disease. Studies demonstrating the role of these uremic toxins in inducing cardiovascular effects or other negative effects, such as coordination disturbances or polyneuritis in dialysis patients, have attracted increasing attention over the past decade ([@bibr28-ijao.5000603]). The molecular weights of these uremic toxins vary widely from very small compounds, such as urea (MW = 60 Da), to globulins (e.g., \[ß2-microglobulin MW = 11,800 Da). Generally, uremic toxins are split into 3 groups: small water-soluble compounds, middle molecules and protein-bound solutes ([@bibr29-ijao.5000603]). In the latest literature research performed by the European Uremic Toxin (EUTox) Work group, 56 newly reported solutes have been found ([@bibr30-ijao.5000603]). Examples of solutes classified into the 3 aforementioned groups are shown in [Table I](#table1-ijao.5000603){ref-type="table"}. ###### Examples of Uremic Solutes and their Molecular Weights, Classified into 3 Groups: Small Water-soluble Compounds, Middle Molecules and Protein-bound Solutes ([@bibr28-ijao.5000603], [@bibr33-ijao.5000603]) ![](10.5301_ijao.5000603-table1) Molecule Molecular weight Example ------------------------------- ------------------ ---------------------------------------------------------------------------------------------------------------------------- Small water-soluble compounds \<500 Da asymmetric dimethylarginine, guanidine, uric acid, oxalate, ethylamine, methylguanidine, neopterin, phenylacetic acid Middle molecules \>500 Da β2-microglobulin, adiponectin,α1-acid glycoprotein, cystatin C, prolactin, osteocalcin, vascular endothelial growth factor Protein-bound solutes Variable p-cresylsulfate, indoxyl sulfate, phenol, indol-3-acetic acid, hippuric acid, homocysteine, carboxymethyllysine, acrolein Small water-soluble compounds are also referred to as low molecular weight molecules, which have molecular weights \<500 Da and are soluble in water. The second group, consisting of middle molecular weight toxins, includes compounds with molecular weights above 500 Da. Protein-bound solutes constitute the third group of uremic toxins, which have molecular weights less than 500 Da but demonstrate strong binding to albumin. Small water-soluble compounds are easily removed, diffusively, across the semipermeable membrane, owing to the concentration gradient between the blood and dialysis fluid. Conventional hemodialysis is an effective treatment mode for the removal of those compounds, whereas it is poorly suited for middle molecules such as (ß2-microglobulin, even with the use of high-flux membranes. This phenomenon is attributable to a decrease in the diffusion coefficient with increasing molecular size. Alternative therapies such as hemofiltration or hemodiafiltration include convective transport mechanisms and the improved removal of both low- and middle-molecular-weight toxins such as (ß2-microglobulin ([@bibr31-ijao.5000603]). The driving force for convective transport is a pressure gradient across the membrane, thus leading to ultrafiltration (UF), in which the molecules are pulled from the blood into the dialysate. An increase in the ultrafiltration rate increases the solute removal of low- and middle-molecular-weight solutes. In hemodialysis, UF is limited to the removal of only excess body fluid from the blood, whereas in convective therapies UF is increased beyond the target weight loss, and fluid balance is preserved via the infusion of a physiological solution into the blood ([@bibr32-ijao.5000603]). In clinical practice, the UF rate is limited by the blood flow rate because only a portion of the volume, usually 25% to 30%, is filtered before the blood cell concentration increases to a level at which cellular damage and dialyzer clotting occur ([@bibr32-ijao.5000603]). Given the ability to obtain ultrapure water and dialysis fluids for use in water treatment systems and online HDF machines, the use of online hemodiafiltration is increasing steadily in Europe ([@bibr34-ijao.5000603]). Online HDF post-dilution is considered the most effective and safest dialysis treatment, owing to its superior blood purification of all uremic toxins and its decreased association with the incidence of cardiovascular events ([@bibr35-ijao.5000603]). However, this therapy is more complex and requires high exchange volume rates of up to 24 L per treatment. The automated machine settings aim to optimize the filtration fraction with the lowest possible rate of machine alarms but do not regulate the 2 primary determinants of the convection volume: the treatment time and the blood flow rate ([@bibr36-ijao.5000603]). Furthermore, in comparison with other dialysis modalities, HDF also provides superior removal of certain protein-bound uremic solutes ([@bibr37-ijao.5000603]). However, the removal of protein-bound solutes via dialysis strategies is still less efficient than the removal of non-protein-bound solutes of similar molecular weights, owing to the resistance induced by the protein binding ([@bibr28-ijao.5000603]). The total removal depends on how rapidly the solute unbinds from its carrier protein as the free concentration decreases ([@bibr32-ijao.5000603]). Investigations into protein-bound uremic toxin removal are often restricted to some representatives, such as indoxyl sulfate, conjugates of p-cresol, p-cresylsulfate, and p-cresylglucoronide, and phenylacetic acid ([@bibr28-ijao.5000603], [@bibr38-ijao.5000603]). With the deployment of adsorptive therapies that were first implemented for the treatment of severe liver failure, the removal of these molecules may be enhanced ([@bibr38-ijao.5000603][@bibr39-ijao.5000603]--[@bibr40-ijao.5000603]). Because the loss of albumin is associated with the removal of protein-bound uremic toxins, protein-leaking membranes have been developed for hemodialysis. Dialysis treatments with such membranes, compared with conventional high-flux dialysis, provide greater clearance of these molecules, but this comes at the cost of a certain degree of albumin loss into the dialysate ([@bibr28-ijao.5000603]). Albumin Removal in Different Treatment Modalities {#section3-ijao.5000603} ================================================= Hypoalbuminemia is associated with mortality in patients with ESRD. Albumin removal during dialysis treatment leads to a decrease in the serum albumin concentration, because it is influenced by the albumin synthesis rate, catabolism, distribution between the intra- and extravascular compartments and external loss under pathological conditions ([@bibr41-ijao.5000603]). The limits of albumin removal across a dialyzer in terms of patient tolerance have been addressed in 2 reviews published in 2003 and 2005, but this issue has not been resolved ([@bibr27-ijao.5000603], [@bibr42-ijao.5000603]). Nevertheless, albumin removal during different treatment modalities has been elucidated: a small but non-significant decrease in the serum albumin concentration has been observed over a period of 24 months in a group of 17 stable peritoneal dialysis patients, despite protein losses in the range of 5 to 7 g per day ([@bibr43-ijao.5000603]). The albumin removal during hemodialysis with conventional high-flux membranes is generally reported to be in the range of 0 to 2 g per 4-hour treatment, depending on the membrane material and the surface area ([@bibr27-ijao.5000603]). The reported albumin removal levels in online-HDF treatments vary widely and depend on the dilution mode, degree of flux across the membrane, type of membrane, and other treatment parameters. The albumin loss per HDF treatment typically ranges from 1 to 4 g and can sometimes reach values of greater than 5 g ([@bibr44-ijao.5000603]). The protein removal levels with the use of reused dialyzers after bleaching procedures are in the range of 10 to 12 g per treatment; these levels are associated with a significant decrease in the serum albumin concentration ([@bibr45-ijao.5000603]). However, comparable average serum albumin concentrations have been reported in another study in which bleach reprocessing was limited, equivalent to an average dialyzer albumin removal rate of 4.3 g per hemodialysis session ([@bibr46-ijao.5000603]). More recent data published as an abstract at the World Congress of Nephrology 2009 have indicated dialysate albumin losses in the range of 0.48 g for Evodial, with 2.2 to 15.5 g per 4-hour session for the FDY 210 dialyzer (Nikkiso) used in hemodiafiltration mode ([@bibr47-ijao.5000603]). The most recent studies examining widely used membranes have reported albumin losses of 3.1 ± 2.4 g ([@bibr48-ijao.5000603]) and 3.0 ± 2.4 to 4.3 ± 3.5 g ([@bibr49-ijao.5000603]). Currently, it is unclear how much albumin loss per treatment session is tolerated in ESRD patients. There has also been discussion on the utility of a certain degree of loss during albumin removal without triggering antioxidant effects versus facilitating the synthesis of new albumin with antioxidant effects ([@bibr50-ijao.5000603]). Medium Cut-off (MCO) Membranes -- a New Generation of Hemodialysis Membranes {#section4-ijao.5000603} ============================================================================ In the years since the development of hemodialysis as a treatment for ESRD, a number of improvements have been made to dialyzer membranes. Despite these advances, the overall clinical outcomes for patients still present challenges. Many observational studies have supported the hypothesis that higher-molecular-weight toxins are responsible for a number of dialysis comorbidities, such as chronic inflammation and related cardiovascular diseases ([@bibr51-ijao.5000603]), immune dysfunctions ([@bibr52-ijao.5000603]), anemia and EPO hyper responsiveness ([@bibr53-ijao.5000603]), thus in turn influencing the mortality risk. However, when high cut-off membranes are used to filter higher-molecular-weight toxins, the patients lose unacceptable amounts of albumin and other essential proteins. In an open, randomized, cross-over, 2-center, controlled, prospective clinical study, dialysis patients have been treated with high cut-off HCO1100 dialyzers (Baxter-Gambro, Hechingen, Germany) in series with the low-flux dialyzer PF14L (HCO/LF-HD) or the high-flux dialyzer PF210H (HF-HD) ([@bibr1-ijao.5000603]). The patients treated with HCO/LF-HD showed significantly greater decrease in multiple immune mediators such as slL-2R, sTNF-R1, sTNT-R2, and FLCs but also a significant decrease in albumin from 36.2 ± 3.5 to 31.0 ± 4.7 g/L after 3 weeks. Thus, hemodialysis treatment with high cut-off membranes achieves a more effective removal of larger uremic toxins, thus decreasing inflammatory activity, at the cost of higher albumin removal. The existing gap in the high selective membrane permeability has now been closed with the development of medium cut-off membranes (MCO). Today, advances in polymer recipes, spinning technologies, and the whole dialyzer-manufacturing concept, including the increased use of high-tech equipment, have resulted in the production of improved, safer and higher-quality dialyzer products. Different approaches in terms of the membrane and dialyzer design, such as fiber undulation, high package density, and improved flow distribution of dialysate fluid as well as decreased fiber diameter to increase internal filtration, have led to improved dialyzer performance but remain insufficient with regard to the removal of middle-molecular-weight products and simultaneous retention of proteins such as albumin. These goals require an extremely narrow pore-size distribution of the dialysis membrane, which has not previously been obtained through the phase inversion method but has recently been achieved through a tailored and well-controlled spinning technology that creates larger, uniformly sized, and densely distributed pores. The newly developed medium cut-off membrane exhibits high selective membrane permeability. After the in vitro characterization of the membrane properties in the context of dextran filtration, Boschetti-de-Fierro et al ([@bibr54-ijao.5000603]) have classified MCO membranes as having permeabilities closest to that of the natural kidney, as compared with other conventional dialysis membranes, including all permeability classes (low- and high-flux membranes, high cut-off, and protein-leaking membranes). This classification has introduced a new term, the molecular weight retention onset (MWRO). According to the authors' definition, the MWRO is the molecular weight at which the sieving coefficient is 0.9. Together with the general known molecular weight cut-off (MWCO), at which the sieving coefficient is 0.1, it is possible to map different types of blood purification membranes. Different dialyzer families can be distinguished on the basis of their membranes, such as low-flux, high-flux, protein-leaking, high cut-off and medium cut-off membranes. Medium cut-off membranes expand the known limits of high-flux membranes. This expansion in membrane permeability and selectivity represents a large step toward the realization of ideal dialysis membrane separation properties. Boschetti-de-Fierro et al have described 4 dialysis membrane groups in their general classification and their typical performances ([@bibr17-ijao.5000603]). The data, including the water permeabilities and sieving coefficients of (ß2-microglobulin and albumin, of these dialysis membrane groups, extended to include the class of MCO membranes, are given in [Table II](#table2-ijao.5000603){ref-type="table"}. ###### General classifications and typical performance of membranes used in dialysis ![](10.5301_ijao.5000603-table2) Dialyzer type Water permeability^[a](#table-fn2-1-ijao.5000603){ref-type="table-fn"}^ Sieving coefficient^[b](#table-fn2-2-ijao.5000603){ref-type="table-fn"}^ ----------------- ------------------------------------------------------------------------- -------------------------------------------------------------------------- ------------ Low-flux 10--20 -- \<0.01 High-flux 200--400 0.7--0.8 \<0.01 Protein-leaking 50--500 0.9--1.0 0.02--0.03 High cut-off 1100 1.0 0.2 Medium cut-off 600--850 1.0 0.008 with 0.9% wt.-% sodium chloride at 37°C ± 1 ° C and QB 100--500 mL/min. according to EN1283. In vitro data suggest expanded toxin removal similar to that observed with high cut-off membranes, with simultaneous retention of albumin, such that medium cut-off membranes are appropriate for regular use in conventional treatment schedules and treatment modes, e.g.,4-hour treatments, 3 times weekly, in Europe. The first 2 studies, which were designed to compare MCO dialyzers with the last generation of high-flux dialyzers, have confirmed this assumption ([@bibr2-ijao.5000603]). In the first randomized hemodialysis study, 3 different MCO prototype versions MCO AA, BB, and CC (Gambro Dialysatoreny, a subsidiary of Baxter International) with different membrane pore sizes/permeabilities AA \<BB \<CC54 were compared with a high-flux FX CorDiax 80 dialyzer (Fresenius Medical Care) ([@bibr55-ijao.5000603]). The study was conducted in the LKH University hospital in Graz and at the LKH in Bruck, both in Austria, and 19 ESRD patients were included. The primary endpoint was the overall clearance of λ-free light chains (λ-FLCs) with a molecular weight of \~45,000 Da. The secondary outcomes were the removal of medium-sized solutes such as κ--Ig (MW \~22,500 Da) free light chains,α1-microglobulin, complement factor D, myoglobin,(ß2-microglobulin and small solutes, as well as the safety of these prototypes. The studied MCO dialyzer prototypes achieved significantly higher overall clearances of κ-FLC and λ-FLC (MCO AA, BB and CC vs. FX CorDiax 80: 8.5 ± 0.54,11.3 ± 0.51,15.0 ± 0.53 vs. 3.6 ± 0.51 mL/min), and the removal of other medium-sized solutes was significantly greater, whereas the total mass of the albumin removal was moderate (medians \[range\] of MCO AA, BB and CC vs. FX CorDiax 80: 2.9 g \[1.5 -- 3.9\], 4.8 g \[2.2 -- 6.7\] and 7.3 g \[1.9 -- 9.7\] vs. \<0.3g\[\<0.3-\<0.3\]). In the second study, the performance of 2 MCO prototype dialyzers (MCO AA and MCO BB) in hemodialysis mode was compared with the performance of 2 high-flux dialyzers, FX CorDiax 80 and FX CorDiax 800 (Fresenius Medical Care) in hemodialysis and high-volume hemodiafiltration modes (post-dilution volume-controlled mode with a target total convective ultrafiltration volume of ≥23 L). The study was conducted at the dialysis center in Elsenfeld am Main, Germany. The primary outcome of this study was the λ-FLC overall clearance of MCO prototypes in hemodialysis mode in comparison to the high-flux dialyzers used in hemodialysis and hemodiafiltration mode. The results of this study indicated greater overall λ-FLC clearance by the MCO dialyzers in hemodialysis mode compared with both the HD and HDF treatments with the last-generation high-flux dialyzers (\[least squares mean (standard error)\]): MCO AA 10.0 (0.57), MCO BB 12.5 (0.57) vs. high-flux HD 4.4 (0.57) and HDF 6.2 (0.58) mL/min). The clearances of α1-microglobulin, complement factor D, κ-FLC, and myoglobin were generally greater for MCO than for high-flux HD and similar or greater than in HDF treatments, whereas the albumin removal was moderate with MCO but greater than that of high-flux HD and HDF (medians \[range\] of MCO AA and BB vs. FX CorDiax 80 and FX CorDiax 800; 3.2 g \[1.9 -- 3.9\] and 4.9 g \[1.1 -- 7.2\] vs. 0.2 g \[0.2 -- 0.9\] and 0.4 g \[0.3 -- 0.8\]). On the basis of both studies, the MCO membranes in hemodialysis mode demonstrate the effective removal of a wide range of middle molecules and exhibit substantially better performance than that of standard high-flux hemodialysis treatment, even exceeding the performance of high-volume postdilution HDF for large solutes, particularly λ-FLC. Because MCO membranes are associated with albumin removal in the range reported in the literature for HDF, the effects of albumin removal with MCO dialyzers on serum albumin levels would be expected to be similarly low, thus providing safe treatment in routine hemodialysis mode. Conclusions {#section5-ijao.5000603} =========== The newest generation of highly selective and permeable MCO membranes meets both requirements for high quality and good performance for dialysis treatment, featuring the removal of large middle-molecules up to a molecular weight of 45,000 Da typical of a high cut-off membrane, and the low removal of albumin as in state-of-the-art high-flux membranes. For uremic solutes in the 15,000- to 45,000-Da size range, MCO membranes offer improved clearance compared with that of high-flux membranes used in HD mode and equivalent clearance to that of high-flux membranes in high-volume HDF mode. The benefit is performance equivalent to that of high-volume HDF without a need for the online production of substitution fluid or for vascular access required for high blood flow rates. Therefore, the use of MCO membranes simplifies the delivery of high-removal treatment for ESRD patients, providing a removal spectrum that extends the current capabilities of the best therapy available. This improvement should allow clinicians to surpass the benefits provided by hemodiafiltration, which requires large amounts of high-quality fluid and a more complex setup, while utilizing regular HD equipment. These special MCO membranes should raise the standard of treatment available for all chronic HD patients, potentially decrease inflammatory responses, and generally improve patient outcomes. Future studies are planned to demonstrate the clinical benefits of this innovative product. Financial support: This work has been financed by Baxter International. Conflict of interest: All authors are employees of the Gambro Dialysatoren GmbH, Hechingen (Germany) or Gambro Lundia AB. Gambro AB (including all direct and indirect subsidiaries) is now part of Baxter International Inc. Baxter is a manufacturer of dialysis devices. None of the authors has proprietary interest. All experimental information is given in great detail to exclude any bias on the results.

{ "pile_set_name": "PubMed Central" }

Introduction {#Sec1} ============ Anthocyanin pigments are important flavonoid compounds that exhibit a wide range of biological functions in plants^[@CR1]--[@CR3]^, including as attractants for pollinators and seed dispersers and in protecting plants against abiotic and biotic stresses^[@CR4],[@CR5]^. More importantly, anthocyanins show beneficial effects to human health and exhibit potential protective functions against cancer and heart disease^[@CR6],[@CR7]^. These properties are partially attributed to their strong antioxidant capacity^[@CR8]^. According to studies in the model plant *Arabidopsis thaliana*, regulatory anthocyanin genes can be mainly divided into positive and negative regulatory genes based on whether they promote or inhibit the expression of structural genes for anthocyanin biosynthesis, respectively^[@CR9]^. Positive regulatory factors primarily include three types of genes: R2R3-MYB^[@CR10]^, basic helix-loop-helix (bHLH), and WD40 transcription factors^[@CR11]^. These genes promote the biosynthesis of anthocyanins. Additionally, there are two main types of negative regulatory genes, including the R3-MYB transcription factor^[@CR10],[@CR12]^ and the nitrogen-induced LBD transcription factor^[@CR13]^. The bioavailability and activity of anthocyanins and their regulatory genes vary widely across plants^[@CR5],[@CR14]^. "Zicaitai" (*Brassica rapa* L. ssp. *chinensis* var. *utilis*), a representative purple variety of *B. rapa*, exhibits purple petioles and flower stalks^[@CR15]^. The purple pigment of Zicaitai has been confirmed as anthocyanins^[@CR4]^. Although several studies have characterized anthocyanins in *Brassica* crops^[@CR16]--[@CR19]^, there is limited information on the genes involved in anthocyanin biosynthesis in Zicaitai. Quantitative trait locus (QTL)-Seq is a method that combines bulked segregant analysis (BSA) and high-throughput whole-genome re-sequencing to detect the major locus of a certain quantitative trait in a segregating population. BSA was proposed in 1991^[@CR20],[@CR21]^. It selects parents that show a contrasting phenotype on a trait of interest to build a segregating population---either F~2~ recombinant inbred lines, double haploid, or backcross populations---and then selects two groups of individual plants, each showing segregation of the trait to one of the parents, as two mixed pools to perform genotype analysis. The genomic position of the polymorphic molecular markers that shows significant segregation of genotypes is the region that harbors the major QTL. Currently, BSA has been updated to QTL-Seq through the replacement of traditional markers such as RAPD (random amplification polymorphic DNA) or RFLP (restriction fragment length polymorphism) to SNP (single nucleotide polymorphism) markers, accompanied by high-throughput re-sequencing and SNP-index analysis^[@CR22]^. QTL-Seq shows much higher efficiency than traditional QTL mapping, which is typically time-consuming and involves labor-intensive genotyping and maintenance of the mapping populations. However, QTL-Seq analysis always locates one locus for a trait at one time, while traditional QTL mapping may locate several loci for one trait in one experiment. QTL-Seq has been widely and successfully used in many crop populations, such as in the mapping of flowering traits in cucumber^[@CR23]^, tomato fruit weight^[@CR14]^, and 100-grain weight and root traits of chickpeas^[@CR24],[@CR25]^. In the present study, we performed QTL-Seq analysis and located different major loci to that obtained by linkage mapping using the same F~2~ population of Zicaitai and "Caixin". A simulated population was generated and analyzed, and the results supported that QTL-Seq and linkage mapping are able to detect different major loci under the combination of different modes of inheritance. We further assembled the draft-genome sequences of Zicaitai and Caixin. By combining the assembled sequences with the locus located by QTL-Seq, we determined the causal genes and their functional mutations responsible for anthocyanin enrichment in the Zicaitai accession. Materials and Methods {#Sec2} ===================== Plant materials and whole-genome re-sequencing {#Sec3} ---------------------------------------------- Zicaitai DH line ZCT095 was used as the receptor parent and Caixin DH line L58DH was used as the donor parent to construct an F~2~ population of 200 plants. The picture of the two parental lines have been shown in previous study^[@CR3]^. 30 plants exhibiting an extreme purple phenotype were selected from the population as the purple group, while 30 non-purple plants were selected as the green group. The frequency distribution on the anthocyanin accumulation in the F~2~ population were also shown in previous report, plants that have more than 0.4 mg/g dry weight of total anthocyanin content were selected and considered as the purple group. DNA from both parents and two groups of 60 samples was extracted from fresh leaves at the six-leaf stage. The DNA samples from the two groups were then combined into two pools for library construction. Following this, 100-bp pair-end reads were generated by the Illumina Solexa sequencing platform from BerryGenomics Biotech Co., Ltd. (Beijing, China). Variant calling with re-sequencing data {#Sec4} --------------------------------------- The paired-end reads of Zicaitai, Caixin, and the two sample pools were aligned to the *B. rapa* Chiifu reference genome version 3.0 using BWA^[@CR26]^ software with the method "mem"^[@CR27]^. Samtools^[@CR28]^ was then used to call the SNP and InDel variants from the aligned reads of Zicaitai and Caixin. Low quality (Q \< 10, DP \< 5), multi-allelic, or heterozygous variants in the VCF files of Zicaitai and Caixin were filtered out using Bcftools^[@CR29]^. Finally, SNP datasets were searched for polymorphic loci between Zicaitai and Caixin. Following this, the genotypes of these polymorphic loci were called out from the aligned reads of the two sample pools. Draft-genome assembly of the two parental lines {#Sec5} ----------------------------------------------- The adaptors, duplicates, and low-quality reads were filtered from the raw Illumina Sollexa sequencing reads produced from three DNA libraries with insert sizes of 180 bp, 350 bp, and 500 bp. The filtered datasets were then submitted for assembly into contig sequences using *SOAPdenovo* (Version r240)^[@CR30]^, implementing the default parameters. And then gaps in the assembly were closed using the tool GapCloser. Finally, the two parental genomes were assembled into scaffolds with N50 lengths of 4.5 kb and 6.5 kb, with a total length of 403.9 Mb and 350.8 Mb for Zicaitai and Caixin, respectively. The sequences of the draft assembly were deposited in BRAD database (<http://brassicadb.org/brad/datasets/pub/Zct_Cx/assemble/>). QTL-Seq analysis {#Sec6} ---------------- The SNP-index was calculated for all SNPs in the two pools of mixed samples, with the genotype of Zicaitai as the reference. During the calculation, we filtered SNPs with a SNP-index \<0.3 or \>0.7, which denotes a co-segregation of certain genotypes in the two pools. A 200 kb sliding window with a 20 kb increment was applied to slide across the genome, and the average value of the SNP-index was calculated in each window. The Δ(SNP-index) of Zicaitai (range from −1 to +1) was calculated by using the SNP-index of the purple pool minus the SNP-index of the green pool. We further repeated these calculations with the genotype of Caixin as the reference to obtain the Δ(SNP-index) of Caixin (range from −1 to +1). Finally, the Δ(SNP-index) of the population (range from −2 to +2) was calculated from the sum of Δ(SNP-index) of Zicaitai and Δ(SNP-index) of Caixin and plotted along the 10 chromosomes of *B. rapa* to show the signals detected by the QTL-Seq analysis. Linkage mapping {#Sec7} --------------- The software MapQTL (<https://www.kyazma.nl/index.php/MapQTL/>) version 6 was used to perform linkage analysis with genotype datasets from both populations of Zicaitai/Caixin and the simulation. The algorithm "MQM mapping" was selected to perform the calculation. The output logarithm of odds (LOD) scores were plotted along the genetic distances of the markers analyzed. Population simulation and analysis {#Sec8} ---------------------------------- The polymorphic positions between Zicaitai and Caixin in 10 chromosomes were used as coordinates to simulate two sets of genotype data. The two genotype datasets were considered as the homologous genotypes of two homozygous parental genomes. The combination of the two sets of genotypes (two parental haplotypes) resulted in the heterozygous status of a simulated F~1~ genome. Based on this F~1~ genome, we further simulated recombination events between the two parental haplotypes. Besides the random combinations of different chromosomes between the haplotypes, we simulated \~25 recombinations with 20% bias for both the pollen and egg haplotypes. It means 2.5 recombination events for each chromosome in average. We simulated 200 such haplotypes in total for both types. One egg haplotype was then randomly combined with one pollen haplotype to generate genotypes for the 200 F~2~ individuals. Six loci (different to that of Zicaitai and Caixin population) were further set to simulate trait-related genetic factors, with one locus showing dominant suppression (epistasis) on the other five loci. While the other five loci directly contribute to the phenotype level of the simulated trait and follow the recessive genetic model, with one locus designated as the major factor contributing 60% to the phenotype, and each of the other four loci contributing 10% to the phenotype. Based on these rules and the simulated genotype data of the 200 F~2~ individuals, we obtained the phenotype levels for all F~2~ individuals. These simulated data were then submitted to both QTL-Seq and linkage analysis. The perl script for the population simulation is available through the link: <http://brassicadb.org/brad/datasets/pub/Zct_Cx/simCode/>. Determination of candidate genes {#Sec9} -------------------------------- Based on the comparative genomic information between *B. rapa* and *A. thaliana*, we identified the pairwise syntenic-gene relationships between the two species. With the functional information of syntenic orthologs in *A. thaliana*, as well as our previous work on the determination of anthocyanin biosynthesis genes in *B. rapa*, we predicted the potential functions of candidate genes located in the QTL regions detected by QTL-Seq. We selected the intersection region of five sliding windows (200 kb sliding window with a 20 kb increment) that have the highest ∆SNP-index as the candidate interval, which are the peak signal in A07. Since the size of the sliding window was 200 kb and the increment was 20 kb, the intersection of five neighbor windows is 120 kb. The start and stop positions of the interval were then refined according to the locations of SNPs around its boundaries. It was found that there was one anthocyanin-synthesis related gene, *BrMYBL2.1*, located at this QTL region of A07. Experimental verification of a large-sequence insertion {#Sec10} ------------------------------------------------------- DNA was extracted from the fresh leaves of both Zicaitai and Caixin, and a pair of primers was designed to amplify the sequence that contains the insertion variant from the two DNA samples. The PCR system constituted 20 μL of PCR mixture with 2 μL of DNA (120 ng/μL), 10 μL of 2 × Rapid Taq Master Mix from the Vazyme Biotech Co., Ltd. (Nanjing, China), 0.8 μL (10 μM) of both forward and reverse primers, and 6.4 μL of ddH~2~O. The reaction mixture was incubated in a thermal cycler (9700, ABI, USA) at 95 °C for 3 min, followed by 35 cycles of 95 °C denaturation with 15 s for each cycle, following by primer annealing at 60 °C for 15 s, and a final extension at 72 °C for 5 min. The PCR products were further separated by electrophoresis on 1% agarose gel, running at 150V for 10 min, and then submitted to silver staining for band analysis. The PCR products were then sent to Majorbio Pharmaceutical Technology Co., Ltd. (Shanghai, China) for Sanger sequencing. The sequencing results were analyzed in BioEdit (<http://www.mbio.ncsu.edu/bioedit/bioedit.html>) and MUSCLE^[@CR31]^. Results {#Sec11} ======= BSA-Seq determines different major loci to those of linkage mapping {#Sec12} ------------------------------------------------------------------- An accession of Zicaitai was crossed with an accession of Caixin to create an F~2~ population of 200 individual plants showing segregation of the trait of anthocyanin enrichment (i.e., phenotypic variation in the purple color observed on the stem and leaf). We further selected two groups of F~2~ plants exhibiting an intense purple color, with one group containing 30 individuals showing intense purple and the other containing 30 samples lacking any purple. The DNA of the two groups of samples was extracted and combined into two respective pools. The two DNA pools together with two DNA samples of the two parents Zicaitai and Caixin were sequenced on the Illumina Solexa sequencing platform. The genomes of the two parents were sequenced to \~10× coverage for each (\~5-Gb Illumina re-sequencing data), while the two pools were sequenced to \~20× coverage for each (\~10-Gb data; Table [1](#Tab1){ref-type="table"}). Using the genome sequences of "Chiifu" as the reference, the reads were aligned and SNP and insertion/deletion (InDel) variants from the genomes of the two parents were called. Polymorphic loci of the SNPs and InDels between the two parents were selected for further analysis. The resequencing data of the two mixed pools were then aligned to the reference genome, and the genotypes of these polymorphic loci between the parents were called out from each of the two pools.Table 1The statistics on sequencing data generated for QTL-Seq and draft-genome assembly.StudiesSampleLibrary size (bp)Clean data (Gb)QTL-SeqZicaitai3505.31Caixin3505.42Pool 135010.99Pool 235011.01AssembleZicaitai18021.26Zicaitai5005.46Caixin18021.40Caixin5005.62 A significant signal was detected in chromosome A07, which differs from our previous linkage mapping results^[@CR3]^. A total of 1.19 million SNPs were identified between the two parental lines. In the F~2~ pool, approximately 1.02 million SNPs were detected at the polymorphic positions between the parental lines. Using the genotype datasets of the two parents and two pools, the SNP-index and Δ(SNP-index) were calculated to locate the QTL locus (see Methods). Figure [1](#Fig1){ref-type="fig"} shows a major locus that was detected in close proximity to the end of chromosome A07, which harbors the genetic factor that regulates the formation of extreme levels of purple color, i.e., contributes the major effect of anthocyanin enrichment. We previously performed linkage mapping with the same F~2~ population on the trait of anthocyanin variation and determined a major QTL on chromosome A09^[@CR3]^, which differs from the current major location identified by QTL-Seq. Additionally, a weak QTL signal was also detected in A07, suggests that these is a genetic factor located at A07 which also contributes to the anthocyanin enrichment in Zicaitai. Together with the information of physical positions of these genetic markers used in linkage mapping, we confirmed that the two methods located distinct major loci for anthocyanin enrichment in one F~2~ population---one locates at A07, while the other one locates at A09. The results suggest the existence of different genetic factors that may adopte different models of inheritance in the regulation of anthocyanin formation in this Zicaitai accession.Figure 1Major QTL loci of anthocyanin enrichment determined from an F~2~ population of Zicantai/Caixin. (**a**) QTL-Seq located at the major QTL locus at chromosome A07; the grey points are the Δ(SNP-index) of each SNP, the purple points are the average values of the Δ(SNP-index) in a 200-kb sliding window with a 20-kb increment across all chromosomes, and the red lines show the confidence intervals of the Δ(SNP-index). (**b**) The LOD scores generated by linkage analysis with the MEM algorithm using software MapQTL6.0. The major QTL locus was located at chromosome A09. Simulation experiments for assessing the inconsistency between the QTL-Seq and linkage analysis {#Sec13} ----------------------------------------------------------------------------------------------- We generated simulation data to explore the potential possibility of the inconsistent chromosomal locations detected by QTL-Seq and linkage analysis. Firstly, two sets of genotype data were simulated on the same positions of the polymorphic SNPs between Zicaitai and Caixin and were considered as genotypes of two in silico homozygous parents. Secondly, we simulated recombination events (cross-overs) between the two parental haplotypes (i.e., half of the homozygous genotypes), as occurs in the sex cells in F~1~ plants, with one paternal haplotype (pollen) or one maternal haplotype (egg) having \~25 cross-overs (\<0.2 bias). With this rule, we generated 200 such paternal and maternal haplotypes. Thirdly, we randomly selected one paternal haplotype and one maternal haplotype, and combined them into the genotype data of one simulated F~2~ plant---the process of selfing of F~1~ plants. We repeated this process and simulated 200 such F~2~ plants. Furthermore, a trait and its causal loci were simulated in one of the parents, and the information of their inheritance in the F~2~ population was surveyed and recorded for further mapping analysis. To design multiple factors regulated trait, we simulated six loci (Supplementary Table [S1](#MOESM1){ref-type="media"}) that were associated to the final trait, with five loci contributing directly to the level of the trait. Among the five loci, one contributed 60% to the final phenotype of the trait and followed the recessive genetic model, while each of the other four loci contributed 10% to the phenotype of the trait and also followed the recessive model. The last locus does not contribute directly to the trait, but shows epistatic effects on all the other five loci and follows the dominant genetic model. Under these rules, one of the parents shows 100% of the trait, while the other one shows 0% F~1~ also shows 0%, which is similar to the purple phenotype in the Zicaitai-Caixin population. Combining these rules with the simulated genotypes of the F~2~ population, we further calculated the trait level in the 200 simulated F~2~ individuals. We performed linkage mapping and QTL-Seq analysis with these simulated genotype datasets and the trait information, following the same analysis pipeline as that used in studies on the real F~2~ population of Zicaitai-Caixin. Figure [2a](#Fig2){ref-type="fig"} shows the inheritance patterns of the parental genomic fragments in each of the 200 F~2~ individuals. Linkage analysis identified a major locus on chromosome A02 (Fig. [2c](#Fig2){ref-type="fig"}), which corresponds to the simulated epistatic locus. Another smaller signal was detected on A04, which corresponds to the simulated major locus (60% contribution) that directly regulates the simulated trait. Additionally, we selected two groups from the F~2~ population, with one group of samples showing \>70% of the trait and the other showing 0% of the trait. Each group contained 25 samples and were recorded as a pool. We further simulated \~30 reads to cover the simulated SNP loci of the 25 samples in each pool and performed QTL-Seq analysis using the same method as that used in the real Zicaitai-Caixin population. The result showed that the QTL-Seq located a major locus at chromosome A04 (Fig. [2b](#Fig2){ref-type="fig"}), corresponding to the simulated major locus (60% contribution)---which is different to the major locus (chromosome A02) mapped by linkage analysis. This simulation analysis generated similar inconsistence between QTL-seq and linakge mapping analysis as that observed in the mapping of anthocyanin enrichment trait in the Zicaitai-Caixin population, supporting that there are two different loci following different genetic models relating to purple color regulation in Zicaitai.Figure 2QTL mappings in a simulated population with 200 F~2~ individuals. (**a**) The inheritance of parental genomic fragments in each of the F~2~ individuals; green fragments are from parent one, orange fragments are from parent two, while grey fragments are heterozygous genotype from both parents. (**b**) QTL--Seq analysis located the major QTL locus at chromosome A04. The colors denote similar objects as that in Fig. [1](#Fig1){ref-type="fig"}. The six black triangles denote the locations of the simulated loci. The bigger triangle at chromosome A04 is epistatic to the other loci. (**c**) Linkage analysis mapped the major QTL locus at chromosome A02. Draft-sequence assembly and fine-mapping of the causal gene {#Sec14} ----------------------------------------------------------- We screened candidate genes in the mapped regions with the annotation information and functional mutations of both genes. We previously predicted the anthocyanin biosynthesis genes in the genome of *B. rapa* by comparison with genes in *A. thaliana*^[@CR32]^. As reported in the linkage analysis, there are two such genes, *BrEGL3.1* (*BraA09g015240.3C*) and *BrEGL3.2* (*BraA09g013410.3C*), in the QTL region of A09^[@CR3]^ (Table [2](#Tab2){ref-type="table"}). They are two homeologs of the EGL gene, which is a positive regulator of anthocyanin biosynthesis in *A. thaliana*. Furthermore, we found one gene, *BrMYBL2.1* (*BraA07g036130.3C*), located almost at the midpoint of the major locus in A07 (Table [2](#Tab2){ref-type="table"}), which was detected by QTL-Seq. *BrMYBL2.1* is a homeolog of the MYBL2 gene, which, in contrast to EGL3, is a negative regulator of anthocyanin biosynthesis. We further assessed the sequencing depth of the *BrMYBL2.1* gene to determine if there was any structural variant that could not be determined by variant calling based on the re-sequencing reads mapping. By comparing the differences in sequencing depth between the two pools of purple and non-purple plants, we detected that in the third exon of *BrMYBL2.1*, a position in the purple pool exists that shows much lower sequencing depth than the non-purple pool (Fig. [3a](#Fig3){ref-type="fig"}).Table 2Candidate anthocyanin-biosynthesis genes in major QTL loci.QTL lociCandidate genesChromosomeStartStopIDStartStopA0725,112,90325,230,669*BrMYBL2.1*25,180,91825,181,774A097,230,3969,242,789*BrEGL3.1*8,904,7648,908,450*BrEGL3.2*7,821,1827,824,280Figure 3A large sequence insertion was found in gene *BrMYBL2* in Zicaitai. (**a**) The depth the reads covered on each nucleotide around the genomic region of gene *BrMYBL2*; purple points denote the reads depth from the sequencing data of purple pool in the QTL-Seq analysis, while green points denote reads depth from that of the green pool. (**b**) The gene model of *BrMYBL2* and the positions of a sequence insertion and four non-synonymous mutations in the third coding exon of this gene in Zicaitai. We assembled draft genome sequences of Zicaitai and Caixin to investigate the structural mutations in the candidate genes of anthocyanin enrichment between the parents. For both Zicaitai and Caixin, we generated \~80× coverage sequencing data in total (Table [1](#Tab1){ref-type="table"}). These data was used to assemble the two genomes. Finally, we obtained draft genome sequences for both Zicaitai and Caixin with assembled sizes of \~403-Mb and \~350-Mb, and scaffold N50 values of 4-kb and 6-kb, respectively (Table [3](#Tab3){ref-type="table"}). Although the two assembled genomes do not possess sufficient sequence contiguity, they are useful for the detection of structural variations in the coding sequence regions of the genes. Assembled sequences were then compared with candidate genes to assess for structural variations. We BLASTed the scaffolds of both Zicaitai and Caixin against the coding sequences of the candidate genes. The structural variations between Zicaitai and Caixin on each of these candidate genes were then observed on these aligned sequences. By doing this, we determined a large sequence insertion (with "N" as a gap in it) in the third exon of *BrMYBL2.1* in Zicaitai compared to that of Caixin, which overlapped with the region that shows lower re-sequencing depth in the purple pool than the non-purple pool (Fig. [3b](#Fig3){ref-type="fig"}). No such structural variations were found in the other candidate genes.Table 3The draft-genome-assembly information of the two parents.SampleAssembled size (Mb)Max length (bp)N50 (bp)Zicaitai403.9598,1824,576Caixin350.8195,9176,543 Mutation verification through Sanger sequencing {#Sec15} ----------------------------------------------- To verify the insertion variation in *BrMYBL2.1* of Zicaitai, we performed polymerase chain reaction (PCR) amplification of the sequence containing the insertion. The electrophoresis products indicated that the size of the amplified sequences differed between the two parents. The sequence amplified from Zicaitai was \~100-bp longer (estimated based on the size of the DNA ladder) than that of Caixin (Fig. [4a](#Fig4){ref-type="fig"}), which is consistent with previous results of the comparisons between the assembled sequences. The PCR products were further submitted to Sanger sequencing, and the results showed that there was a sequence insertion containing "GGGAATCGATCCAACTTTGTTTC" linked with poly-"A"s to the gene in Zicaitai (Fig. [4b](#Fig4){ref-type="fig"}) in comparison with Caixin. This confirmed the sequence insertion detected by the sequence assembly. Sanger sequencing peaks denoting different types of nucleotides were observed to be overlapping with each other following the poly-A and could not be easily recognized (high level of heterozygous peaks) for determining the accurate size and sequences of the insertion variation. However, the \~100-bp sequence insertion in the third coding exon of *BrMYBL2.1* changed the coding protein in Zicaitai. The gene *BrMYBL2.1* in chromosome A07 is a negative regulator of anthocyanin biosynthesis, while *BrEGL3.1*, located at the QTL region of A09, is a paralog of EGL3, which is a positive regulator of anthocyanin biosynthesis. The functional mutations of the two genes both contributed to the variation in anthocyanin enrichment of Zicaitai. The different roles of the two genes in the biosynthesis of anthocyanins explains the different results of the QTL-Seq and linkage mapping, even when using the same population.Figure 4Experimental verification of a sequence insertion in gene *BrMYBL2*. (**a**) The electrophoretogram of the PCR products of the target sequence that contains the insertion variant. Three replicates for both Caixin and Zicaitai were labeled as one to three. (**b**) The alignment of the target sequences from the Chiifu reference genome (Chiifu_V3.0), draft sequence assembly of Caixin (Caixin-ass) and Zicaitai (Zicaitai-ass), and Sanger sequencing results (Caixin-F and Caixin-R, Zicaitai-F and Zicaitai-R, F denotes forward sequence, R denotes reverse sequence). The red color denotes the inserted sequence. Discussion {#Sec16} ========== QTL-Seq and linkage analysis using the same segregating population can identify different major loci. In this work, we performed QTL-Seq analysis on the purple phenotype (i.e., the trait of anthocyanin enrichment) in an F~2~ population of Zicaitai-Caixin and compared the result with our previous linkage-analysis study using the same population. The two methods located different major anthocyanin enrichment loci in Zicaitai, which indicated that two major genetic factors contribute to the trait and have different roles and/or follow different genetic models in the regulation of anthocyanin enrichment. In order to test this, we generated a simulated F~2~ population with two major loci regulating a trait under different genetic models. The first directly contributes to the phenotype under a recessive model, while the second is epistatic to the first locus and follows a dominant model. QTL-Seq and linkage mapping were then conducted on the simulated population using the same pipeline as that applied to that of the Zicaitai-Caixin population. The results showed that QTL-Seq and linkage mapping located different major loci from that observed in the Zicaitai/Caixin population. QTL-Seq located the major locus that regulates the trait directly, while the linkage mapping located the locus that is epistatic to the previous locus. It is reasonable that the sampling of two pools with opposite and extreme traits will cause the QTL-Seq to detect a locus associated with a high level of the trait, while linkage analysis will locate the gene that shows a clearer inheritance pattern. Inspired by the simulation analysis, we speculated that these two major factors identified through QTL-Seq and linkage mapping in Zicaitai-Caixin F~2~ populations should have different regulation pathways and/or follow different genetic models relating to anthocyanin biosynthesis. Actually, the *BrMYBL2* gene detected by QTL-Seq function as a negative regulator and may contribute directly to the purple trait in Zicaitai. While in the linkage analysis, the positive regulator gene *EGL3* with potentially epistatic function was localized to mediate the anthocyanin biosynthesis. The complexity of the regulation mechanisms on anthocyanin biosynthesis in *Brassica*s were highlighted by recent studies. Many efforts have been devoted to dissect the genes that contribute to the purple traits in different *Brassica* crops. In *Brassica napus*, *BnAPR2*, encoding an adenosine 5'-phosphosulfate reductase, at the end of A03 chromosome was identified as an incomplete dominant regulatory gene through map-based cloning^[@CR33]^. Another gene *BrMYB73* was also mapped to the end of A03 in *B. rapa*^[@CR34]^. *BrMYB73* was predicted to encode a R2R3-MYB transcription factor, and one deletion and one SNP were found in this gene in purple-leaf parent. Recently, Li *et al*. identified a *bHLH49* transcription factor (*BrbHLH49*) on A07 that might positively regulate the anthocyanin accumulation in Zicaitai based on a specific-locus amplified fragment sequencing method^[@CR19]^. These studies revealed the genetic diversity of anthocyanin biosynthesis related genes in *Brassica* crops. QTL mapping together with draft genome assembly is an efficient means of fine-mapping causal genes and mutations from segregating populations. In QTL analysis, considering the limited recombination events in a given population, the mapping resolution is not always sufficient to locate the causal genes, and dozens or hundreds of candidate genes can make the fine-mapping of the gene difficult. Furthermore, with variants called from the re-sequencing data, only small-scale variants, such as SNPs and short InDels, can be analyzed in these candidate genes, which results in the major loss of large-scale functional mutations during the screening of candidate genes. However, the draft genome assembly of parental genomes and pairwise comparisons of coding sequences of candidate genes between parents will capture these large-scale mutations, thus resulting in the rapid fine-mapping of the causal genes and mutations of the traits. In this work, we performed draft genome assembly of Zicaitai and Caixin. Together with the mapping results from the QTL-Seq, we located a large sequence insertion in the coding exon of an anthocyanin biosynthesis-related gene. The insertion changed the sequence of the translated protein of the genes in Zicaitai in comparison to Caixin. Considering that high-throughput sequencing is becoming increasingly affordable, QTL mapping accompanied by the draft sequence assembly of parental genomes could constitute an efficient means of fine-mapping the genes in segregation populations. Low-depth re-sequencing of individual samples from a segregating population is a superior option for analyzing a trait that is regulated by more than one major locus. In this study, we located two genomic regions that contribute to the enrichment of anthocyanins in Zicaitai; one being a positive regulator of anthocyanin biosynthesis and the other being a negative regulator of anthocyanin biosynthesis. The two genes have different roles in anthocyanin enrichment in Zicaitai. In similar cases, QTL-Seq as well as linkage mapping may not be able to locate all of these major loci in one experiment. Therefore, we propose a strategy of low-depth re-sequencing of all F~2~ individuals. With the re-sequencing data, a genetic map with a high density of bin markers (combined markers in a local region) can be constructed, allowing for efficient and simple linkage mapping. More importantly, different samples can be repeatedly combined into two extreme pools based on different rules, and QTL-Seq can be performed as many times as needed. This strategy should help to locate multiple major loci that contribute to one or more traits in a segregating population. Supplementary information ========================= {#Sec17} Supplementary Table S1. **Publisher's note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Xin Zhang and Kang Zhang. Supplementary information ========================= is available for this paper at 10.1038/s41598-020-58916-5. The work was supported by the National Program on Key Research Project (2016YFD0100307), the National Natural Science Foundation of China (NSFC grants 31301771 and 31722048), the Science and Technology Innovation Program of the Chinese Academy of Agricultural Sciences, and the Key Laboratory of Biology and Genetic Improvement of Horticultural Crops, Ministry of Agriculture, P.R. China. F.C. and X.W. designed the experiments. F.C. and X.Z. analyzed the data. X.Z., J.W., N.G. and J.L. carried out the experiment. K.Z. participated in data analysis and interpretation. F.C., X.Z. and K.Z. wrote this manuscript. The authors declare no competing interests.

{ "pile_set_name": "PubMed Central" }

Introduction ============ Severe Acute Respiratory Syndrome (SARS)-CoV and Middle East Respiratory Syndrome (MERS)-CoV are two closely related zoonotic coronaviruses. Both have successfully crossed the species barrier to allow animal-to-human transmission, and further to allow human-to-human transmission ([@evw246-B42]; [@evw246-B37]). The SARS outbreak in 2003 had a mortality rate of 10% ([@evw246-B1]), and SARS-CoV was considered the most aggressive coronavirus compared to other human coronaviruses that commonly cause mild to moderate infection in their hosts ([@evw246-B45]). MERS-CoV is the cause of an ongoing outbreak of the respiratory illness MERS ([@evw246-B10]). At the time of writing, 1791 MERS cases have been confirmed with a mortality rate of approximately 35% ([@evw246-B48]). Both MERS and SARS have higher mortality rates in elderly and immunosuppressed populations ([@evw246-B19]). The host changes by MERS-CoV and SARS-CoV suggest that other coronaviruses can potentially cross the species barrier, become zoonotic, and enable human-to-human transmission, ultimately causing high morbidity and mortality. SARS-CoV and MERS-CoV exploited mechanistically different approaches to overcome the human species barrier, but these two viruses have a lot in common ([@evw246-B29]). Here, we aim to identify the vulnerable regions in the proteomes of coronaviruses that neither SARS-CoV nor MERS-CoV nor their contemporary and forthcoming relatives can proliferate without, and address how to mobilize a defense against the present and future coronaviruses by targeting these regions. SARS-CoV and MERS-CoV are positive (+)-strand RNA viruses encoding approximately 25 protein products. The MERS-CoV proteome is primarily composed of two polyproteins, ORF1a and ORF1ab; the latter is generated by a -1 ribosomal slippage frameshift. These proteins are cleaved into 16 nonstructural proteins (NSPs). NSPs 1--10 are products of both polyproteins, whereas NSPs 12--16 are only yielded by ORF1ab. NSP11 is unique to ORF1a ([@evw246-B44]). Structural proteins envelope (E), spike (S), membrane (M), and nucleocapsid (N) are elements of the physical structure that encloses the viral genome and come from distinct reading frames, unlike ORF1a and ORF1ab, which come from overlapping reading frames. Additionally, the structural proteins are the product of subgenomic mRNAs that are joined during discontinuous negative RNA strand synthesis ([@evw246-B44]). Finally, NS3 protein (NS3), NS4A protein (NS4A), NS4B protein (NS4B), NS5 protein (NS5), and Orf8b protein encompass the remainder of the proteome and also arise from distinct reading frames ([@evw246-B44]). Our approach utilizes genomic sequence data, which is readily available for viruses known to cause disease. However, because most viruses pose no major threat to their host, they pass by unnoticed leaving the majority of virus genome space uncharted. With the availability of cost-efficient genome sequencing technology, and recent developments in the field of viral metagenomics, large-scale identification of viral genome space is on the rise ([@evw246-B39]; [@evw246-B34]). By exploring viral diversity, critical components constituting a viral genus' fitness can be evaluated. Examples such as the common influenza virus illustrate the rapidity of viral gene mutation and in order to maintain immune protection, an annual flu vaccination is recommended. Underway efforts aim to generate broadly neutralizing vaccines whose design accounts for the genomic sequences of multiple types of influenza virus to eliminate frequent re-vaccination against the flu ([@evw246-B16], [@evw246-B17]). Development of broadly neutralizing vaccines often relies on the consensus or ancestral sequences of extant viral sequences in order to provide greater coverage for related viruses ([@evw246-B26]). Unfortunately, consensus sequences can be misleading, and ancestral sequence reconstruction is error-prone for quickly diverging sequences ([@evw246-B32]). In addition, viruses with compact genomes often express proteins with structural disorder that may undergo structural transformations. Although these transformer proteins, like VP40 in Ebola, are masters at changing their structure, and thus expanding their functional repertoire as needed for the life cycle of the virus ([@evw246-B4]), flexible regions are potentially important in rewiring protein--protein interactions between the virus and its host ([@evw246-B5]; [@evw246-B35]; [@evw246-B18]). The flexibility trait of many viral proteins is a complicating factor in vaccine development. For instance, Dengue virus exhibits serotype-specific antibody affinity that causes antibody-dependent enhancement, an obstacle in the development of Dengue vaccines that protects against all four serotypes ([@evw246-B15]). To overcome the hurdle posed by structural flexibility, we propose an additional screening step in identifying potential vaccine or antiviral targets that considers the structural flexibility of the viral proteins. The Structural Genomics Initiatives increased their success rate by excluding proteins predicted to be structurally disordered ([@evw246-B41]). A similar approach can perhaps benefit vaccine development. Furthermore, to make this approach robust to potential mutations, minimizing loss in efficacy or resistance, the evolutionary context of sequence and structure must be considered. Thus, we suggest expanding the concept of broadly neutralizing vaccines/antivirals by increasing the diversity of viruses considered if possible. Sites conserved for sequence, structure, and with low disorder propensity among diverse virus protein homologs are very likely to be constrained from 1) changing sequence on evolutionary time scales and 2) undergoing real-time structural transitions. These sites have potential as targets for broad-specificity antivirals or vaccines because conservation makes them broad-specificity and low dynamics avoids targeting a conformational ensemble, which is not only difficult ([@evw246-B51]), but that may change as the sequence diverges ([@evw246-B40]). A recent large-scale study of structural disorder in \>2,000 viral genomes in 41 viral families found the amount of disorder in different virus families varying from 2.9% to 23.1% ([@evw246-B36]). It was reported that *Coronaviridae* has very low disorder content (mean disorder 3.68%) ([@evw246-B36]). *Coronaviridae* contains two subfamilies: *Coronavirinae* and *Torovirinae*. SARS-CoV and MERS-CoV are part the *Coronavirinae* subfamily, from here on referred to as coronavirus (CoV). The lack of disorder is intriguing because it may be important for rewiring interactions between viral proteins and host proteins ([@evw246-B35]) and providing opportunities to acquire novel functional sequence motifs ([@evw246-B18]). Structural disorder has also been proposed to be important for viral viability, enabling multifunctionality and vigor in response to changes in the environment ([@evw246-B49]). Given the low fraction of structural disorder reported across *Coronaviridae*, we set out to investigate the conservation of structural disorder and secondary structure across CoV. Sites identified as conserved for structure and lacking disorder can be considered to be vulnerable and druggable in the proteomes of coronaviruses. The structural divergence capacity of these regions is limited, leaving a wider range of the present and emergent coronaviruses susceptible to the effects of potential broadly neutralizing anti-CoV therapies targeting these sites. We will refer to these sites as target sites. Materials and Methods ===================== Protein Family Reconstruction ----------------------------- Protein sequences were identified by individual BLAST searches with MERS-CoV (Taxonomy ID: 1335626) proteins ORF1ab (YP_009047202.1; polyprotein), S protein (YP_009047204.1), M protein (YP_009047210.1), E protein (YP_009047209.1), and N protein (YP_009047211.1) against coronaviruses. BLAST searches of the ORF1ab protein were performed, using start and end positions as detailed in the ORF1ab NCBI Reference Sequence file, against the refseq_protein database. The sequences retrieved from the BLAST output maintained the following cutoff: \>30% sequence identity and \>50% coverage relative to MERS-CoV sequence query. The 30% sequence identity and 50% query coverage cutoff strikes a balance between alignment quality and at least 10 sequences for most protein families. NSP1 (YP_009047202.1; 1-193), NSP2 (YP_009047202.1; 194-853), NS3 (YP_009047205.1), NS4A (YP_009047206.1), NS4B (YP_009047207.1), NS5 (YP_009047208.1), ORF8b protein (YP_009047212.1), and NSP11 (YP_009047203.1; 4378-4391) are not included in this study due to \<10 BLAST hits. Multiple sequence alignments were constructed for the selected BLAST hits using MAFFT ([@evw246-B25]). Phylogenetic trees were constructed using MrBayes 3.2.2 with a four category gamma distribution and the mixed model for amino acid substitution ([@evw246-B20]; [@evw246-B38]). Each tree ran for five million generations, with a sample frequency of 100. The final tree was constructed from the last 75% of samples, discarding the first 25% of samples as the default burnin, and using the half-compatible parameter, to avoid weakly supported nodes (i.e., with a posterior probability \<0.5). All trees were midpoint rooted. For every protein family, the amino acid substitution rate per site in its multiple sequence alignment was calculated using empirical Bayesian estimation as implemented in Rate4Site ([@evw246-B31]). Substitution rates were calculated using 16 gamma categories, the JTT substitution matrix ([@evw246-B24]), and the reconstructed phylogenies. The rates were normalized per protein family with an average across all sites equal to zero and SD equal to 1. This means that sites with a rate \<0 are evolving slower than average, whereas sites with a rate \>0 are evolving faster than average. Prediction of Intrinsic Disorder Propensity and Secondary Structure ------------------------------------------------------------------- Intrinsic disorder propensity was inferred using two different predictors: IUPred (default settings; "long" option) ([@evw246-B11], [@evw246-B12]) and DISOPRED2 ([@evw246-B46]) for all proteins. For IUPred, the site-specific continuous disorder propensities for each protein were mapped onto their corresponding position in the multiple sequence alignment as raw disorder propensities and as binary states, order or disorder, using two cutoffs of 0.4 and 0.5. Disorder propensities below the cutoff were assigned order and disorder propensities at the cutoff or above were assigned disorder. For the DISOPRED2 predictions that were inferred using the nr database, the continuous disorder propensities for every site in a protein were mapped onto their corresponding position in the multiple sequence alignment as raw disorder propensities and as binary states, order or disorder, using a cutoff of 5. Consequently, for every protein family (a multiple sequence alignment and its corresponding phylogenetic tree), two continuous matrices and three binary matrices resulted: IUPred 0.4, IUPred 0.5, and DISOPRED2. An additional matrix was generated to indicate sites where the binary order and disorder assignments differ between IUPred 0.4 and DISOPRED2. A similar methodology was employed to analyze secondary structure predicted by PSIPRED ([@evw246-B33]) and JPred ([@evw246-B13]). For both predictors, the uniref90 database was used and sites were classified as loops, alpha helices, or beta strands and mapped back onto their corresponding sites in the multiple sequence alignment. This resulted in two three-state matrices for each protein family alignment, one for each predictor, and two binary matrices displaying secondary structure elements (alpha helix and beta strand) or loops. An additional matrix was generated to indicate sites where the secondary structure assignments differ between PSIPRED and JPred. For every protein family, the binary matrices resulting from the different disorder predictions and from the different secondary structure predictions were analyzed in the corresponding evolutionary context using GLOOME. GLOOME (Gain-Loss Mapping Engine) analyzes binary presence and absence patterns in a phylogenetic context ([@evw246-B8]). In this study, the Rate4Site option in GLOOME was used to analyze the binary matrices (IUPred 0.4, IUPred 0.5, DISOPRED2, PSIPRED, and JPred) with the corresponding phylogenetic trees to map change of state across sites in each individual protein phylogeny ([@evw246-B9]; [@evw246-B8]). GLOOME was run with 16 gamma categories and a substitution matrix set to equal rates within each state and transitions between states treated equally. From the binary disorder and order matrices, transition rates between disorder and order or vice versa (DOT) were estimated. From the binary structure and loop matrices, transition rates between structure and loop or vice versa (SLT) were estimated. Similar to Rate4Site, the rates were normalized per protein family with an average across all sites equal to zero and SD equal to 1. This means that sites with a rate \<0 are evolving slower than average, while sites with a rate \>0 are evolving faster than average. Protein Family Visualization ---------------------------- Protein families were visualized in an integrative manner with a phylogenetic tree, any matrix (multiple sequence alignment or predictor based) displayed as a heatmap, and site-specific sequence transition rates using Python packages ETE3 ([@evw246-B21]) and Matplotlib ([@evw246-B22]). Statistical Analysis of Amino Acid Evolutionary Rate Distributions ------------------------------------------------------------------ Amino acid evolutionary rates (SEQ) for all sites across all alignments were aggregated and binned into four possible categories characterized by the distribution of PSIPRED predicted secondary structure at each site. Sites predicted to have a loop across all sequences are "conserved loops; C(L)" and sites predicted to have a helix across all sequences or a strand across all sequences are "conserved helix-strand; C(HS)" ([table 3](#evw246-T3){ref-type="table"}). Sites predicted to have all three states (helix, strand, and loop) or any combination of loop and one other state are "non-conserved helix, loop, strand; NC(HLS)" and sites predicted to have a mixture of helix and strand are "non-conserved helix-strand; NC(HS)" ([table 3](#evw246-T3){ref-type="table"}). In all cases, gaps were ignored when classifying combinations of secondary structure at a site or if secondary structure conservation exists at a particular site. Results ======= Phylogenies ----------- Phylogenies were built for all protein products encoded in the MERS-CoV single-stranded RNA genome, except for NSP1, NSP2, NS3, NS4A, NS4B, ORF8b protein, and NSP11, all of which had insufficient sequence data (\<10 sequence hits with BLAST). NSP12 is often used as a measure for newly identified coronaviruses. According to the International Committee of Taxonomy of Viruses, a major criterion in determining if a coronavirus is considered novel is pairwise sequence identity below 90% for NSP12 in all comparisons to previously known coronaviruses ([@evw246-B3]). Four main clades, alphacoronavirus, betacoronavirus, gammacoronavirus, and deltacoronavirus ([fig. 1](#evw246-F1){ref-type="fig"}), are identified in agreement with the taxonomic classifications described by the ICTV ([@evw246-B23]). Coronaviruses not listed by the ICTV are assumed to be a part of the clade in which representatives with known classifications are situated in our NSP12 phylogeny. ![CoV representative phylogeny. (*A*) NSP12 is a representative for the CoV protein phylogenies, colored by clade (alphacoronavirus or FIPV, green; betacoronavirus has four different subclades: SARS, blue; MERS, gray; HKU1, pink; EQU, purple; gammacoronavirus or SW1, yellow; deltacoronavirus or HKU19, white.) Posterior probability indicating node support is shown in red. (*B*) Protein family distribution across coronavirus based on the given cutoff (\>30% sequence identity and \>50% coverage relative to MERS-CoV sequence query). Clade color applied throughout the remaining figures. Areas shaded in gray with an arrow indicate that the protein family is not identified for that clade with the given cutoffs, but is found from the arrow tip. See [Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online, for the remaining phylogenetic trees.](evw246f1p){#evw246-F1} The MERS clade and SARS clade are sister clades in the NSP12 phylogeny. The HKU1 clade and EQU clade are also sister clades. Together these four clades form the Betacoronavirus clade, in accordance with the ICTV classification ([@evw246-B23]). Betacoronavirus is represented in all phylogenies although the order of the individual subclades varies. Alphacoronavirus is often found as the sister clade or outgroup to betacoronavirus. Deltacoronavirus or gammacoronavirus are the most distantly related to the betacoronavirus. In the nucleocapsid phylogeny, gammacoronavirus is the first outgroup clade to betacoronavirus, and alphacoronavirus is the most distant outgroup. Most NSP trees exhibit some unresolved nodes at junctures immediately preceding terminal nodes. As an effect of the 50% majority rule, most of the 546 resolved nodes are well supported with posterior probability \>0.9 for 82% and \>0.99 for 68% ([Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online). Most trees follow the NSP12 topology for the main clades, with minor clade rearrangements. It should be noted that for NSP5, the entire alphacoronavirus clade is placed within the betacoronavirus clade, as a sister clade to the MERS clade ([Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online). This may be due to increased sequence divergence rates or due to recombination. Recombination events are rather frequent in coronaviruses ([@evw246-B43]), and the MERS clade potentially underwent multiple recombination events as part of the host change ([@evw246-B52]). The phylogenies for membrane protein, spike protein, NSP5, and NSP8--NSP16 demonstrate (with the given BLAST cutoffs) recoverable protein homologs such that all coronaviruses are represented (i.e., all coronaviruses represented in the NSP12 phylogeny). Nucleocapsid, NSP4, and NSP7 have recoverable homologs in all clades except deltacoronavirus. NSP3 and NSP6 homologs are too divergent in deltacoronavirus and/or gammacoronavirus relative to MERS-CoV. Envelope appears specific to betacoronavirus ([fig. 1](#evw246-F1){ref-type="fig"}), but it is a short protein that has been found to diverge rapidly and is likely present outside betacoronavirus ([@evw246-B14]). Because different protein families yield slightly different phylogenies, for the remaining evolutionary analyses, every protein family was analyzed in the context of its own phylogeny. Intrinsic Disorder Is Rarely Conserved -------------------------------------- For all protein families, structural disorder propensities were predicted using IUPred ([@evw246-B11], [@evw246-B12]) and DISOPRED2 ([@evw246-B46]). To verify the robustness of the binary IUPred and DISOPRED2 predictions, the binary assignments were compared on a site-by-site basis ([table 1](#evw246-T1){ref-type="table"}). When converted to binary (i.e., two states per site disordered or ordered) IUPred 0.4 and IUPred 0.5 are in good agreement with the larger differences seen for NSP8, NSP9, and nucleocapsid (7.5%, 6.5%, and 19.0%, respectively) ([table 1](#evw246-T1){ref-type="table"}). Comparing IUPred 0.4 or IUPred 0.5 to DISOPRED2, large differences are in particular seen for nucleocapsid (38.7% and 29.7% respectively) and NSP8 (23.5% and 25.9%, respectively) ([table 1](#evw246-T1){ref-type="table"}). For nucleocapsid, regions that are found to be disordered by IUPred 0.4 are found to be ordered by IUPred 0.5 and DISOPRED2 ([fig. 2](#evw246-F2){ref-type="fig"} and [Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online). For NSP8, regions that are only slightly disordered in a few sequences according to IUPred 0.4 and IUPred 0.5, DISOPRED2 predicts disorder to be conserved for all sequences ([fig. 3](#evw246-F3){ref-type="fig"}). ###### Protein Family Wide Disagreement of Disorder and Secondary Structure Predictions --------------------- ![](evw246ie1p.jpg) --------------------- Tukey boxplot constructed using the IUPred 0.4 predicated disorder fraction (number of disordered sites/total sites) per sequence per protein. Green dots represent outliers; red diamond are the mean and red lines are the median values. ![The evolutionary context of intrinsic disorder in nucleocapsid. The phylogenetic tree was built using the multiple sequence alignments for nucleocapsid. Here, the multiple sequence alignment is colored by disorder propensity (with gaps in gray): (*A*) IUPred 0.4, blue-to-white-to-red shows disorder propensity according to the scale for IUPred 0.4. (*B*) IUPred 0.5, blue-to-white-to-red shows disorder propensity according to the scale for IUPred 0.5. (*C*) DISOPRED2, blue-to-white-to-red shows disorder propensity according to the scale for DISOPRED2. Above the heat maps, the normalized evolutionary rates per site for amino acid substitution (SEQ) and the DOT for the binary transformations of *A*--*C* are shown. Heat maps visualized with the Python packages ETE3 ([@evw246-B21]) and Matplotlib ([@evw246-B22]).](evw246f2p){#evw246-F2} ![The evolutionary context of intrinsic disorder in NSP8. The phylogenetic tree was built using the multiple sequence alignments for NSP8. (*A*) The multiple sequence alignment is colored by amino acid according to scale, arranged based on TOP-IDP disorder promoting propensity of the amino acids ([@evw246-B7]), and gray denotes gaps. (*B*) IUPred disorder propensity per site in the multiple sequence alignment. Blue-to-white-to-red shows disorder propensity according to the scale for IUPred 0.4. (*C*) IUPred disorder propensity per site in the multiple sequence alignment. Blue-to-white-to-red shows disorder propensity according to the scale for IUPred 0.5. (*D*) DISOPRED2 disorder propensity per site in the multiple sequence alignment. Blue-to-white-to-red shows disorder propensity according to the scale. Above the multiple sequence alignment, the normalized evolutionary rates per site for amino acid substitution (SEQ) and the DOT for the binary transformations of *B*--*D* are shown. Heat maps visualized with the Python packages ETE3 ([@evw246-B21]) and Matplotlib ([@evw246-B22]). See [Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online for additional graphics for every protein family.](evw246f3p){#evw246-F3} To quantify the fraction of disordered sites per protein family, we report the IUPred 0.4 results only for simplicity ([table 1](#evw246-T1){ref-type="table"}). In general, IUPred 0.4 predicts more disorder than DISOPRED2, but several protein families have almost no disordered sites. NSP3 and NSP8-10 have some variation in disorder content for different viruses. Based on the fraction of disorder, nucleocapsid is the only highly disordered protein among the CoVs in this study, even if NSPs 8--10 have outliers that are \>20% disordered. To compare the disorder-to-order transition rates (DOT) for all protein families where the binary matrices of disorder and order include both states, the quadrant count ratio (QCR) was estimated as a measure of association in assigning slower than average vs. faster than average transition rates. For IUPred 0.4 vs. IUPred 0.5, for IUPred 0.5 vs DISOPRED2, and for IUPred 0.4 vs. DISOPRED2, the QCRs were 0.76, 0.69, and 0.63, respectively. This shows a strong positive association for site-specific DOT for all methods and cutoffs, with IUPred 0.4 vs. IUPred 0.5 being the strongest ([table 2](#evw246-T2){ref-type="table"}). For nucleocapsid and NSP8, the positive associations are weaker, suggesting that many sites have IUPred disorder propensity in the 0.4 to 0.5 range and large differences between IUPred and DISOPRED2, in accordance with the large disagreement between the binary assignment of these predictors ([tables 1](#evw246-T1){ref-type="table"} and [2](#evw246-T2){ref-type="table"}). ###### QCR[^a^](#evw246-TF2){ref-type="table-fn"}**for DOT and SLT** Rate -------------- -------------------------------------------- ------ ------ ------ NSP3 0.75 0.68 0.61 0.51 NSP4 N/A[^b^](#evw246-TF3){ref-type="table-fn"} N/A 0.58 0.61 NSP5 0.72 0.84 0.73 0.65 NSP6 N/A N/A N/A 0.70 NSP7 N/A N/A 0.9 0.66 NSP8 0.86 0.43 0.38 0.67 NSP9 0.76 0.55 0.62 0.67 NSP10 0.93 0.67 0.68 0.42 NSP12 0.96 0.89 0.86 0.57 NSP13 0.76 0.70 0.59 0.36 NSP14 0.87 0.85 0.8 0.58 NSP15 0.82 0.85 0.67 0.53 NSP16 0.93 0.86 0.85 0.59 Envelope N/A N/A 0.58 0.53 Membrane 0.54 0.57 0.67 0.55 Nucleocapsid 0.33 0.43 0.31 0.61 Spike 0.81 0.62 0.49 0.55 All 0.76 0.69 0.63 0.55 QCR: Quadrant Count Ratio measures the association for the same site-specific rate with different predictors or cutoffs. N/A: at least one of the rates in the comparison could not be estimated due to the lack of any disordered state in the binary state matrix ([Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online). Secondary Structure Prediction and Structure-to-Loop Transitions ---------------------------------------------------------------- For all protein families, secondary structure elements were predicted using PSIPRED ([@evw246-B33]) and JPred ([@evw246-B13]). For most protein families, the disagreement between secondary structure predictors is greater than for the disorder predictors ([table 1](#evw246-T1){ref-type="table"}). In fact, 15 of the 17 protein families compared disagree at more than 10% of alignment sites, and two of these disagree at more than 20% of sites. To compare the binary structure-to-loop transitions (SLT), QCR was estimated as a measure of association for SLT based on the different predictors. In general, there is a moderate positive association between SLT for PSIPRED vs. SLT for JPred that is weaker than for the different DOT comparisons ([table 2](#evw246-T2){ref-type="table"}). It should be noted that SLT does not differentiate between alpha helix and beta strand, but considers both as "structure." This is a correct assumption if protein structure is conserved and consistently predicted, but for some protein families that is not the case. Four protein families (NSP3, NSP12, NSP13, and SPIKE) have more than 40% of their sites found within the NC(HLS) category with non-conserved helix, strand, and loop (two or three states present at the same site) ([table 3](#evw246-T3){ref-type="table"}). For NSP13, JPred predicts 72% of all sites to be a mixture of helix, strand, and loop, or any combination of loop and one other structural element ([fig. 4](#evw246-F4){ref-type="fig"}). Envelope and NSP6 have 13% and 12% of their respective sites in the NC(HS) category. Considering only the PSIPRED predictions, the NC(HS) category has 245 sites across all 17 protein families. That is one-tenth the size of the next smallest set which is C(HS) with 2275 sites. Next, C(L) has 3344 sites, and the largest category is NC(HLS) with 4257 sites. Comparing the evolutionary sequence rates for the sites in the different categories, based on PSIPRED predictions only, reveals that sites in the C(HS) category are evolving at a slower rate than all other categories. NC(HS) is only just significantly different (*P* = 4.62E−03) from C(HS), and is not significantly different from NC(HLS) and C(L) (*P* = 1.85E−02 and *P* = 8.33E−01, respectively). However, NC(HLS) and C(L) are significantly different from each other, and both are significantly different from C(HS) (*P* = 1.82E−46 and *P* = 2.33E−21, respectively) ([fig. 5](#evw246-F5){ref-type="fig"}). ![The evolutionary context of secondary structure in NSP13. The phylogenetic trees were built using the multiple sequence alignments for NSP13. (*A*) The multiple sequence alignment is colored as in [fig. 3](#evw246-F3){ref-type="fig"}. (*B*) PSIPRED secondary structure prediction per site in the multiple sequence alignment, color coded according to the scale. (*C*) JPred secondary structure prediction per site in the multiple sequence alignment, color coded according to the scale. Above the multiple sequence alignment, the normalized evolutionary rates per site for sequence substitution (SEQ) and SLT based on the binary transformations of B-C are shown. Heat maps visualized with the Python packages ETE2 ([@evw246-B21]) and Matplotlib ([@evw246-B22]). See [Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online for a complete set of graphics for every protein family.](evw246f4p){#evw246-F4} ![Comparison of SEQ at sites characterized by secondary structure. All pairwise rate distributions, except NC(HS) vs. NC(HLS) and NC(HS) vs. C(L), are significantly different (*P* \< 0.05, after Bonferroni correction: *P* \< 0.008). For a summary of the U statistic and two-tailed *P* values for each pairwise comparison see [Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online.](evw246f5p){#evw246-F5} ###### Structural Conservation of Sites Per Protein Family --------------------- ![](evw246ie2p.jpg) --------------------- Identifying Target Sites ------------------------ For regions with five or more consecutive sites that were 100% conserved in sequence across 1) all CoV or 2) across the MERS and SARS clades, the information of structural disorder prediction from IUPred and DISOPRED2 was used to identify all ungapped sites that were consistently predicted to have 100% conserved order. Next, the information of secondary structure prediction from PSIPRED and JPred was used to narrow down this list further by only including sites that are not changing their predicted secondary structure state for both predictors. Applying the aforementioned filters to the initial 10,000 sites resulted in one (1) region of five residues or more conserved across all CoV within the N-terminal domain of NSP12: DNQDL ([table 4](#evw246-T4){ref-type="table"}). Interestingly, this region is in the vicinity of sites found important for nucleotidylating activity across the order *Nidovirales* ([@evw246-B27]). ###### Sites Conserved in Sequence and Structural Property Protein family PfamA domain Conserved sites in the MSA[^a^](#evw246-TF4){ref-type="table-fn"} ------------------------------------------------------- ----------------- ------------------------------------------------------------------- NSP5 *Peptidase_C30* 149-GSCGS-153[^b^](#evw246-TF5){ref-type="table-fn"} 213-AWLYAA-218[^b^](#evw246-TF5){ref-type="table-fn"} NSP7 *Replicase* 7-KCTSVVLL-14[^b^](#evw246-TF5){ref-type="table-fn"} 16-VLQQL-20[^b^](#evw246-TF5){ref-type="table-fn"} NSP12 *RPol N-term* 228-L**[DNQDL]{.ul}**NG-235 239-DFGDF-243 *RdRP_1* 521-DKSAG-525 588-MTNRQ-592 677-LANECAQVL-685 800-**[GGT]{.ul}**SSGD-706 *C-term* 853-YPDPSR-858 871-KTDGT-875 889-**[YPL]{.ul}**TK-893 NSP13 *N-term* 10-SQTSLR-15 *AAA_30* 362-NALPE-366 402-DPAQLP-407 *AAA_12* 539-SSQGS-543 NSP14 *NSP11* 281-AHVAS-285[^b^](#evw246-TF5){ref-type="table-fn"} 290-MTRCLA-295[^b^](#evw246-TF5){ref-type="table-fn"} 438-HAFHT-442[^b^](#evw246-TF5){ref-type="table-fn"} 494-CNLGG-499[^b^](#evw246-TF5){ref-type="table-fn"} Sites conserved across all clades in the protein family are underlined and in BOLD font. All other sites are conserved across the SARS and MERS clades. Experimentally determined structures are available in Protein Data Bank ([@evw246-B2]).. Considering only the sequences in the SARS and MERS clades, 21 sequence regions of five residues or more were found in seven protein families ([table 4](#evw246-T4){ref-type="table"}). For NSP5, NSP7, and NSP14, experimentally determined structures show that most regions are surface accessible ([fig. 6](#evw246-F6){ref-type="fig"}). Some of the identified target sites are known for their functional importance. For instance, C145 in the middle of GSCGS in NSP5 is part of the catalytic dyad in the NSP5 protease ([@evw246-B50]). For NSP12 and NSP13, which have the majority of all sites, no structures are available. The sites adjacent to DNQDL are also conserved in the SARS and MERS clades, and five additional target sites, conserved for the SARS and MERS clades, are found in the C-terminal direction relative to the DNQDL motif ([table 4](#evw246-T4){ref-type="table"}). Continuing into the RNA-dependent RNA polymerase domain (RdRP) in NSP12, four additional regions of target sites are found, and the last three regions are found in the C-terminal part. Importantly, in RdRP and in the C-terminal part are sites that are also conserved across all CoVs in this study. NSP13 has four regions of target sites distributed across the protein. ![Target sites shown in 3D context. **(***A*) NSP5 dimer, based on PDB id 1UK4 ([@evw246-B50]). (*B*) NSP7, based on PDB id 5F22 (unpublished). (*C*) NSP14, based on PDB id 5C8T ([@evw246-B30]). Protein structure visualized with Bioviva Discovery Studio .](evw246f6p){#evw246-F6} Discussion ========== We have analyzed the protein evolution of the genetic components that make up the MERS-CoV proteome. As previously established, MERS-CoV has the same genomic makeup as HKU4-CoV and HKU5-CoV in the MERS clade ([@evw246-B47]). Some protein products are only found in the MERS clade, and these were excluded from this study due to insufficient data. Furthermore, for other protein products, some clades may not be represented in our protein families if their proteins were too divergent. This was an important factor in determining the applied BLAST hit cutoffs, as relaxing cutoffs produced alignments with more gaps and increasing stringency reduced the representative pool. Because alignment quality is important due to the sensitivity of both Rate4Site and for phylogenetic reconstruction, the chosen cutoffs are suitable. We note some clade-specific differences in recoverable homologs between different CoV, but many components are shared among them ([fig. 1](#evw246-F1){ref-type="fig"}). Viral proteins often possess multifunctionality, mediated by a conformational change in response to environment-specific factors ([@evw246-B49]). Although conformational flexibility is important for function, it also offers flexibility in what sequence motifs are on display. If these sequences are rapidly diverging, different sequence motifs will be displayed, reinforcing the notion that flexible regions are potentially important in rewiring protein--protein interactions between virus and host ([@evw246-B18]). Although most CoV proteins have almost no intrinsic disorder, several CoV protein families have homologous sites that display loop in some sequences, helix in others and strands in some ([table 3](#evw246-T3){ref-type="table"}, [Supplementary Data](#sup1){ref-type="supplementary-material"}, [Supplementary Data](#sup1){ref-type="supplementary-material"} online). These sites are not necessarily disordered but they may be conformationally flexible in real-time (with secondary structure transitions in the same sequence, making them difficult to predict) or on evolutionary time-scales (so that different secondary structure elements actually are present in different sequences). The C(HS) and C(L) sites make up approximately 50--80% of most multiple sequence alignments. With the common expectation that protein structure is more conserved than sequence these numbers are surprisingly low. Neither PSIPRED nor JPred consistently predicts the same state for 20--50% of all sites in these multiple sequence alignments. The accuracy of PSIPRED and JPred's secondary structure predictions are about 80% ([@evw246-B6]; [@evw246-B13]). PSIPRED has been found to rarely predict an alpha helix instead of a beta strand and vice versa, and most of the PSIPRED errors are due to secondary structure not being predicted ([@evw246-B28]). When secondary structure is not conserved for the same site in a multiple sequence alignment, it suggests that the secondary structure prediction may be 1) inaccurate, 2) not predicted with high confidence, or 3) the regions are indeed metamorphic; they can transition from one element to another. Although (1) is difficult to address without experimentally determined structures for all sequences, (2) and (3) are not necessarily incompatible interpretations because low confidence secondary structure prediction could indicate metamorphic secondary structure regions. Metamorphic secondary structure regions have interesting consequences for conformational and functional flexibility. It should be noted that, despite the low amount of disordered sites in most CoV proteins, several regions are not conserved in disorder propensity across all sequences, but sometimes the different predictors disagree as in the case of NSP8. Clade-specific disordered regions resulting from indel events suggest that they are not essential to the critical functions of the protein, but could cause gain-and-loss of interactions with its hosts. However, when disorder propensity is only mildly fading for a region that is present across the protein family, it may be important for the fundamental function of the protein. The virus structural proteins that interact to form the virion commonly include an envelope protein, a membrane protein, and a capsid protein that together form the machinery that encases, transports, and releases the virus. The interactions between the structural proteins are often regulated by conformational changes like VP40 in Ebola ([@evw246-B4]) and Envelope protein from Dengue virus ([@evw246-B53]). Conformational changes in these proteins are needed for the virus life cycle. For CoVs, nucleocapsid is the only structural protein that is highly disordered. Yet, rapid evolutionary dynamics of disorder is present in nucleocapsid using two different IUPred cutoffs (0.4 and 0.5) and with DISOPRED2. Even if the different predictors and cutoffs disagree somewhat where regions with rapid evolutionary dynamics are present, these patterns suggest that nucleocapsid may be rapidly changing from one virus to another. It should also be noted that two MERS clade specific inserts around position 241 and toward the C-terminal are consistently predicted to be highly disordered. With inserts and changing structural dynamics between clades or viruses, the questions become 1) which sequence motif are displayed and 2) to what extent are these sequence motifs displayed? Furthermore, based on the inconsistent prediction of secondary structure elements, the possibility that CoVs are more conformationally flexible than their intrinsic disorder content implies is noteworthy. Altogether, this suggests that various mechanisms for rewiring conformational and functional space are operating in the coronaviruses studied here. If regions symptomatic of conformational and functional flexibility can be avoided in order to identify broad-specificity antiviral targets with potential to be effective against coronaviruses of today and in the future, coronaviruses as a group may become more attractive drug targets for the pharmaceutical industry in the event an additional coronavirus changes host to include humans or increase its virulence. Supplementary Material ====================== ###### Click here for additional data file. We thank Joseph Ahrens, Janelle Nunez-Castilla, and Helena Gomes Dos Santos for assistance in the lab and for helpful discussions. The authors would also like to acknowledge the Instructional & Research Computing Center (IRCC) at Florida International University for providing HPC computing resources that have contributed to the research results reported within this article, web: <http://ircc.fiu.edu>. Supplementary Material ====================== [Supplementary Data](#sup1){ref-type="supplementary-material"} and [Supplementary Data](#sup1){ref-type="supplementary-material"} are available at Genome Biology and Evolution online ([http://www.gbe. oxfordjournals.org/](http://www.gbe.oxfordjournals.org/)). [^1]: **Associate editor**: Dr. Chantal Abergel

{ "pile_set_name": "PubMed Central" }

Introduction ============ The latest Global Burden of Disease Study showed that 1.25 billion people had migraine in 2017.[@ref1] Migraine was responsible for 45.1 million years of life lived with disability.[@ref2] In the 15-49 year age group, migraine was the top cause of life lived with disability.[@ref3] Although some authors have suggested that prophylactic drugs should be recommended for migraineurs who have at least four headache days per month, only 13% of patients reported current use of preventive drugs.[@ref4] Ineffectiveness of and/or contraindications to migraine medication affect 10-15% of people with migraine.[@ref5] Hence, a need exists to investigate non-drug interventions. Acupuncture has emerged as a prophylactic treatment for migraine because of its potential efficacy.[@ref6] [@ref7] [@ref8] Clinically, two acupuncture interventions are widely used: manual acupuncture and electrical acupuncture. Manual acupuncture is commonly used by traditional acupuncturists and is highly effective for pain relief.[@ref9] Besides insertion of penetrating needles into specific acupuncture points, manual manipulation (that is, intermittent rotation as well as lift and thrust) of the needle may enhance its clinical benefit.[@ref10] [@ref11] However, clinical evidence for the benefit of manual acupuncture for migraine prophylaxis remains scarce. Several randomised clinical trials found no differences between manual and sham acupuncture.[@ref12] [@ref13] [@ref14] [@ref15] [@ref16] These negative results might be ascribed to inappropriate placebo control settings. Blinding is challenging in trials of non-drug treatments. An appropriate placebo acupuncture design should be both physiologically inert and indistinguishable from true acupuncture. So far, only one randomised clinical trial of acupuncture in migraine has reported successful blinding.[@ref15] Furthermore, some authors have suggested that sham acupuncture interventions were not totally inert, often being associated with moderately large non-specific effects.[@ref17] [@ref18] [@ref19] This could make detection of additional specific effects of manual acupuncture difficult. In particular, sham acupuncture involving penetrating needles should be avoided in clinical trials.[@ref20] However, penetrating sham acupuncture has been commonly used in previous randomised clinical trials in migraine.[@ref12] [@ref13] [@ref14] [@ref15] [@ref16] Therefore, we designed a randomised clinical trial among patients with migraine who were naive to acupuncture, using non-penetrating sham control and assessment of blinding to determine the efficacy of manual acupuncture and quantify the true placebo response in the prophylaxis of episodic migraine without aura. Methods ======= Study design ------------ This randomised, single blind, three arm clinical trial was conducted in seven centres in China between 5 June 2016 and 15 November 2018. The protocol and the statistical analysis plan are available in appendix 1. The total trial period was 24 weeks, including four weeks of baseline assessment, eight weeks of treatment after randomisation, and 12 weeks of follow-up. We obtained written informed consent from each patient before randomisation. Participants ------------ A neurologist made the diagnosis of episodic migraine without aura on the basis of the International Classification of Headache Disorders, 3rd edition β version (ICHD-3β).[@ref21] Other inclusion criteria were age between 15 and 65 years, history of migraine without aura for more than 12 months, initial onset of migraine before the age of 50 years, between two and eight migraine attacks during the baseline phase, naivety to acupuncture, and ability to give informed consent. Exclusion criteria included all other types of primary and secondary headaches, history of a clinically significant disorder (for example, severe mental illness), pregnancy or breast feeding, and non-adherence to the baseline diary. In addition, all patients were instructed not to take any other analgesics and to avoid starting other interventions. Randomisation and masking ------------------------- After the baseline assessment, an independent investigator randomised eligible patients in a 2:2:1 ratio to receive manual acupuncture, sham acupuncture, or usual care. We used the 2:2:1 ratio to facilitate recruitment and increase patients' adherence. Treatments were allocated in a centralised manner via an interactive web response system with stratification by centre and a block size of five. To maintain blinding in patients, outcome assessors, and statisticians, the allocation sequence was concealed until the end of the study. Study interventions ------------------- Fourteen licensed acupuncturists delivered treatments. They each had more than five years of clinical experience and attended centralised training before recruitment. All patients received 20 sessions of 30 minute acupuncture treatments or usual care over eight weeks. They were treated every other day to fulfil a 10 session treatment course and received another 10 session course after a nine day break. In both the manual and sham acupuncture group, acupuncture treatment began after randomisation. In the usual care group, patients received acupuncture for free after waiting 24 weeks. Details of acupuncture (location of acupuncture points/non-acupuncture points, depth of insertion) are shown in tables 1 and 2 of appendix 1. Acupuncturists applied manual acupuncture at 10 obligatory acupoints, included bilateral L14, LR3, EX-HN5, GB20, and GB8. Additional acupoints were selected on the basis of meridian diagnosis and the patient's symptoms: bilateral ST8 for Yangming meridian headache, BL10 for Taiyang meridian headache, and DU20 for Jueying meridian headache. Streitberger acupuncture needles (0.30 mm in diameter and 30 mm in length) were used. After sterilisation, the sharp needles were inserted into the deep tissue layers of acupoints. Acupuncturists then tried to elicit acupuncture de-qi sensation by the manual manipulation of needles. Over a 30 minute period, manual manipulation for each acupoint lasted 10 seconds and was repeated four times with intervals of 10 minutes. Non-penetrating sham acupuncture was performed at four bilateral non-acupuncture points, which were all located on the back and in different segments from the headache area. After sterilisation, Streitberger placebo needles with blunt tip were used.[@ref22] When they were fixed on the skin through plastic rings, patients felt a pricking sensation, simulating a puncture of the skin. However, instead of penetrating the skin, the needles retracted up into the shaft when they were pressed against the skin. We formulated and followed standardised step-by-step instructions and operations to use the same rituals in the manual acupuncture and sham acupuncture groups as far as possible (appendix 2). In accordance with the migraine guidelines developed by the Canadian Headache Society Prophylactic Guidelines Development Group in 2012,[@ref23] we provided usual care to all three groups in terms of lifestyle changes and migraine self-management. We taught patients about lifestyle factors that may contribute to an increased frequency of migraine, identifying migraine triggers or precipitants through the keeping of a detailed headache diary, pacing, pain communication, sleep hygiene, and regular exercise. In case of severe pain (visual analogue score \>8), diclofenac sodium enteric coated tablets (25 mg/tablet; maximal tolerated dose 200 mg/day) were allowed as a rescue medication. Outcome measures ---------------- Independent research assistants supervised patients to complete their headache diaries in paper and pencil format from baseline to week 20 and evaluated their outcomes. The primary outcomes were change in the mean number of migraine days and migraine attacks per four week cycle during weeks one to 20 after randomisation compared with baseline (the four weeks before randomisation). Secondary outcomes included the proportion of patients achieving at least a 50% reduction in the mean number of migraine days or migraine attacks during weeks 17 to 20 and changes in the severity of migraine as measured by a visual analogue scale, the Migraine-Specific Quality-of-Life Questionnaire (MSQ),[@ref24] the Pittsburgh Sleep Quality Index (PSQI),[@ref25] the Migraine Disability Assessment Score (MIDAS),[@ref26] the Beck Anxiety Inventory (BAI) scale,[@ref27] the Beck Depression Inventory II (BDI-II) scale,[@ref28] and the mean dose of used rescue medication from baseline to week 20. We also measured patients' expectations of of positive outcomes from acupuncture by using the Acupuncture Expectancy Scale.[@ref29] At the end of the study, we determined the maintenance of blinding of patients by asking them whether they thought the needles had penetrated the skin. Patients recorded adverse events after each treatment, including bleeding, subcutaneous haemorrhage, serious pain, palpitation, fainting, and local infection. Statistical analyses -------------------- On the basis of a pilot study, we anticipated a mean difference of 1.8 (SD 2.8) days and 1.2 (1.8) attacks for manual versus sham acupuncture and a difference of 2.2 (2.8) days and 1.5 (1.8) attacks for manual acupuncture versus usual care. To drive all the primary hypothesis tests, we used PROC POWER in SAS to calculate that a sample of 135 evaluable patients (54 in manual acupuncture, 54 in sham acupuncture, and 27 in usual care) would be needed to provide 90% power at a two sided significance level of 5%. Therefore, we planned to recruit 150 patients into this study, with an anticipated 10% dropout rate. We did efficacy analyses in the full analysis set, which included all randomised patients who received at least one session of treatment and had at least one efficacy measure. We imputed missing data by using the last observation carried forward method.[@ref30] We did sensitivity analyses in the per protocol set, which included all randomised patients who had no major protocol deviation. We did safety analyses in the safety set, which included all randomised patients who received at least one session of acupuncture. We assessed the poolability of treatment effects across centres by including an interaction term of treatment group by centre in the model for the primary analysis. If P values associated with the interaction term of both primary endpoints were greater than 0.15,[@ref31] the treatment effect was homogenous across centres. We presented continuous variables as the mean with standard deviation or median and interquartile range. For tests across groups, we used analysis of variance or the Kruskal-Wallis test when relevant. We described categorical variables as numbers and percentages and used the χ^2^ test or Fisher's exact test. For comparisons of the primary endpoints across the groups, we used analysis of covariance with the baseline migraine days or migraine attacks adjusted. Superiority of manual acupuncture over sham acupuncture or usual care could be claimed only when both co-primary endpoints were statistically significant. We used a fixed sequence procedure for multiple comparisons among groups, which would not inflate the type I error, and comparisons were made with an α value of 0.05. We also evaluated multivariate adjusted treatment effects across the groups on the basis of a linear mixed effects model. For the secondary endpoints, we made pairwise multiple comparisons on the basis of Bonferroni adjusted P values and confidence intervals, if a significant difference was noted across the groups. An independent statistician used SAS statistical software for all statistical analyses. For both continuous and categorical variables, 95% confidence intervals were calculated as appropriate. All the statistical comparisons were two sided with P\<0.05 considered as significant. Patient and public involvement ------------------------------ No patients were involved in the design or conduct of the study, in setting the research question, or in determining the outcome measures of this study, nor did they have any input on data analysis, interpretation, or writing up of results. Results of the trial will be shared with all participants by web or email. Results ======= Study participants ------------------ Between 5 June 2016 and 15 November 2018, we screened 210 patients, of whom 150 were randomised ([fig 1](#f1){ref-type="fig"}). The main reason for exclusion was not meeting the inclusion criterion of between two and eight migraine attacks during the baseline phase (table A in appendix 3). The baseline characteristics of all randomised patients were well balanced among the three groups ([table 1](#tbl1){ref-type="table"}). Similar baseline characteristics were observed in the full analysis set population (table B in appendix 3). ![Flowchart showing trial group assignments, loss to follow-up, treatment completion, and protocol deviations](xus053060.f1){#f1} ###### Baseline characteristics of randomised population. Values are mean (SD) unless stated otherwise Characteristics Manual acupuncture (n=60) Sham acupuncture (n=60) Usual care (n=30) --------------------------------------------------- --------------------------- ------------------------- ------------------- Age, years 36.6 (12.0) 36.0 (10.9) 37.3 (11.7) No (%) women 47 (78) 50 (83) 26 (87) Median (IQR) body mass index 20.4 (19.5-23.1) 22.0 (20.3-23.4) 20.8 (19.5-22.4) Median (IQR) disease duration, years, 10.0 (5.0-19.5) 10.0 (6.0-14.0) 16.5 (7.8-23.0) Accompanying symptoms: No (%) nausea or vomiting 53 (88) 51 (85) 25 (83) No (%) photophobia or phonophobia 52 (87) 52 (87) 23 (77) Patient personality: Neuroticism 31.8 (7.1) 31.9 (6.6) 31.3 (7.1) Extraversion 37.4 (7.0) 38.0 (6.6) 37.0 (6.4) Openness 37.6 (4.9) 37.4 (4.8) 35.9 (2.8) Agreeableness 43.2 (4.6) 43.1 (4.9) 43.6 (5.2) Conscientiousness 41.2 (5.6) 41.8 (6.2) 41.5 (5.5) No (%) chronic migraine 1 (2) 2 (3) 0 (0) Acupuncture Expectancy Scale score 10.9 (3.7) 11.1 (3.5) NA Days with migraine 5.8 (2.6) 6.3 (3.8) 5.8 (3.0) No of migraine attacks 3.8 (1.4) 4.1 (2.6) 3.9 (1.7) Mean VAS score 5.1 (1.3) 5.3 (1.3) 5.2 (1.8) Migraine-Specific Quality-of-Life Questionnaire: Role restrictive subscale 55.2 (17.6) 54.4 (15.5) 56.0 (15.0) Role preventive subscale 60.6 (23.3) 59.9 (19.5) 64.0 (19.0) Emotional subscale 61.6 (23.3) 65.3 (20.3) 62.9 (21.6) Pittsburgh Sleep Quality Index 5.7 (2.7) 5.6 (2.7) 5.3 (3.5) Median (IQR) Migraine Disability Assessment Score 27.5 (12.0-49.8) 33.5 (15.3-57.3) 27.5 (13.8-44.8) Median (IQR) doses of rescue medication 0.0 (0.0-1.8) 0.0 (0.0-3.5) 0.0 (0.0-2.0) Median (IQR) Beck Depression Inventory II 6.5 (3.0-13.0) 7.0 (2.3-14.0) 7.0 (1.0-13.0) Median (IQR) Beck Anxiety Inventory 10.0 (4.0-15.0) 7.5 (4.0-14.0) 7.0 (3.8-9.5) IQR=interquartile range; NA=not applicable; VAS=visual analogue scale. After randomisation, six (4%) patients were lost to follow-up (three in the manual acupuncture group, one in the sham acupuncture group, and two in the usual care group; P=0.3025, χ^2^ test). Of these patients, three were lost without one primary outcome measure and another three did not complete all the visits. Therefore, 147 patients were included in the full analysis set and 144 in the per protocol set. Primary outcomes ---------------- [Table 2](#tbl2){ref-type="table"} and [figure 2](#f2){ref-type="fig"} show the primary analyses. The reduction of migraine days and migraine attacks per four weeks from baseline was significantly greater in the manual acupuncture group than in the usual care group during weeks 1 to 20. Compared with sham acupuncture, manual acupuncture resulted in a significantly greater reduction in migraine days during weeks 13 to 20, with a group difference of −1.4 (95% confidence interval −2.4 to −0.3; P=0.005) days at weeks 13 to 16 and −2.1 (−2.9 to −1.2; P\<0.001) days at weeks 17 to 20, and a significantly greater reduction in migraine attacks at weeks 17 to 20, with a group difference of −1.0 (−1.5 to −0.5; P\<0.001) attacks. Sham acupuncture also resulted in a significantly greater reduction in migraine attacks during weeks 5 to 20 compared with usual care. We observed similar results in the per protocol analysis population (table C in appendix 3). In addition, we developed mixed effects models including participants, centres, acupuncturists, baseline, study visits, treatment groups, and the interaction term of treatment groups by study visits, the results of which were consistent with those of the primary analyses (table D in appendix 3). The poolability of both primary efficacy endpoints across centres was verified (P values for the interaction effects were \>0.15). ###### Primary outcome measurements of full analysis set population. Values are mean (SD) unless stated otherwise Time point Manual acupuncture (n=58) Sham acupuncture (n=60) Usual care (n=29) P value\* Pairwise comparisons---mean (95% CI) --------------------------------------------------------------- --------------------------- ------------------------- ------------------- ----------- -------------------------------------- --------- --------------------- --------- --------------------- ------- **Change from baseline in mean number of days with migraine** Weeks 1-4 −2.2 (2.0) −1.6 (3.0) −0.7 (2.9) 0.003 −0.9 (−1.8 to 0.0) 0.052 −1.5 (−2.5 to −0.4) 0.004 −0.6 (−1.7 to 0.5) 0.589 Weeks 5-8 −3.1 (2.6) −2.7 (3.6) −1.3 (2.5) 0.002 −0.7 (−1.6 to 0.2) 0.198 −1.7 (−2.8 to −0.5) 0.001 −1.0 (−2.1 to 0.2) 0.119 Weeks 9-12 −3.5 (2.6) −2.9 (3.8) −1.3 (2.7) \<0.001 −1.0 (−2.0 to 0.1) 0.077 −2.1 (−3.4 to −0.8) \<0.001 −1.1 (−2.4 to 0.1) 0.100 Weeks 13-16 −3.5 (2.5) −2.4 (3.4) −1.0 (2.4) \<0.001 −1.4 (−2.4 to −0.3) 0.005 −2.4 (−3.7 to −1.2) \<0.001 −1.1 (−2.3 to 0.2) 0.136 Weeks 17-20 −3.9 (3.0) −2.2 (3.2) −1.4 (2.7) \<0.001 −2.1 (−2.9 to −1.2) \<0.001 −2.4 (−3.5 to −1.4) \<0.001 −0.4 (−1.4 to 0.7) 0.495 **Change from baseline in mean number of migraine attacks** Weeks 1-4 −1.1 (1.3) −1.0 (2.5) −0.2 (1.5) 0.032 −0.3 (−0.9 to 0.3) 0.643 −0.8 (−1.6 to −0.1) 0.026 −0.5 (−1.3 to 0.2) 0.305 Weeks 5-8 −1.8 (1.4) −1.7 (2.5) −0.7 (1.3) 0.002 −0.3 (−0.9 to 0.2) 0.526 −1.0 (−1.7 to −0.4) \<0.001 −0.7 (−1.4 to 0.0) 0.032 Weeks 9-12 −1.9 (1.3) −1.8 (2.7) −0.4 (1.3) \<0.001 −0.4 (−1.0 to 0.2) 0.410 −1.5 (−2.3 to −0.8) \<0.001 −1.1 (−1.9 to −0.4) 0.001 Weeks 13-16 −2.1 (1.5) −1.8 (2.7) −0.7 (1.3) \<0.001 −0.6 (−1.2 to 0.1) 0.092 −1.3 (−2.1 to −0.6) \<0.001 −0.8 (−1.5 to 0.0) 0.043 Weeks 17-20 −2.3 (1.7) −1.6 (2.5) −0.4 (1.3) \<0.001 −1.0 (−1.5 to −0.5) \<0.001 −1.8 (−2.4 to −1.2) \<0.001 −0.8 (−1.4 to −0.2) 0.008 P values, differences, and confidence intervals were based on analysis of covariance adjusted for days with migraine (or number of migraine attacks) at baseline. According to fixed sequence procedure planned for primary analyses, pairwise comparisons at weeks 17-20 were made at nominal α level without further adjustment. For visits other than weeks 17-20, multiple pairwise comparisons were adjusted by Bonferroni approach. ![Primary outcomes throughout trial. Top: number of migraine days per 4 weeks through study period. Bottom: frequency of migraine attacks per 4 weeks through study period](xus053060.f2){#f2} Secondary outcomes ------------------ [Table 3](#tbl3){ref-type="table"} summarises the secondary analyses. Overall, the responder rates in the manual acupuncture group were significantly higher than in the two control groups during weeks 17 to 20. All subscales of the Migraine-Specific Quality-of-Life Questionnaire were improved significantly more in the manual acupuncture group than in the two control groups at weeks 20. The visual analogue scale showed a greater reduction in score for the manual acupuncture group than for the two control groups at weeks 20. ###### Secondary outcome measurements of full analysis set population Time point Manual acupuncture (n=58) Sham acupuncture (n=60) Usual care (n=29) P value\* Pairwise comparison -------------------------------------------------------------------------------------- --------------------------- ------------------------- ------------------- ----------- --------------------- --------- ----------------------- --------- ---------------------- --------- **Proportion of participants with ≥50% reduction in number of migraine days (%)** Weeks 17-20 82.5 45.8 17.9 \<0.001 36.7 (17.0 to 56.4) \<0.001 64.6 (43.5 to 85.7) \<0.001 27.9 (4.6 to 51.2) 0.012 **Proportion of participants with ≥50% reduction in number of migraine attacks (%)** Weeks 17-20 78.9 44.1 14.3 \<0.001 34.9 (14.7 to 55.0) \<0.001 64.7 (44.2 to 85.1) \<0.001 29.8 (7.6 to 51.9) 0.004 **Mean (SD) change from baseline in visual analogue scale** Weeks 20 −2.2 (2.5) −0.9 (1.9) −0.0 (1.4) \<0.001 −1.3 (−2.2 to −0.3) 0.005 −2.1 (−3.3 to −1.0) \<0.001 −0.9 (−2.0 to 0.3) 0.224 **Mean (SD) change from baseline in MSQ, role restrictive subscale** Weeks 20 24.4 (21.4) 12.8 (16.0) 5.8 (14.2) \<0.001 11.6 (3.5 to 19.8) 0.002 18.6 (8.5 to 28.7) \<0.001 7.0 (−3.1 to 17.0) 0.285 **Mean (SD) change from baseline in MSQ, role preventive subscale** Weeks 20 22.5 (25.1) 10.1 (14.1) 4.5 (15.5) \<0.001 12.5 (3.7 to 21.2) 0.002 18.1 (7.2 to 29.0) \<0.001 5.6 (−5.2 to 16.4) 0.630 **Mean (SD) change from baseline in MSQ, emotional subscale** Weeks 20 19.5 (26.8) 8.4 (19.0) 8.8 (20.9) 0.020 11.2 (0.9 to 21.4) 0.027 10.7 (−2.0 to 23.4) 0.129 −0.4 (−13.1 to 12.2) \>0.999 **Mean (SD) change from baseline in global score of PSQI** Weeks 20 −1.8 (2.3) −0.7 (2.5) −0.1 (2.6) 0.007 −1.1 (−2.2 to 0.0) 0.058 −1.7 (−3.0 to −0.3) 0.011 −0.6 (−2.0 to 0.8) 0.894 **Median (IQR) change from baseline in score of MIDAS** Weeks 20 −15.0 (−36.0-−5.0) −6.0 (−30.0-2.0) −0.5 (−16.0-6.5) 0.007 −8.0 (−17.0 to 2.0) 0.139 −14.0 (−28.0 to −3.0) 0.006 −6.0 (−22.0 to 4.0) 0.523 **Median (IQR) change from baseline in doses of rescue medication** Weeks 17-20 0.0 (−1.0-0.0) 0.0 (−1.0-0.0) 0.0 (−1.5-0.0) 0.745 NA NA NA NA NA NA **Median (IQR) change from baseline in BDI- II** Weeks 20 −2.0 (−6.0-0.0) −1.0 (−5.0-1.0) 0.0 (−2.5-1.0) 0.081 NA NA NA NA NA NA **Median (IQR) change from baseline in BAI** Weeks 20 −4.0 (−8.0-0.0) −1.0 (−5.0-1.0) 0.0 (−3.5-1.5) 0.005 −2.0 (−5.0 to 0.0) 0.051 −4.0 (−7.0 to −1.0) 0.007 −1.0 (−4.0 to 1.0) 0.980 BAI=Beck Anxiety Inventory; BDI-II=Beck Depression Inventory II; IQR=interquartile range; MIDAS=Migraine Disability Assessment Scores; MSQ=Migraine-Specific Quality-of-Life Questionnaire; NA=not applicable; PSQI=Pittsburgh Sleep Quality Index. Pairwise comparison include differences in means, medians, and proportions and corresponding confidence intervals. For continuous variables, analysis of variance or Kruskal-Wallis test was used for comparisons among treatment groups. If Kruskal-Wallis test was significant, pairwise comparisons were made using Wilcoxon rank sum test. Difference in median is estimated on basis of Hodges-Lehmann estimator of location shift. Asymptotic (Moses) confidence intervals are provided. For categorical variables, linear probability model was used for comparisons among treatment groups. For all pairwise comparisons, P values and confidence intervals were adjusted by Bonferroni approach. The total scores on the Pittsburgh Sleep Quality Index and Migraine Disability Assessment Score were significantly lower in the manual acupuncture group than in the usual care group at week 20. However, we found no significant difference in these scores between manual acupuncture and sham acupuncture. We also found no significant difference in the mean dose of rescue medication or in Beck Anxiety Inventory and Beck Depression Inventory II scores among the three groups at week 20. At the end of the study, we assessed the credibility of blinding of treatment. We found no significant difference between the manual acupuncture and sham acupuncture groups for patients' ability to correctly guess their allocation status (P=0.891) (table E in appendix 3). Adverse events -------------- In the manual acupuncture group, five (8%) patients reported at least one acupuncture related adverse event compared with no patients in the sham acupuncture group. One patient withdrew from the study because of an adverse event. No patients in either the manual or sham acupuncture group had severe adverse events (table F in appendix 3). Discussion ========== In this study in acupuncture naive patients with episodic migraine without aura, 20 sessions of manual acupuncture produced a relatively long lasting reduction in migraine days and migraine attacks compared with sham acupuncture and usual care. We found that, compared with the usual care group, both the reductions in migraine days and migraine attacks per four week period from baseline in the manual acupuncture group were significantly larger from the first four week period after the start of treatment, and the significant reductions lasted until the last four week period of follow-up (weeks 17-20), with an apparent increasing trend. Compared with the usual care group, only in migraine attacks per four week period from baseline did the sham acupuncture group show significant reductions from the second four week period after treatment to the last four week period of follow-up. In the last four week period, that reduction fell a little. In comparisons between the manual acupuncture group and the sham acupuncture group, reductions in migraine days were significantly larger in the manual group at weeks 13-16 and 17-20, and reductions in migraine attacks were significantly larger in the manual group at weeks 17-20. Overall, the therapeutic effects in the manual acupuncture group occurred earlier, were larger, and might last longer. Comparison with other studies ----------------------------- The latest Cochrane meta-analysis,[@ref32] published in 2016, found that up to both eight weeks/two months and three to four months after randomisation, acupuncture was associated with a small but statistically significant reduction in migraine attacks compared with sham. The effect size was −0.35 (95% confidence interval −0.57 to −0.13) at up to eight weeks/two months after randomisation and −0.32 (−0.53 to −0.10) at up to three to four months after randomisation. Since then, results have been published for only one randomised clinical trial.[@ref30] It found that electrical acupuncture resulted in a long term reduction in migraine attacks. At week 16 after randomisation, a greater reduction was observed in the true electrical acupuncture group than in the sham acupuncture group (difference of 1.1 (0.4 to 1.9) attacks; P=0.002) and in the true electrical acupuncture group than the no acupuncture group (difference of 1.8 (1.1 to 2.5) attacks; P\<0.001). The effect size for manual acupuncture over sham acupuncture observed in our study was similar to the effect size seen with electrical acupuncture,[@ref30] but larger than the effect size seen in the latest Cochrane meta-analysis.[@ref32] We also updated the meta-analysis to include our results (appendix 4). Additionally, the effect size for manual acupuncture was larger than the effect size seen with prophylactic drugs, such as monthly subcutaneous injections of fremanezumab (225 mg or 675 mg) or galcanezumab (120 mg or 300 mg).[@ref33] [@ref34] However, our findings are not completely consistent with those of several randomised clinical trials,[@ref12] [@ref13] [@ref14] [@ref15] [@ref16] which found no significant differences between true acupuncture and sham acupuncture in reducing migraine headaches. The inconsistency might be driven predominantly by differences in treatments received by the control group. Comparison between manual and electrical acupuncture ---------------------------------------------------- In acupuncture practice, manual acupuncture and electrical acupuncture are the two most common interventions for the treatment of pain conditions. Manual acupuncture emphasises the occurrence of "de-qi" feelings, which can be induced by proper and effective manual manipulation of the needle. The intensity of de-qi feelings plays a key role in the analgesic effects of manual acupuncture.[@ref35] The importance of manipulation of the needle is no less than that of the selection of the appropriate acupoint.[@ref36] The effects of electrical acupuncture depend on current parameters (frequency, intensity, and duration of pulses). Different manual manipulations or different electrical parameters may produce different therapeutic effects. The peripheral and central mechanisms for pain relief produced by manual acupuncture and electrical acupuncture are homologous, but some differences exist. The effects of manual acupuncture may be mediated by all types of afferents, whereas the effects of electrical acupuncture might be preferentially mediated by Aβ and parts of Aδ type afferents.[@ref9] In addition, subtle differences exist between the areas of the brain activated by manual and electrical acupuncture.[@ref37] So far, only a few small trials for several conditions have directly compared the effects of manual and electrical acupuncture.[@ref38] [@ref39] [@ref40] [@ref41] [@ref42] The findings of these studies were inconsistent and inconclusive. The choice of different manual manipulations or different electrical parameters may need to be based on a specific disease or condition and should be guided by more high quality randomised controlled trials. Strengths and limitations of study ---------------------------------- An advantage of this trial is the design of a reasonable sham comparator. Previous acupuncture research has often used penetrating sham acupuncture, involving needling non-acupuncture points, needling irrelevant acupuncture points, or superficial needling.[@ref43] [@ref44] However, whether the needle is inserted into an acupuncture point or a non-acupuncture point, it could produce a physiological effect.[@ref44] This is in part due to the activation of the pain inhibiting system in the spinal cord and diffuse noxious inhibitory controls.[@ref45] [@ref46] [@ref47] In addition, sham acupuncture applied at points beside the real acupoints might be an active control, because the area of the acupoint could be enlarged by increased expression of nociceptive substances during various pain conditions.[@ref48] In this study, we applied non-penetrating sham acupuncture at heterosegmental non-acupuncture points, thereby avoiding segmental analgesia and minimising any physiological effect in the sham acupuncture group. Another strength of this trial is the successful blinding. Previous trials lacked assessment of blinding or had ineffective blinding,[@ref12] [@ref13] [@ref14] [@ref16] so comparisons between true acupuncture and sham acupuncture might be biased by unsuccessful blinding. Ideal blinding should be psychologically credible yet physiologically inert. To ensure successful blinding, we recruited acupuncture naive patients, used non-penetrating needles as the control, and designed the same procedures to perform the same rituals as far as possible in the manual and sham acupuncture groups. These factors made the sham acupuncture as close to a real placebo as possible. At the end of the study, we asked patients to guess whether they thought the needles penetrated the skin. Satisfyingly, we found no significant difference between the two treatment groups in patients' ability to correctly guess their allocation status. Successful blinding showed the reliability of the difference between manual acupuncture and sham acupuncture, as well as between sham acupuncture and usual care. Furthermore, successful blinding increased the adherence of patients, especially those allocated to the sham acupuncture group. Only one patient in the sham acupuncture group was lost to follow-up. The study does, however, have limitations. Firstly, the lack of baseline prophylaxis is not typical, but the latest meta-analysis indicated that acupuncture reduced the frequency of migraine significantly more than did drug prophylaxis and was less likely to lead to dropout due to adverse effects and to reporting of adverse effects.[@ref32] In addition, we should respect the opinions of patients who are reluctant to use drug treatment; the decision to treat with drugs and the choice of a prophylactic drug should be made together with the patient.[@ref49] Secondly, the timeframe of this study may be not have been long enough to see a long lasting effect, although we observed that the treatment effects of manual acupuncture could persist for 12 weeks after treatment. In the future, trials with longer follow-up periods are needed to assess how long the effects of manual and sham acupuncture last. Clinical relevance ------------------ A considerable number of patients do not respond well to drug treatment, cannot tolerate the adverse effects of drugs, or have contraindications, which can lead to low adherence, chronification of headache, and acute drug overuse.[@ref50] Our results show that acupuncture can be recommended as an effective alternative treatment. Although sham acupuncture produced a minor benefit in reducing migraine attacks, the differences between manual acupuncture and sham acupuncture support the specific effects of manual acupuncture. On the basis of our results, clinical acupuncturists should pay careful attention to the specific components of acupuncture, including acupuncture points, needle manipulation, and acupuncture sensations, and the non-specific effects of acupuncture (acupuncture ritual or expectation for efficacy[@ref51] [@ref52]) should also not be neglected in clinical practice. In randomised controlled trials of acupuncture, the sham acupuncture group should not use penetrating needles. Conclusion and policy implications ---------------------------------- Among acupuncture naive patients with episodic migraine without aura, treatment with manual acupuncture, compared with sham acupuncture or usual care, resulted in a significantly greater reduction in the frequency of migraine days and migraine attacks. Acupuncture can be recommended as a prophylactic treatment when updating the guidelines and insurance coverage for migraine and its related substance misuse. When discussing prophylactic treatment strategies with patients, clinicians should provide them with information about acupuncture as an option. What is already known on this topic =================================== 1. Clinical evidence for the benefit of manual acupuncture in migraine prophylaxis remains scarce 2. Several recent clinical trials found that manual acupuncture was superior to no acupuncture 3. Comparisons between manual acupuncture and sham acupuncture found only minor or even no differences What this study adds ==================== 1. Manual acupuncture was more effective than sham acupuncture and usual care in reducing migraine headaches We thank Fengcai Zhu and Jingxin Li of the Jiangsu Provincial Center for Disease Control and Prevention, for guidance of the statistical analyses. We also thank the research assistants, acupuncturists, supporters of this study, and all of the patients who participated in this study. Extra material supplied by authors ###### Appendix 1: protocol ###### Appendix 2: acupuncture ritual ###### Appendix 3: supplementary tables ###### Appendix 4: meta-analysis Contributors: SX, LY, and XL contributed equally to the study. SX, LY, XL, HW, and WW conceived and designed the study. SX, LY, XL, HW, and WW had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. SX, LY, and XL drafted the manuscript. SX, LY, XL, MW, GC, QZ, ZZ, JS, HJ, HW, and WW contributed to the acquisition, analysis, and interpretation of data. All authors contributed to the critical revision of the manuscript for important intellectual content and gave final approval of the version to be published. WL did the statistical analysis. HW and WW obtained funding. GC, ZZ, JS, HJ, GH, FL, HW, and WW provided administrative, technical, or material support. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. HW and WW supervised the study and are the guarantors. Funding: This study was supported by a grant from the National Natural Science Foundation of China (61327902), and a project grant from Hubei University of Chinese Medicine/Hubei Provincial Collaborative Innovation Center of Preventive Treatment by Acupuncture and Moxibustion. Study sponsors were not involved in study design, data interpretation, writing the article, or the decision to submit the article for publication. Competing interests: All authors have completed the ICMJE uniform disclosure form at [www.icmje.org/coi_disclosure.pdf](http://www.icmje.org/coi_disclosure.pdf) and declare: funding for the submitted work as described above; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work. Ethical approval: This study was approved by the Clinical Trial Ethics Committee of Tongji Medical College, Huazhong University of Science and Technology (2016S009), the Medical Ethics Committee of Wuhan No.1 Hospital (WuWeiYiYuanLunShen\[2016\]14), the Ethics Committee of Hubei Province Hospital of Traditional Chinese Medicine (HBZY2016-C25-01), the Ethics Committee of Xiangyang No.1 Hospital (YN20171220), the Ethics Committee of The Second Hospital of Huangshi (HEYLWJY1701), and the Ethics Committee of Hubei University of Chinese Medicine (\[2018\] IEC No. 2). The trial was overseen by an independent trial steering committee and data monitoring and ethics committee. All patients gave written informed consent. Data sharing: The raw trial data after de-identification can be shared on individual request to the corresponding author at wwang\@vip.126.com. Transparency: The lead author affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained. Dissemination to participants and related patient and public communities: Participants will be informed of the results of the study by telephone. Dissemination to the public will be achieved through media outreach.

{ "pile_set_name": "PubMed Central" }

Introduction {#Sec1} ============ Percutaneous sacroiliac (SI) screw fixation is a minimally invasive and effective method for the stabilization of the posterior pelvic ring \[[@CR1]--[@CR6]\]. Screw placement under two-dimensional (2D) fluoroscopic control is the gold standard \[[@CR4], [@CR6]--[@CR8]\]. CT-controlled and navigated SI screw fixation techniques are demanding procedures and, therefore, not widely available beyond trauma centers \[[@CR6], [@CR9]--[@CR15]\]. Major complications of fluoroscopically controlled percutaneous SI screw fixation include nerve and vessel lesions following screw misplacement. The relative risk of lesions of the fifth lumbar nerve (L5), for example, is up to 8 % \[[@CR6], [@CR16]--[@CR20]\]. Variations in the morphology of the upper sacrum, widely known as ''sacral dysmorphism'', \[[@CR21], [@CR22]\] primarily account for this relatively high complication rate. Three-dimensional (3D) models of the upper sacrum have been described in the literature to assess safe corridors for SI screw placement even in the presence of sacral dysmorphism \[[@CR23]--[@CR25]\]. These techniques, however, require particular planning software and are typically not applicable to the surgeon during preoperative assessment. The aim of this study, therefore, was to present a simple method for the preoperative assessment of the sacral morphology using CT scans with widely available tools. In addition, it was our goal to classify the morphology of the upper sacrum based on the risk of iatrogenic nerve lesions during percutaneous SI screw fixation. Materials and methods {#Sec2} ===================== CT scans of a consecutive series of 1000 patients were analyzed in this study. The patients were included irrespective of gender, health conditions, or premorbidities. Exclusion criteria were:Age \<18 years.Recent or consolidated pelvic fractures.Osteolytic pelvic lesions. All patients underwent abdominal CT scans, which include the osseous pelvis, in a supine position with hip and knee joints extended. CT scans were performed during clinical routine. Indications for the CT scans were different from the aim of this study. Accordingly, the patients were not exposed to additional radiation for the purpose of this study. A 64-section CT scanner (LightSpeed VCT, GE Healthcare, Chalfont St. Giles, UK) was used to perform the CT scans. The CT images were taken with 0.6 mm slice depth and saved in slices of 1.25 mm in thickness. 2D reconstructions and measurements were performed with both a standard picture archiving and communication system (PACS) and software (Impax EE R20 XIV SU2, AGFA diagnostic software, Ridgefield Park, NJ). True inlet, outlet, and lateral views of the sacrum were obtained by using a two-dimensional reconstruction tool in a standardized and reproducible manner. In a first step, the midsagittal plane was selected in the lateral view (Fig. [1](#Fig1){ref-type="fig"}a). The axes were adjusted parallel to the lower endplate of S1 (red lines) and along the posterior boarder of S1 (blue line). In a next step, both axes were adjusted in the outlet view (Fig. [1](#Fig1){ref-type="fig"}b) to the vertical axis of the sacrum (green line) and parallel to a line connecting the cranial borders of the SI joint (red line). In a last step, the adjustment in the inlet view (Fig. [1](#Fig1){ref-type="fig"}c) included alignment in a sagittal axis from the middle of the promontorium to the spinous process (green line) and in the coronal plane parallel to posterior wall of the body of S1 (blue line). The same procedure was performed in S2 as well.Fig. 1**a** Alignment in the *lateral view* in a midsagittal plane: parallel to the lower endplate of S1 in the upper, middle, and lower quartiles of the body S1 (*red lines*) and along the posterior boarder of S1 (*blue line*). **b** Alignment in the *outlet view*: along the vertical axis of the sacrum (*green line*) and parallel to a line connecting the upper borders of the SI joint (*red line*). **c** Alignment in the *inlet view*: from the middle of the promontorium to the spinous process (*green line*) and parallel to the posterior wall of the body of S1 (*blue line*). **d** Assessment of the width of the corridor in the *inlet view* at the SI joint (diameter *1*), in the middle of the distance between SI joint and midsagittal plane (diameter *2*) and in the midsagittal plane (diameter *3*) The assessment of the width of the corridor in the inlet view (inlet corridor) is shown in Fig. [1](#Fig1){ref-type="fig"}d. A line parallel the posterior wall of S1 (blue line) served as a reference for the measurement. The width of the corridor was defined as the perpendicular distance between the anterior cortex of the sacrum and this line. The corridor was assessed on both sides at the SI joint (diameter 1), in the middle of the distance between SI joint and midsagittal plane (diameter 2) and in the midsagittal plane (diameter 3) on the right side and left side. In addition, the corridor was assessed in the upper quartile, in the middle and in the lower quartile of the craniocaudal extension of the body S1 by parallel translation of the red axis in the sagittal view (Fig. [1](#Fig1){ref-type="fig"}a). The same procedure was performed in S2 as well. Subsequently, the diameters were used to calculate different types of inlet corridors. In a first step, the corridors were either defined as ascending, horizontal, or descending corridors. In an ascending corridor, the diameter 1 (SI joint) is smaller than the diameter 3 (midsagittal, Fig. [2](#Fig2){ref-type="fig"}a). In a descending corridor, the diameter 1 is larger than the diameter 3 (Fig. [2](#Fig2){ref-type="fig"}b). In a horizontal corridor, the diameters 1 and 3 are equal (Fig. [2](#Fig2){ref-type="fig"}c). A threshold of 3 mm was applied to eliminate marginal differences between the diameters, which are not clinically relevant. Accordingly, the diameter 3 in an ascending corridor should be minimum more than 3 mm larger than the diameter 1 (and vice versa in a descending corridor).Fig. 2**a** Ascending type. **b** Descending type. **c** Horizontal type In a second step, the presence of a ''notch'' was assessed. A notch is an indentation of the anterior cortex of the sacrum between the SI joint and the midsagittal plane with proximity to the nerve root L5, which, therefore, may be harmed during screw misplacement. The arithmetic mean between diameter 1 and 3 was calculated, which is equivalent to diameter 2 in the presence of a straight-lined anterior cortex between SI joint and the midsagittal plane. If the measured diameter 2 is more than 3 mm smaller than the calculated diameter 2, the presence of a notch was defined (Fig. [3](#Fig3){ref-type="fig"}a). Figure [3](#Fig3){ref-type="fig"}b shows an example of a horizontal corridor without notch.Fig. 3**a** Horizontal type with notch. **b** Horizontal type without notch SPSS Statistics 21.0 (SPSS, Chicago, IL, USA) was used for the statistical analysis. Metric scaled data are reported as arithmetic mean ± standard deviation and categorical data as absolute frequency and percentage distribution. Correlations for metric scaled data were quantified using the Pearson coefficient and for non-parametric samples using the Spearman coefficient. The Kolmogorov--Smirnov test was used for the determination of the distribution form. The probability level was set at *p* \< 0.05. Results {#Sec3} ======= CT scans of 1000 patients were analyzed in this study. Measurements were performed both on the right side and left side resulting in 2000 measurements. The mean age of the patients was 68.8 ± 21.5 years. There were 623 female (62.2 %) and 387 male patients (37.8 %) included. The results of the assessment of the different corridor types are shown in Table [1](#Tab1){ref-type="table"}. While the ascending type was the most frequent corridor type in the upper quartile of S1 (71 %), its relative frequency declined to 5 % in the lower quartile. There were no ascending types in S2. Descending types were found in 15 % in the upper quartile of S1 and in 86 % in the lower quartile. In S2, descending types were, by far, the most frequent corridor type. Horizontal types were less frequent with a relative incidence between 2 and 14 % (Table [1](#Tab1){ref-type="table"}). The results of the assessment of the presence of a notch are shown in Table [2](#Tab2){ref-type="table"}. We found a notch in the majority of cases with relative frequencies ranging from 69 % (upper quartile of S1) to 95 % (upper quartile of S2) (Table [2](#Tab2){ref-type="table"}). The mean depths of the notch are shown in Table [3](#Tab3){ref-type="table"}. There were no significant differences between the different groups.Table 1Distribution of the relative frequency of the different corridor types in the upper, middle, and lower quartiles of S1 and S2![](402_2016_2528_Figa_HTML.gif){#d29e530}S1 upper quartile (%)S1 middle quartile (%)S1 lower quartile (%)S2 upper quartile (%)S2 middle quartile (%)S2 lower quartile (%)Ascending type71215000Horizontal type14129225Descending type156686989895 Table 2Distribution of the relative frequency of the presence of a notch in the upper, middle, and lower quartiles of S1 and S2S1 upper quartileS1 middle quartileS1 lower quartileS2 upper quartileS2 middle quartileS2 lower quartileNumber (*n* = 2000)138014951878189716041711Percent69 %75 %94 %95 %80 %86 % Table 3Depth of the notch in the upper, middle, and lower quartiles of S1 and S2 (in mm)Mean (±SD)S1 upper quartile9.9 (±7.0)S1 middle quartile8.5 (±5.2)S1 lower quartile9.4 (±4.3)S2 upper quartile11.8 (±7.6)S2 middle quartile6.1 (±2.3)S2 lower quartile7.5 (±5.1) Discussion {#Sec4} ========== Percutaneous SI screw fixation under fluoroscopic control is an effective and widely used method for the stabilization of the posterior pelvic ring. Screw misplacement with consecutive nerve lesions, mainly of the nerve root L5, is a major complication of this technique \[[@CR6], [@CR16]--[@CR20]\]. Interindividual variations in the morphology of the upper sacrum, the so-called sacral dysmorphism, particularly account for these findings. Sacral dysmorphism has been described by several authors. Most of these studies, however, are anatomical studies describing the morphological variance of the upper sacrum without assessing their relevance for percutaneous SI screw fixation \[[@CR26]--[@CR28]\]. Other authors assess the sacral morphology considering its clinical relevance for SI screw fixation. Carlson, for example, described the three-dimensional shape of the corridor for SI screw fixation (vestibule concept) in a CT-based study, \[[@CR23]\], while Mendel showed a CT-based 3D model of secure bone corridors (safe zones) and optimal trajectories for sacroiliac screws \[[@CR24]\]. The authors, additionally, provided recommendations for the angulation of SI screws in different planes. These recommendations, however, differ significantly between the authors, which, obviously, is not surprising given the variations in sacral morphology. The angle in a caudocranial direction (outlet view), for example, ranged from 2° to 18°, while the angle in a posteroanterior direction (inlet view) ranged from 2° to 16° \[[@CR23], [@CR24]\]. In conclusion, a ''one-size-fits-all'' approach to percutaneous SI screw fixation is not feasible given the diversity of the sacral morphology. To avoid screw misplacement in percutaneous SI screw fixation, CT guided and navigated procedures have been developed \[[@CR9]--[@CR12], [@CR14], [@CR15], [@CR29]\]. These procedures allow for secure screw placement even in the presence of narrow osseous corridors and sacral dysmorphia, which was demonstrated in the above-mentioned studies. Besides a higher radiation exposure \[[@CR14]\], the major drawback of these techniques, however, is their limited availability beyond trauma centers. Accordingly, SI screw fixation under fluoroscopic control is and will continue to be an important technique for minimally invasive stabilization of the posterior pelvic ring. SI screw fixation under fluoroscopic control not only relies on sound intra-operative imaging, but also on the surgeon´s ability to interpret the two-dimensional images in a 3D context. Several authors have provided different approaches for the interpretation of these images in an inlet, outlet, and lateral views \[[@CR4], [@CR30]--[@CR32]\]. Routt, for example, reported that abnormal morphological patterns of the upper sacrum could be easily identified using mainly the outlet view and the lateral view \[[@CR30]\]. We agree that the outlet view allows for secure screw placement above the first sacral foramina in S1 even in the presence of dysmorphic foramina. The lateral view allows for the assessment of the upper, lower, and posterior borders of the osseous corridor. However, we disagree that the lateral view is valuable for the assessment of the anterior border of the corridor. First, the anterior border is not strict parallel to the projection in the lateral view. Our data show that the anterior border may be divergent (ascending type), horizontal, or convergent (descending type) relative to the posterior border, which highly influences its radiological appearance in the lateral view. Second, indentations of the anterior cortical border of the corridor (notches) may not be assessable in a view axial to this border. The techniques of corridor assessment described by Mendel \[[@CR31]\] and Noojin \[[@CR32]\] neither assess the ''notch'' problem as well. Miller \[[@CR33]\] summarized the three key points of fluoroscopically controlled SI screw fixation in the presence of sacral dysmorphism in a recent review article as follows: First, dysmorphic sacral neuroforamina are assessed in the outlet view. Second, the posterior wall and variances in the sacral alar slope are assessed in the lateral view. Third, notches of the anterior cortex are assessed in the inlet view. According to our own experience, however, the inlet view may not be appropriate to reliably detect notches of the anterior cortex in all cases. The inlet view is a two-dimensional projection with an overlap of the anterior cortex of the vertebral bodies S1 and S2. Our data, however, show that the anterior cortex of the upper sacrum is not a flat surface in a three-dimensional space, which would result in a single line in a two-dimensional projection. The corridor type, for example, changes very frequently within the vertebral body of S1 (Table [1](#Tab1){ref-type="table"}). In addition, the presence or absence of a notch is not detectable in an inlet view in the majority of cases and further limits the assessment of the anterior cortex of the upper sacrum in this view. The fluoroscopic control during SI screw fixation is adequate when drilling in a strict transverse direction. Transverse corridors of adequate width, however, are present in a minority of patients only. Ascending corridors with a notch have a very narrow intraosseous corridor in a transverse direction. These types require a more posterior entry point (even posterior to the posterior wall of S1 in a lateral view) and a more ascending drilling direction, which increases the risk of nerve lesions particularly in the additional presence of a notch. The aim of our study, therefore, was to develop a simple method for the preoperative assessment of the corridor type and the presence of a notch with widely available tools as shown above. We consider horizontal and descending corridor types without a notch to be feasible for fluoroscopically controlled SI screw fixation, while ascending types with notch bear a higher risk of iatrogenic nerve lesions. In these cases, we recommend navigated SI screw fixation. One interesting finding of our study was the high frequency of ascending corridors in the upper quartile of S1 with a much lower frequency in the middle quartile and lower quartile of S1. An ascending corridor in the upper quartile is mainly determined by the inclusion of the sacral promontory in the corridor analysis. In the upper quartile, there was an additional notch in 69 % of the cases as well. Accordingly, we consider a screw trajectory to the promontory to be associated with a higher risk of L5 nerve root lesions than more flat screw trajectories. One limitation of our study has to be noted. It was our goal to classify the morphology of the upper sacrum based on the risk of iatrogenic nerve lesions during percutaneous SI screw fixation. Accordingly, we focused on the corridor morphology in the inlet view and defined different inlet corridor types as well as the presence of a notch. For a safe overall screw placement additional parameters, such as the absolute diameter of the corridor in mm and corridors in the outlet view are relevant as well. The assessment of these parameters, however, was not the goal of this study. In conclusion, our concept obviously is not a "one-size-fits-all'' concept neither. It aims at three goals: (a) to increase the awareness of the surgeon for the problem of sacral dysmorphism, (b) to emphasize the importance of a preoperative assessment of the osseous corridor, and (c) to provide a simple method for the preoperative assessment with widely available tools. A correction to this article is available online at <https://doi.org/10.1007/s00402-018-2928-7>. Open access funding provided by Medical University of Innsbruck. Conflict of interest {#FPar1} ==================== The authors declare that they have no conflict of interest.

{ "pile_set_name": "PubMed Central" }

All relevant data are within the paper. Introduction {#sec005} ============ Epilepsy and intellectual/developmental disabilities (ID/DD) are both common pediatric neurological disorders. ID/DD is one of the main comorbidities of epilepsy \[[@pone.0141782.ref001]\] with about a quarter of epileptic children having ID/DD. The prevalence is much higher in children presenting with early-onset (before three months after birth) epilepsy \[[@pone.0141782.ref002]--[@pone.0141782.ref004]\]. Although frequent and refractory seizures may cause cognitive and motor regression, common pathophysiological mechanisms may be responsible for the high rate of co-occurrence of epilepsy and ID/DD \[[@pone.0141782.ref005]\]. The etiologies of these two disorders are complex and diverse with the majority being unknown. Genetic factors play a major role in the etiologies of epilepsy and ID/DD, especially in pediatric patients, who are highly heterogeneous. Defects in many genes have been reported as shared underlying mechanisms of epilepsy and ID/DD \[[@pone.0141782.ref006]--[@pone.0141782.ref009]\]. These genes are involved in different pathways. Ion channel genes, which are particularly relevant to epilepsy, account for a significant proportion \[[@pone.0141782.ref010]\]. However, the phenotypes related to these genes are difficult to differentiate clinically, and the detection of mutations in suspected genes is always a challenge. Seizures were always intractable and indicated a poor prognosis when co-occurring with ID/DD. Therefore, knowing the genetic background and pathogenesis of epilepsy and ID/DD is valuable not only for diagnosis and prognosis, but also for genetic counseling and treatment. Many genes related to epilepsy and ID/DD have been reported. However, for individual patients with non-syndromic epilepsy, it has been difficult to isolate the causative gene mutations from a large number of possible candidate genes using conventional Sanger sequencing. The rapidly developing targeted next-generation sequencing (NGS) has been proved to be a fast, economic and accurate approach for screening gene mutations in disorders with both genetic and phenotypic heterogeneity, including epilepsy and ID/DD \[[@pone.0141782.ref011]\]. In this study, we used targeted NGS to investigate 300 candidate genes related to epilepsy and ID/DD in 253 Chinese children with unexplained epilepsy and ID/DD. We aim to make genetic diagnosis for these patients and find clues to help us explain the common genetic background of epilepsy and ID/DD. Methods {#sec006} ======= Ethics statement {#sec007} ---------------- Written informed consent was obtained from the parents of all the patients. This study was approved by the Institutional Review Boards of Peking University First Hospital. All data of this study were analyzed anonymously. Patients {#sec008} -------- A total of 253 Chinese children with unexplained epilepsy and ID/DD were recruited from the Department of Pediatrics, Peking University First Hospital from 2006 to 2014. All patients were clinically diagnosed as having epilepsy and ID/DD of unknown origin, including 65 patients diagnosed as early-onset (before three months after birth) epilepsy. Nevertheless, it was strongly suspected that the etiology of these patients' diseases was genetic, owing to the following evidences: (1) no definite perinatal brain injury; (2) no hypoxia, ischemia, infection of the central nervous system or cranial trauma; (3) no evidence of typical inherited metabolic disorders or specific neurodegenerative disorders based on clinical features, neuroimaging or blood/urinary metabolic diseases screening; (4) normal routine karyotyping and (5) the detection of chromosome sub-telomeric rearrangements with multiplex ligation-dependent probe amplification (MLPA) showing no abnormalities \[[@pone.0141782.ref012]\]. 253 cases of our cohort included cases from 246 trios, of which father or mother of four trios (1649, 5165, 5168, and 5237) had epilepsy history, and parents of all the other 242 trios had no epilepsy and any related history. Other seven cases were probands of seven families with more than one patient, of which six families (3604, 5750, 6047, 6364, 6526, and 6636) consisted of two affected children and their unaffected parents, and one family (5240) with five affected patients (proband, father, uncle, aunt and grandfather) accorded with autosomal dominant inheritance. All genomic DNA used in experiments were extracted from peripheral leukocytes. Targeted next-generation sequencing {#sec009} ----------------------------------- In accordance with the literatures searched within online databases, a total of 300 candidate genes associated with epilepsy and ID/DD were selected as the genes of interest. We used a custom-designed gene panel, synthesized using the Agilent Sure-Select Target Enrichment technique (Zhongguancun Huakang Gene Institute, China), to capture the coding regions from the 300 genes, including their exons and exon-intron boundaries (1.285M bp in total). The following NGS was performed on an Illumina GAIIx platform (Illumina, San Diego, CA, U.S.A.) using paired-end sequencing of 110 bp. Bioinformatics analysis of the raw data included the following steps: (1) image analysis using RTA software (real-time analysis, Illumina); (2) base calling using CASAVA software v1.8.2 (Illumina); (3) filtered out duplicate and low base quality score reads using the Genome Analysis Tool kit (GATK); (4) aligned clean paired-end reads to the human reference genome build hg19 using BWA software (Pittsburgh Supercomputing Center, Pittsburgh, PA, U.S.A.); (5) identified insertion-deletions (indels) and single-nucleotide polymorphisms (SNPs) using the GATK and annotated using ANNOVAR; (6) performed in silico pathogenicity prediction of novel missense variations using Polyphen2 \[[@pone.0141782.ref012], [@pone.0141782.ref013]\]. The sequencing depth was more than 5X (range of 5X-185X; average of 136X), and the mean coverage was 98.56%. On average, 423 variations within the 300 genes were found in each patient. We then formulated the following filtering criteria to determine every possible pathogenic variation from the large amount of initial variations: (1) insertion/deletion variations; (2) premature/delayed termination codon variations; (3) splice site variations including substitution at nucleotide +1/-1 of intron; (4) missense variations predicted by polyphen2 as probably/possibly damaged or benign. The variation meeting any one of the above criteria was considered to be a candidate for pathogenic variations and was selected for further analysis \[[@pone.0141782.ref012]\]. According to the HGMD Professional database, the 1000 Genomes Browser, PubMed and the UCSC database, we marked the reported pathogenic mutations and excluded known polymorphisms. Finally, on average 17 possible pathogenic variations were identified in each patient (range 5--27). We chose variations which to validate according to the known inheritance pattern of the involved genes. Heterozygous variations of genes with autosomal or X-linked dominant heredity, homozygous or compound heterozygous variations of genes with autosomal recessive heredity, and hemizygous variations of genes with X-linked recessive heredity were regarded as likely causative variations. We performed validation and parental origin analyses for these variations by conventional Sanger sequencing, and confirmed causative mutations according to parental origin of the variations and clinical features of the patients \[[@pone.0141782.ref012]\]. Protein structure modelling {#sec010} --------------------------- The homology modelling server SWISS-MODEL \[[@pone.0141782.ref014]--[@pone.0141782.ref016]\] was used to predict the tertiary structure of *KCNAB1* protein. We chose the protein crystal structure 3EAU \[[@pone.0141782.ref017]\] of *KCNAB2* from Protein Data Bank (PDB) \[[@pone.0141782.ref018]\] as the template. For the homology of *KCNAB1* and *KCNAB2*, the structure from 89 to 413 amino acids of *KCNAB1* protein (UniProtKB ID: Q14722) were predicted. Results {#sec011} ======= We identified causative mutations within 24 genes in 46 patients of our cohort, including two likely pathogenic mutated genes in two patients. The total detection rate of our study was 18% (46/253) in the whole group and 26% (17/65) in the early-onset epilepsy group. The detected mutations included 32 novel and 16 reported mutations. Nineteen of the mutations were severe, including eight premature/delayed termination codon mutations, ten insertion/deletion mutations and one splicing site mutation. The remaining 29 mutations were non-synonymous, including 27 mutations predicted to be "probably damaging" and two mutations (*AFF2* p.Gly547Asp and *RELN* p.Val3426Ile) that were predicted to be "possibly damaging" by Polyphen2. In two patients (5240 and 6189) with distinct epilepsy and ID/DD phenotypes, two mutated genes were regarded as likely pathogenic due to the unmatched phenotype or the unavailability of segregation analysis in the family. An overview of the clinical features of patients and their mutations is described in [Table 1](#pone.0141782.t001){ref-type="table"}. 10.1371/journal.pone.0141782.t001 ###### Overview of clinical features of the patients and the mutations. ![](pone.0141782.t001){#pone.0141782.t001g} Gene Case Sex Study age (y) Seizures (onset age) ID/DD Nucleotide substitution Amino acid substitution Parental origin Novel/ reported Final diagnosis ----------- ------------------------------------------- ----- --------------- ------------------------------------------------------- -------------------- ------------------------- -------------------------- ----------------- ----------------- ----------------- *SCN1A* 152 M 8 GTCS, FS (6 m) Severe c.4547C\>A p.Ser1516Ter De novo Reported DS 2038 M 10 GTCS, PS (4 m) Severe c.2134C\>T p.Arg712Ter De novo Reported DS 5791 M 3 PS, GTCS (3 m) Severe c.4942C\>T p.Arg1648Cys De novo Reported DS 6047 M 3 PS, FS, Myoclonus, Absences (8 m) Severe c.2589+3A\>T \- De novo Reported DS 6207 F 15 PS. GTCS, FS (3 m) Severe c.3733C\>T p.Arg1245Ter De novo Reported DS 6222 F 0.5 PS, SE (16 d) Severe c.659T\>A p.Val220Asp De novo Novel MMPSI 6300 F 3 PS, Myoclonus, GTCS, SE, FS (6 m) Severe c.3372delT p.Phe1124LeufsTer4 De novo Novel DS 6492 M 4 PS, FS (7 m) Severe c.2488G\>T p.Glu830Ter De novo Novel DS *SCN8A* 3129[\*](#t001fn002){ref-type="table-fn"} M 9.5 GTCS, FS (11 m) Severe c.2668G\>A p.Ala890Thr De novo Novel 5487[\*](#t001fn002){ref-type="table-fn"} M 3.5 GS (6 m) Severe c.4850G\>A p.Arg1617Gln De novo Reported EE 6219[\*](#t001fn002){ref-type="table-fn"} F 1.5 PS, GTCS, Myoclonus, IS (3 d) Severe c.1221G\>C p.Leu407Phe De novo Novel EIEE 6325[\*](#t001fn002){ref-type="table-fn"} F 1 PS, GTCS (2.5 m) Moderate to severe c.2549G\>A p.Arg850Gln De novo Novel EIEE 6908[\*](#t001fn002){ref-type="table-fn"} M 1.5 GTCS (4 m) Moderate to severe c.4787C\>G p.Ser1596Cys De novo Novel EE YL F 3 PS (6 m) Severe c.4935G\>A p.Met1645Ile De novo Novel *KCNQ2* 597 M 3 GS (3 d) Severe c.365C\>T p.Ser122Leu De novo Reported EIEE 2202 M 2 PS, GTCS, FS (3 d) Severe c.956A\>C p.Lys319Thr De novo Novel EIEE 2312 M 1.5 PS, FS (4 h) Severe c.830C\>T p.Thr277Ile De novo Novel EIEE 5630 M 1 PS, IS (3 d) Severe c.1655A\>C p.Lys552Thr De novo Reported EIEE *IQSEC2* 3481 M 9 PS (3 y 6 m) Severe c.88_90delATC p.Ile30del De novo Novel 3604 M 7 PS, FS (6 m) Severe c.1049C\>A p.Ala350Asp Mother Novel 5292 M 2 PS, IS, Myoclonus (1 y 4 m) Severe c.2846_2852delCCCAGGT p.Ser949CysfsTer7 De novo Novel *CDKL5* 1613 F 1 PS (40 d) Severe c.2314delA p.Lys772ArgfsTer12 De novo Novel EIEE 5057 F 1.5 PS (1 m) Severe c.528G\>A p.Trp176Ter De novo Novel EIEE *DYRK1A* 2091 M 7 GTCS, FS, SE (1 y 6 m) Severe c.859G\>T p.Asp287Tyr De novo Novel 2959 F 3 GS (2 y) Severe c.946C\>T p.Gln316Ter De novo Novel *GABRB3* 1538 M 1 PS, GTCS (6 m) Severe c.914C\>T p.Ala305Val De novo Novel SYH M 1.5 PS (9 m) Severe c.509T\>G p.Leu170Arg De novo Novel *PCDH19* 6526 F 9 PS (5 m) Severe c.1091delC p.Pro364ArgfsTer4 Father Reported LXX F 3 PS, FS (9 m) Moderate c.370G\>A p.Asp124Asn Father Novel *STXBP1* 527 M 2 IS (7 d) Severe c.568C\>T p.Arg190Trp De novo Reported EIEE MYS M 2 IS (3 d) Severe c.568C\>T p.Arg190Trp De novo Reported EIEE *AFF2* 6636 M 11 PS (6 y) Moderate c.1640G\>A p.Gly547Asp Mother Novel *ALDH7A1* 5921 M 4 PS (3 m) Moderate to severe c.1553G\>C c.1061A\>G p.Arg518Thr p.Tyr354Cys FatherMother Novel Reported PDE *ATP1A2* 5871 M 1.5 PS, FS (1 m) Mild c.2563G\>A p.Gly855Arg De novo Reported *CASK* 2584 F 6 IS (1 y) Severe c.2141delC p.Ala714GlufsTer13 De novo Novel MICPCH *FOXG1* 2539 M 2 GTCS, PS, Laugh attack (6 m) Severe c.738C\>A p.Tyr246Ter De novo Novel *GRIN2A* 6245 F 6 PS, ESES (5 y) Severe c.2191G\>A p.Asp731Asn De novo Reported *GRIN2B* 1503 M 2 PS, IS, Myoclonus, Tonic, Startle (6 m) Severe c.1985A\>C p.Gln662Pro De novo Novel *KCNAB1* HY F 4 PS, GTCS (10 d) Severe c.1062dupCA p.Leu355HisfsTer5 De novo Novel EIEE *PRRT2* 5240 F 1 GS, IS (3 m) Severe c.649dupC p.Arg217ProfsTer8 Father Reported *RELN* 6235 F 7 Myoclonus (3 y) Severe c.10276G\>A c.2252A\>C p.Val3426Ile p.Lys751Thr Father Mother Novel Novel *SHANK3* ZXT M 3 IS (1 y 10 m) Severe c.3598G\>C p.Ala1200Pro De novo Novel *SLC2A1* 1649 M 5.5 GTCS (4 y 8 m) Mild c.1477T\>C p.Ter493ArgextTer56 Mother Novel *SYNGAP1* 5828 F 5 Atypical absences, Myoclonus, Atonic, GTCS, FS (10 m) Severe c.829dupC p.Pro277ProfsTer7 De novo Novel *UPF3B* 6189 M 2 IS (8 m) Severe c.883T\>A p.Leu295Met Mother Novel *ZEB2* 4620 M 3 PS (?) Severe c.1426_1427insA p.Met476AsnfsTer6 De novo Reported M, male; F, female; y, years; m, months; d, days; h, hours; PS, partial seizures; FS, febrile seizures; SE, status epilepticus; DS, Dravet syndrome; EIEE, early infantile epileptic encephalopathy; EE, epileptic encephalopathy; IS, infantile spasms; GS, generalized seizures; PDE, pyridoxine-dependent epilepsy; MMPSI, malignant migrating partial seizures in infancy; ESES, electrical status epilepticus in sleep; MICPCH, mental retardation and microcephaly with pontine and cerebellar hypoplasia \*have been reported in \[[@pone.0141782.ref012]\]. Among the 46 cases with a causative mutation, patients with an *SCN1A* mutation accounted for the largest proportion of 17% (8/46), followed by *SCN8A*, *KCNQ2* and *IQSEC2* of 13% (6/46), 9% (4/46) and 7% (3/46) respectively. A total of 38% (9/24) of the mutated genes (*SCN1A*, *SCN8A*, *KCNQ2*, *IQSEC2*, *CDKL5*, *DYRK1A*, *GABRB3*, *PCDH19*, and *STXBP1*) reoccurred at least two times in this study, and 63% (15/24) of the mutated genes (*ALDH7A1*, *AFF2*, *ATP1A2*, *CASK*, *FOXG1*, *GRIN2A*, *GRIN2B*, *KCNAB1*, *PRRT2*, *RELN*, *SHANK3*, *SLC2A1*, *SYNGAP1*, *UPF3B* and *ZEB2*) occurred only one time. All mutated genes detected in this study along with the number of patients in whom each gene was detected were shown in [Fig 1](#pone.0141782.g001){ref-type="fig"}. ![Pathogenic (red) and likely pathogenic (green) mutated genes identified in 253 patients with unexplained epilepsy and intellectual/developmental disabilities.](pone.0141782.g001){#pone.0141782.g001} Notably, we detected a novel *de novo* heterozygous mutation (c.1062dupCA p.Leu355HisfsTer5) within *KCNAB1* in one patient with EIEE. This mutation led to a premature termination codon. The patient with the mutation was a four-year-old girl with early-onset seizures (onset at ten days after birth). The frequent partial seizures were followed by intractable generalized tonic-clonic seizures (GTCS). Until now, she has used seven antiepileptic drugs (AEDs). Sodium Valproate, Levetiracetam, Topiramate and Clonazepam had no obvious effect, while Oxcarbazepine, Lamotrigine and Zonisamide reduced the seizures; among them Zonisamide add-on had the best effect with a 75% reduction of the seizures. Her parents are planning to consent for her to receive vagal nerve stimulation (VNS) to treat her intractable epilepsy. She also has severe ID/DD now (non-verbal, limited interaction with her parents, but can walk independent from 2 years 3 months), with occasional panic attacks. Her prenatal history was normal and neurological examination was unremarkable. One of her electroencephalograph (EEG) recordings showed a slow spike-wave in the left frontotemporal region, while another EEG recorded frequent sharp waves in the left temporooccipital region. Her cranial magnetic resonance imaging (MRI) was normal. *KCNAB1* encodes the beta-1 member of the shaker-related family of voltage-gated potassium channels. This member includes three isoforms (Kvβ1.1- Kvβ1.3) of the *KCNAB1* gene \[[@pone.0141782.ref019]--[@pone.0141782.ref023]\]. Kvβ1.1, the longest isoform, is restricted expressed in brain \[[@pone.0141782.ref024]\]. The shaker-related voltage-gated K+ (Kv) channels consist of alpha and beta subunits \[[@pone.0141782.ref025]\]. The beta subunits modulate the gating properties of the alpha-subunit potassium channels. Voltage-dependent potassium channel proteins are responsible for the electrical properties of excitable cells and play physiological roles in non-excitable cells \[[@pone.0141782.ref026]\]. To further study the pathogenicity of p.Leu355HisfsTer5 mutation, protein tertiary structures of wild type and p.Leu355HisfsTer5 mutation of *KCNAB1* were predicted using the SWISS-MODEL. Because the template 3EAU was a homotetramers crystal structure, we predicted both monomer and tetramer of *KCNAB1* protein. According to the monomer prediction ([Fig 2A](#pone.0141782.g002){ref-type="fig"}), the mutated *KCNAB1* protein lost C-terminal helices when compared to wild type. According to the tetramer prediction ([Fig 2B](#pone.0141782.g002){ref-type="fig"}), the wild type was able to bind to NADP^+^ by sharing the same binding domains of *KCNAB2* protein (data from: <http://www.uniprot.org/uniprot/Q14722>). But the mutated *KCNAB1* protein was not able to bind to NADP^+^ for losing a NADP^+^ binding domain (375--381 amino acids), though might still be able to form a tetramer. Previous study of experimental point mutation within NADP^+^ binding domain of *KCNAB1* protein showed significant effects on Kv1 channel trafficking and axonal targeting \[[@pone.0141782.ref027]\]. Therefore, the p.Leu355HisfsTer5 mutation that we identified is probably pathogenic. ![Structure modeling of wild type and p.Leu355HisfsTer5 mutation of *KCNAB1* with SWISS-MODEL.\ A: Monomer view: comparing with wild type, the mutant lost C-terminal structure of the protein; B: Tetramer view: comparing with wild type, the mutant might lose the ability of binding NADP^+^.](pone.0141782.g002){#pone.0141782.g002} To understand a possible common genetic mechanism of epilepsy and ID/DD, we classified 24 mutated genes identified in our study into groups according to gene function. We found that ion channel genes had the largest percentage of occurrence, 33% (8/24) with genes related to synapse coming second at 21% (5/24). Other genes also identified were classified as having functions in transcriptional regulation, protein kinase modulation, cell metabolism and cell-cell interaction. The classification and the number of genes of each group were shown in [Table 2](#pone.0141782.t002){ref-type="table"}. 10.1371/journal.pone.0141782.t002 ###### The classification of mutated genes. ![](pone.0141782.t002){#pone.0141782.t002g} Gene Function Gene (n) Mutated Genes -------------------------- ---------- ----------------------------------------------------------------------------- Ion channel 8 *SCN1A*, *SCN8A*, *KCNQ2*, *GABRB3*, *ATP1A2*, *GRIN2A*, *GRIN2B*, *KCNAB1* Synapse 5 *IQSEC2*, *STXBP1*, *PRRT2*, *SHANK3*, *SYNGAP1* Transcription regulation 4 *AFF2*, *FOXG1*, *UPF3B*, *ZEB2* Protein kinase 3 *CDKL5*, *DYRK1A*, *CASK* Cell metabolism 2 *ALDH7A1*, *SLC2A1* Cell-cell interaction 2 *PCDH19*, *RELN* Discussion {#sec012} ========== In this study, we discovered 32 novel and 16 reported mutations within 24 genes in 46 patients of our cohort, including two likely pathogenic mutated genes in two patients. The total detection rate of our study was 18% (46/253) in the whole group and 26% (17/65) in the early-onset epilepsy group. Early-onset epilepsy had the relatively higher detection rate. We made genetic diagnosis for these 46 patients. This was critical for them to improve further management and genetic counseling for their epilepsy. We expanded the phenotype and mutation spectrum of the 24 genes identified in our study. This provided more information for further understanding of these disease-causing genes. Patients with an *SCN1A* mutation accounted for the largest proportion, 17% (8/46), of which seven patients were diagnosed as Dravet syndrome and one patient was diagnosed as MMPSI. MMPSI cases with a *SCN1A* mutation have been reported and MMPSI is regarded as the most severe phenotype of *SCN1A* to date \[[@pone.0141782.ref028], [@pone.0141782.ref029]\]. Our MMPSI patient with a novel *SCN1A* mutation provided further evidence that *SCN1A* defects play an important role in MMPSI. A total of 63% (15/24) of the mutated genes occurred only one time in our study; therefore, it seems that epilepsy and ID/DD are phenotypes that occur as a consequence of brain dysfunction caused by highly diverse mutated genes, most of which are isolated and fit the rule of common disease rare variations. In addition, the spectrum of mutated genes in our study is rather different from those reported in other similar studies \[[@pone.0141782.ref011], [@pone.0141782.ref030]\]. Population diversity and different inclusion criteria (both for patients and candidate genes) may account for this inconsistency. In this study, we identified a novel *de novo* heterozygous mutation (c.1062dupCA p.Leu355HisfsTer5) within *KCNAB1* in one patient with EIEE. *KCNAB1* has been reported as a susceptibility gene for epilepsy, particularly temporal lobe epilepsy (TLE), but no pathogenic mutation has been reported. An association study of 2717 epileptic patients reported that numerous SNPs located within *KCNAB1* contributed to the susceptibility to epilepsy. These patients manifested various forms of epilepsy \[[@pone.0141782.ref031]\]. Furthermore, *KCNAB1* was regarded as a candidate gene for lateral temporal epilepsy (LTE) because of its functional interaction with *LGI1* \[[@pone.0141782.ref032]\], the disease-causing gene of autosomal dominant LTE (ADLTE). However, sequencing of *KCNAB1* in ADLTE families without *LGI1* mutations failed to identify any mutations. This suggested that *KCNAB1* does not act as a major disease-causing gene in ADLTE \[[@pone.0141782.ref033]\]. Nevertheless, another association study of 142 LTE patients suggested that *KCNAB1* may be a susceptibility gene of LTE \[[@pone.0141782.ref034]\]. In addition; a genome-wide scan study was conducted on a TLE family. Linkage analysis identified a locus on chromosome 3q25-q26. *KCNAB1* was one of the highest priority genes in this region, but sequencing of *KCNAB1* was unable to identify any mutations \[[@pone.0141782.ref035]\]. In summary, although several previous studies have supported the association of *KCNAB1* with epilepsy, no *KCNAB1* mutations have been reported in patients with this disease previously. However, a mouse model of *KCNAB1* disruption showed significant alterations in hippocampal learning and memory functions \[[@pone.0141782.ref036]\], supporting a possible relationship between *KCNAB1* defects and brain dysfunction. Here we reported an epileptic patient with a *KCNAB1* mutation, which supports the relationship between *KCNAB1* dysfunction and epilepsy, and interestingly, the epileptic discharges of this patient located mostly at temporal region. We also found four patients with a *KCNQ2* mutation. Because *KCNQ2* \[[@pone.0141782.ref037]\] and *KCNT1* \[[@pone.0141782.ref038]\] have already been reported to be common and important genes for epileptic encephalopathy, adding the recently reported *KCNA2* \[[@pone.0141782.ref039]\], *KCNH1* \[[@pone.0141782.ref040]\], *KCNC1* \[[@pone.0141782.ref041]\] genes and now the *KCNAB1* gene, this led us to pay more attention to the potassium channel genes as a group in epilepsy, especially epileptic encephalopathy. Two mutated genes were regarded as likely having pathogenic mutation in our study. First, we detected a heterozygous mutation (p.Arg217ProfsTer8) within *PRRT2* in one family (5240). The proband had severe ID/DD and infantile spasms, but other individuals with the same mutation in this family had benign epilepsy during infancy and normal intelligence during adulthood. We hypothesized that the *PRRT2* mutation in the proband may only increase the risk of epilepsy, while another undiscovered mutated gene may instead contribute to the severe phenotype. In addition, we discovered a hemizygous mutation (p.Leu295Met) in *UPF3B* which have been reported as a causative gene of X-linked recessive mental retardation. The previously reported cases with mutation in *UPF3B* had no seizures. The mutation we found is novel, but probably damaging predicted by Polyphen2. Unfortunately, DNA samples of other male maternal family members were unavailable to make sure the pathogenicity of this mutation. To understand a possible common genetic mechanism of epilepsy and ID/DD, we classified the mutated genes identified in our cohort according to gene function. We have found that ion channel genes had the largest percentage of occurrence. This suggests that ion channels play a vital role in the pathogenesis of epilepsy and ID/DD. Activation of neurotransmitter receptor ion channels at synapses promotes synaptic plasticity during brain development. Therefore, abnormal ion transport may affect neural excitability and brain development, resulting in epilepsy and ID/DD \[[@pone.0141782.ref042]\]. Further, synapse formation and normal function are essential in the signaling and the formation of neural networks. Genes related to synapse formation and function were also closely related to epilepsy and ID/DD. In addition, some factors in transcriptional regulation, protein kinase modulation, cell metabolism and cell-cell interaction may also participate in the common pathogenesis of epilepsy and ID/DD. However, relevant details remain unclear. We believe further study of the common pathogenesis of epilepsy and ID/DD are urgently needed. In our study, with the detection rate of 18%, the targeted NGS is certainly supposed to be an efficient and precise approach to screen monogenic mutations in patients with highly heterogeneous disorders such as epilepsy and ID/DD. However, according to our experience, some limitations of this approach and tips for best performance should be discussed here. First, owing to false positive results, conventional Sanger sequencing is definitely required for the validation of the variations supposed to be significant, especially when the targeted regions have insufficient coverage. Second, on the other hand, false negative results may also occur and may lead to the loss of crucial data. This might be one of the reasons that our study failed to detect gene mutations in the other 207 patients. Third, DNA samples of the parents and other affected or even unaffected members of families are essential to analyze the pathogenicity of the variations. Availability of almost all parental DNA samples in our study played a significant role in data analysis. However, the unavailability of other members in a few families hampered further confirmation of their etiology. Finally, precise clinical data is a prerequisite, without which the genetic diagnosis cannot be made. For example, patient 5871 who carried a *de novo* mutation in *ATP1A2* also had an inherited homozygous mutation (p.Ile105Val) in *CLN3*. Although the *CLN3* nonsynonymous mutation was predicted to be "probably damaging", we still excluded its pathogenicity according to his phenotype, not like neuronal ceroid lipofuscinoses clinically. In summary, we used targeted NGS to investigate causative gene mutations in Chinese children with unexplained epilepsy and ID/DD. We established genetic diagnosis for 46 patients of our cohort and expanded the phenotype and mutation spectrum of 24 genes associated with epilepsy and ID/DD. This study is the first to identify a *KCNAB1* mutation in a patient with EIEE. More cases with mutations in this gene are needed to confirm and clarify its role in epilepsy. We thank our patients for participating in this study. We thank Dr. Macdonald (Department of Neurology, Vanderbilt University Medical Center) for his kindly critical reading and valuable suggestion to this manuscript. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: YJ. Performed the experiments: YZ WK. Analyzed the data: YZ YJ WK YG KG HX. Contributed reagents/materials/analysis tools: YJ XL YW YZ JW FG XW. Wrote the paper: YJ YZ.

{ "pile_set_name": "PubMed Central" }

1. Introduction {#sec1-animals-08-00027} =============== Food guarding (FG) is a natural behavior for Canids, yet can result in adverse consequences when domestic dogs exhibit this behavior in animal shelters. FG can occur when a dog has a consumable item that it wants to retain. Guarding behaviors, such as growling and snapping towards a perceived competitor, can be seen in puppies as young as two to three weeks of age \[[@B1-animals-08-00027]\]. Most, if not all, standardized behavior assessments currently used in shelters \[[@B2-animals-08-00027],[@B3-animals-08-00027],[@B4-animals-08-00027]\] include an evaluation of FG behavior. These assessments typically evaluate the extent of FG by using a fake hand, made of rubber or plastic on a dowel, to pull the food bowl away from the dog and touch their cheek while eating. Different assessments vary the time and degree of contact between the dog and the fake hand. In 2005, The American Society for the Prevention of Cruelty to Animals^^®^^ (ASPCA^^®^^) \[[@B5-animals-08-00027]\] conducted a survey of US animal shelters. Eighty-nine percent of responding organizations reported they conducted a behavior assessment in their shelter and almost all assessed for guarding around a food bowl. For the sake of concision in this manuscript, the specific FG component of the behavior assessment is referred to as the FG item. Two recent studies indicated that assessing for FG in the shelter was a poor predictor of FG behavior in the home. Marder et al. \[[@B6-animals-08-00027]\] looked at 97 dogs adopted from an animal shelter and compared their FG behavior during assessment and in-home. All dogs were assessed using Match-Up II Shelter Dog Rehoming program \[[@B4-animals-08-00027]\]. Twenty (21%) dogs exhibited FG in the assessment and of those, 11 (55%) showed FG in-home. Interestingly, of the 77 dogs that did not display FG during the assessment, 17 (22%) did display FG in-home. Whether or not dogs guarded food during the shelter assessment had little bearing on the FG behavior they displayed in-home. The second study \[[@B5-animals-08-00027]\] followed 96 dogs that exhibited FG during the Safety Assessment for Evaluating Rehoming (SAFER™) assessment \[[@B2-animals-08-00027]\]. Dogs that exhibited FG were allowed free-access to dry dog food in their kennel, and then subsequently adopted into homes. Three weeks after adoption, only one of the 96 dogs was reported to exhibit FG behavior in the home. By three months, no dogs were reported showing FG behavior in-home. The discrepancy between these two studies in the rate of FG in-home could be a result of the free feeding regimen the dogs experienced while in the shelter, substantially different durations of follow-up, support given to adopters, fundamental differences in the shelter dog populations, or other unidentified variables. However, the conclusions regarding the inability of the FG item to reliably predict FG in-home is consistent across the two studies. Both of these studies call into question the value of assessing FG while the dog is in the shelter. It appears that dogs showing FG guarding during the assessment are no more or less likely to exhibit FG in a home. Equally as important, the Marder study \[[@B6-animals-08-00027]\] found that adopters did not perceive FG behavior as a problem in keeping their dog, which suggests that standardized FG assessments may be unnecessary and irrelevant. In 2016, Patronek and Bradley \[[@B7-animals-08-00027]\] considered the usefulness of standardized behavior assessments in shelters and noted that "reliably predicting problematic behaviors in future adoptive homes is vanishingly unlikely." They argued that if a dog bit or threatened to bite during a standardized assessment that whether or not the dog would do so in an adoptive home "would be at best, not much better than flipping a coin". Although the authors were referring to aggression assessments in general, and not just the FG portion of the assessment, their primary argument is that standardized assessments result in an unacceptably high incidence of false positives (which are those dogs that exhibit aggression during the assessment and not in-home). A high rate of false positives is particularly alarming when it comes to FG, because FG was one of the top reasons shelter staff reported a dog would be deemed unadoptable \[[@B5-animals-08-00027]\] and euthanized \[[@B6-animals-08-00027]\]. Even when a shelter chose to not euthanize dogs with FG behavior, those dogs were more likely to have adoption restrictions (such as experienced owners only or families without children) thereby reducing the pool of potential adopters for the dogs \[[@B6-animals-08-00027]\]. Since the publication of the Marder et al. \[[@B6-animals-08-00027]\] and the Mohan-Gibbons et al. \[[@B5-animals-08-00027]\] studies, members of the animal welfare field have shared their interest and concerns with the ASPCA^^®^^ about omitting the FG item from their assessment process. Their primary concerns were that injuries to staff or adopters would increase and that dogs would be returned more often. Thus, the primary goals of this study were to compare the outcomes, injuries, and returns after the FG item was removed from the assessment. 2. Materials and Methods {#sec2-animals-08-00027} ======================== 2.1. Inclusion Criteria {#sec2dot1-animals-08-00027} ----------------------- In order to be included in this study, shelters had to already be using a standardized behavior assessment that included a FG item and be willing to discontinue the FG item for the investigation phase. The FG item involved presenting the dog with a bowl of food and then using a fake hand on a dowel to touch the dog and the bowl while the dog was eating the food. Only FG toward people was addressed; FG directed towards other dogs was not a part of any standardized assessment in this study. Any dog exhibiting FG was included, regardless of breed or age (as long as they were at least 6 months old, per assessment guidelines). Shelters refrained from making any other program or adoption changes during the study that would have affected the research. 2.2. Recruitment of Shelters {#sec2dot2-animals-08-00027} ---------------------------- Animal shelters were recruited for this study through a convenience sample. The authors had prior relationships with some of the shelters and spoke directly with the staff to ask for their participation. Other organizations volunteered themselves for the study following a presentation on guarding behaviors in the shelter given by Drs. Emily Weiss and Amy Marder at the Society of Animal Welfare Associations conference in 2014. The authors also sought recommendations from other leaders in the field to identify organizations that might be interested in participating. Any organization that expressed interest in participating was sent an email providing a summary of the project. A total of 11 shelters enrolled in the study. In the first month, 2 shelters were excluded: 1 shelter pre-screened dogs for FG as part of their relocation program, rendering them inadmissible and the other shelter withdrew because they felt they could not meet the data collection requirements. The shelters were all non-profit humane societies except for one large municipal animal control organization. These shelters were located in seven states: Colorado, Florida, Mississippi, New York, North Carolina, and South Carolina. There was one shelter in each of these states except for Florida and New York, which each had two. Seven shelters used the SAFER™ assessment and two shelters used their own standardized assessment. 2.3. Protocol {#sec2dot3-animals-08-00027} ------------- Nine shelters completed the 5-month study from 1 September 2015 to 31 January 2016. In the first 2 months (baseline period), shelters were instructed to continue with their regular process for assessing behavior, then they were to omit the FG item for the next 2 months (investigation phase). In the investigation phase, shelters were advised to conduct all other components of their assessment (minus the FG item) and to proceed with adoptions as per normal. A data collection form was issued to each shelter to collect the following information for each month of the baseline phase and the investigation phase for the entire dog population: (1) total intake, (2) total returns of dogs adopted within the past 30 days, (3) average length of stay, (4) total number of dogs adopted, transferred, and euthanized; and the total number of bites and injuries (5) observed while in-shelter and (6) reported in-home post-adoption. Those data were also collected for the dogs that were classified as food guarders in both the baseline and investigation phases. During the last month of the study, in January, the only data collected were the number of returns and reports of bites or injuries that occurred in-home (\#2 and \#6 above). Bites and injuries were collected as a single category, defined as any type of injury to a human, in any context, while in the shelter or after adoption. This information was gleaned from reports in the shelter and from conversations when owners contacted the shelter to discuss their dog post-adoption. Shelters were asked to report all incidents for any reason and, as a result, an incident could occur during any interaction with the dog and were not restricted to interactions around food. For example, a dog in the FG group may have caused harm by pulling someone off their feet while on a walk was still counted as an injury, even though no food was involved. This ensured a conservative approach and that all incidents were captured for this research. On a monthly basis, completed forms were collected from each shelter by email. At the end of the study, the authors conducted an informal conference call with all of the shelters to learn more about their experiences during the study. 2.4. Identification of FG {#sec2dot4-animals-08-00027} ------------------------- Shelters were instructed to identify FG in one of three ways: (1) during the FG item of the assessment in the baseline phase, (2) from interviews conducted with the person who brought the animal to shelter, or (3) through observations made by staff, volunteers, and potential adopters while the dog was housed at the shelter (separate from the assessment). As noted in [Section 2.3](#sec2dot3-animals-08-00027){ref-type="sec"}, the FG item was performed only during the baseline phase but the other ways of identifying FG were utilized over the entire study period. Once staff saw a dog exhibiting FG behavior, they recorded on the standardized data collection form how that behavior was identified. The researchers provided shelters with descriptions of FG to enable them to categorize the level of severity of the guarding behavior. These descriptions were labeled as mild, moderate or severe ([Appendix A](#app1-animals-08-00027){ref-type="app"}). The behaviors could be seen in various contexts including during the assessment, in their kennel, or reported from the owner. Staff documented the level of severity for each dog regardless of how the behavior was identified. 2.5. Grants {#sec2dot5-animals-08-00027} ----------- Once enrolled in the study, each shelter was issued a \$2500 grant to support the work needed to complete the study. Grant funds were used to train staff, create literature on FG for adopters and staff, start a behavior hotline, or recruit shelter volunteers. 2.6. Statistical Analysis {#sec2dot6-animals-08-00027} ------------------------- Characteristics of the groups were described using frequencies and percentages calculated with Microsoft Excel. Chi-square tests were performed to compare the baseline and investigation phase rates of bites/injuries and returns for the entire general dog population, as well as specifically for the group of FG dogs. Chi-square was also used to compare the percentage of food guarders and non-food-guarders euthanized during the four-month study period. StataSE 13.1 (StataCorp LP, College Station, TX, USA) was used for these analyses, and *p* \< 0.05 was considered to be statistically significant. 3. Results {#sec3-animals-08-00027} ========== 3.1. Population {#sec3dot1-animals-08-00027} --------------- A total of 14,180 dogs entered the nine shelters across the four-month study period. The highest intake at a participating shelter was 3247 dogs and the lowest intake was 596 dogs. During the baseline phase, 7112 dogs entered the shelters and 7068 dogs entered during the investigation phase. During the baseline phase, only 49% of the dogs were assessed upon intake. The other dogs were returned to their owners before an assessment could be completed or were considered adoptable without being assessed, due to their desirable appearance, size, or breed. Over the whole study period, 5% (778/14,180) of the dogs either exhibited FG or had a history of FG. Further details of which dogs were identified with FG during each phase can be found in [Figure 1](#animals-08-00027-f001){ref-type="fig"}. 3.2. Risk of Injury {#sec3dot2-animals-08-00027} ------------------- [Figure 1](#animals-08-00027-f001){ref-type="fig"} shows total intake of dogs between the two study phases and color groupings for those who displayed FG (green) vs non-FG (orange). The number of bites/injuries was low in both phases. FG dogs, either in the shelter or in the home, were no more likely to inflict injuries once the assessment item was stopped. A within-phase study analysis was also conducted. In the home, adopters were significantly more likely to report bites/injuries from dogs in the FG group compared to the non-FG group, in both baseline (*p* = 0.006) and investigate stages (*p* = 0.01). In the home, there were only 8 reported bites/injuries for the FG group and 35 for the non-FG group. In the shelter, there was no association of bites/injuries related to the FG or non-FG groups, in baseline (*p* = 0.5) or investigative phases (*p* = 0.6). 3.3. Outcomes for FG Dogs {#sec3dot3-animals-08-00027} ------------------------- Out of the 14,180 dogs entering the shelter, 14% (2051/14,180) were still in the shelter at the termination of the study and could not be assigned an outcome. Over the course of the study, 56% (7961/14,180) of all dogs were adopted and 18% (2576/14,180) were transferred to other organizations for adoption. Fifty-four percent of dogs (3867/7112) were adopted during the baseline phase and 58% (4094/7068) during the investigation phase. Adoption of FG dogs was 39% in the baseline (223/571) and 45% the investigation phase (93/207) (*p* = 0.14, X^2^ = 2.17, 1 df). Across both study phases, dogs that showed FG behavior stayed in the shelter longer, a mean of 13.6 days (range = 4.7--44 days), than the general dog population, a mean of 9.9 days (range = 3.6--20 days). A total of 11.2% of dogs (1592/14,180) were euthanized during the study. Significantly more FG dogs (15.4%, 120/778) were euthanized than non-FG dogs (10.9%, 1472/13,402), *p* \< 0.001, X^2^ = 0.09, 1 df. Of the 120 dogs euthanized in the FG group, 74% (*n* = 89) had behavior listed as a reason for euthanasia and 45% of those had FG listed as a reason for euthanasia. The other 55% were euthanized for other behavioral issues such as aggression toward other animals or people, high arousal, or problematic behavior in the kennel. ### Degree of FG and Impact If dogs exhibited FG in-shelter, the majority of the time it was mild in severity ([Table 1](#animals-08-00027-t001){ref-type="table"}). Another key finding is the percentage of dogs identified with severe FG was unchanged, even once the FG item was omitted. [Figure 2](#animals-08-00027-f002){ref-type="fig"} shows outcomes of adoption, euthanasia, and transfers, categorized by the severity of FG. During both baseline and investigation phases, dogs with severe FG behavior were less likely to be adopted and more likely to be euthanized than moderate or mild food guarders. 3.4. Returns {#sec3dot4-animals-08-00027} ------------ Dogs in the FG group were no more likely to be returned (*p* = 0.7, X^2^ = 0.11, 1 df) in the investigation phase (11%, 10/93) than they were during the baseline phase (9%, 21/223). There were significantly more returns of all dogs (*p* \< 0.001, X^2^ = 18.33, 1 df) during the investigation phase (13%, 526/4094) than during the baseline phase (10%, 379/3867). However, even though statistically significant, a difference of 3% in returns is too small to be meaningful to the field. 4. Discussion {#sec4-animals-08-00027} ============= This study documented the impact of omitting the FG item from standardized shelter behavior assessments. During the baseline phase, 8% of dogs were classified as food guarders. During the investigation phase, when dogs were no longer assessed for FG, only 3% were classified as food guarders. Other studies \[[@B5-animals-08-00027],[@B6-animals-08-00027]\] reported higher percentages of FG (14--21%), however they looked at the number of FG dogs based on the number assessed, not on overall intake numbers. If this study were to calculate FG dogs based on those assessed (which could only be done during baseline phase), the percentage aligns with the other studies (16.5%). However, since many shelters assess less than half their dog population, these reported percentages likely over-represent the prevalence of FG behavior in the shelter dog population. When designing this study, it was not anticipated that such a large number of dogs would not be assessed during baseline. While it is not known what each individual staff person did within each shelter (relative to the dogs not assessed), shelters did not always assess dogs who were easily adoptable (i.e., healthy, small, obviously social, and friendly dogs). It is not clear how much bias this brings and in what direction the bias would be, due to the lack of assessments among a subset of the population of dogs. One key factor used to improve validity was each shelter was its own baseline. This study was designed to be applied research in functioning shelters, constrained by shelter realities. Since this was a convenience sample, there could be a bias in those shelters that chose to participate. The shelters varied in size (number of dogs and staff), one was a municipal shelter while most were non-profit organizations, and they varied in their location around the country. We deliberately aimed for a large sample size from a variety of shelters to minimize biases in the results and improve external validity. The safety of staff, volunteers, and adopters was identified as a critical concern when shelters considered removing the FG item from their assessments. However, few bites or injuries were reported during this study, whether or not dogs were assessed for FG. It is interesting that the rate of bites and injuries was so low, even in the baseline phase, when a significant number of dogs were not assessed. This further confirms the conclusion that testing for FG does not improve safety in a shelter setting. Reported injuries in the home did not increase once the FG item was stopped, however there was a higher percentage of injuries by FG dogs compared to the non-FG dog group. However, there was a low prevalence of bites/injuries in the home for all dogs ([Figure 1](#animals-08-00027-f001){ref-type="fig"}) and the reporting included all bites and injuries, even when not involving food. Since many shelters advised adopters that these dogs might FG in the home, it is possible that new owners were more likely to report bites or injuries thus biasing the numbers. It is plausible that dogs who FG in the shelter may also have other behavior problems that make them more likely to cause injuries post-adoption, but this study did not yield that data. Another limitation of this study, which was also beyond the scope, was a lack of systematic follow-up in the home, like the protocol used in the Mohan-Gibbons study \[[@B5-animals-08-00027]\]. As a result, we could not ask for more information on the situations where the bite or injury occurred in the home. We did not ask shelters to separate actual bites from other types of injuries, nor did we ask for detail about the incidences. Although collecting that information would have provided more behavioral context, the concern was it would introduce unwanted bias because staff and adopters would need to correctly interpret every situation if FG was present or not. By tracking all incidences of bites and other injuries, our conclusions are more conservative. In this study, dogs with food guarding behavior stayed in the shelter longer than the general population and were more likely to be euthanized. An earlier survey \[[@B5-animals-08-00027]\] found that FG behavior was one of the top reasons why shelters did not place dogs into adoption. In this study, while only 15.4% of FG dogs were euthanized, over half of these were euthanized for behavior concerns unrelated to FG (i.e., medical problems and other severe behavior challenges). Most of the dogs that showed FG behavior in both the baseline and investigation phases were classified as mild FG. In the investigation phase, even though fewer dogs were identified with food guarding behavior, the percentage of dogs classified as severe remained unchanged (17% in both phases). Thus, it appears that dogs exhibiting more serious FG behavior are likely to be identified without using a standardized assessment. Relinquishing owners may report that their dog displays FG when it is more serious and staff may be more likely to observe severe guarding behavior during daily interactions in the shelter. This confirms the assertion made by Patronek & Bradley \[[@B7-animals-08-00027]\] that more serious forms of aggression are highly likely to be identified without putting the dog through a behavior assessment. The animal welfare community had expressed concerns that returns to the shelter would increase if the FG item was eliminated. Our study showed that return rates for FG dogs were not significantly different from the baseline phase, however the return rate for the general population was. Even though those returns presumably included dogs that guarded food in the home, a difference of 3% is not large enough to be meaningful in the context of shelter returns. The statistical significance is driven by the large number of dogs adopted from the nine shelters. A dog returned from the FG group may not have been returned for food guarding. The reasons given for returns were insufficient to provide any level of detail. It is challenging to gather quality data when those data are based on owners relaying details to shelter personnel who then capture that information into restrictive shelter software. Future work could consider learning more about the reasons for dogs being returned to the shelter. If shelters wish to further reduce returns, we recommend that they have conversations with all adopters about normal canine behavior, such as FG, and that any dog might exhibit this behavior in the home. Staff should encourage adopters to call the shelter if they encounter problems and need help. Other research has shown that when pet owners are given realistic expectations, they are more likely to keep their pets \[[@B8-animals-08-00027]\]. Perhaps providing adopters with basic information about FG and offering post-adoption support will prove to be a key to successful adoptions for this at-risk population. Shelters did not make large organizational changes during the course of the study. Shelter staff was aware when the FG was stopped. We encouraged shelters to talk with owners relinquishing dogs to obtain a history of FG, and some shelters changed their intake forms to reflect this. We encouraged adoption counselors to tell potential adopters that any dog might guard their food and some added this information to their adoption paperwork. Many shelters reported that staff became more diligent about recognizing and reporting FG behavior in the investigative phase because the dogs were not going to be assessed for FG. One shelter even uncovered that staff would not report observations of a dog guarding food in the kennel (during the baseline phase) because they believed "the behavior staff would catch it in the FG item during the assessment". It is crucial for shelters to establish a clear communication system for staff and volunteers to share pertinent behavior information about the dogs in their care, especially when considering omitting the FG item from their assessment. In previous research, a significant number of dogs that were observed guarding food during the assessment did not exhibit this behavior in the home \[[@B5-animals-08-00027],[@B6-animals-08-00027]\]. These "false positive" dogs are more likely to stay longer in the shelter and to be euthanized. In addition, some dogs do not guard food while in the shelter but will at home. These "false negative" dogs are often incorrectly deemed as "safe" by shelter staff and sent home with adopters who are not adequately prepared to encounter this behavior. Any dog can exhibit food guarding behavior. Our research, paired with previous research referenced in this manuscript, substantiate that the existence of food guarding behavior in the shelter does not always predict the same behavior in the home. The incidence of injury to shelter staff and adopters is low and assessing dogs for FG does not further reduce the already low injury rate. 5. Conclusions and Recommendations {#sec5-animals-08-00027} ================================== Previous research suggested that many dogs exhibiting FG in the shelter can be safely placed into homes because FG is often not exhibited in the home nor was it a problem for adopters when it was seen. In light of the current results, the authors recommend that shelters need not conduct the FG item of a standardized assessment because it results in a longer stay in the shelter, increased likelihood of euthanasia, and the incidence of false positives is likely unacceptably high. FG was not prevalent in the participating nine shelters and, of the dogs that did exhibit the behavior, most showed only mild FG behavior. Once the FG item was omitted, more of those dogs were adopted and fewer were euthanized. Severe FG behavior was found regardless of the assessment. Discontinuing the FG assessment item did not result in an increase in injuries to shelter staff or adopters, and there was no meaningful change in returns post-adoption. Our recommendation is for shelters to collect as much information about the dog from all sources (such as prior owner and staff observations) to best support each dog in finding a home. For example, if a dog shows guarding behavior over his food while in the kennel, the shelter should have procedures to ensure staff are trained so that they can stay safe while interacting with the dog. However, they should not assume the dog will show this behavior in his new home. Shelters should not ignore FG behavior. Rather, staff should be transparent with adopters about what they know from prior history and behavior in-shelter. Staff should set adopters up for success by explaining that FG is a normal behavior for dogs, it has a low occurrence, and if seen, they should contact the shelter promptly for assistance. There are techniques for managing and modifying FG behavior in the home when the owners are both concerned as well as able to implement these techniques. The authors deeply appreciate the following shelters for their participation in the study and for their time in the data collection process; Humane Society of Boulder Valley (CO), SPCA Tampa Bay (FL), Pet Alliance of Greater Orlando (FL), Southern Pines Animal Shelter (MS), SPCA of Wake County (NC), Nebraska Humane Society (NE), Erie County SPCA (NY), Animal Care Centers of New York (NY), Charleston Animal Society (SC). Processing costs for this paper were covered by the Special Issue sponsors, Maddie's Fund®, Found Animals, and The Humane Society of the United States (HSUS). Emily Weiss, Margaret R. Slater, Pamela Reid, Heather Mohan-Gibbons, and Hugh Mulligan conceived and designed the project; Margaret R. Slater, Hugh Mulligan, and Heather Mohan-Gibbons analyzed the data; Heather Mohan-Gibbons, Emily D. Dolan, Margaret Slater, Emily Weiss, and Pamela Reid wrote the manuscript. The authors declare no conflict of interest. Below are three categories with descriptions of guarding behaviors that may be observed during the food portion of an assessment. Please choose the broad category that is the best fit for each dog. Vocal behaviors, such as growling or barking, are not listed, as they can be observed in any category. ![Behavioral Categories for Food Guarding.](animals-08-00027-g0A1){#animals-08-00027-f0A1} Appendix A.1. Mild ================== Dog's head, neck and/or body become stiff/rigid but the dog does not escalate to Moderate or Severe categories. May show one or more of these behaviors: eats more quickly as food assessment progressesholds a crouched position over bowl when person is nearmoves body and/or head to block the bowl from the assessor's approach or the fake handmomentarily freezes or stops eating (without lifting head from bowl) when assessor approaches or reaches with fake hand Appendix A.2. Moderate ====================== The guarding behavior remains in or near the food bowl. Dogs may show any of the behaviors in the Mild category plus one or more of these behaviors: may lift lips/bare teethbites in the direction of the food or at the fake hand while head is in the bowl and/or while touching the dog's cheek with the fake hand or person is nearalthough the dog may bite the fake hand, the dog does not lift his head from the bowl to bite toward the assess-a-hand or pursue the hand as it is retracted Appendix A.3. Severe ==================== The guarding behavior is directed towards the assessor or handler. Dogs may show any of the Mild or Moderate behaviors, plus at least one of the following: multiple bites to the fake hand and/or bites that move up the fake handleaves the bowl to bite the fake hand (either while the hand is advancing towards the bowl or as it is withdrawn)any attempts to lunge, snap, or bite towards the person's arm or body or the handler during the food bowl itemdogs that exhibit guarding behavior either before or after the food guarding assessment (for example, the bowl is placed down and the dog immediately shows any guarding behavior before assessor has attempted their initial approach to the bowl or dogs that continue to guard the area after the food bowl is removed). ![Total intake of dogs between the two study phases: percentage of dogs who were assessed, those dogs identified with food guarding behavior, and incidence of injury/bites in each phase. Non-FG groups are in orange and food guarding (FG) groups in green, with corresponding *p*-values displayed between the baseline and investigative phases. In baseline phase, there are blue boxes and dotted lines that show the percentage of dogs that had the standardized shelter assessment and of those assessed, how many were identified with FG. This was added so the reader can visualize the large percentage of dogs that did not have a standardize assessment performed and to clarify the two ways a shelter can arrive at incidence of FG (17% is only of those assessed while 8% is all intake).](animals-08-00027-g001){#animals-08-00027-f001} ![Outcomes of the FG dogs, by severity of FG behavior, are shown for the (**a**) Baseline phase (**b**) Investigation phase. Some dogs in each severity category did not have an outcome by the end of the study period (no outcome). In the baseline, 20% of FG was identified by history and 80% by the assessment. During the investigation phase, 71% of FG was identified by history and 29% from shelter observations.](animals-08-00027-g002){#animals-08-00027-f002} animals-08-00027-t001_Table 1 ###### Severity of Food guarding (FG) behavior identified in all FG dogs comparing the baseline (8%, 571/7112) and investigation (3%, 207/7068) phases. There was an overall significant difference in the categories of severity between the baseline and investigative phase (*p* = 0.001, X^2^ = 14.86, 2 df). In the baseline, 20% of FG was identified by history and 80% by the assessment. During the investigation phase, 71% of FG was identified by history and 29% from shelter observations. Severity of FG Behavior Baseline Investigation --------------------------------- ---------- --------------- --------- -------- **Total Dogs with FG Behavior** **571** **8%** **207** **3%** Mild 391 68% 117 57% Moderate 85 15% 55 27% Severe 95 17% 35 17% **Total** 571 100% 207 100%

{ "pile_set_name": "PubMed Central" }

1. Introduction {#sec1-toxins-09-00255} =============== The genus *Fusarium* is implicated in devastating plant diseases. Of primary importance is the *sambucinum* species complex, which contains species causing Fusarium Head Blight (FHB) on a number of small-grain cereals such as wheat \[[@B1-toxins-09-00255]\]. The economic impact of this disease is exacerbated by the deposition in the cereal matrix of mycotoxins, which ultimately end up in the food and feed chain with significant health risks for humans and animals \[[@B2-toxins-09-00255]\]. Many FHB-associated species may co-occur on the host and the composition of this population is variable throughout and between growing seasons \[[@B3-toxins-09-00255]\]. *F. graminearum* is the predominant species within the FHB complex \[[@B1-toxins-09-00255]\]. In addition, *F. poae*, has been described as one of the most prominent FHB species in several studies \[[@B4-toxins-09-00255],[@B5-toxins-09-00255],[@B6-toxins-09-00255],[@B7-toxins-09-00255]\]. Its occurrence is often correlated with the incidence of other *Fusarium* species and its preponderance increased with the application of azole fungicides \[[@B4-toxins-09-00255],[@B8-toxins-09-00255],[@B9-toxins-09-00255]\]. The most important class of mycotoxins associated with FHB is the trichothecenes. They are produced by a large number of *Fusarium* species and can be divided mainly into type A (diacetoxyscirpenol (DAS), neosolaniol (NEO), and T-2 toxin, HT-2 toxin) and type B trichothecenes (deoxynivalenol (DON), nivalenol (NIV) and fusarenon-x (FUS-X), differing at the functional group occupying the carbon 8 position \[[@B10-toxins-09-00255]\]. Genes for trichothecene biosynthesis are organized in different loci and clusters. Homologs with different functionality in different *Fusarium* species determine which end products will be produced \[[@B11-toxins-09-00255]\]. The main determinant for type A vs. type B production resides in different alleles of the *Tri1* gene \[[@B12-toxins-09-00255]\]. Surprisingly however, several authors have simultaneously detected both type A and type B trichothecenes from isolates of *F. poae* \[[@B13-toxins-09-00255],[@B14-toxins-09-00255],[@B15-toxins-09-00255]\]. The specific lifestyle of fungal pathogens greatly influences their epidemiology, and their potential for adaptation and evolution. Species that combine sexual and asexual reproduction, and that harbor active transposable elements (TEs) have been theorized to possess the greatest potential for genetic variation and hence pose the highest threat to durable disease management \[[@B16-toxins-09-00255]\]. The occurrence of "cryptic" stages may complicate the identification of the precise lifestyle of fungi. This is the case when a sexual cycle is presumed to occur, but has never been witnessed, such as observed for several *Fusarium* species \[[@B17-toxins-09-00255]\]. Recombination is a clear indication that a sexual cycle is ongoing in fungi \[[@B18-toxins-09-00255]\], and genome-driven discovery of meiosis-specific markers may uncover the footprints of a sexual cycle \[[@B19-toxins-09-00255]\]. The occurrence of repeat-induced point mutation (RIP), a defense mechanism against TEs, is another line of evidence for meiosis, as it only occurs during the pre-meiotic stage \[[@B20-toxins-09-00255]\]. Subsequently, degenerated remnants of TE copies remain in the genome after RIP passage \[[@B20-toxins-09-00255]\]. The activity of RIP has recently been extensively proven for *F. poae* \[[@B19-toxins-09-00255]\]. The mating locus (MAT1) is another important genetic fingerprint for a sexual cycle. This locus can be occupied by two idiomorphs MAT1-1 and MAT1-2 and the equal distribution of both idiomorphs in fungal populations has been considered as an indication of an active sexual cycle. The occurrence of both MAT1-1 and MAT1-2 has been confirmed for *F. poae* isolates \[[@B17-toxins-09-00255],[@B21-toxins-09-00255]\]. One of the benefits of sexual reproduction is meiosis, which introduces diversity via recombination. Many studies have been carried out which investigate the population diversity of *F. poae*. Kerenyi et al. \[[@B22-toxins-09-00255]\] detected a highly complicated composition of vegetative compatibility groups (VCGs) in 50 geographically diverse isolates. Somma et al. \[[@B23-toxins-09-00255]\] found a high level of intra-species variability in a geographically narrow population using amplified fragment length polymorphism (AFLP) \[[@B24-toxins-09-00255]\], which was also found for a worldwide *F. poae* collection \[[@B21-toxins-09-00255],[@B25-toxins-09-00255]\] with AFLP and inter simple sequence repeats (ISSR). Consistently, geographic origin could only partially explain this large diversity \[[@B22-toxins-09-00255]\]. Contrary to the results in genome-wide approaches, multi-locus barcoding consistently leads to the detection of rather low intra-species variability in *F. poae* \[[@B26-toxins-09-00255],[@B27-toxins-09-00255]\]. This discrepancy may be in part explained by the specific genome composition of *F. poae*. Isolates of this species have four core chromosomes (the "core genome") and a set of supernumerary chromosomes, highly variable in size, composition, and number between individuals \[[@B19-toxins-09-00255],[@B28-toxins-09-00255]\]. It has been shown that their presence leads to significant rearrangements within the core chromosomes \[[@B19-toxins-09-00255]\]. Two major translocations from the supernumerary genome into core chromosome 3 of *F. poae* isolate 2516 were found, each exceeding several 100 kb. Finally, the supernumerary genome harbors numerous active TEs (where they are protected from RIP), which can migrate from the supernumerary to the core genome. Species that are capable of combining both sexual and asexual reproduction, and species that possess an active arsenal of TEs, are considered to be the biggest challenge for durable disease management \[[@B16-toxins-09-00255]\] and a key trait of a successful pathogens \[[@B29-toxins-09-00255]\]. We set out to investigate both features in this study. We started by assessing the genetic diversity using a genome wide approach in a representative *F. poae* collection from different geographic origins. We searched for hallmarks of sexual and asexual reproduction, and indications for how genomes evolve in a population. In addition, we investigated the genetic variability in the trichothecene gene clusters and tried to link this genetic variability with the highly variable mycotoxin profiles of *F. poae* comprising both type A and type B trichothecenes. 2. Results {#sec2-toxins-09-00255} ========== 2.1. Genetic Diversity of Fusarium poae {#sec2dot1-toxins-09-00255} --------------------------------------- For this study, we used 69 available *F. poae* isolates ([Table 1](#toxins-09-00255-t001){ref-type="table"}) and AFLP analysis was performed to assess their genetic diversity. A total of 247 markers were scored of which 201 were polymorphic. As a control *F. poae* isolate 9125 was scored independently three times in the analysis. The clustering of these three fingerprints shows the reproducibility of the AFLP method but also illustrates the variability obtained for a perfectly clonal isolate. The genotyping error rate calculated from these three repetitions was 3.3--3.8% similar to levels described in literature (2--5%) \[[@B30-toxins-09-00255],[@B31-toxins-09-00255]\]. Based on the AFLP analysis, the *F. poae* isolates are divided into sub-clusters with no clear correspondence to either host or geographic origin. This is exemplified by isolates from maize (L24, K46, S46, Q57, F49), isolates from barley (6114, 175, 182, 185, bfb0176) and isolates from Italy (designation 9\#\#\#) which are scattered throughout the tree. However, the only Chinese isolate in the study (bfb0173) is clearly distinct from all other *F. poae* isolates analyzed ([Figure 1](#toxins-09-00255-f001){ref-type="fig"}). Isolates which share \>90% genetic similarity as determined by the Dice coefficient are grouped by the same color in [Figure 1](#toxins-09-00255-f001){ref-type="fig"}. Several other genetic markers were included in [Figure 1](#toxins-09-00255-f001){ref-type="fig"}, including the translocations of supernumerary sequences to chromosome 3 \[[@B19-toxins-09-00255]\]. In isolate 2516, two sequence blocks, each several 100 kb in size, translocated from the supernumerary chromosomes to core chromosome 3. Seven isolates in the collection are found to have the first insertion, and three of these also have the second insertion. This likely occurred in a sequential fashion: starting with a \>204 kb translocation closest to the telomere of chromosome 3, followed by a 464 kb event further away from the telomere (883,738 bp). [Figure 1](#toxins-09-00255-f001){ref-type="fig"} shows that an entire sub-cluster of the AFLP tree is made up of isolates containing the first insertion of supernumerary sequence. 2.2. Fusarium poae Likely Combines Sexual and Asexual Reproduction {#sec2dot2-toxins-09-00255} ------------------------------------------------------------------ It has been shown that RIP is associated with sexual reproduction \[[@B20-toxins-09-00255]\]. Since RIP functions on the core genome of *F. poae* \[[@B19-toxins-09-00255]\], it stands to reason that meiosis should be ongoing in this species. Moreover, the presence and conservation of the "meiotic toolbox" was shown \[[@B19-toxins-09-00255]\]. We expanded these authors' investigation into the distribution of MAT1-1 and MAT1-2 in *F. poae*. These results are visualized in [Figure 1](#toxins-09-00255-f001){ref-type="fig"} and [Table 1](#toxins-09-00255-t001){ref-type="table"}. The MAT1-1 idiomorph is predominantly present (81% MAT1-1, 19% MAT1-2; one undetermined, but both idiomorphs do co-occur in the same location ([Table 1](#toxins-09-00255-t001){ref-type="table"})). The presence of detectable recombination events dictates that *F. poae* should be a sexual species, even if the sexual stage has never been observed. All *F. poae* isolates were screened for the presence of two major translocations from the supernumerary genome to chromosome 3 (see also \[[@B19-toxins-09-00255]\]). Eight isolates containing the first translocation of supernumerary sequence (closest to the telomere of chromosome 3) were collected from different locations across Flanders, Belgium ([Table 1](#toxins-09-00255-t001){ref-type="table"}). It can be assumed that this genotype has spread clonally from one location to the other two. 2.3. Transposable Element Proliferation between Near-Clonal Isolates {#sec2dot3-toxins-09-00255} -------------------------------------------------------------------- To uncover if TE insertions are shared between isolates, the location of TE insertions in isolates 2516 and 2531 were compared. According to [Figure 1](#toxins-09-00255-f001){ref-type="fig"}, these isolates are near-clonal (comparable to the AFLP pattern of the three repetitions of isolate 9125). Reads from an Illumina HiSeq run of DNAs from five different isolates, including isolate 2531 were mapped onto the reference assembly of isolate 2516 and 73 of the 135 TE copies on the core genome of 2516 did not have read support from the DNA read mix, including absence of coverage by reads originating from isolate 2531 ([Table 2](#toxins-09-00255-t002){ref-type="table"}). These TEs represent copies uniquely present in isolate 2516, and they are absent in the same location in isolate 2531. 2.4. Assessing the Genetic Chemotype of F. poae Isolates {#sec2dot4-toxins-09-00255} -------------------------------------------------------- All three thrichothecene loci were positioned on the assembly of *F. poae* isolate 2516 within their exact genomic coordinates. The *Tri1* gene is located on chromosome 1 (55,207--56,959 bp). The organization of the *Tri1* region in isolate 2516 is identical to what was described previously by Proctor et al. \[[@B32-toxins-09-00255]\]. The main trichothecene biosynthesis cluster (Tri5) is located on chromosome 2 (5,545,692--5,575,616 bp) and the *Tri101* gene can be found on chromosome 4 at (4,906,101--4,907,486 bp). These positions are similar to those for the three loci in *F. graminearum*. To study the organization of the main trichothecene cluster, we applied the diagnostic PCR developed by Dinolfo et al. \[[@B33-toxins-09-00255]\] to predict the genotype of all *F. poae* isolates towards NIV production. This assay is based on the *Tri7* gene. Using this approach, all isolates were genetically chemotyped to be of the NIV genotype. In a next step, the main trichothecene gene cluster was extracted from chromosome 2 of isolate 2516 (NCBI accession LYXU01000002.1). [Figure 2](#toxins-09-00255-f002){ref-type="fig"} depicts the organization of the main cluster and the comparison with the gene clusters in *F. sporotrichioides* (isolate NRRL 3299, T-2 toxin chemotype, NCBI accession AF359360.3) and *F. graminearum* (isolate 88-1, NIV chemotype, NCBI accession AF336365.2). The overall organization of the main trichothecene biosynthesis cluster is highly similar to that previously described in *F. sporotrichioides* and *F. graminearum*. The largest difference is the intergenic distance between *Tri5* and *Tri6* that is expanded in isolate 2516. A low complexity (low GC %) region with similarity to a *pogo* TE accounts for an additional 2.1 kb of sequence in this region. The best Blastx hit for this element is "*pogo* transposable element with KRAB domain from *F. oxysporum* f. sp. *cubense* race 1" (NCBI accession ENH68388.1). The genome sequences of isolates 2548, 7555 and bfb0173 were consulted to find differences in the organization of the main trichothecene biosynthesis cluster with isolate 2516 and these isolates. The main trichothecene biosynthesis cluster of isolates 2548, 7555 and bfb0173 can be found on contigs 77, 419 and 361 respectively. The RIPped *pogo* TE between *Tri5* and *Tri6* is also present in isolate bfb0173 and is highly conserved (three SNPs (single-nucleotide polymorphisms) between 2516 and bfb0173). It is not present in isolates 2548 and 7555. The *F. poae* collection was screened for the presence of this element ([Figure S1](#app1-toxins-09-00255){ref-type="app"}). Thirty-eight isolates contain the *pogo* element at the same location as isolates 2516 and bfb0173. [Figure 1](#toxins-09-00255-f001){ref-type="fig"} visualizes the distribution of these isolates across the AFLP tree. Downstream of *Tri14* a RIPped retrotransposon occurs in isolates 2548 and bfb0173. In isolate bfb0173, it consists of 3.7 kb of low complexity sequence, while in isolate 2548 it is only partially assembled, with 1.2 kb of sequence that is almost identical to the sequence of isolate bfb0173 (10 SNPs). The best Blastx hit for this retrotransposon is a protein from *Claviceps purpurea* (NCBI accession CCE29311.1). In isolates 2516 and bfb0173, there is a 400 bp repetitive element without Blastx hits between *Tri7* and *Tri3* that is not present in isolates 2548 and 7555. The three insertions in and around the main trichothecene biosynthesis cluster (between *Tri7* and *Tri3*, between *Tri6* and *Tri5* and bordering *Tri14*) show a specific pattern, as different combinations of two insertions are not shared between the same isolates. Finally, variation at the *Tri1* locus, on chromosome 1, is known to be responsible for chemotype diversity in *Fusarium* species \[[@B12-toxins-09-00255],[@B34-toxins-09-00255]\]. The *Tri1* genes in isolates 2516 and bfb0173 are identical to one another and to the published sequence for *FpTri1* (NCBI accession GQ915520), and 97% sequence identity to the *Tri1* gene of isolates 2548 and 7555 (respectively 51 and 52 (=51 + 1) SNPs across the 1753 bp gene). Respectively 24 and 25 of these SNPs occur in the exons of the gene, while 87% of the gene is exonic (1527 bp). Remarkably, only two (out of 24) and three (out of 25) of these SNPs are non-synonymous mutations. Two of these amino acid substitutions occur at the N-terminal end of the protein within the first 14 amino acids. [Figure S2](#app1-toxins-09-00255){ref-type="app"} shows the read mapping of HiSeq reads from isolate bfb0173 and isolate 2548 on the published *FpTri1* type, and shows that a disproportionate amount of SNPs occurs in the introns of the genes. A subset of 34 isolates of the *F. poae* collection was selected for sequencing of a 1100 bp fragment of *Tri1*. The existence of two major *Tri1* types within the collection was confirmed, with additional variation at the level of individual isolates ([Figure 3](#toxins-09-00255-f003){ref-type="fig"}). 2.5. Assessing the Phenotypic Chemotype of *F. poae* Isolates {#sec2dot5-toxins-09-00255} ------------------------------------------------------------- [Figure 4](#toxins-09-00255-f004){ref-type="fig"} shows for 61 *F. poae* isolates the trichothecenes that were detected after growth in the trichothecene biosynthesis inducing medium. The values were log transformed as the amounts of DAS were higher than those of the other toxins. DAS is also the most common mycotoxin, with 59 of 61 isolates being positive for DAS, ranging from the LOD to 22804 ng/mL. Of these 59 isolates, 55 produced NEO in the range of the LOD to 375 ng/mL. Of this set, 24 isolates produced FUS-X in the range of the LOD to 172 ng/mL and, of these FUS-X producers, 14 isolates also produced NIV in the range of 88 ng/mL to 122 ng/mL. The hierarchical nature of these chemotypes is visualized in [Figure 4](#toxins-09-00255-f004){ref-type="fig"}A. Finally, there is no significant effect of the presence/absence of the *pogo* element on the chemotype, for any of the four detected trichothecenes (Mann--Whitney U test, *p* values \> 0.05) ([Figure 1](#toxins-09-00255-f001){ref-type="fig"} and [Table 1](#toxins-09-00255-t001){ref-type="table"}). [Figure 4](#toxins-09-00255-f004){ref-type="fig"} summarizes the trichothecene production in vitro for all isolates. DAS production is about 50-fold higher compared to the production of NEO, FUS-X or NIV. Note that two isolates show outlier values for DAS and NEO production (isolate 30702 and isolate 9194). Concordantly, [Figure 4](#toxins-09-00255-f004){ref-type="fig"}B shows that production of DAS and NEO is correlated. [Table S1](#app1-toxins-09-00255){ref-type="app"} shows the result of two biological repetitions of the chemotyping experiment for 28 isolates. Compounds with a plus sign were detected consistently, while compounds indicated with "X" were only detected in one of the two repetitions. For 20 of the isolates, the chemotype previously shown in [Table 1](#toxins-09-00255-t001){ref-type="table"}, moved "further downstream", with FUS-X or even NIV production where there was none in another iteration. This hierarchical nature of production was never violated in any of the strains in either of the repetitions. The chemotype of seven isolates was identical in the different repetitions: for example isolate 7555 only produced DAS, and isolate 2569 produced all four compounds. 3. Discussion {#sec3-toxins-09-00255} ============= In this paper, we assessed genetic variation, sexual and asexual reproduction, the trichothecene clusters and the chemotype of a set of *F. poae* isolates. AFLP analysis on the *F. poae* isolates to assess genetic variation among isolates from geographic different areas was performed. The isolates show \>75% similarity, suggesting that they are part of a single monophyletic lineage, which is similar to what was found before for *F. poae* \[[@B35-toxins-09-00255],[@B36-toxins-09-00255]\]. No obvious connection was found between geography and the AFLP sub-clustering, which is in accordance with previous studies \[[@B21-toxins-09-00255],[@B25-toxins-09-00255]\]. Still, isolate bfb0173 from China which is the most distant, clustered separately from all other isolates and the dominant genotype among the Italian isolates did not occur in Belgium. In accordance, Kerenyi, Taborhegyi, Pomazi and Hornok \[[@B22-toxins-09-00255]\] reported a link between geographic origin and VCG/RAPD profile (Random amplification of polymorphic DNA). We subsequently researched whether there was a link between the position of isolates in the AFLP tree and the host plant. Five isolates from maize were included in the study, aside from isolates from wheat and barley. The maize isolates are scattered throughout the tree, suggesting that no significant host specialization has occurred within *F. poae*. This is in agreement with literature reports that suggest that most isolates have a broad host range \[[@B22-toxins-09-00255]\]. However, a host preference cannot be excluded. As species capable of combining both sexual and asexual reproduction, and species with an active arsenal of TEs, are considered to be the biggest challenges for durable disease management \[[@B16-toxins-09-00255]\] and key trait of successful pathogens \[[@B29-toxins-09-00255]\], we investigated both features in the *F. poae* isolates. A sexual cycle has never been shown for *F. poae*, but is highly likely to occur. However, the skewed mating type distribution encountered among the Belgian *F. poae* isolates indicates that meiosis does not occur at high frequency \[[@B19-toxins-09-00255]\]. Although in oomycetes, such as *Phytophtora infestans*, an unbalanced distribution of the two mating types has been related to fitness costs, in *F. poae* there is no evidence for a similar link. Opposed to what was described recently by Dinolfo, Castanares and Stenglein \[[@B21-toxins-09-00255]\], we did not find any evidence for homothallism in *F. poae*. To confirm the presence of a sexual cycle, RIP was studied in isolate 2516. This analysis showed that meiosis and RIP likely did not happen very recently, allowing the accumulation of intact TE on the core genome during asexual propagation \[[@B19-toxins-09-00255]\]. The tight correlation between RIP and meiosis should offer opportunities for understanding the balance between clonal and sexual reproduction, by using TE insertions as markers for "track and trace" studies and monitoring when/if they become RIPped. We showed that near-clonal isolates can differ heavily in TE proliferation, which may occur over very short evolutionary periods. It seems likely that such TE mobilization is obscuring the nearly clonal nature of some isolates. The occurrence of TE integrations during asexual reproduction, such as in in vitro cultures, may also account for the instability of single spore cultures reported previously for *F. poae* \[[@B22-toxins-09-00255]\]. Moreover, novel genotypes that are generated through the presence of the supernumerary genome show large genetic divergence from other *F. poae* isolates when analyzed by a genome-wide method such as AFLP. These findings may explain the discrepancy in intra-species variability between genome-wide and multi-locus barcoding approaches that has been witnessed for *F. poae*. Genome-wide studies such as those employing RAPD \[[@B22-toxins-09-00255]\], AFLP \[[@B21-toxins-09-00255],[@B23-toxins-09-00255]\] and ISSR \[[@B21-toxins-09-00255],[@B25-toxins-09-00255]\] consistently detect medium to high genetic diversity, which is in sharp contrast to multi-locus barcoding analyses \[[@B26-toxins-09-00255],[@B27-toxins-09-00255]\]. The genome dynamics and the active TEs arsenal that are present in *F. poae*, may accelerate its evolution and lead to elevated levels of genetic diversity, without interfering with barcode diversity. In *Fusarium* species within the *F. sambucinum* clade, trichothecene production is orchestrated by three loci comprising the Tri 5 gene cluster, the Tri1-Tri16 cluster and the Tri 101 gene. Within the trichothecene type B producing FGSC (*Fusarium graminearum* species complex), it is common to define the chemotype of an isolate by its genotype, determined by a diagnostic PCR of a differential trichothecene biosynthesis gene \[[@B37-toxins-09-00255]\]. A similar method has been developed for the NIV chemotype in *F. poae* using the *Tri7* gene \[[@B33-toxins-09-00255]\]. All isolates in the study of Dinolfo et al. \[[@B21-toxins-09-00255]\], all isolates in the study by Covarelli et al. \[[@B7-toxins-09-00255]\], and all isolates in our study were genetically determined to have the NIV chemotype. Another important locus determining trichothecene production is *Tri1*. Variations in *Tri1* sequence are responsible for recent shifts in trichothecene production within *F. graminearum* this species \[[@B34-toxins-09-00255]\]. Remarkably in *F. poae*, different haplotypes of the *FpTri1* gene can also be found although most SNPs are synonymous mutations. The existence of two major *Tri1* types within the collection was confirmed, with additional variation at the level of individual isolates. As both types code for highly similar enzymes, this variability cannot explain the variable phenotypic chemotype. Finally, we studied the *Tri5* gene cluster in *F. poae*. In *F. poae* isolate 2516, this main trichothecene cluster and neighboring genes are similar in organization to the trichothecene clusters in *F. graminearum* and *F. sporotrichioides* \[[@B38-toxins-09-00255]\]. However, uniquely for a *Fusarium* species, but in line with the unique genome dynamics of *F. poae*, the main trichothecene biosynthesis cluster and its environment are invaded by TEs. The insertion of repetitive sequences into the genome is a useful tool for track and trace studies in fungal populations, as every such insertion can be considered a unique event. In fact, all sequenced genome contain a unique pattern of TE insertions in the trichothecene biosynthetic gene cluster. Using the three insertions across the trichothecene biosynthesis cluster, a recombination event was uncovered illustrating the presence of a sexual cycle in *F. poae*. In *F. graminearum*, it was shown that the main trichothecene cluster, with similar genomic coordinates as in *F. poae*, is situated in a region of the genome with an elevated recombination rate \[[@B39-toxins-09-00255]\]. Ward et al. \[[@B40-toxins-09-00255]\] have shown that the composition of the main trichothecene cluster in type B producing species is the result of multiple recombination events. O'Donnell et al. \[[@B41-toxins-09-00255]\] have shown that the main trichothecene biosynthesis cluster in *F. aethiopicum* (a member of the FGSC) is the result of inter-chemotype recombination. That cluster is the result of recombination between different haplotypes in these species independently, but it remains to be seen what the functional consequences for trichothecene production have been. The localization of a RIPped *pogo* element between *Tri6* and *Tri5* in isolates 2516 and bfb0173 is remarkable. No intact copies of this element could be found in the core genome, and it was likely RIPped to extinction in isolate 2516 \[[@B19-toxins-09-00255]\]. In the *F. equiseti* complex, this space is occupied by an additional Zn2Cys6 transcription factor \[[@B32-toxins-09-00255]\]. There is a strong link between transposable elements and transcription factors (TF). Many TF-associated DNA binding sites are derived from ancient transposable elements and TEs have played a defining role in the formation of complex regulatory networks \[[@B42-toxins-09-00255]\]. Moreover, some TF families are derived from ancient TE domestication \[[@B43-toxins-09-00255]\]. There is no similarity between the *pogo* element in *F. poae* and the Zn2Cys6 TF in *F. equiseti* (data not shown), so an evolutionary link between the two seems unlikely. Nevertheless, insertions into the *Tri6-Tri5* intergenic region have been shown to influence trichothecene production \[[@B44-toxins-09-00255]\]. To investigate whether the variability in trichothecene genes influenced the production of trichothecenes (i.e., the phenotypic chemotype), we monitored the production of both type A (DAS and NEO) and type B (FUS-X and NIV) trichothecenes. Remarkably, there was no correlation between any of the variations at the level of the *Tri* genes (Tri1, *pogo* element between *Tri5* and *Tri6*) and the production of type A and Type B trichothecenes. Other researchers addressed this phenotypic chemotype of *F. poae*. Pasquali et al. \[[@B45-toxins-09-00255]\] used agmatine rather than L-arginine as a nitrogen source in their trichothecene biosynthesis inducing medium and detected FUS-X and NIV production from four *F. poae* isolates, with large inter-isolate differences. Previous studies have utilized autoclaved cereal substrates \[[@B15-toxins-09-00255],[@B23-toxins-09-00255]\] and in planta inoculations \[[@B46-toxins-09-00255]\]. The discrepancies in chemotype depending on the matrix or medium (and nitrogen source) used, may lead to an underestimation of certain toxicological aspects of *F. poae*. The Italian isolates in our study were previously chemotyped by growth on autoclaved wheat. Three that were at that time found to produce no type A trichothecenes, produced both DAS and NEO in the mineral medium. This is an important point to make, as, at least in certain assays, type A trichothecenes are more toxic than type B trichothecenes \[[@B47-toxins-09-00255],[@B48-toxins-09-00255]\]. Despite all the variability in trichothecene production between isolates, the chemotype was shown to be strictly hierarchical in all isolates, implying that when NIV is detected in *F. poae*, Type A trichothecenes are also synthesized. Inoculations in cereal matrices seemingly shift the balance more towards type B production (NIV), while the in vitro mineral medium with [l]{.smallcaps}-arginine mainly stimulated DAS production \[[@B46-toxins-09-00255]\]. However, trichothecenes of the DAS-related group such as mono-acetyl scirpenol and scirpenol are often overlooked in mycotoxin surveys, while they may be important compounds in nature, particularly associated with *F. poae* \[[@B49-toxins-09-00255],[@B50-toxins-09-00255]\]. Finally, although *F. poae* is often reported as T-2 toxin and HT-2 toxin producing species, our data on the TRI genes detected in *F. poae* confirmed previous reports on the lack of the key gene related to the production of both mycotoxins in the trichothecene cluster \[[@B32-toxins-09-00255]\]. This achievement is extremely important for a correct risk assessment of the *F. poae* occurrence on agro-food important crops, since T-2 toxin and HT-2 toxin are the most acute poisoning mycotoxins produced by Fusarium species and both are under evaluation at European level for their possible limits in food products. 4. Conclusions {#sec4-toxins-09-00255} ============== Using a combined approach of AFLP and MAT locus analysis, we identified hallmarks of both sexual recombination and clonal spread of isolates in the population. Based on Tri7 sequence analysis, all *F. poae* isolates had the NIV chemotype, although not all isolates produced NIV in vivo. *Tri* gene clusters showed two layers of genetic variability. Firstly, the Tri1 locus was highly variable with mostly synonymous mutations and mutations in introns. Secondly, in a subset of isolates, the main trichothecene gene cluster was invaded by a transposable element between *Tri5* and *Tri6*. Nevertheless, these sources of variability could not explain variations in the phenotypic chemotype. 5. Materials and Methods {#sec5-toxins-09-00255} ======================== 5.1. *Fusarium* Collection {#sec5dot1-toxins-09-00255} -------------------------- For the purpose of this study, a broad collection of *Fusarium* isolates was gathered ([Table 1](#toxins-09-00255-t001){ref-type="table"}). Of the 69 *F. poae* isolates, forty-one isolates were collected from fields in Flanders, Belgium. Ten Italian isolates were previously described \[[@B23-toxins-09-00255]\]. Four isolates (Norwegian isolates) were donated by Dr. Anne van Diepeningen. One isolate (PD93/1780) originated from the *Fusarium* collection at Wageningen university (WUR). Twelve isolates came from the MUCL culture collection (MUCL, Louvain-La-Neuve, Belgium) and originated from maize and wheat. One Chinese isolate was included in the study (bfb0173). Eight *F. graminearum* isolates and five *F. culmorum* isolates were obtained from Belgian fields as described above and the *F. graminearum* 8/1 isolate was kindly provided by Dr. Karl-Heinz Kogel. The *F. sporotrichioides* isolate (MUCL6133) and *F. langsethiae* isolate (MUCL34988) were purchased from the MUCL collection. One *F. langsethiae* isolate (CBS11324) was purchased from the CBS collection (Utrecht, The Netherlands). The other *F. langsethiae* isolates were kindly donated by Dr. Simon Edwards (Harper Adams College, UK) and Dr. Ingerd Hofgaard (Bioforsk, Norway) respectively. For a detailed overview of these isolates, we refer to [Table S3](#app1-toxins-09-00255){ref-type="app"}. Field samples, presumed to consist of a mix of species and genotypes, were obtained and confirmed to contain *F. poae* as described in Audenaert, Van Broeck, Bekaert, De Witte, Heremans, Messens, Hofte and Haesaert \[[@B5-toxins-09-00255]\]. These were purified to a single spore level according to a method by Dr. Susanne Vogelgsang (Agroscope, Zürich, Switzerland, personal communication). Mycelium plugs were placed on a PDA (potato dextrose agar) plate for 7 days under a light regime of UV/darkness for sporulation (12 h (365 nm 10 W)/12 h). Sporulated mycelium was resuspended in 9 mL of distilled water. After vortexing and diluting, the resulting conidial dilutions were poured on plates containing water agar. Conidia were allowed to settle briefly after which the suspension was poured off. Plates were incubated for 16 h in the dark in a slanted position. Finally, typical *F. poae* spores were isolated using a Pasteur pipette. DNA extraction of isolates was performed as described by Audenaert et al. \[[@B5-toxins-09-00255]\]. Isolates confirmed to be *F. poae* with primers Fp82F/R ([Table S2](#app1-toxins-09-00255){ref-type="app"}) \[[@B51-toxins-09-00255]\] were preserved at −80 °C as spores in a 20% glycerol solution. 5.2. Phylogenetic Analyses {#sec5dot2-toxins-09-00255} -------------------------- DNA for AFLP analyses was extracted with the Invisorb Spin Plant MiniKit (Invitek, Berlin, Germany) according to the manufacturer's instructions. DNAs were quantified with the Nanodrop 1000 system (ThermoFisher Scientific, Merelbeke, Belgium) and AFLP analysis was performed as described by Vos et al. \[[@B24-toxins-09-00255]\]. For selective amplification, four primer pairs, EcoRI-AC/MseI-CC, EcoRI-AC/MseI-CA, EcoRI-AC/MseI-CG and EcoRI-GG/MseI-CA were used. Fragments were analyzed on a CEQ 2000 Genetic Analysis System (Beckman Coulter, Fullerton, California) and visualized with the Genographer 1.6.0 software (Benham, Montana State University, Bozeman, Montana). Bands were scored visually and data was processed as described in Scauflaire et al. \[[@B52-toxins-09-00255]\]. From the visual scoring, a binary matrix was constructed and UPGMA cluster analysis was performed based on the Dice similarity coefficients between the isolates. To simulate the AFLP pattern of a clonal isolate, and to determine the genotyping error rate, three DNA subsamples from one isolate were included in the experiment. 5.3. *Fusarium poae* Genome Data {#sec5dot3-toxins-09-00255} -------------------------------- The sequence of the trichothecene biosynthesis loci in isolates bfb0173, 2548 and 7555 was extracted from their assemblies and compared with the genome sequence of isolate 2516 (NCBI accession LYXU00000000.1 \[[@B19-toxins-09-00255]\]). The annotation of isolate 2516 was checked to ascertain that genes were correctly annotated, in accordance with literature on related *Fusarium* species. Similarities between amino acid sequence of the predicted proteins, and between nucleotide sequence of the intergenic regions, were determined with ClustalW2 \[[@B53-toxins-09-00255]\]. To visualize differences between the different Tri1 types, the HiSeq reads of isolates bfb0173 and 2548 were mapped to the only *Tri1* type already published (NCBI accession GQ915520) with CLC Genomics Workbench 7.5. Reads were mapped at 0.5 length fraction and 0.8 similarity fraction. DNA from five *F. poae* isolates (6127, 6114, 42824, 30702 and 2531) was pooled in a 1:1:1:1:1 ratio. This DNA mix was sequenced with Illumina technology as outlined by Vanheule et al. (2016) \[[@B19-toxins-09-00255]\]. The resulting HiSeq reads cannot be separated and are therefore a hybrid pool of five isolates. Investigation into identical integration of TEs was performed as outlined by Vanheule et al. \[[@B19-toxins-09-00255]\]. 5.4. Trichothecene Production Analyses {#sec5dot4-toxins-09-00255} -------------------------------------- A trichothecene biosynthesis inducing medium described by Gardiner et al. \[[@B54-toxins-09-00255]\] was used. The general composition of the medium was identical as described in their study, but Phytagel was excluded and NaNO~3~ was exchanged with [l]{.smallcaps}-arginine (Duchefa Biochemie, Haarlem, The Netherlands) at 5 mM as Gardiner et al. \[[@B54-toxins-09-00255]\] showed that this nitrogen source induces trichothecene biosynthesis in *F. graminearum*. Medium was prepared in double concentration. Conidia were harvested by adding distilled water amended with 0.01% Tween80 (Merck, Darmstadt, Germany) to the fully grown PDA plates and by rubbing the mycelium with a spatula. Conidia were counted with a Bürker counting chamber and diluted to a final concentration of 2 × 10^6^ conidia/mL. Finally, 0.5 mL of double concentrated medium and 0.5 mL of double concentrated conidial suspension were added together in 24 well plates (leading to 1× concentration of medium and 1× concentration of conidia). The 24 well plates were closed with a lid and the fungus grew stationary under 16 h light/8 h dark regime at 22 °C. The medium and fungal mass was removed from the wells. After seven days, supernatant was extracted from the cultures by centrifugation at 2750 g for 10 min and immediately processed for LC-MS/MS analysis. Sample preparation and LC-MS/MS followed an "evap and shoot" principle, without extensive sample cleanup. Individual mycotoxin solid standards (1 mg) of DON, NIV, NEO, FUS-X, T-2 toxin, HT-2 toxin, DAS and deepoxydeoxynivalenol (DOM) were supplied by Coring System Diagnostics (Gernsheim, Germany) as certified solutions. All mycotoxin solid standards were dissolved in acetonitrile (1 mg/mL), and were storable for a minimum of 1 year at −18 °C \[[@B55-toxins-09-00255]\]. Working solutions of 10 ng/μL for DON, NIV, NEO, FUS-X, T-2 toxin, HT-2 toxin, DAS and DOM were prepared in methanol and stored at −18 °C. From the individual working solutions, a mixture was prepared in methanol, stored at −18 °C and renewed monthly with the following concentrations: the mycotoxin mix (mycotoxins, 10 ng/μL) and the internal standard (DOM, 10 ng/μL). An aliquot of the trichothecene biosynthesis inducing medium was transferred in a 10 mL glass tube and the internal standard was added (DOM, 10 μL, final concentration 100 ng/mL). To construct a calibration curve, five blank aliquots were spiked with the mycotoxin mix in an increasing concentration range (2 μL (20 ng/mL), 5 μL (50 ng/mL), 10 μL (100 ng/mL), 15 μL (150 ng/mL) and 20 μL (200 ng/mL), respectively). Samples were vortexed for 2 min (Labinco, Breda, The Netherlands) and spiked medium was evaporated to dryness under a gentle N~2~-stream at 60 °C using the Turbovap^®^ LV Evaporator (Biotage, Uppsala, Sweden). The dried residue was redissolved in 100 μL injection solvent, which consisted of 70% mobile phase A (water/methanol (95/5, *v*/*v*) + 5 mM ammonium acetate, 0.1% glacial acetic acid) and 30% mobile phase B (water/methanol (95/5, *v*/*v*) + 5 mM ammonium acetate, 0.1% glacial acetic acid). Prior to injection, the sample was vigorously vortexed for 2 min, collected in a 0.22 μm Ultrafree-MC centrifugal device (Millipore, Bedford, MA, USA) and centrifuged for 10 min at 10,000 *g*. LC-MS/MS methodology was as detailed in Vanheule, Audenaert, De Boevre, Landschoot, Bekaert, Munaut, Eeckhout, Hofte, De Saeger and Haesaert \[[@B8-toxins-09-00255]\]. LODs and LOQs for the trichothecenes produced by *F. poae* were, respectively, 30 and 61 ng/mL (NIV), 31 and 63 ng/mL (FUS-X), 39 and 78 ng/mL (NEO) and 41 and 82 ng/mL (DAS). Certain samples showed peaks which according to the calibration curve corresponded to concentrations between the LOD and the LOQ for DAS, NEO and FUS-X. These samples were assigned the LOD concentration as they could not be reliable quantified. Samples with concentrations out of the range of the calibration curve were diluted and reanalyzed. To circumvent the possible matrix effect issue, a new calibration curve was constructed with a blank sample. This blank sample was also diluted, using the same protocol as the unknown diluted samples. Internal standards and spiked concentrations were then added anew. Data acquisition and processing was performed with MassLynx 4.1 and QuanLynx 4.1 software respectively (Micromass, Manchester, UK). Pearson correlations between the levels of trichothecenes were determined with Microsoft Excel 2013. The concentration range of the trichothecenes was visualized in Box-Whiskers plots, constructed with SPSS22 (IBM, Armonk, NY, USA). The effect of the *pogo* insertion on trichothecene production was analyzed with the Mann--Whitney U test in SPSS. 5.5. Diagnostic PCRs and Amplicon Sequencing {#sec5dot5-toxins-09-00255} -------------------------------------------- All primers are listed in [Table S2](#app1-toxins-09-00255){ref-type="app"}. Mating type was determined as described by Kerenyi et al. (2004). For the two insertions of supernumerary sequence into the core chromosomes (INS1 and INS2), primers were used as described in Vanheule et al. (2016). A 1100 bp fragment of the *Tri1* gene was amplified as described in Proctor et al. (2009) for a set of 34 *F. poae* isolates selected from different backgrounds and chemotypes. The presence of a *pogo* TE in the Tri cluster was investigated with three primers as detailed in [Table S2](#app1-toxins-09-00255){ref-type="app"}. PCR reactions were done with Promega GoTaq G2 Polymerase according to the manufacturer's instructions (Promega, Leiden, The Netherlands). Annealing temperatures were chosen, based on Promega's Biomath Calculator. PCRs were performed in an Applied Biosystems GeneAmp PCR System 9700, amplicons were separated on 1.5% (*w*/*v*) agarose gels stained with 0.1 μg/mL ethidium bromide, and visualized with a Biorad Gel Doc XR+. PCR products were purified with the EZNA Cycle-Pure kit (VWR Chemicals, Haasrode, Belgium) and sequenced in both directions by Macrogen Inc. (Amsterdam, The Netherlands). Forward and reverse sequences were curated with CLC Main Workbench 7. Maximum parsimony analysis was performed with PAUP version 4.0b10 \[[@B56-toxins-09-00255]\]. The *Tri1* fragment from *Fusarium sp.* NRRL 36351 (NCBI accession GQ915523.1) was used as the outgroup, however for construction of the phylogenetic tree mid-point rooting was used. Bootstrapping was performed with PHYLIP based on 1000 replicates \[[@B57-toxins-09-00255]\]. We would like to acknowledge the financial support received for this study from MYCOKEY, a project funded within the Horizon2020 Research and Innovation program of the European Commission (project number 678781). The following are available online at [www.mdpi.com/2072-6651/9/9/255/s1](www.mdpi.com/2072-6651/9/9/255/s1), Figure S1: Diagnostic PCR of the pogo insertion between Tri5 and Tri6, Figure S2: Read mapping of reads from isolate bfb0173 (A and C) and isolate 2548 (B and D) on NCBI accession GQ915520, which is the only sequence of *FpTri1* available \[[@B32-toxins-09-00255]\], Table S1: Chemotypes of 28 *F. poae* isolates in two biological repeats, Table S2: List of all primers used in this study, Table S3: Isolates from additional Fusarium species collected for this study. ###### Click here for additional data file. K.A., C.W. and T.v.d.L. conceived and designed the experiments; A.V., M.D.B., A.M., J.S., F.M., B.B. performed the experiments; A.V., T.v.d.L. and M.D.B. analyzed the data; J.S. and F.M. contributed reagents/materials/analysis tools; A.V., S.D.S., G.H., C.W., T.v.d.L. and K.A. wrote the paper. The authors declare no conflict of interest. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results. ![AFLP tree of *F. poae* isolates, and superposition of genetic markers. UPGMA (Unweighted Pair Group Method with Arithmetic Mean) cluster analysis was performed based on the Dice similarity coefficients between the isolates. In total, 247 markers were scored of which 201 were polymorphic. Three technical replicates of isolate 9125 are also included in the graph (blue bar). Clusters of isolates that share more than 90% genetic similarity according to the Dice similarity coefficient are grouped with the same colors. Bootstrap values exceeding 70% are given at the nodes, based on 1000 replications. Four genetic markers that were determined throughout this study are listed (*MAT1*, *pogo*, INS1, and INS2). White entries are markers for which no data could be collected. *MAT1*: yellow is MAT1-1, green is MAT1-2. *Pogo*: green indicates presence of the *pogo* element between *Tri5* and *Tri6* in the major gene cluster, red indicates absence. INS1 and INS2 are the major insertions of supernumerary sequence into chromosome 3 of a subset of isolates. Yellow indicates the absence of the insertion, green indicates the presence of the insertion. The isolate for which reference genomes is available is shown with asterisks.](toxins-09-00255-g001){#toxins-09-00255-f001} ![Organization of the main trichothecene biosynthesis cluster in *F. poae* (isolate 2516; NCBI accession LYXU01000002.1), which is identical to the organization in *F. graminearum* (isolate 88-1, NIV chemotype, NCBI accession AF336365) and *F. sporotrichioides* (isolate NRRL 3299, NCBI accession AF359360). The difference in Tri6-Tri5 intergenic region between *F. poae* and *F. graminearum/F. sporotrichioides* is explained by the presence of a RIPped pogo transposable element (yellow). This element is also present at the same location in isolate bfb0173. In isolates 2516 and bfb0173, there is a 400 bp repetitive element between Tri7 and Tri3 (blue). Finally, in isolates 2548 and bfb0173, there is a RIPped retrotransposon downstream of Tri14 (shaded box).](toxins-09-00255-g002){#toxins-09-00255-f002} ![Maximum parsimony phylogenetic tree that was built from the 1100 bp fragments of the Tri1 gene with PAUP \* 4.0b10. The tree is based on 74 parsimony informative characters. Bootstrap values were calculated with the PHYLIP package and those exceeding 70% are shown on the tree. *Fusarium* sp. NRRL 36351 (NCBI accession GQ915523.1) was used as outgroup, however, for construction of the phylogenetic tree, mid-point rooting was used.](toxins-09-00255-g003){#toxins-09-00255-f003} ![Summary of trichothecene biosynthesis in *F. poae* isolates. (**A**), simplified representation of the chemotype of DAS producing *F. poae* isolates. The chemotypes detected were strictly hierarchical, i.e., when compounds lower on the scheme are detected, the compounds higher on the scheme have been produced as well. (**B**), Pearson correlations between the four trichothecenes that are produced by *F. poae* isolates. (**C**), overview as Box-Whisker plots of the concentration ranges of the four trichothecenes. Note that the range of DAS is two orders of magnitude higher than the range for NEO, FUS-X and NIV concentrations. The boxes for every mycotoxin show the lower and upper quartile (delineating the boxes) of the measured concentrations (ng/mL) and the median (thick line within the boxes). The whiskers represent the minimum and maximum values. Circles and asterisks are outliers and extreme values which fall respectively outside of one-and-a-half additional box lengths and three additional box lengths counted from the upper quartile limit.](toxins-09-00255-g004){#toxins-09-00255-f004} toxins-09-00255-t001_Table 1 ###### List of *F. poae* isolates (*n* = 69) used in this study. Location, host, year of isolation and mating type are shown. The ID is the identification number in our local collection. Concentrations of mycotoxins are expressed as ng/mL. ND: not detectable (\<LOD). \>LOD: amount above the LOD but below the LOQ. NCA: no chemotype available (not determined). DAS: diacetoxyscirpenol; NEO: Neosolaniol; FUS-X: Fusarenon-X; NIV: Nivalenol. Absence (-) or presence (+) of two supernumerary sequence insertions in chromosome 3 (INS1: \>204 kb and INS2: 464 kb) are indicated in the table. Availability of an AFLP profile is indicated by "+" or "-". -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ID Location Host Year Reference Mating Type INS1 INS2 AFLP DAS NEO FUS-X NIV ----------- ----------------------------- ---------------- ------ ------------------------------ ------------------- ---------------------------- ---------------------------- ---------------------- ----------------------------------------------------------------- ------ ------- ----- 175 Aas, Norway barley 1996 Sundheim L. unpubl. MAT1-2 \- \- \- 355 55.1 ND ND 177 Norway wheat 1996 Sundheim L. unpubl. MAT1-1 \- \- \+ NCA 182 Norway barley 1996 Sundheim L. unpubl. MAT1-1 \- \- \+ NCA 185 Norway barley 1996 Sundheim L. unpubl. MAT1-1 \- \- \+ ND ND ND ND 1879 Bottelare, Belgium wheat 2010 this study MAT1-1 \- \- \+ 2160 66 25 88 2004 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 89 ND ND ND 2019 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 231 54 ND ND 2022 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 1588 62 18 ND 2023 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 44 12 ND ND 2028 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 3650 79 77 93 2031 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 109 ND ND ND 2033 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 1306 59 ND ND 2041 Zwevegem, Belgium wheat 2010 this study MAT1-2 \- \- \+ 60 12 \>LOD 108 2043 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ 3971 72 ND ND 2044 Zwevegem, Belgium wheat 2010 this study MAT1-2 \- \- \+ 144 9 ND ND 2056 Zwevegem, Belgium wheat 2010 this study MAT1-1 \- \- \+ NCA 2371 Bottelare, Belgium wheat 2011 this study MAT1-2 \- \- \+ 6923 101 ND ND 2375 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 15 10 ND ND 2377 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 31 11 \>LOD 103 2380 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 1743 64 ND ND 2381 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 222 54 ND ND 2390 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 3027 67 ND ND 2392 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 196 14 ND ND 2395 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 193 55 ND ND 2410 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 9562 84 ND ND 2411 Bottelare, Belgium wheat 2011 this study MAT1-1 \- \- \+ 238 54 ND ND 2424 Koksijde, Belgium wheat 2011 this study MAT1-1 \- \- \+ 2098 59 ND ND 2476 Poperinge, Belgium wheat 2011 this study MAT1-1 \- \- \+ 488 75 172 103 2491 Poperinge, Belgium wheat 2011 this study MAT1-2 \- \- \+ 12,336 99 ND ND 2514 Zwevegem, Belgium wheat 2011 this study MAT1-1 \- \- \+ 2800 73 83 116 2516 Zwevegem, Belgium wheat 2011 this study MAT1-1 \+ \+ \+ 1126 57 ND ND 2517 Zwevegem, Belgium wheat 2011 this study MAT1-1 \- \- \+ 641 58 ND ND 2519 Zwevegem, Belgium wheat 2011 this study MAT1-2 \- \- \+ 9766 139 ND ND 2521 Zwevegem, Belgium wheat 2011 this study MAT1-1 \+ \- \- ND ND ND ND 2524 Zwevegem, Belgium wheat 2011 this study MAT1-1 \- \- \+ 380 17 ND ND 2525 Zwevegem, Belgium wheat 2011 this study MAT1-1 \- \- \+ 2195 61 ND ND 2531 Zwevegem, Belgium wheat 2011 this study MAT1-1 \+ \+ \+ 803 59 22 92 2532 Zwevegem, Belgium wheat 2011 this study MAT1-1 \- \- \+ 1940 66 25 ND 2547 Zwevegem, Belgium wheat 2011 this study no amplicon \- \- \+ NCA 2548 Zwevegem, Belgium wheat 2011 this study MAT1-1 \- \- \+ 67 ND ND ND 2565 Zuienkerke, Belgium wheat 2011 this study MAT1-1 \+ \- \+ 69 18 \>LOD 115 2569 Zuienkerke, Belgium wheat 2011 this study MAT1-1 \+ \- \+ 4910 44 \>LOD 122 2570 Zuienkerke, Belgium wheat 2011 this study MAT1-1 \+ \- \+ 6630 66 25 ND 2571 Zuienkerke, Belgium wheat 2011 this study MAT1-1 \+ \- \+ NCA 2671 Linter, Belgium wheat 2011 this study MAT1-1 \+ \+ \+ 2457.8 75 37 100 6127 Wageningen, The Netherlands wheat 1964 MUCL MAT1-2 \- \- \- 10,864 303 75 ND 6114 Denmark barley 1964 MUCL, Hennebert G.L. unpubl. MAT1-2 \- \- \+ 135 55 ND ND 7555 Heverlee, Belgium wheat 1965 MUCL MAT1-1 \- \- \- 1067.8 ND ND ND 9125 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-1 \- \- \+ 5096 73 ND ND 9139 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-2 \- \- \+ 8 ND ND ND 9181 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-2 \- \- \+ NCA 9186 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-1 \- \- \+ 8024 77 24 ND 9189 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-1 \- \- \+ 6342 98 ND ND 9192 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-1 \- \- \+ 10,415 132 46 91 9194 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-1 \- \- \+ 13,404 270 81 ND 9196 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-2 \- \- \+ 3610 115 34 ND 9203 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-1 \- \- \+ 115 55 ND ND 9209 Ferrara, Italy wheat 2005 Somma et al. (2010) MAT1-1 \- \- \+ 5502 115 28 ND 11456 Heverlee, Belgium barley 1968 MUCL, Meyer J.A. unpubl. MAT1-2 \- \- \+ 264 20 \>LOD 103 30702 unknown in vitro plant 1990 MUCL, Marchand D. unpubl. MAT1-2 \- \- \+ 22,804 375 57 ND 15926 Quebec, Canada wheat 1970 MUCL, Hennebert G.L. unpubl. MAT1-1 \- \- \+ NCA 42824 Belgium wheat 2000 MUCL MAT1-1 \- \- \- 2120 73 ND ND bfb0173 China barley 2005 Yang et al. (2008) MAT1-1 \- \- \+ NCA F49 Ath, Belgium maize 2007 MUCL (Scaufflaire J.) MAT1-1 \- \- \+ 5720 76 ND ND K46 Ath, Belgium maize 2007 MUCL (Scaufflaire J.) MAT1-1 \- \- \+ 5339 100 ND ND L24 Buissenal, Belgium maize 2007 MUCL (Scaufflaire J.) MAT1-1 \- \- \+ 162 55 ND ND Q57 Buissenal, Belgium maize 2007 MUCL (Scaufflaire J.) MAT1-1 \- \- \+ 916 56 ND ND S46 Villeroux, Belgium maize 2007 MUCL (Scaufflaire J.) MAT1-1 \- \- \+ 1208 59 ND ND PD93/1780 The Netherlands carnation 2003 WUR, Waalwijk et al. (2003) MAT1-1 \- \- \+ 4008 66 23 ND *n* = 69 1 not determined\ 8 isolates containing INS1 3 isolates containing INS2 AFLP for 64 isolates chemotype for 61 isolates; no chemotype analysed for 8 isolates 13 MAT1-2\ 55 MAT1-1 -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- toxins-09-00255-t002_Table 2 ###### Transposable elements at unique locations in isolate 2516. List of TEs found on the core genome of isolate 2516, as well as those that do not have any read support from reads of a DNA mix consisting of five isolates, including isolate 2531. These are TE copies that are not present in the same location in isolate 2531, and this is a lower bound as there is no way to discern in which isolate(s) the remaining TEs are present (reads cannot be traced back to isolates). \#, indicates the number of copies. \# On the Core Genome of 2516 \# Present in 2516 but Not in 2531 (No Read Support) ------------------ ----------------- ---------------- ------------------------------- ------------------------------------------------------ Retrotransposons RLG\_*Maggy* Gypsy/Ty3 like 27 1 RLG\_*Skippy* Gypsy/Ty3 like 5 2 DNA transposons DTF\_*Fot4* Pogo 1 0 DTF\_*Fot5-A* Pogo 40 26 DTF\_*ESP4-B* Pogo 12 0 DTF\_*Drogon* Pogo 33 33 DTF\_*Viserion* Pogo 8 3 DTM\_*Hop7* Mutator 8 8 DTM\_*Hop4* Mutator 1 0 Sum 135 73 [^1]: These authors have equal contributions.

{ "pile_set_name": "PubMed Central" }

1.. INTRODUCTION ================ Vector-borne diseases like Chagas disease and leishmaniasis are caused by parasitic human pathogens and are a major health burden in many developing countries. Current therapies are not very effective and suffer from toxic side effects \[[@R1]-[@R2]\]. In addition, the emergence of drug-resistant strains has been reported \[[@R3]-[@R6]\]. These vector-borne diseases have been recognized by the World Health Organization (WHO) as Neglected Tropical Diseases (NTD) -- chronic infectious diseases endemic mainly in underdeveloped countries, and even though millions of people are affected and thousands die every year, there are no effective cures \[[@R7]\]. In recent years, significant research efforts have been focused on NTD due to policies and research programs implemented by the WHO and other governmental and private organizations \[[@R8]\]. Fungi from *Aspergillus*species cause a series of broncho-respiratory infections collectively known as aspergillosis \[[@R9]-[@R10]\]. Infections by *Aspergillus fumigatus* are the most common in immuno-compromised individuals. Once infection has been established, the mortality rate can be close to 50% \[[@R11]\]. Therefore, new effective anti-fungal drugs are urgently needed. A possible mode of intervention against these parasitic and fungal pathogens is to inhibit the activity of enzymes that aid in cell wall biosynthesis and/or host-pathogen interactions \[[@R12]-[@R13]\]. It has recently been shown that galactofuranose (Gal*f*), a sugar not found in humans, plays an important role in cell wall biosynthesis in *A. fumigatus* and many bacteria and is a major component of the cell surface matrix of *Trypanosoma cruzi* and *Leishmania major*, the causative agents of Chagas disease and leishmaniasis, respectively. In these parasites, Gal*f* plays a major role in virulence \[[@R14]\]. The flavoenzyme UDP-galactopyranose mutase (UGM) is a unique enzyme not present in humans and is essential in the biosynthesis of Gal*f*. Here, we provide an overview of the biosynthesis of Gal*f* and its role in pathogenesis in eukaryotic pathogens with a focus on recent studies on eukaryotic UGMs from *T. cruzi*, *L. major*, and *A. fumigatus*. 1.1. Neglected Tropical Diseases Caused by*T. cruzi*and*L. major* ----------------------------------------------------------------- Chagas disease (or American trypanosomiasis) is endemic throughout Central and South America. It is caused by the protozoan parasite *T. cruzi* and is usually transmitted through a sylvatic cycle from an infected triatomine ("kissing bug") vector that lays parasite-laden feces on wounds and mucous membranes, as well as in conjunctivas \[[@R15]\]. In addition, there have been reports of infection via blood transfusion and orally through ingesting infected mother's milk, raw and undercooked meat, or other food infected by triatomines and/or their feces \[[@R16]\]. Very often in its early stages, Chagas disease is asymptomatic. If left untreated, parasite invasion becomes a serious health risk; symptoms can develop 10-20 years later when the disease becomes chronic and has high mortality rates, typically due to the parasitosis of the heart, causing myocarditis \[[@R17]\]. It is estimated that approximately 16-18 million people have Chagas disease and approximately 50,000 of them die annually; however, these numbers could be higher, since infections are often misdiagnosed due to the very limited, or sometimes complete lack of symptoms \[[@R15],[@R18]\]. Leishmaniasis is a vector-borne systemic disease caused by a trypanosomatid protozoa from the *Leishmania spp.*, which invade human macrophages and replicate intracellularly after being transmitted to humans by infected sandflies (genera *Phlebotomus* and *Lutzomyia*) \[[@R19]\]. Depending on the particular parasitic species, leishmaniasis can develop into three forms: cutaneous, mucocutaneous, or visceral leishmaniasis \[[@R20]\]. *L. major* is the causative agent of cutaneous leishmaniasis, which manifests as a severe skin infection that often causes disfigurement and is endemic in developing countries in the tropics, subtropics, and the Mediterranean basin, with thousands reported new cases annually \[[@R7]\]. 1.2. Infections Caused by *A. fumigatus* ---------------------------------------- Fungi of the genus *Aspergillus* are responsible for several human diseases ranging from allergic reactions and lung infections to sepsis and death \[[@R9]\]. There are hundreds of members of the *Aspergillus*genus,and some are pathogenic to humans, with *A. fumigatus*and*A. niger*being the most common \[[@R21]-[@R23]\]. Among the diseases related to *Aspergillus* infection, allergic bronchopulmonary aspergillosis (ABPA) and invasive pulmonary aspergillosis (IPA) represent a significant health threat to both immuno-competent and immuno-compromised persons \[[@R9],[@R24]\]. IPA infections are commonly observed in patients receiving chemotherapy, organ transplants, and in late-stage AIDS \[[@R25]-[@R26]\]. An increase from 0.3% to 5.8% in IPA infections in patients admitted to intensive care units (ICUs) has been reported in recent years, and has been accompanied by a high mortality rate (50-70%) \[[@R27]-[@R28]\]. This demonstrates the need for new anti-fungal drugs to combat *Aspergillus* infections. 2.. ROLE OF GALACTOFURANOSE IN VIRULENCE ======================================== Galactose is a hexose and a C-4 epimer of glucose (Fig. **[1](#F1){ref-type="fig"}**). In mammals, galactose exists only in the pyranoside form (Gal*p*) linked to other carbohydrates as an essential component of glycolipids and glycoproteins \[[@R29]\]. The main source of galactose in humans comes from consumption of dairy products and its metabolism occurs through the Leloir or Isselbacher pathways \[[@R30]-[@R31]\]. Galactose in the furanoside (Gal*f*) form is not found in mammals; however, Gal*f* is an important building block of glycans of the cell wall and cell surface in several pathogenic organisms and, therefore, its biosynthesis is a strategic target in the discovery of anti-microbial treatments \[[@R14]\]. The specific role of Gal*f*-containing molecules in *Leishmania spp.*, *T. cruzi*, and *A. fumigatus* is described in this section. 2.1. Galactofuranose in *T. cruzi* ---------------------------------- In *T. cruzi*, *β*-Gal*f* is found in glycoinositolphospholipids (GIPLs) and glycosylphosphatidylinositol (GPI) anchor proteins \[[@R32]-[@R33]\]. These glycoconjugates are highly expressed throughout the life cycle of *T. cruzi* and are the main component of the parasite dense surface coat, which has a protective function in parasite survival in the hydrolytic and digestive environment inside their hosts and are important for proliferation \[[@R34]-[@R36]\]. For instance, a 45 kDa GPI-mucin is expressed only in invasive trypomastigotes and not in non-invasive amastigotes \[[@R37]\]. Using specific monoclonal antibodies against this protein prevented adhesion of *T. cruzi* to heart myoblasts \[[@R37]\]. These results suggest that Gal*f*-containing glycoconjugates are involved in the mechanism of myocardial invasion by *T. cruzi.* 2.2. Galactofuranose in *Leishmania spp* ---------------------------------------- In *L. major*, Gal*f* is found in the oligosaccharide core of lipophosphoglycans (LPG) and glycoinositolphospholipids (GIPL) that are essential for parasite survival in the midgut of the vector insect and for parasite transmission to the mammalian host \[[@R38]-[@R40]\]. GIPL-1 from *L. major* has been shown to contribute to the infection process \[[@R41]-[@R42]\]. LPG deletion mutants in *L. major* showed LPG involvement in resistance to oxidative stress and evasion of the human immune system \[[@R39]-[@R40]\]. 2.3. Galactofuranose in *A. fumigatus* -------------------------------------- Of the vast *Aspergillus*genus that includes over 185 species,*A. fumigatus* and *A. niger* are among the \~20 reported human fungal pathogens that cause a variety of opportunistic diseases facilitated by the suppression of the immune system \[[@R43]\]. Gal*f* has been identified in both organisms and is an important component in the fungal cell wall assembly, where it was found in galactomannan, glycoproteins, sphingolipids, and lipid-linked glycans \[[@R44]-[@R48]\]. In *A. fumigatus*, Gal*f* accounts for up to 5% of the dry weight, and is important for fungal growth and cell wall biosynthesis, cell morphogenesis and wall architecture, hyphal adhesion, spore development, and pathogenesis \[[@R22]-[@R23],[@R49]\]. 3.. UDP-GALACTOPYRANOSE MUTASE: AN ATTRACTIVE DRUG TARGET AGAINST EUKARYOTIC HUMAN PATHOGENS ============================================================================================ UDP-Galactopyranose mutase (UGM) is a flavin-dependent enzyme that catalyses the isomerization of UDP-Gal*p* to UDP-Gal*f* through a unique type of flavin-dependent catalysis (Fig. **[2](#F2){ref-type="fig"}**) \[[@R13],[@R50]-[@R52]\]. The gene encoding for UGM (*glf*) was first identified in prokaryotes in 1996 while studying the*Escherichia coli* K12 O antigen \[[@R53]\]. In the following years, it was identified in other pathogens including the eukaryotes *L. major*, *T. cruzi*, and *A. fumigatus* \[[@R54]\]. Deletion of the UGM gene leads to attenuated virulence in *L. major* \[[@R55]\]. In *T. cruzi*, the role of Gal*f* in binding to mammalian cells has been shown; however, deletion of the UGM gene in this parasite have not been performed. Deletion of the UGM gene in *A. fumigatus*, in addition to causing attenuated virulence, leads to cell-wall morphology defects, increased sensitivity to anti-fungal drugs, and growth reduction \[[@R21]-[@R22]\]. These results validate UGM as a potential target for the development of drugs against these eukaryotic pathogens. 3.1. Primary Structure of UGMs ------------------------------ The polypeptide chain lengths of eukaryotic UGMs are generally about 100 amino acid residues longer than those of the prokaryotic enzymes (Fig. **[3](#F3){ref-type="fig"}**). Sequence alignment reveals a moderate identity (47-60%) among the eukaryotic UGMs from *A. fumigatus* (*Af*UGM), *L. major* (*Lm*UGM), and *T. cruzi* (*Tc*UGM), and a slightly lower (37-44%) sequence identity among the prokaryotic homologs from *Escherichia coli* (*Ec*UGM), *Mycobacterium tuberculosis* (*Mt*UGM), *Klebsiella pneumoniae* (*Kp*UGM), and *Deinococcus radiodurans* (*Dr*UGM). However, the sequence identity between eukaryotic and prokaryotic UGM groups is surprisingly low (14-18%) (Tab. **[1](#T1){ref-type="table"}**). Conserved among all UGMs is the GxGxxG motif that is necessary for FAD binding. Only partial conservation of active site residues is observed (Fig. **[3](#F3){ref-type="fig"}**). The low amino acid conservation and the extra amino acid sequence in eukaryotic UGMs endows these enzymes with unique structural features that are important for enzyme function; these are discussed in the next section. Interestingly, an obvious NAD(P)H binding domain or motif is not found in this family of enzymes. This is intriguing since this class of enzymes has been shown to function only in the reduced state. 3.2. 3-Dimensional Structure of UGMs ------------------------------------ Whereas several crystal structures of bacterial UGMs have been determined \[[@R56]-[@R58]\], among the eukaryotic enzymes, only the structure of *Af*UGM is known at this time (Tab. **[2](#T2){ref-type="table"}**) \[[@R59]-[@R60]\]. *Af*UGM is a mixed α/β fold protein containing three structural domains (Fig. **[4](#F4){ref-type="fig"}**). Domain 1 includes a Rossmann fold core and participates in FAD binding. Domains 2 and 3 function in substrate binding \[[@R59]\]. This general 3-domain architecture is also found in the bacterial enzymes; however, the eukaryotic enzymes have extra structural elements that are important in oligomerization and substrate recognition, as summarized below. The conformations of the flavin and flavin-protein interactions are highly conserved between prokaryotic and eukaryotic UGMs (Tab. **[3](#T3){ref-type="table"}**). The isoalloxazine ring of the oxidized enzyme is planar, which is typical for flavoenzymes. Characterization of the reduced FAD conformation is important for understanding the chemical mechanism because the reduced FAD functions as a nucleophile in the UGM reaction. This function places certain structural restrictions on the flavin isoalloxazine. In particular, steric considerations suggest that the reduced isoalloxazine should be nonplanar with the wings of the isoalloxazine bending away from the substrate. Indeed such a conformation is observed in reduced *Af*UGM and *Dr*UGM. In both cases, the isoalloxazine exhibits a butterfly-like deviation from planarity in which the pyrimidine ring bends \~7º away from the substrate site such that the *si* face is concave \[[@R56],[@R59]\]. Curiously, bending of the isoalloxazine by \~13º in the opposite direction is observed in reduced *Kp*UGM; the relevance of this conformation is uncertain since it appears to be inconsistent with nucleophilic attack \[[@R61]-[@R62]\]. Comparison of the structures of *Af*UGM and bacterial UGMs complexed with UDP-Gal*p* reveals conserved motifs and important differences. In all the complex structures (Tab. **[2](#T2){ref-type="table"}** and Tab. **[3](#T3){ref-type="table"}**), the OH-4 of the Gal*p* moiety interacts with the flavin O-4 through hydrogen bonding, and the anomeric carbon of the sugar is placed within a short distance from the flavin N-5 (Fig. **[5](#F5){ref-type="fig"}**). Also, several Arg and Tyr residues are conserved and participate in electrostatic interactions with the pyrophosphate portion of UDP-Gal*p* (Tab. **[3](#T3){ref-type="table"}**). In contrast to bacterial UGMs, in *Af*UGM the OH-6 of Gal*p* is rotated by 110º. There is also a variation in the conformation of bound UDP. In *Af*UGM, UDP is displaced by \~4 Å and rotated by about 90º with respect to bacterial *Kp*UGM and *Dr*UGM. This allows for the hydrogen bonding of uracil with Gln107, a residue that is not present in bacterial UGMs. These structural differences in substrate recognition between bacterial and eukaryotic UGMs could have implications for inhibitor discovery. In particular, it seems unlikely that compounds that target the uridine site of bacterial UGMs will be effective against eukaryotic UGMs. Large protein conformational changes (\>10 Å movements) accompany substrate binding in UGMs. *Af*UGM, and presumably other eukaryotic UGMs, exhibit larger conformational changes. Comparison of the structures of the substrate-free and substrate-bound forms revealed two flaps (residues 179--187 and 203--209) that close down over the substrate like the flaps of a box top (Fig. **[5](#F5){ref-type="fig"}**). Whereas the 180s flap is analogous to the mobile loop of bacterial UGMs, the 200s flap is unique to eukaryotic UGMs. The dramatic closing of the active site in bacterial and eukaryotic UGMs is an important aspect of the catalytic mechanism. These movements result in the assembly of the constellation of residues that position the substrate for nucleophilic attack by the FAD. Furthermore, the closing of the active site prevents diffusion of intermediates, such as UDP, out of the active site during the catalytic cycle. Various oligomeric states have been observed for UGMs in solution. The oligomeric states of several UGMs have been determined from size exclusion chromatography, small-angle X-ray scattering (SAXS), and analysis of protein-protein interfaces in crystal lattices (Tab. **[4](#T4){ref-type="table"}**). Bacterial UGMs tend to form dimers in solution. *Ec*UGM, *Kp*UGM, and *Mt*UGM form a semicircular dimer \[[@R58],[@R63]\]. The fact that this structure is formed by multiple UGMs in different crystal lattices attests to its veracity. The oligomeric state of *Dr*UGM is less certain. The classic UGM semicircular dimer is not found in the *Dr*UGM lattice, and solution studies of the oligomeric state have not been performed on the enzyme. The *Dr*UGM crystal lattice implies decameric and dimeric assemblies, but clearly additional work is needed to determine the oligomeric state and quaternary structure of *Dr*UGM. In contrast, the oligomeric state and quaternary structure of *Af*UGM have been unequivocally determined using a combination of SAXS and X-ray crystallography \[[@R59]\]. These studies have shown that *Af*UGM is unique among UGMs in that it forms a tetramer in solution \[[@R59],[@R64]\]. The *Af*UGM tetramer is a dimer-of-dimers assembly (Fig. **[6](#F6){ref-type="fig"}**). Unique structural features of *Af*UGM that are absent in the bacterial enzymes enable tetramerization. These extra elements include a longer C-terminus, an extra helix in domain 2, and extension of another helix of domain 2 (Fig. **[4](#F4){ref-type="fig"}**). Because the reduced FAD is essential for catalysis, the mechanism by which the enzyme is activated by flavin reduction is an important aspect of UGM biochemistry. Insight into the structural underpinnings of this mechanism has been obtained by comparing crystal structures of oxidized and reduced UGMs. Inspection of the bacterial enzyme structures reveals little difference between the oxidized and reduced conformations, aside from the bending of the isoalloxazine described above. Initial results for *Af*UGM potentially reveal a much more complex mechanism for activating eukaryotic UGMs (Fig. **[7](#F7){ref-type="fig"}**). Two crystal forms of *Af*UGM have been described, a *P*6~5~22 form reported by us \[[@R59]\], and a *P*1 form reported by Sanders' group \[[@R60]\]. Although the interpretation of these structures is complicated by the binding of sulfate ion in the*P*6~5~22 form, and weak electron density in the *P*1 form, the structures tantalizingly imply large conformational changes involving the conserved histidine loop (G61-G62-H63). The structures show that, in the oxidized enzyme, conserved His63 is near the pyrimidine ring of the isoalloxazine and the carbonyl of Gly62 points away from the isoalloxazine (Fig. **[7A](#F7){ref-type="fig"}** and **[7B](#F7){ref-type="fig"}**), which is unprecedented for UGMs. The structures further indicate that flavin reduction induces a crankshaft rotation of the loop backbone, which reverses the orientation of the Gly62 carbonyl bond vector and moves the imidazole of His63 by over 5 Å. These changes bring the carbonyl of Gly62 within hydrogen bonding distance of the N5 of the reduced flavin and move His63 to the *si* face of the isoalloxazine where it stacks in parallel with the isoalloxazine and forms a hydrogen bond with the OH-2' of the ribityl (Fig. **[7C](#F7){ref-type="fig"}**). These interactions between the histidine loop and the flavin help stabilize the reduced state of the enzyme and are found in all other UGM structures. The presence of two Gly residues in the loop is unique to eukaryotic UGMs and probably accounts for the large conformational changes seen in *Af*UGM compared to the bacterial enzymes. More research is needed to validate these conformational changes for *Af*UGM and determine whether other eukaryotic UGMs exhibit analogous movements. 3.3. Chemical Mechanism of Eukaryotic UGMs ------------------------------------------ Despite structural differences, the unique chemical mechanism utilized by UGMs is conserved among the members of this enzyme family. For all UGMs, only the reduced form of the enzyme is active \[[@R63]\], although the reaction does not involve a net gain or loss of electrons, which is common among many other classes of FAD-dependent enzymes \[[@R65]-[@R67]\]. The reported steady-state values with UDP-Gal*f* as substrate and dithionite as the reductant show only minor differences in *k~cat~*and *k~cat/~K~M~* among members of the UGM family (Table **[5](#T5){ref-type="table"}**). The enzymatic reaction has been shown to involve cleavage of an anomeric bond and the formation of a Gal*p*-FAD adduct (Fig. **[8](#F8){ref-type="fig"}**) \[[@R68]\]. This process was initially thought to involve one of three mechanisms: a single-electron transfer from the reduced flavin to a postulated oxocarbenium intermediate of Gal*p* \[[@R69]-[@R70]\] or a nucleophilic substitution via an S~N~1 or S~N~2 mechanism, both leading to the formation of a Gal*p*-FAD adduct \[[@R71]-[@R72]\]. While the Gal*p* is bound to the FAD it undergoes ring opening and closing rearrangement and, after nucleophilic attack by UDP^-^, the UDP-Gal*f* is produced. Formation of the FAD-sugar adduct has been demonstrated by chemical quenching, trapping, and characterization by mass spectrometry in both eukaryotic and prokaryotic UGMs \[[@R71],[@R73]\]. Rapid reaction kinetic analysis with reduced *Tc*UGM and UDP-Gal*f* did not show the presence of a transient flavin semiquinone, inconsistent with a single electron transfer step. Instead, absorbance changes consistent with the formation of a flavin iminium ion, were observed and occur very fast \[[@R73]\]. The structures of *Af*UGM and prokaryotic UGMs in complex with UDP-Gal*p* clearly show that the substrate binds in a conformation optimal for direct attack by the flavin N5. Furthermore, linear free energy relationship (LFER) studies with prokaryotic UGM, reconstituted with various FAD analogs, show changes in *k~cat~* values that correlate linearly with changes in the nucleophilicity of the flavin N5 (slope of ρ = --2.4 ± 0.4), which is consistent with an S~N~2 mechanism \[[@R72]\]. Viscosity effect studies showed that product release was not rate limiting in the case of *Tc*UGM \[[@R73]\]. *In vivo*, all UGMs function in an aerobic environment. Therefore, a system for the generation and maintenance of the reduced flavin must exist in the cell. Despite not having found a canonical NAD(P)H binding motif in the primary sequence of eukaryotic UGMs, NAD(P)H was identified as an effective electron donor for the reduction of the flavin cofactor in *Tc*UGM \[[@R73]\]. Kinetic analyses show that there is preference for NADPH, as it reduces the flavin 7 times faster and binds 5 times tighter than NADH (Tab. **[6](#T6){ref-type="table"}**). In contrast, *Mt*UGM is unable to effectively react with reduced coenzymes \[[@R73]\]. It has been previously reported that the activity of *Kp*UGM was enhanced by the addition of NADH or NADPH \[[@R74]-[@R75]\]. The binding affinities or rates of reduction were not reported, however, the rate enhancement was observed at concentrations greater than 20 mM NADH and at incubation times longer than 10 minutes \[[@R74]\]. Taking into account that NADPH is capable of reducing *Tc*UGM with a rate constant in the second time scale and it binds with micromolar affinity, it is clear that relative to eukaryotic UGM, the bacterial enzymes are not effective NAD(P)H oxidases. The mechanism shown in (Fig. **[8](#F8){ref-type="fig"}**) was recently proposed for *Tc*UGM \[[@R73]\]. Although, as mentioned above, the initial steps in the catalysis, NAD(P)H binding and subsequent FAD reduction, occur much less effective in prokaryotic UGMs, the steps leading to the conversion of Gal*f* are conserved in these enzymes \[[@R71],[@R73]\]. 4.. METHODS FOR HIGH THROUGHPUT SCREENING FOR UGM INHIBITORS ============================================================ With the exponential advance of robotics, data collection, and analysis methods, high throughput screening (HTS) provides an effective and relatively fast preliminary analysis of chemical libraries composed of thousands of chemical compounds for the search of potential chemotherapeutics \[[@R76]-[@R80]\]. Whether the goal is to find an effective inhibitor for a well-explored enzyme or to match existing drugs to new macromolecular targets, HTS provides the rational starting point in the drug discovery process. Elimination of ineffective drug candidates early on using HTS is essential and saves time and resources during later stages of drug development, since libraries can contain thousands of compounds with a 0.1-0.2% probability of identifying positive hits \[[@R81]-[@R82]\]. Thus, the development of a successful assay for HTS is extremely important. Standard methods used to assay UGMs include: HPLC analysis, UV/Vis and stopped-flow spectroscopy, redox potentiometry, fluorescence polarization, and radiochemical detection \[[@R64],[@R69],[@R71],[@R83]-[@R84]\]. The HPLC method has been adopted by many groups, as it easily allows one to measure the activity of UGMs both qualitatively and quantitatively. In general, the assay monitors the reverse reaction, UDP-Gal*f* to UDP-Gal*p* conversion. Both the substrate and the product are easily detected at 262 nm, which corresponds to the absorbance maxima of UDP. Despite the broad utilization of the HPLC method by many research groups, this assay is not suitable for screening large chemical libraries because of the lengthy HPLC run times and because it is not suitable for running multiple measurements at once. A radioactive assay based on the generation and monitoring of tritiated formaldehyde, from the radioactive UDP-Gal*f* degradation product, was used in the screening of 1,300 potential inhibitors against prokaryotic *Mt*UGM. However, the poor sensitivity of the assay due to the equilibrium of the reaction not favoring the formation of UDP-Gal*f* isomer was an essential drawback of this approach for high throughput screening applications \[[@R84]\]. Other reported assays used in HTS against UGMs are based on fluorescence polarization (FP) \[[@R83],[@R85]-[@R86]\]. FP relies on changes in the tumbling of a chromophore as it transitions from the enzyme-bound to the free-state due to competition by an inhibitor. This assay is simple, fast, and can be done on a small scale. Various fluorescent probes based on UDP were synthesized to develop a FP assay for both prokaryotic and eukaryotic UGMs (Fig. **[9](#F9){ref-type="fig"}**). Chromophore **1** was shown to effectively bind prokaryotic UGMs from *M. tuberculosis* and *K. pneumoniae* and used in HTS with the library strategically derivatized from a thiazolidinone core \[[@R85]\]. Contrary to prokaryotic UGMs, the fluorescein fluorophore was not as effective with *Af*UGM, and TAMRA analog **2** was developed instead. Chromophore **2** was shown to bind to *Af*UGM with relative high affinity, thus, yielding a potential tool for HTS in search of UGM inhibitors in eukaryotes \[[@R83]\]. The binding of UDP-chromophore to other eukaryotic UGMs is much less effective. For instance, the K~d~ value of chromophore **2** for *Tc*UGM is \>20 μM, and similar low affinity is observed for *Lm*UGM (Qi and Sobrado, unpublished results). However, since the active sites of eukaryotic UGMs are highly conserved, inhibitors of *Af*UGM might also be effective against the other eukaryotic UGM homologs. 5.. *IN SILICO* DRUG DESIGN AND ITS APPLICATIONS TO UGM ======================================================= Protein crystal structures serve as the blueprints for computer-guided molecular recognition, design, and virtual-screening of drug-like molecules and diagnostic probes \[[@R87]\]. There are three extensively used strategies for *in silico* drug design: *de novo* design, fragment-based drug discovery \[[@R88]-[@R89]\], and virtual screening \[[@R90]\]. The first two are very similar in their algorithms and concepts: both are based on design "from scratch", involving the screening of small pharmacophoric chemical blocks (or fragments) within the three-dimensional active site of the target enzyme. At later stages of the experiment, these fragments are further expanded ("grown") upon other moieties or directly joined together through a chemical bond or a linker. Virtual screening deals with the vast libraries of small chemical compounds utilizing high-throughput docking and pharmacophore-based searching algorithms and can be classified into two broad categories: ligand-based or structure-based docking and scoring \[[@R90]-[@R91]\]. Recently, a virtual screening using various computational tools toward the identification of inhibitors against *Ec*UGM (also *Kp*UGM and *Mt*UGM) was applied to a small-molecule library comprised of 84,000 compounds (LeadQuest,Tripos, Inc.) \[[@R57]\]. A total of 13 compounds (0.015% of the library) were identified as positive hits and tested for inhibitory activity toward *Kp*UGM and *Mt*UGM. Only three compounds were shown to be effective inhibitors and had comparable affinity with the best previously published prokaryotic UGM inhibitors (IC~50~ 7.2 - 62 µM, \[[@R84]-[@R86],[@R92]-[@R93]\]). The effective application of *in silico* screening to bacterial UGMs suggests that a similar approach can be applied to eukaryotic enzymes. Furthermore, it is expected that an *in silico* screening approach with the structure of *Af*UGM will also identify potential inhibitors for *Tc*UGM and *Lm*UGM. 6.. CONCLUDING REMARKS ====================== Gal*f* is important for cell wall biosynthesis and cell surface glycan structures in bacteria, fungi, and parasites. Gal*f* is either essential for growth or important for pathogenesis, making enzymes in its biosynthetic pathway potential drug targets. In this pathway, UGM is an ideal target for drug discovery because this enzyme is absent in humans, and its structure and chemical mechanism are unique. During the past decade, the catalytic mechanism was fully elucidated and the structural differences between prokaryotic and eukaryotic UGM characterized. Recent discoveries in the field of eukaryotic UGMs set the stage for the identification of inhibitors that might lead to drugs for the treatment of neglected diseases like Chagas disease, leishmaniasis, and fungal infections caused by *Aspergillus spp*. In principle, every conformation along the catalytic cycle is a potential design target, including both the active, reduced enzyme and the inactive, oxidized one. Strategies toward developing effective drugs can include the design of a small-molecule competitive inhibitor with much higher binding affinity to UGM with respect to UDP-Gal*p*/*f*, or even molecules that do not bind to the active site but interact with the mobile loops to prevent proper binding of the substrate. This work was supported by NIH grant R01 GM094469. CONFLICT OF INTEREST ==================== The authors confirm that this article content has no conflicts of interest. UDP : = Uridine diphosphate UGM : = UDP-galactopyranose mutase Gal*p* : = Galactopyranose Gal*f* : = Galactofuranose WHO : = World Health Organization NTD : = Neglected Tropical Diseases *A. fumigatus* : = *Aspergillus fumigatus* *A. niger* : = *Aspergillus niger* *T. cruzi* : = *Trypanosoma cruzi* *L. major* : = *Leishmania major* *Leishmania spp.* : = *Leishmania species* ABPA : = Allergic bronchopulmonary aspergillosis IPA : = Invasive pulmonary aspergillosis AIDS : = Acquired Immune Deficiency Syndrome ICU : = Intensive care unit GIPLs : = Glycoinositolphospholipids GPI : = Glycosylphosphatidylinositol FAD : = Flavin adenine dinucleotide NAD(P)H : = Nicotinamide adenine dinucleotide (phosphate) SAXS : = Small-angle X-ray scattering LFER : = Linear free energy relationship HTS : = High throughput screening HPLC : = High performance liquid chromatography UV/Vis : = Ultraviolet/visible FP : = Fluorescence polarization TAMRA : = Tetramethylrhodamine ![Structures of *β*-D-galactopyranose (**A**) and *β*-D-galactofuranose (**B**).](CPD-19-2561_F1){#F1} ![Reaction catalyzed by UDP-galactopyranose mutase.](CPD-19-2561_F2){#F2} ![Sequence alignment of eukaryotic (group 1) and prokaryotic (group 2) UGMs. The residues conserved in all of the sequences are shown in red shaded boxes. Those conserved only in one group are shown in red color. Active site residues are marked with a star, and those interacting with flavin are marked with triangles. The α-helix (spiral) and β-sheets (arrows; TT -- strict β-turns) of *Af*UGM are depicted on top. ClustalW was used to generate the alignment and ESPript 2.2 to create the figure. (The color version of the figure is available in the electronic copy of the article).](CPD-19-2561_F3){#F3} ![Protomer structure of reduced *Af*UGM complexed with UDP-Gal*p*. FADH^-^ and UDP-Gal*p* are colored gray and green, respectively. The flexible active site flaps are colored red. (The color version of the figure is available in the electronic copy of the article).](CPD-19-2561_F4){#F4} ![Close-up view of the active site of reduced *Af*UGM complexed with UDP-Gal*p* highlighting flap closure. The flaps of the ligand-free reduced enzyme are colored magenta, while those of the UDP-Gal*p* complex are colored cyan. The black arrows denote the direction of flap closure induced upon substrate binding. The red arrow denotes the direction of nucleophilic attack by the flavin N5 on the anomeric C atom of the substrate. (The color version of the figure is available in the electronic copy of the article).](CPD-19-2561_F5){#F5} ![Structure of the *Af*UGM tetramer. Each protomer has a different color. The yellow protomer has the same orientation as the protomer in Fig. **[4](#F4){ref-type="fig"}**. (The color version of the figure is available in the electronic copy of the article).](CPD-19-2561_F6){#F6} ![Conformations of the histidine loop in *Af*UGM structures. **A**, oxidized *P*6~5~22 form (PDB code 3UTE); **B**, oxidized P1 form (PDB code 3UKH); **C**, reduced enzyme (PDB code 3UTF).](CPD-19-2561_F7){#F7} ![Proposed mechanism for *Tc*UGM. The oxidized enzyme binds and reacts with NADPH. UDP-Gal*p* binds to the reduced enzyme and a flavin-sugar adduct is formed rapidly by the direct attack of the flavin. Formation of the flavin iminium ion leads to opening of the sugar ring. Attack of the UDP to form the UDP-Gal*f* and its release occur rapidly. The reaction can occur for several more cycles (\~1000) before the enzyme is oxidized by molecular oxygen. The rate limiting step is proposed to be the closing of the sugar ring \[[@R73]\].](CPD-19-2561_F8){#F8} ![Fluorescent probes used in HTS assays for the identification of UGM inhibitors.](CPD-19-2561_F9){#F9} ###### Primary Structure Comparison of UGMs from Different Organisms. *Tc*UGM *Lm*UGM *Ec*UGM *Kp*UGM *Mt*UGM *Dr*UGM --------- ---------- ---------- --------- ---------- ---------- ---------- *Af*UGM **47.0** **49.4** 14.0 15.3 15.2 17.8 *Tc*UGM **60.1** 16.8 18.3 14.5 15.2 *Lm*UGM 15.8 16.8 14.2 15.4 *Ec*UGM **38.7** **44.4** **37.4** *Kp*UGM **42.1** **39.0** *Mt*UGM **37.6** Percentage Identity Shown in Bold Corresponds to the UGMs from the Same Class (Prokaryotic or Eukaryotic). ClustalW Program was Used to Calculate Percentage Identity ###### Available UGM Crystal Structures in the Protein Data Bank (PDB) UGM Active site ligand PDB code \[[@R94]\] Ref. ----------------- ----------------------- --------------------- -------------------------------------- *Af*UGM~red~ 3UTF \[[@R59]\] *Af*UGM~red~ UDP 3UTG \[[@R59]\] *Af*UGM~red~ UDP-Gal*p* 3UTH \[[@R59]\] *Af*UGMox sulfate 3UTE \[[@R59]\] *Af*UGM~red~ UDP-Gal*p* 3UKF \[[@R60]\] *Af*UGM~red/ox~ UDP-Gal*p* 3UKH \[[@R60]\] *Af*UGM~ox~ UDP 3UKL \[[@R60]\] *Ec*UGM~ox~ 1I8T \[[@R63]\] *Kp*UGM~ox~ FMN 3KYB n.a.[\*](#T2F1){ref-type="table-fn"} *Kp*UGM~ox~ UMP, UDP-Glc*p* 3GF4 \[[@R61]\] *Kp*UGM~ox~ UDP-Gal*p* 3INR \[[@R62]\] *Kp*UGM~red~ UDP, UDP-Gal*p* 3INT \[[@R62]\] *Kp*UGM~red~ 1WAM \[[@R58]\] *Kp*UGM~ox~ 2BI7 \[[@R58]\] *Kp*UGM~red~ 2BI8 \[[@R58]\] *Mt*UGM~ox~ 1V0J \[[@R58]\] *Dr*UGM~ox~ UDP 3HE3 \[[@R56]\] *Dr*UGM~red/ox~ UDP-Gal*p* 3HDY \[[@R56]\] *Dr*UGM~ox~ UDP-Gal*p* 3HDQ \[[@R56]\] *Dr*UGM~ox~ UDP, UDP-CH~2~-Gal*p* 3MJ4 \[[@R95]\] Gruber TD, Dimond MC, Kiessling LL, Forest KT, Structure of UDP-galactopyranose mutase bound to flavin mononucleotide. Unpublished results. ###### Ligand Interactions with UGMs ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Type of interaction *Af*UGM~red~ *Ec*UGM~ox~ *Kp*UGM~red~ *Dr*UGM~red~ --------------------------------------------------- -------------------------- ---------------------------- -------------------------------------- ---------------------------- ---------------------------- UDP-Gal*p* contacts π-π stacking with uracil Y104, F158 n.a.[\*](#T3F1){ref-type="table-fn"} F152, Y155 F176, Y179 H-bonding with uracil F106, Q107 N270 F175, N296 Interactions with diphosphate R182, Y317\ R174, Y185, R280, Y314 R198, Y209\ R327, Y419\ R305, Y335, Y370 Y453 H-bonding with Gal*p* R182, N207\ N84, Y349 H109, R305 N457 Other important amino acids for substrate binding N163, W167 W160 T180, W184 FAD contacts H-bonding with ribose H63, G456, S461 N39, Y347 H60, L350, T355 H85\ Y371 Interaction with pyrophosphate T18, L46\ F12, N39\ F13, S14, N41, R343 F39, A40\ R447 R340 N67, R364 π-π stacking with isoalloxazine H63 H56 H60 H85 Interaction with adenine D38, S39, V242 E31, K32, D212,F213 F219 R60, D242\ Y243 H-bonding with isoalloxazine V64, Q458 I57, M349, Y346 I61, M352 I86, M373 H-bonding with N5 G62 A55 P59 P84 H-bonding with ribose D38 E31 D59 FAD-substrate contacts OH-4(Gal*p*) and CO-4(FAD) n.a.[\*](#T3F1){ref-type="table-fn"} OH-4(Gal*p*) and CO-4(FAD) OH-4(Gal*p*) and CO-4(FAD) ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Not available. ###### Amino Acid Composition and Molecular Weight of UGMs from Different Organisms Organism Oligomeric state in solution Number of amino acids MW of monomer, Da Ref. ------------------- ------------------------------------------------------------------------- ----------------------- ------------------- --------------------------- *A. fumigatus* Tetramer[a](#T4F1){ref-type="table-fn"},[b](#T4F2){ref-type="table-fn"} 510 56,820 \[[@R59], [@R64]\] *T. cruzi* Monomer[a](#T4F1){ref-type="table-fn"} 480 54,690 \[[@R73]\] *L. major* Monomer[a](#T4F1){ref-type="table-fn"} 491 54,970 \[[@R96]\] *E. coli* Dimer[c](#T4F3){ref-type="table-fn"},[d](#T4F4){ref-type="table-fn"} 367 42,970 \[[@R63], [@R97]\] *K. pneumoniae* Dimer[d](#T4F4){ref-type="table-fn"} 384 44,460 \[[@R58], [@R61]-[@R62]\] *M. tuberculosis* Dimer[d](#T4F4){ref-type="table-fn"} 399 45,820 \[[@R58]\] *D. radiodurans* Not determined 397 45,700 \[[@R56], [@R98]\] Determined by size exclusion chromatography. Determined by SAXS. Determined by light scattering. Inferred from analysis of protein-protein interfaces in the crystal lattice. ###### Steady State Kinetic Parameters of UGMs from Different Organisms ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Organism k~cat~, s^-1^ K~M~, µM k~cat~/K~M~, µM^-1^ s^-1^ Ref. ------------------- ----------------------------------------------- --------------------------------------------- -------------------------------------------------- -------------------- *A. fumigatus* 72 ± 4*[a](#T5F1){ref-type="table-fn"}* 110 ± 15*[a](#T5F1){ref-type="table-fn"}* 0.65 ± 0.09*[a](#T5F1){ref-type="table-fn"}* \[[@R59], [@R64]\] *T. cruzi* 13.4 ± 0.3*[a](#T5F1){ref-type="table-fn"}*;\ 140 ± 10*[a](#T5F1){ref-type="table-fn"}*;\ 0.093 ± 0.006*[a](#T5F1){ref-type="table-fn"}*;\ \[[@R73]\] 11.5 ± 0.4*[b](#T5F2){ref-type="table-fn"}*;\ 200 ± 20*[b](#T5F2){ref-type="table-fn"}*;\ 0.056 ± 0.005*[b](#T5F2){ref-type="table-fn"}*;\ 8.4 ± 0.9*[c](#T5F3){ref-type="table-fn"}* 690 ± 150*[c](#T5F3){ref-type="table-fn"}* 0.012 ± 0.001*[c](#T5F3){ref-type="table-fn"}* *L. major* 5 ± 0.2*[a](#T5F1){ref-type="table-fn"}* 87 ± 11*[a](#T5F1){ref-type="table-fn"}* 0.057 ± 0.006*[a](#T5F1){ref-type="table-fn"}* \[[@R96]\] *E. coli* 27*[a](#T5F1){ref-type="table-fn"}* 22*[a](#T5F1){ref-type="table-fn"}* 1.22*[a](#T5F1){ref-type="table-fn"}* \[[@R99]\] *K. pneumoniae* 5.5 ± 0.7*[a](#T5F1){ref-type="table-fn"}* 43 ± 6*[a](#T5F1){ref-type="table-fn"}* 0.12 ± 0.02*[a](#T5F1){ref-type="table-fn"}* \[[@R100]\] *M. tuberculosis* 8 13 0.62 \[[@R101]\] *D. radiodurans* 66 ± 2.4*[a](#T5F1){ref-type="table-fn"}* 55 ± 7*[a](#T5F1){ref-type="table-fn"}* 1.18*[a](#T5F1){ref-type="table-fn"}* \[[@R56]\] ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Reduced with 5-20 mM dithionite. Reduced with 0.5 mM NADPH. Reduced with 2.5 mM NADH. ###### Kinetic Parameters of *Tc*UGM Reduction with NAD(P)H \[[@R73]\]. Substrate *k* ~red~, s^-1^ *K* ~d~, µM *k* ~red~/*K*~d~, µM^-1^ s^-1^ ----------- ------------------ ------------- -------------------------------- NADH 0.085 ± 0.0006 550 ± 10 0.00015 ± 0.000002 NADPH 0.600 ± 0.006 98 ± 3 0.0061 ± 0.0001

{ "pile_set_name": "PubMed Central" }

I. INTRODUCTION {#acm20151-sec-0001} =============== Multi‐channel CT has been widely accepted for diagnostic applications as well as for radiotherapy virtual simulation purposes. The effect of CT acquisition parameters on image quality is important to all clinical applications, though the goals of the applications are quite different. A few studies^(^ [^1^](#acm20151-bib-0001){ref-type="ref"} ^--^ [^6^](#acm20151-bib-0006){ref-type="ref"} ^)^ regarding optimizing acquisition parameters for certain applications in diagnostic imaging (DI) using multi‐channel CT scanners partially addressed the effects of machine settings on image quality. However, a more thorough study is needed regarding image quality for newer scanners for diagnostic and therapeutic applications. CT technology continues to develop and there is a vast difference in acquisition parameter options between manufacturers. Though there is some information available regarding technical details, such as image thickness for CT radiotherapy simulation,[^(7)^](#acm20151-bib-0007){ref-type="ref"} the optimal CT image acquisition parameter settings for radiotherapy simulation have not yet been thoroughly investigated. The purpose of this study was to address this gap, and determine the optimal image acquisition parameter settings for conventional CT simulation protocols, including pitch (defined as couch movement per gantry rotation divided by beam width), detector configuration, and rotation time using several different CT scanners in our radiation oncology (RO) department. Routine radiation dose values were also collected in order to help understand image quality characteristics between scanners. II. MATERIALS AND METHODS {#acm20151-sec-0002} ========================= A. Rationale for CT Simulation in Radiation Oncology {#acm20151-sec-0003} ---------------------------------------------------- The implementation of CT simulators in radiation oncology (RO) is somewhat different from CT scanners in DI, though image quality is a main concern for both. Unlike the regular practice in DI, breath holding, contrast injection, variable image thickness, and gantry tilt are generally discouraged in RO. Because most patients undergoing CT simulation in RO have diagnosed cancers, and generally only one scan is used for CT simulation, radiation dose is of less concern than it is in DI. Instead, image quality has singular importance, as accuracy in RO treatment planning is particularly vulnerable to artifacts. In order to simulate radiation treatment conditions, free or gated breathing is required during the CT scan. Fast scanning is generally not desirable because a snapshot of one breathing phase does not represent the average anatomy position during treatment delivery. A scan time of 1 to 2 minutes is considered acceptable. For CT simulation, patient skin surface must be included in the image for treatment planning. Image truncation is intolerable; therefore a larger display field of view (DFOV) is used, usually 50 cm or 65 cm. For a head scan, a body size DFOV is generally used, because part of the shoulders must be scanned as part of patient positioning for most procedures. For CT simulation, the choice of image thickness involves balancing the conflicting considerations of the increased resolution of digitally reconstructed radiographs (DRR) obtained from thin images and increased physician effort required to contour multiple images. The standard choice of the physicians and physicists in our institution is 2.5 mm or 3 mm image thickness and 2.5 mm interval, which agrees with general recommendations.[^(7)^](#acm20151-bib-0007){ref-type="ref"} B. Phantoms and Scanners {#acm20151-sec-0004} ------------------------ A Catphan phantom[^(8)^](#acm20151-bib-0008){ref-type="ref"} (The Phantom Laboratory, Salem, New York) was primarily used to evaluate image quality. Three modules were used in this study: 1) CTP528, high‐contrast resolution bar patterns; 2) CTP515, 1.0% low‐contrast objects in the Supra Slice portion; 3) CTP401, artifact evaluation. The high‐contrast resolution section of the GE CT daily QA phantom[^(9)^](#acm20151-bib-0009){ref-type="ref"} (GE Medical Systems, Milwaukee, Wisconsin) was used for additional artifact evaluation. The high contrast region was chosen to evaluate artifacts because objects with high‐edge gradients are more likely to produce helical artifacts. Both phantoms were positioned by suspending each from an appropriate bracket, following the manufacturers\' standard setup recommendations. Five different CT scanners were evaluated: General Electric (GE) LightSpeed (LS) RT (4‐channel, 80 cm bore)LS RT16 (16‐channel, 80 cm bore)LS16 (16‐channel, 70 cm bore) Philips Medical Systems AcQsim CT (1‐channel, 80 cm bore)MX8000 (16‐channel, 70 cm bore). C. Evaluation Methods {#acm20151-sec-0005} --------------------- Various pitch, detector configuration, and gantry rotation time (1 sec and less) settings were used to acquire images, while maintaining the following constant settings: 120 kVp, 240 effective mAs (defined as tube current $\left( \text{mA} \right) \times \text{rotation}$ time (sec) / pitch; where $\text{pitch} = \text{table}$ movement per gantry rotation / total beam collimation), 1s rotation time unless specified otherwise, 25 cm display field of view (DFOV), standard reconstruction filter, 2.5 mm image interval, and 2.5 mm image thickness for GE or 3 mm image thickness for Philips. On multi‐channel CT scanners, there are multiple data channels that route the signals collected on the detector surface to the reconstruction computer. These data channels can be utilized in a flexible manner to produce more than one thin image during one rotation of the gantry, or fewer thicker images during one rotation of the gantry. The combination of the number of active channels and the detector width (z‐direction) assigned to each channel is called the detector configuration. The product of these two values represents the nominal x‐ray beam width (z‐direction). In general, a combination of fine detector spacing (also called small channel width) and a wide beam width are the most desirable in CT imaging. Fine detector spacing produces a denser helical dataset from which to interpolate, and a wide beam allows for higher table speeds and faster exams. The phantom images were evaluated from three aspects: 1) high‐contrast resolution; 2) low‐contrast resolution; and 3) the presence of helical artifacts. A single observer primarily evaluated all the images. Zoom, display window and level were adjusted to best display the image region of interest for high‐contrast and low‐contrast resolution evaluation. For low‐contrast resolution, since CTP 515 module spans 70 mm, the images with best resolution were selected for evaluation. Three levels were used to score the presence of artifact: "Yes", "Slight", and "No". "Yes" meant the artifact was clearly visible, "No" meant no artifact was present, and "Slight" meant the artifact was present in the phantom images, but it was very subtle and most likely invisible in clinical images. To examine the effect of gantry rotation time, 1 sec and sub‐second rotations were evaluated for scanners that had these options. The fastest gantry rotation for GE RT and Philips AcQsim was 1 sec. Therefore they were not included in the evaluation of gantry rotation time effect. The single‐channel Philips AcQsim CT is limited by prolonged scan time and the effects of excessive tube heating, so only pitches greater than 1 were evaluated for this single‐channel CT. In order to interpret results such as low‐contrast resolution, the standard annual radiation dose assessment made for each scanner was also reported on a normalized (per 100 mAs) basis. These dose evaluations were obtained by using the standard CT Dose Index (CTDI) approach.[^(10)^](#acm20151-bib-0010){ref-type="ref"} III. RESULTS {#acm20151-sec-0006} ============ A. Artifact evaluation {#acm20151-sec-0007} ---------------------- For all CT scanners, helical artifacts were more severe when the channel width was equal to the image thickness and a relatively high pitch was used. For example, for the GE LS16 (Fig.[1](#acm20151-fig-0001){ref-type="fig"}), a detector configuration of $8 \times 2.5\,\text{mm}$ and a pitch of 1.675 resulted in 2.5 mm thick images with severe artifacts Fig. [1(a)](#acm20151-fig-0001){ref-type="fig"}, while a $16 \times 1.25\,\text{mm}$, 1.75 pitch setting showed slight artifacts Fig. [1(b)](#acm20151-fig-0001){ref-type="fig"}, and a $16 \times 1.25\,\text{mm}$, 0.562 pitch had no artifacts in the images Fig. [1(c)](#acm20151-fig-0001){ref-type="fig"}. ![Artifacts in 2.5 mm thick images of the Daily QA phantom for GE LS16: (a) $8 \times 2.5\,\text{mm}$, pitch 1.675; (b) $16 \times 1.25\,\text{mm}$, pitch 1.75; (c) $16 \times 1.25\,\text{mm}$, pitch 0.562.](ACM2-9-151-g001){#acm20151-fig-0001} For GE scanners, a channel width setting equal to half of the image thickness was sufficient to produce artifact‐free images for pitch values close to 1. For GE LS RT, $4 \times 1.25\,\text{mm}$ and a pitch of 0.75 Fig. [2(a)](#acm20151-fig-0002){ref-type="fig"} offered the best image quality for 2.5 mm thick images, while $4 \times 2.5\,\text{mm}$ and 1.5 pitch Fig. [2(d)](#acm20151-fig-0002){ref-type="fig"} settings caused severe helical artifacts. There was no substantial difference in the severity of artifacts for the settings $4 \times 1.25\,\text{mm}$ and pitch 1.5 Fig. [2(b)](#acm20151-fig-0002){ref-type="fig"}, and $4 \times 2.5\,\text{mm}$ and pitch 0.75 Fig. [2(c)](#acm20151-fig-0002){ref-type="fig"}, though the former produced better low‐contrast resolution. ![Artifacts in 2.5 mm thick images of the Catphan phantom for GE LS RT: (a) $4 \times 1.25\,\text{mm}$, pitch 0.75; (b) $4 \times 1.25\,\text{mm}$, pitch 1.5; (c) $4 \times 2.5\,\text{mm}$, pitch 0.75; and (d) $4 \times 2.5\,\text{mm}$, pitch 1.5.](ACM2-9-151-g002){#acm20151-fig-0002} The Philips CT scanners behaved quite differently from the GE CT scanners (Fig. [3](#acm20151-fig-0003){ref-type="fig"}). For the Philips MX8000, $16 \times 1.5\,\text{mm}$ detector configuration for 3 mm image thickness setting (i.e., a channel width equal to half image thickness) caused severe artifacts for all available pitches Fig. [3(D)](#acm20151-fig-0003){ref-type="fig"}. A channel width of 1/4 of the image thickness with pitch less than 1 was needed to eliminate helical artifacts Fig. [3(a)](#acm20151-fig-0003){ref-type="fig"} on this scanner. The other two settings, $16 \times 0.75\,\text{mm}$, pitch 1.438 Fig. [3(b)](#acm20151-fig-0003){ref-type="fig"} and $16 \times 1.5\,\text{mm}$, pitch 0.688 Fig. [3(c)](#acm20151-fig-0003){ref-type="fig"}, resulted in less severe and different artifacts; the setting for Fig. [3c](#acm20151-fig-0003){ref-type="fig"} were better than those for Fig. [3b](#acm20151-fig-0003){ref-type="fig"}. For Philips AcQsim CT, helical artifacts increased with pitch (Fig. [4](#acm20151-fig-0004){ref-type="fig"}). ![Artifact evaluation of 3 mm thick images of the Catphan phantom for Philips MX8000: (a) $16 \times 0.75\,\text{mm}$, pitch 0.688; (b) $16 \times 0.75\,\text{mm}$, pitch 1.438; (c) $16 \times 1.5\,\text{mm}$, pitch 0.688; (d) $16 \times 1.5\,\text{mm}$, pitch 1.313. Note: Windows and levels were adjusted separately for optimal display.](ACM2-9-151-g003){#acm20151-fig-0003} ![Artifact evaluation of 3 mm thick images of the Catphan phantom for Philips AcQsim single‐channel CT. (a) pitch 1.0; (b) pitch 1.5.](ACM2-9-151-g004){#acm20151-fig-0004} Helical artifacts appeared to be scanner/model specific, not only different across manufacturers, but also different across models made by a single manufacturer. In Fig.[5](#acm20151-fig-0005){ref-type="fig"}, we compared the images from GE LS RT16 Fig. [5(a)](#acm20151-fig-0005){ref-type="fig"}, [5(b)](#acm20151-fig-0005){ref-type="fig"} with those from GE LS16 Fig. [5(c)](#acm20151-fig-0005){ref-type="fig"}. For the same pitch (1.375) and detector configuration ($16 \times 1.25\,\text{mm}$), the GE LS RT16 produced more severe artifacts than the GE LS16 Fig. [5(b)](#acm20151-fig-0005){ref-type="fig"}, [5(c)](#acm20151-fig-0005){ref-type="fig"}. The combinations of the largest channel width and largest pitch that produced artifact‐free images for all scanners are listed in Table [1](#acm20151-tbl-0001){ref-type="table-wrap"}. ![GE 16‐channel CT comparison, 2.5 mm thick images, for detector configuration: $16 \times 1.25\,\text{mm}$. GE LS RT16, pitch 0.562 (a), pitch 1.375 (b), and GE LS16, pitch 1.375 (c).](ACM2-9-151-g005){#acm20151-fig-0005} ###### Combination of the largest channel width and largest pitch that produced artifact‐free images. *Model* *Image thickness (mm)* *Pitch* *Detector configuration (mm)* *Head* $\text{CTDI}_{W}/100\text{mAs}$ *(mGy)* *Body* $\text{CTDI}_{W}/100\text{mAs}$ *(mGy)* ---------------- ------------------------ --------- -------------------------------------------------------------------- ------------------------------------------------ ------------------------------------------------ Philips MX8000 3 0.688 $16 \times 0.75 = 12$ 13.2 4.2 Philips AcQsim 3 1 $1 \times 3.00 = 3$ 13.0 3.5 GE LS16 2.5 1.375 $16 \times 1.25 = 20$ 14.4 7.3 GE LS RT16 2.5 0.938 $16 \times 1.25 = 20$ 12.4 6.8 GE LS RT 2.5 0.75 $4 \times 1.25 = 5$ 23.8 10.4 *2.5* *0.75* $4 \times 2.5 = 10$ [^\*^](#acm20151-tbl-note-0001){ref-type="fn"} *17.7* *6.4* Not artifact‐free for helical scan, but for dose comparison. The helical streak artifact appears to depend on the combination of detector configuration and pitch, and that the relationships between these characteristics are not consistent between or even within vendors. The severity of artifacts increased with increasing pitch. The artifact evaluation in Tables [2](#acm20151-tbl-0002){ref-type="table-wrap"}, [3](#acm20151-tbl-0003){ref-type="table-wrap"} and [4](#acm20151-tbl-0004){ref-type="table-wrap"} based on the high‐contrast section of the Catphan phantom gives additional information for specific scanner settings. ###### Image quality results for GE LS RT16. *Rotation Time (s)* *Mode / Pitch* *High Contrast Resolution (lp/cm)* *1% Low Contrast Diameter (mm)* *Artifact Present?* --------------------- ---------------- ------------------------------------ --------------------------------- --------------------- 0.5 Axial 7 4 No 1.0 Axial 7 4.5 No 0.5 0.562 7 4 No 0.5 0.938 7 4 No 0.8 0.938 8 4 No 0.5 1.375 7 4.5 Yes 0.8 1.375 7 4 Yes Technique: effective $\text{mAs} = 240$, image thickness 2.5 mm, detector configuration $16 \times 1.25\,\text{mm}$. Low contrast was defined as the diameter of the smallest circular target detectable. ###### Image quality results for GE LS16. *Rotation Time (s)* *Pitch* *High Contrast Resolution (lp/cm)* *1% Low Contrast Diameter (mm)* *Artifact Present?* --------------------- --------- ------------------------------------ --------------------------------- --------------------- 0.5 0.562 8 3.5 No 0.8 0.562 8 4 No 0.8 0.938 7 4 No 0.8 1.375 7.5 3.5 Slight 1 1.75 7 3 Yes Technique: effective $\text{mAs} = 240$, image thickness 2.5 mm, detector configuration $16 \times 1.25\,\text{mm}$. Low contrast was defined as the diameter of the smallest circular target detectable. ###### Image quality results for Philips MX8000. *Rotation Time (s)* *Pitch* *High Contrast 1% Resolution (lp/cm)* *Low Contrast Diameter (mm)* *Artifact Present?* --------------------- --------- --------------------------------------- ------------------------------ --------------------- 0.4 0.564 8 4 No 0.5 0.563 8 3.5 No 0.75 0.563 8 3 No 1.5 0.563 8 3 No Technique: effective $\text{mAs} = 240$, image thickness 3 mm, detector configuration $16 \times 0.75\,\text{mm}$. Low contrast was defined as the diameter of the smallest circular target detectable. B. Resolution evaluation {#acm20151-sec-0008} ------------------------ Image resolution (low and high contrast) was evaluated using the phantom sections shown in Fig. [6](#acm20151-fig-0006){ref-type="fig"}. For these GE scanners (Tables [2](#acm20151-tbl-0002){ref-type="table-wrap"} and [3](#acm20151-tbl-0003){ref-type="table-wrap"}), the image resolution did not appear to be affected by pitch and rotation time, provided that mA was adjusted appropriately. For high‐contrast resolution, the variation was within 1 line pair (lp) /cm around the average value of $7.3\text{lp}/\text{cm}$. For low contrast resolution, the variation was within 0.6 mm over the average diameter of 3.9 mm for 1% detectable low‐contrast objects. These variations may be due to quantum and statistical fluctuations. However, GE LS16 had systematically better high‐ and low‐contrast resolution than GE LS RT16. GE LS16 has a maximum tube current limit of 440 mA, thus a pitch of less than 1 had to be used to achieve 240 effective mAs for 0.5s gantry rotation. Of these two GE models, the larger bore CT scanner had lower image quality and was more prone to helical artifacts. ![Sample image sections used for resolution evaluation: (a) Catphan CTP515 mode, the marked area is 1% contrast which was used for low‐contrast resolution evaluation; (b) Catphan CTP528 mode, use for high‐contrast resolution evaluation. Techniques: GE LightSpeed RT16, pitch 0.938, rotation time 0.8 sec, $16 \times 1.25\,\text{mm}$ detector configuration, 2.5 mm image thickness (display not optimized).](ACM2-9-151-g006){#acm20151-fig-0006} A similar approach was used for images obtained using Philips MX8000 (Table [4](#acm20151-tbl-0004){ref-type="table-wrap"}). A clear trend was observed that short rotation time (0.5 sec or less) caused slight low‐contrast resolution degradation, though high‐contrast resolution remained unaffected. The image resolution was best for Philips MX8000, followed by GE LS16, and GE RT16. C. Radiation Dosimetry {#acm20151-sec-0009} ---------------------- The CTDIW was normalized for 100 mAs using the standard head and body phantoms [^(11)^](#acm20151-bib-0011){ref-type="ref"} and represents the most recent annual dosimetry evaluations. CTDIW / 100 mAs varied by nearly a factor of 2 from 12.4 mGy to 23.8 mGy among the five scanners for the head phantom. For the head CTDI phantom, normalized dose (CTDIW / 100 mAs) was highest for GE RT, followed by GE LS16, Philips MX8000, and Philips AcQsim, and was lowest for GE LS RT16. For the body CTDI phantom, CTDIW / 100 mAs ranged from 3.5$- 10.4$mGy; this value was highest for GE LS RT, followed by GE LS16, GE LS RT16, and Philips MX8000, and was lowest for Philips AcQsim. For this study, the head CTDI dose results would be more relevant than the body CTDI values. The normalized body CTDI results were included for completeness. IV. DISCUSSION {#acm20151-sec-0010} ============== We found that for GE scanners, a channel width setting equal to half of the image thickness was sufficient to produce artifact‐free and good quality images for pitch values close to 1. Our result with GE RT 4‐channel CT agrees with Takahashi, et al.[^(3)^](#acm20151-bib-0003){ref-type="ref"} who determined that for this scanner model, the settings that achieved the best IQ were found to be $4 \times 1.25\,\text{mm}$ detector configuration and a pitch of 0.75 (though that study was only focused on depicting arterial stenosis, and imaging artifact was not specifically evaluated). For GE RT, the detector setting $4 \times 1.25\,\text{mm}$ resulted in a high CTDIW of 2.38 cGy (Table [1](#acm20151-tbl-0001){ref-type="table-wrap"}), while $4 \times 2.5\,\text{mm}$ setting lowered CTDIW to 1.77cGy due to improved dose efficiency. If a faster scan speed is needed, we would suggest using $4 \times y2.5\,\text{mm}$ and pitch 0.75 rather than $4 \times 1.25\,\text{mm}$ and pitch 1.5, though they have the same table speed and similar image quality. For Philips MX8000, a channel width of 1/4 of the image thickness with a pitch of less than 1 was needed to eliminate helical artifacts. However, the MX8000 had overall somewhat better high‐ and low‐contrast resolution than GE scanners (Tables [2](#acm20151-tbl-0002){ref-type="table-wrap"}, [3](#acm20151-tbl-0003){ref-type="table-wrap"}, [4](#acm20151-tbl-0004){ref-type="table-wrap"}). Differences between these two manufacturers may be due to the different reconstruction algorithms or different dose values (matching technique parameters does not necessarily result in equivalent radiation output or quality between scanners). We observed a larger than expected difference in normalized dose (CTDIW / 100 mAs) among these scanners; when using the head CTDI phantom, there was a range of a factor of 2 in CTDIW / 100 mAs. When using the body CTDI phantom, we observed a difference between vendors by a factor of 3. In general, when spatial resolution was enhanced, artifacts present in the image were also enhanced. In helical CT, data re‐binning and interpolation are used to generate planar transverse images from helical attenuation data. This process yields better quality images if helical data is more finely spaced. The choice of detector configuration is generally a much more powerful method of reducing helical artifact than decreasing scan pitch. Sometimes, the selection of the maximum number of active channels and the finest channel width must be balanced against the need for a faster couch speed. The small differences in spatial resolution shown in Tables [2](#acm20151-tbl-0002){ref-type="table-wrap"} to [4](#acm20151-tbl-0004){ref-type="table-wrap"} are not likely to be detectable in clinical images, because the DFOV of 50 cm or greater over the $512 \times 512$ matrix used in CT simulation limits the achievable spatial resolution. Some difference in contrast resolution is expected due to the different bore sizes, detector designs and reconstruction algorithms. Increasing exposure, or effective mAs, will increase low‐contrast resolution across all scanners. Though phantom studies are appropriate for comparative evaluations, they are not accurate simulations of patient scans. A major difference between a phantom and a patient is that a patient often moves (voluntarily or involuntarily) during a CT acquisition, especially for those of the chest and abdomen. The resulting blur is not present on the stationary phantom images. Faster rotation time reduces motion‐blurring artifact[^(12)^](#acm20151-bib-0012){ref-type="ref"} and improves patient image quality, though it should not affect the phantom image quality. Because fast scan time (breath‐hold) is not desired in CT simulation, a reasonably low pitch, fast rotation and fine detector configuration should be used to improve image quality and also to acquire patient images in a more averaged position rather than a snapshot of one moment. In general, we recommend using fast gantry rotation for chest and abdomen scans, 0.5$- 0.8$ sec for example, if the options are available. Image resolution evaluation can be used as a reference when setting up scan protocols for radiotherapy simulation. Tube current modulation should not affect the appearance of helical artifacts. However, tube current modulation may affect low‐contrast resolution, because effective mAs will vary in the x/y plane, along the axis, or both. V. CONCLUSIONS {#acm20151-sec-0011} ============== In conclusion, we have quantified the effects of changes in various parameters used for RO CT simulation on specific image metrics. Given the differences we found between CT machines in our own study, choosing optimal image acquisition parameter settings for CT simulation purposes may require a phantom evaluation for each scanner model to be used.

{ "pile_set_name": "PubMed Central" }

In the past 5 years, as methicillin-resistant *Staphylococcus aureus* (MRSA) has emerged as a community pathogen, awareness of the role of animal exposure from pets or farming as sources of MRSA has increased ([@R1]--[@R3]). We identified a clone of *S. aureus* previously associated with outbreaks of infections in animals and in humans who work with animals in 2 unique collections of *S. aureus* isolates. The first was from a population-based study of *S. aureus* colonization among residents of northern Manhattan in New York, NY, USA; the second was from isolates obtained from the Dominican Republic. This clone does not digest with the restriction enzyme *Sma*I, which is generally used for pulsed-field gel electrophoresis (PFGE). Consequently, the clone is identified by multilocus sequence typing as sequence type 398 (ST398). Both methicillin-resistant and methicillin-susceptible isolates of *S. aureus* have been reported ([@R4]). ST398 has been found primarily in Europe, where it has been isolated from pigs and pig farmers in the Netherlands and France and from dogs, pigs, horses, and humans in Germany and Austria ([@R5]*--*[@R8]). Colonization with MRSA ST398 has recently been reported in pigs and pig farmers in Canada ([@R9]). The Study ========= The community-based study was conducted from 2004 through 2007 in the northern section of Manhattan, a borough of New York City. Northern Manhattan contains a large, medically underserved population that has close ties to the Dominican Republic. Participants were recruited by using random-digit dialing. Consenting persons and household members were subsequently interviewed and screened for *S. aureus* colonization. A total of 321 eligible households containing 914 household members participated. In 9 households, 13 participants were found to be colonized with *S. aureus* isolates that were *Sma*I resistant. Digestion with the *Cfr9*I, an isoschizomer of *Sma*I, yielded identical PFGE profiles ([Figure](#F1){ref-type="fig"}). Subsequent multilocus sequence typing confirmed the ST398 identification (allelic profile 3--35--19--2-20--26--39). All strains were methicillin susceptible. A representative strain was *spa*-typed as type t571 (allelic profile 8--16--2-25--2-25--34--25, eGenomics type 109); it was Panton-Valentine leukocidin negative. ![Pulsed-field gel electrophoresis profiles of sequence type 398 isolates from the Dominican Republic (D); northern Manhattan (M), New York, NY, USA; and Canada (C) (provided by Scott Weese). Strains within households in which \>1 person was colonized are identified numerically. The dendrogram shows the percent similarity of the isolates. A similarity [\>]{.ul}70% indicates closely related or identical strains.](08-0609-F){#F1} Characteristics of persons colonized with ST398 were similar to those of persons in the community-based study and with northern Manhattan census characteristics ([Table](#T1){ref-type="table"}). The 13 isolates were from 9 different families; 1 family had 4 members colonized with ST398 at either nasal or axillary sites. The mean age of those colonized was 33.4 years; only 1 child (7 years of age) was colonized. Two persons from different families were colonized with ST398 at multiple sites, none of which were confirmed as infections. ###### Characteristics of persons colonized with *Staphylococcus aureus* ST398, northern Manhattan, New York City, NY, USA, 2004--2007, compared with study population and 2000 census population for area\* Characteristics ST398 subset, no. (%) Overall study population, no. (%) 2000 census population, no. (%) ----------------------- ----------------------- ----------------------------------- --------------------------------- Race/ethnicity Latino 11 (84.6) 813 (89) 173,755 (68) Non-Hispanic white 2 (15.4) 90 (9.8) 65,449 (25.6) African American 0 11 (1.2) 53,514 (20.9) Asian 0 0 5,370 (2.1) Sex Male 4 (30.8) 362 (39.6) 120,866 (47.3) Female 9 (69.2) 552 (60.4) 134,723 (52.7) Age group, y \<5 0 84 (9.2) 17,878 (7.0) 5--17 1 (7.6) 238 (26.2) 49,196 (19.3) 18--44 10 (76.9) 297 (32.5) 112,195 (44.0) [\>]{.ul}45 2 (15.4%) 290 (31.9) 76,320 (29.9) Occupational exposure 5 (38.5) 58 (6.4) NA Travel outside USA 7 (53.8) 171 (18.7) NA Daycare exposure 2 (15.4) 87 (9.5) NA Total population 13 (100) 914 (100) 255,589 (100) \*ST, sequence type; NA, not available. No household reported owning pets, although 2 reported animal contact. Of the 12 adults, 5 (41.7%) reported possible job exposure to *S. aureus*, including 1 who worked in a healthcare-associated field. No household reported patronizing *viveros*, or live poultry markets, which are common in the Latino communities of northern Manhattan and the Bronx. Two households reported having children who attended day care, although none of these children were colonized with *S. aureus*. Although 15% of the Dominican population in the study reported travel to the Dominican Republic within 6 months of their interview, none of the colonized participants reported recent travel to the Dominican Republic. No contact among the different households was reported. A second collection of *S. aureus* isolates was gathered during 2007 and 2008 from a convenience sample of 89 anonymous infection and colonization isolates received from the Dominican Republic. Six isolates were identified as methicillin-susceptible *S. aureus* clone ST398. Strains were provided by 1 hospital in Santo Domingo, Dominican Republic (n = 53), and 1 microbiology laboratory, the Laboratorio Referencia (n = 28), which serves as a reference laboratory for the country. Four isolates from the hospital and 2 from Laboratoria Referencia were identified as methicillin-susceptible ST398. Sociodemographic data on these persons were limited. Two of the ST398 isolates were from women with infections living in Santo Domingo, and the remaining 4 were colonization samples. Of the 6 isolates, 5 were found to be *spa*-type t571 (eGenomics type 109), and 1 was found to be type t3625 (eGenomics untyped). Pairwise similarity scores for the isolates were calculated by the Dice coefficient, and an overall similarity score was calculated by using the unweighted pair group method with arithmetic mean. Comparing the isolates by using a dendrogram-based similarity score [\>]{.ul}70%, we found that the strains from northern Manhattan and from the Dominican Republic were closely related, although they contrasted with ST398 isolates from Canada (provided by Scott Weese, Ontario Veterinary College, University of Guelph) ([Figure](#F1){ref-type="fig"}) ([@R10]). Conclusions =========== Identification of the *S. aureus* clone ST398 in northern Manhattan and in the Dominican Republic suggests its introduction into the United States by travelers between the 2 countries. The largely Latino population of northern Manhattan is composed mainly of immigrants or first-generation families from the Dominican Republic; travel between the 2 regions is common. Alternatively, northern Manhattan may contain reservoirs, such as live poultry markets, which may serve as a means of strain transmission. Colonization or infection with the *S. aureus* clone ST398 has been associated with exposure to pigs, pets, and other animals ([@R5]*,*[@R7]--[@R9]), and the *S. aureus* clone ST398 has been isolated from meat products ([@R2]). However, transmission is not limited to animal exposures. Person-to-person spread has occurred among household members and in the hospital setting ([@R6]*,*[@R8]*,*[@R11]). For example, a dramatic increase in persons colonized as well as infected with MRSA clone ST398 was recently reported in a Dutch hospital ([@R12]). The presence of this strain among several household members in our study reinforces earlier observations of the potential for horizontal transmission of this clone after it is introduced into an appropriate setting. Although information about the persons from the Dominican Republic was limited, the 2 groups provided identical strain profiles, suggesting a possible link between the 2 countries. Given ST398's history of rapid dissemination in the Netherlands, its potential for the acquisition of methicillin resistance, and its ability to cause infections in both community and hospital settings, monitoring the prevalence of this strain in northern Manhattan and the Dominican Republic will be important to understand more about its virulence and its ability to spread in these communities. *Suggested citation for this article*: Bhat M, Dumortier C, Taylor B, Miller M, Vasquez G, Yunen J, et al. *Staphylococcus aureus* ST398, New York City and Dominican Republic. Emerg Infect Dis \[serial on the Internet\]. 2009 Feb \[*date cited*\]. Available from <http://www.cdc.gov/EID/content/15/2/285.htm> We gratefully acknowledge the advice and suggestions of Scott Weese. This study was supported by a grant from the US Centers for Disease Control and Prevention. C.D. was supported by a Fulbright Scholar grant made possible by the US Department of State, by the Franco-American Commission for Educational Exchange, and by a grant from the Bourse Collery de l\'Académie Nationale de Médecine. Ms Bhat is a researcher in the Infectious Diseases Division, Department of Medicine, Columbia University. She is currently investigating the molecular epidemiology of *S. aureus* transmission in communities and hospitals.

{ "pile_set_name": "PubMed Central" }

Introduction ============ The father of modern orthodontics, Edward Hartley Angle, in 1899 classified malocclusions in Class I, Class II, and Class III based on permanent first maxillary and mandibular molars relationship and alignment (or lack of it) of teeth with reference to the line of occlusion.[@b1-ccide-10-099] Gradually, Angle's classification was modified and additional information such as jaw relationship and the pattern of growth were also included. Thus, a Class III jaw relationship suggests that the mandible has acquired a more mesial position in relation to the maxilla and/or cranial base.[@b2-ccide-10-099],[@b3-ccide-10-099] Occasionally, due to dental compensation, sometimes there is Class I dental relationship on the Class III skeletal base. Further, Charles Henry Tweed classified Class III malocclusions as a pseudo Class III malocclusion with normal mandible and underdevelopment of maxilla (category A) and skeletal Class III malocclusion with prognathic mandible or an underdevelopment of maxilla (category B).[@b4-ccide-10-099] Class III problems may arise due to deficient growth of maxilla in the downward and forward direction and more forward growth or reduced downward growth of mandible. Hence, a hypodivergent growth pattern accentuates the Class III problem due to more growth rotation of the mandible in the upward and forward direction, while a vertical growth pattern alleviates it due to downward and backward rotation, provided that excessive facial height does not become the problem instead.[@b5-ccide-10-099] Nonsurgical treatment of Class III problems remains a challenge in our profession. However, prompt diagnosis and early intervention of Class III malocclusion may be helpful to reduce the extent of burden for severe Class III malocclusion in late adolescence.[@b6-ccide-10-099] This paper presents an overview of developing Class III malocclusion, with the emphasis on challenges and their solutions based on the best current available evidence. Prevalence ========== Existing literature regarding the global prevalence of Class III malocclusions has shown that its prevalence varies greatly among and within different races, ethnic groups, and geographic studied regions ([Tables 1](#t1-ccide-10-099){ref-type="table"} and [2](#t2-ccide-10-099){ref-type="table"}). There is a wide range of reported prevalence, even with conflicting results, and the discrepancies in the prevalence rate might be attributed to the variation among samples, the timing of investigation, and type of analysis performed.[@b7-ccide-10-099] A recent systematic review[@b7-ccide-10-099] reported a global prevalence of Angle Class III malocclusion within the interval of 0%--26.7% for different populations. Prevalence rates of 15.80%, 15.69%, and 16.59% were revealed for Southeast Asian countries, Chinese, and Malaysian groups, respectively. Among Japanese it was around 14%, for Koreans 9%--19%, and about 1.65% for Taiwanese. For Indian children aged from 5 to 15 years, the prevalence varied within 0%--4.76%.[@b7-ccide-10-099] Further, from a global viewpoint, Indians had the lowest prevalence of 1.19% among all other racial groups. A prevalence of 10.18% was reported for Middle Eastern populations, and among them, for Israeli Arabs it was 1.3%, Iranians about 15.2%, Turkish about 10.30%--11.5%, and Egyptians showed a rate from 4% to 11.38%. Regarding African countries,[@b7-ccide-10-099] the prevalence rate was found to be 4.59% and varying for Kenya, Tanzania, and Nigeria (between 1% and 16.8%). Class III malocclusions have been found to be more prevalent in Hispanic than in African or Caucasian groups. Prevalence of about 9.1% and 8.3% were reported for Americans and Mexican Americans, respectively.[@b8-ccide-10-099] Factors such as the method of malocclusion study and the age group studied may influence the varying prevalence in Caucasians between 3% and 5%.[@b9-ccide-10-099]--[@b15-ccide-10-099] Prevalences of \~5% and from 2% to 6% have been found in Latin and European populations, respectively.[@b8-ccide-10-099] Furthermore, the White population in United Kingdom and Scandinavia had a Class III incidence of about 3%--5%,[@b16-ccide-10-099] and about 6% for Sweden.[@b17-ccide-10-099] For Americans, the prevalence was found to be about 5%.[@b15-ccide-10-099],[@b18-ccide-10-099] Studies on US African-American population groups found the prevalence in the range of 3%--6%.[@b19-ccide-10-099]--[@b21-ccide-10-099] Similar studies conducted on other nationalities revealed a Class III malocclusion prevalence of about 3% for Brazilian,[@b22-ccide-10-099] 14% for Syrian,[@b23-ccide-10-099] and 9.4% for Saudi Arabian individuals.[@b24-ccide-10-099] Etiology of Class III malocclusion ================================== Similar to most of the malocclusions and dentofacial deformities, the etiology of Class III malocclusion is multifactorial. It results from a distortion of normal development, rather than from any pathological process. Expressions of Class III malocclusion are results of interaction between innate factors or genetic hereditary with environmental factors.[@b25-ccide-10-099]--[@b27-ccide-10-099] Studies of human inheritance have provided sufficient evidence to establish the fact that mandibular growth is mainly affected by heredity.[@b27-ccide-10-099]--[@b31-ccide-10-099] Familiar genetic inheritance has a strong influence on skeletal craniofacial dimensions contributing to Class III malocclusion and a significantly higher incidence of this malocclusion has been found to have a familial occurrence between members of many generations.[@b32-ccide-10-099],[@b33-ccide-10-099] The best known example of familial inheritance is Habsburg Jaw, in which mandibular prognathism recurred over multiple generations in the European royalty.[@b34-ccide-10-099],[@b35-ccide-10-099] The pattern of transmission of Class III malocclusion still remains an issue of controversy. According to some authors, the transmission is autosomal recessive, and according to others, it is autosomal dominant with complete or incomplete penetrance; yet, some others support the polygenic transmission mode.[@b36-ccide-10-099],[@b37-ccide-10-099] Environmental factors known to contribute and influence this malocclusion include wrong postural habits of the mandible which pathologically alter the mandibular condyle positioning within the fossa and as a result the final mandibular spatial position expressed with a forward slide of the mandible. Various factors such as growth stimulus, history of prolonged sucking or resting tongue habits, atypical swallowing, nasal airway obstruction, mouth breathing, functional mandibular shifts because of respiratory needs, tongue size and pharyngeal airway shape and size altered (enlarged tonsils, large tongue, adenoids), hormonal imbalances and disturbances such as gigantism or pituitary adenomas, trauma, premature loss of primary teeth, congenital anatomic defects (ie, cleft lip, cleft palate), and muscle dysfunction alone or in combination with other environmental factors play a definitive etiological role.[@b38-ccide-10-099]--[@b43-ccide-10-099] Component of Class III malocclusion =================================== Class III malocclusion represents a complex three-dimensional facial skeletal imbalance between maxillary and mandibular growth along with varying degrees of dentoalveolar and soft tissue compensations which can be expressed in many morphological ways.[@b44-ccide-10-099] Class III malocclusion may be associated with maxillary growth deficiency (and/or maxillary retrognathia), mandibular growth excess (and/or mandibular prognathism), or a combination of both along with vertical and transverse malformations.[@b45-ccide-10-099]--[@b49-ccide-10-099] Based on the position of the maxilla relative to the craniofacial skeleton, Park and Baik classified Class III malocclusions into three basic types: true mandibular prognathism type A -- individual with normal maxilla and prognathic mandible; type B -- individual with excessive growth of maxilla and mandible, but with relatively more growth of mandible; type C -- individual with maxillary hypoplasia, obtuse nasolabial angle, and concave facial profile. Type C individuals can easily be camouflaged orthodontically by dentoalveolar compensation.[@b50-ccide-10-099] Common skeletal features such as shortened anterior (N-S) and posterior cranial base (S-Ar/Ba), reduced saddle angle (N-S-Ar), and an increased gonial angle (Ar-Go-Gn) were identified to lead to a more forward positioning of the glenoid fossa resulting in Class III malocclusion.[@b51-ccide-10-099]--[@b53-ccide-10-099] Studies about the skeletal and dental components of Class III malocclusions have revealed the establishment of a facial pattern at early childhood which has a tendency to worsen with growth.[@b54-ccide-10-099]--[@b58-ccide-10-099] Skeletal Class III malocclusions can be a result of various factors: prognathic and/or macrognathic mandible with a normal maxilla both in position and in size;retrognathic and/or micrognathic maxilla with a normal mandible both in position and in size;combination of retrognathic and/or micrognathic maxilla with prognathic and/or macrognathic mandible;normal skeletal jaw relationship with reverse overjet in the presence of centric relation (CR)--centric occlusion (CO) discrepancy, also known as a "pseudo" Class III relationship. Dental features of Class III individuals include Class III molar and canine relationship, maxillary incisors protrusion and mandibular incisors retrusion with edge-to-edge bite or anterior crossbite. Based on various combinations of skeletal components, patients with Class III malocclusion exhibit a wide range of underlying skeletal and craniofacial features similar to the prevalence of Class III malocclusion, which can vary among different racial and ethnic groups as shown by comparative studies. For example, Mongoloid populations (Japanese, Koreans, and Chinese) with Class III phenotypes present with characteristic features such as acute anterior cranial base angle and a prominent and elongated mandible with a short and hypoplastic maxilla, while normal maxillary size and position were observed for Caucasians.[@b51-ccide-10-099],[@b59-ccide-10-099] Differential diagnosis of Class III malocclusion ================================================ The specter of problems of Class III malocclusion ranges in gravity from dentoalveolar problems with functional anterior shift of the mandible to true skeletal problems with serious maxillomandibular discrepancies, which leads to its highly challenging diagnosis.[@b60-ccide-10-099] Differential diagnosis of Class III malocclusions and anterior crossbite in a step by step approach has been presented here based on the work of authors Ngan et al,[@b6-ccide-10-099] Battagel,[@b61-ccide-10-099] and Turley.[@b62-ccide-10-099] History of the mandibular prognathism or anterior crossbite ----------------------------------------------------------- The first question that should be asked to the patient or parents (caregiver) should concern mandibular prognathism or anterior crossbite in the family and close relatives. A history of it indicates a genetic cause of Class III malocclusion. Examination of any functional shift ----------------------------------- The second most important step is to access the relationship of maxilla and mandible in CR and CO to determine any functional shift. Molar and incisor relationship in this group of malocclusion is one of the most important diagnostic criteria. Class III malocclusion with positive overjet or edge-to-edge incisors relation accompanied by lingually inclined mandibular incisors represents a compensated form of Class III malocclusion. Anterior repositioning of the mandible may be due to abnormal tooth contact in CR that forces the mandible forward in CO.[@b63-ccide-10-099] These pseudo Class III malocclusion individuals have a Class I skeletal pattern, orthognathic profile, and Class I molar relation in CR, but a Class III skeletal and dental pattern in CO.[@b64-ccide-10-099] Elimination of the CO or CR discrepancy should show whether it is a simple Class I malocclusion or a compensated Class III malocclusion.[@b65-ccide-10-099] In these individuals early correction proves to be a favorable environment for future growth. Clinical assessment ------------------- This should be carried out while the patient is sitting upright in natural head position (NHP) to evaluate the sagittal and vertical facial proportions. The profile of the patient should be evaluated in NHP using "a line down from the bridge of the nose to the base of the upper lip and a second one extending from that point downward to the chin. A straight or concave profile in young patients indicates a skeletal Class III jaw relationship."[@b66-ccide-10-099] Similarly, the transverse dimension should be assessed to evaluate any facial or dental asymmetries. Examination of the temporomandibular joint, oral musculature, and intraoral soft and hard tissue should also be performed. Evaluation of the anterior crossbite aimed to differentiate a true Class III malocclusion from a pseudo Class III malocclusion has been described by Ngan et al,[@b60-ccide-10-099],[@b65-ccide-10-099] and the given diagnostic scheme can be adapted ([Figure 1](#f1-ccide-10-099){ref-type="fig"}). Lateral cephalogram analysis ---------------------------- Best analyses for cephalometric assessment of Class III malocclusion are those that correlate the maxilla to the mandible and each of them to the anterior cranial base. These are ANB (2°), Wits (0 mm), maxillomandibular differential (linear measurement from condylion to point A and condylion to gnathion: 23 mm for 12 years old), nasion perpendicular to point A (+2.3 mm), and nasion perpendicular pogonion (0 mm). Values of each of the cephalometric measurements presented here are for normal individuals. It has been found in discriminant analysis that "Wits" appraisal is the most important factor in the decision making from orthodontic camouflage treatment to surgical options.[@b66-ccide-10-099] A "Wits" appraisal from 0 to −5 mm may be suggestive of a Class III problem being resolved by means of orthodontic camouflage treatment with facemask or chin cup therapy.[@b67-ccide-10-099] A "Wits" appraisal between −4 and −12 mm requires further growth treatment response vector (GTRV) analysis using serial cephalometric radiographs before a decision can be made of whether to camouflage or wait for comprehensive growth before surgical treatment.[@b68-ccide-10-099] Growth trends of Class III malocclusion ======================================= For effective treatment planning and realistic prediction of stability of treatment outcomes in Class III growing patients, knowledge of growth trends is of paramount importance. Concern regarding early treatment and the need for interceptive care in the case of Class III treatment has always been a dilemma for clinicians as well as patients, knowing that not all problems will be solved in these cases until growth is further completed, and the outcomes in long term irrespective of treatment approaches may depend on the growth tendency of an individual. Turpin[@b69-ccide-10-099] in his editorial published in the American Journal of Orthodontics and Dentofacial Orthopaedics has emphasized that a deep understanding of the individual patient's growth weighs much more than the treatment timing and treatment mechanics used on a specific Class III malocclusion growing patient whose timing of treatment is considered to be controversial. In essence, he summarized the studies of Mitani et al[@b70-ccide-10-099] on Class III growth assessment over the years, as "the basic pattern of mandibular prognathism is established before puberty and does not change fundamentally. However, their total growth increments were about the same as those with a normal mandible after the pubertal growth peak." The points raised by Turpin suggest that in spite of a controversial treatment timing the clinician should continue to have an interest in early interceptive treatment of Class III malocclusion as and when required. Assessment of growth and prognosis in Class III patients ======================================================== The prognosis of orthopedic treatment for skeletal Class III malocclusion is favorable when treatment is administered before the pubertal growth peak.[@b71-ccide-10-099]--[@b74-ccide-10-099] However, a Class III malocclusion may worsen due to growth if a patient is left untreated. Therefore, early treatment is recommended for skeletal Class III malocclusion to obtain a balanced skeletal relationship and, by doing this, there is a possibility to minimize the need of further future complicated treatment such as orthognathic surgery.[@b71-ccide-10-099]--[@b75-ccide-10-099] Studies have suggested that the posttreatment outcome of orthopedic treatment may not be stable depending on the residual growth.[@b76-ccide-10-099] When unfavorable growth is expected, treatment would not be initiated in the early phase or would be delayed until completion of growth, because a discrepancy between maxillary, mandibular, and skeletal base growth during the pubertal phase can result in relapse of the corrections which have been achieved earlier,[@b77-ccide-10-099] and as a result some patients ultimately may require orthognathic surgery at a later stage. This is a disappointing situation, not only for patients but also for clinicians, and may be prevented if accurate prediction of the eventual prognosis of early orthopedic treatment for skeletal Class III malocclusion would have been possible before commencing and executing treatment. As a result, patient selection and decision making regarding reliability of decisions and the timing of treatment would be much easier.[@b78-ccide-10-099] Thus, the million dollar question which arises here is: Is the prediction of growth of mandible in growing Class III patient possible with optimal accuracy and reliability? Several investigators have taken this challenge and attempted to predict the prognosis of Class III malocclusion based on the evaluation of the patient's single cephalogram for morphological characteristics and cephalometric analysis.[@b79-ccide-10-099]--[@b83-ccide-10-099] Björk,[@b79-ccide-10-099] based on single cephalogram analysis, defined seven structural signs of extreme growth rotation of mandible in the early developmental stage. These seven signs are "the inclination of the condylar head, curvature of the mandibular canal, shape of the lower border of the mandible, width of the symphysis, interincisal angle, intermolar angle, and anterior lower face height." Even though seven structural signs of growth rotation have been reported,[@b79-ccide-10-099] growth prediction of the mandible still remains a challenge, probably because mandibular growth shows wide variations in terms of amount, direction, and time or because the prediction procedure is based on individual reasoning. Several cephalometric variables have been identified based on mandibular morphology by various studies to predict the results of early treatments.[@b77-ccide-10-099],[@b80-ccide-10-099]--[@b83-ccide-10-099] These studies reported long-term success rates of 50.0%--71.4% for orthopedic treatment of skeletal Class III malocclusion. Because of treatment success criteria, patient characteristics, and time points of outcome evaluation differed among studies; the reported success rates ranged widely and cannot be generalized to all patients treated for skeletal Class III malocclusion. In a recent study, Choi et al[@b78-ccide-10-099] evaluated the long-term success of orthopedic treatment in growing skeletal Class III malocclusions, and verified previously reported success rates and prediction models. This research group found the same results published by previous researchers and concluded that no particular method or factor can predict the long-term success of orthopedic treatment for growing skeletal Class III malocclusions. Assessment of direction and magnitude of mandibular and maxillary growth was proposed by Musich by using a serial cephalograms (D Musich, Growth treatment response vector analysis, personal communication, November 1, 2001). He proposed the GTRV analysis to predict the possibility of excessive mandibular growth after early interceptive orthopedic treatment in Class III patients. Based on immense clinical and extensive research experience in Class III malocclusion, Ngan described the use of serial cephalometric radiographs for GTRV analysis to predict excessive mandibular growth.[@b84-ccide-10-099]--[@b86-ccide-10-099] GTRV ==== According to Ngan,[@b84-ccide-10-099]--[@b86-ccide-10-099] "the horizontal growth changes of the maxilla and mandible are determined by locating the A and B points on the posttreatment lateral cephalometric radiograph. Another cephalogram should be taken during 2--4 years of follow-up visits after the treatment of Class III malocclusion with facemask therapy." The incisal tip of maxillary incisor and mesiobuccal cusp tip of maxillary molar are used as a landmark to construct the occlusal plane (O). Point A and point B should be marked on the lateral cephalogram and connecting points A and B perpendicular to the occlusal plane would construct the lines AO and BO. Using stable landmarks on the midsagittal cranial structure, the nasion (N), and sella turcica (S), post-facemask treatment tracing (first tracing) is superimposed on the follow-up radiograph and the lines AO and BO are again constructed on the follow-up radiograph on the occlusal plane of the first tracing. The distance between the A points and point B of the two tracings along the occlusal plane represents the growth changes of the maxilla and mandible, respectively ([Figure 2](#f2-ccide-10-099){ref-type="fig"}). After the measurements are completed, the GTRV ratio for each patient can be determined by using the formula: GTRV = horizontal growth changes of the maxilla/horizontal growth changes of the mandible. Hence, GTRV is defined as "the horizontal growth changes at A point divided by the horizontal growth changes at B point on the post-facemask and follow-up lateral cephalogram." For an individual with the age range of 6--16 and with normal growth pattern, the GTRV ratio is 0.77. This suggests that the horizontal growth of mandible exceeds 23% as compared to the maxilla to maintain a normal skeletal relationship. In studies by Ngan[@b86-ccide-10-099] and Youssef et al,[@b87-ccide-10-099] in patients treated with early interceptive orthopedic treatment, the GTRV ratios were significantly different for successful cases and unsuccessful cases. For successful cases, GTRV was in the range of 0.33--0.88 (mean 0.45) and for unsuccessful cases it was in the range of 0.06--0.38 (mean 0.22). This suggests that subjects with mild to moderate Class III malocclusion can be camouflaged orthodontically after successful early interceptive treatment, if the GTRV ratio ranges between 0.33 and 0.88. Class III patients with a GTRV ratio between 0.33 and 0.38 can be considered as borderline cases which can either be treated successfully with facemask or turn out to be surgical cases at the end and would better be warned of future need for surgical intervention, if the GTRV ratio is \<0.38. Hence, the GTRV analysis as suggested by Ngan[@b84-ccide-10-099]--[@b86-ccide-10-099] is helpful to the clinicians to assess the growth of the mandible to predict the prognosis after early interceptive orthopedic treatment of Class III malocclusion. Treatment timing ================ It is an accepted fact that skeletal Class III malocclusion establishes itself early in life, is not a self-correcting disharmony,[@b70-ccide-10-099],[@b88-ccide-10-099] and is often associated with maxillary constriction. Intervention at an early stage, such as deciduous dentition, or prepubertal growth phase has been recommended.[@b73-ccide-10-099],[@b89-ccide-10-099] In particular, the prepubertal treatment of Class III malocclusion by means of rapid palatal expansion and facemask protraction yields favorable growth corrections both in maxilla and in the mandible.[@b73-ccide-10-099] In a controlled long-term study, it has been found that patients who have been treated before the pubertal growth phase showed a stable increment in the maxillary skeletal width, maxillary intermolar width, and lateronasal width, while patients treated after the pubertal growth phase showed only dentoalveolar effects after the follow-up of \~8 years.[@b90-ccide-10-099] The proper timing of interventions may therefore rely on chronological age[@b91-ccide-10-099] and phases of dentition[@b92-ccide-10-099] for very young patients, and on other radiological indicators, such as cervical vertebral maturation and/or hand and wrist maturation methods for older children.[@b93-ccide-10-099] A summary of the timing of interventions and main indicators for Class III malocclusion has been presented in [Table 3](#t3-ccide-10-099){ref-type="table"}. Treatment of developing Class III malocclusion ============================================== It is very critical to make a decision for developing Class III malocclusion on whether to treat or wait for further growth and dental development. Although a Class III malocclusion may be identified in the developing dentition, a decision needs to be made as to whether it is better to treat it at this stage or wait for further dental development and growth. The timing of early treatment is crucial for a successful outcome. Some studies have reported that treatment should be carried out in patients \<10 years of age to enhance the orthopedic effect.[@b71-ccide-10-099],[@b94-ccide-10-099]--[@b96-ccide-10-099] In contrast, some studies found that age of the patient had little influence on treatment response and outcome.[@b97-ccide-10-099],[@b98-ccide-10-099] There is no strong evidence to support the notion that early treatment would be beneficial. The main goals of early intervention are to create a more favorable environment for growth and to improve the occlusal relationship: for example, correcting the crossbite and facial esthetics.[@b71-ccide-10-099] Hence, interceptive treatment of Class III malocclusions should be undertaken if it prevents damage to the oral tissues and prevents or significantly reduces the amount, or severity, of future orthodontic and surgical intervention. Turpin[@b99-ccide-10-099] developed a list of positive and negative factors that helped decision making on developing Class III malocclusions ([Table 4](#t4-ccide-10-099){ref-type="table"}) and these guidelines were reviewed by Campbell for deciding when to intercept Class III malocclusion.[@b71-ccide-10-099] Turpin suggested that early interceptive treatment of Class III malocclusion should be considered for patients who presented with positive characteristics as mentioned. The author also recommended that individuals with negative characteristics should delay treatment until the completion of growth. He also suggested that patients should be warned that surgery may be needed in future, even after an early successful interceptive treatment. Common cephalometric predictors for successful Class III camouflage for the evaluation of the maxillary and mandibular position include: ANB (\<−2° to −3°);Wits appraisal (−2 to −6 mm for nonsurgical treatment, −6 to −9 mm for a compromised orthodontic result);maxillomandibular differential and gonial angle within the normal range. The most important factor is clinical assessment to evaluate the need to optimize the facial esthetic. The combination of clinical and cephalometric information will identify which type of Class III malocclusion can be treated in the mixed dentition and help in deciding the best interceptive approach. A Class III patient with mild to moderate Class III skeletal patterns with a GTRV ratio between 0.33 and 0.88 can be successfully camouflaged orthodontically later after initial interceptive treatment in mixed dentition, and a GTRV ratio \<0.38 should be warned. However, the GTRV ratio requires a serial lateral cephalogram to be obtained; so, this tool may not be suitable for decision making for new patients.[@b84-ccide-10-099] In mixed dentition from the treatment point of view, there are effectively three types of Class III malocclusions:[@b100-ccide-10-099],[@b101-ccide-10-099] dental: incorrect inclination or position of maxillary or mandibular incisors;pseudo: anterior positioning of the mandible as a result of premature dental contacts deflecting the mandible anteriorly to allow the patient to achieve full intercuspation;skeletal: true skeletal discrepancies in the maxilla and/or mandible. Treatment of simple dentoalveolar anterior crossbites ----------------------------------------------------- A simple anterior crossbite can be corrected with either a removable appliance or a fixed appliance. The percentage of success is increased if there is a minimal existing proclination of the upper incisors and adequate overbite to maintain correction at the end of treatment. Types and indications of both types of appliances are provided below. Removable appliances: there are three types of removable appliances used to intercept the developing crossbite: inclined plane;modified inclined plane;active Hawley appliance.Fixed appliance: 2 by 4 appliance. Inclined plane -------------- An inclined plane is a good treatment choice in deciduous dentition or early mixed dentition.[@b102-ccide-10-099] It is indicated in patients with: retroclined maxillary anterior teeth with an anterior crossbite with or without functional shift;well-aligned mandibular anterior teeth without proclination;normal to deep overbite;average to horizontal growth pattern patients. The inclined plane is fixed onto the lower anterior teeth with temporary cement. Appropriate angulation between the inclined plane and the upper anterior teeth in crossbite should be determined by considering the vertical discrepancy between the teeth in crossbite and the adjacent teeth, as well as the degree of overbite of the teeth in crossbite by adjusting the different contact angulations. Most anterior dental crossbites can be corrected within 3--4 weeks using an inclined plane. Modified inclined plane ----------------------- Modified inclined plane is indicated in patients with an anterior dental crossbite, with lingually inclined maxillary incisors and labially inclined mandibular incisors; for this reason, an inclined plane is contraindicated because it may cause mandibular incisors proclination.[@b103-ccide-10-099],[@b104-ccide-10-099] A modified inclined plane is similar to that of a Hawley appliance plus an inclined plane on the anterior part. The inclined plane portion covers the lower anterior teeth up to their incisal third. When the patient bites, the inclined plane portion raises the bite and proclines the upper anterior teeth labially. The metal wire parts consist of a labial bow and Adam's clasps on the first permanent molars for better appliance retention and stability. The labial bow is placed on the labial side of the lower incisors, near the cervical third, to ensure that the force application is closer to the center of resistance of the lower anterior teeth. The acrylic resin lingual to the lower incisors may be trimmed to allow lingual movement of the lower anterior teeth when the labial bow is activated. Anterior crossbite problems can be corrected within 3--4 weeks using a modified inclined plane. After the correction of anterior crossbite, the same appliance may be used as retainer by trimming of the anterior inclined plane portion. Active Hawley appliance ----------------------- A removable Hawley appliance is indicated in patients with anterior crossbite in mixed dentition stage.[@b100-ccide-10-099] This appliance has an active component anteriorly to procline the upper anterior tooth or teeth to correct the anterior crossbite. This active component can either be a palatal "Z" spring, which is activated by the clinician, or a screw, which the patient activates. Usage of the Jack expansion screw is recommended to procline maxillary anterior teeth at the rate of one turn at every third day till the correction of anterior crossbite. The turning of the screw by a quarter turn (90°) brings about 0.18 mm of linear movement depending upon the pitch of the screw. Correction of 2 mm of overjet with this appliance can be achieved in 5--6 weeks. The appliance also incorporates retentive components (preferably Adam's clasp) to keep the appliance in place and possibly posterior bite plate to disclude the occlusion to aid in uninterrupted proclination of anterior teeth. A removable appliance can only tip the teeth; so it should be only used if simple tipping movements of the upper anterior teeth are required. Fixed appliance --------------- This appliance is often referred to as a "2 by 4" or "2 by 6" appliance as it is only bonded on the two upper first permanent molars and the four upper incisors or six anterior teeth. Fixed appliances mostly use late mixed dentition or early permanent dentition. An open coil NiTi spring often compressed between the molars and the incisors to procline the incisors or a 0.016-inch stainless steel stoppered arch wire may be used to increase the arch length. Glass ionomer cement may be placed temporarily as a posterior fixed bite plane on the molars if disclusion is required. Fixed appliances allow tipping, bodily movement, and correction of rotations as and when required. It has been seen that both types of appliances work well and the results are equally stable. Fixed appliance treatment is quicker and cheaper and has less effect on the patient's speech than a removable appliance, but patients may complain of slightly more difficulty in chewing and biting initially with the fixed appliance.[@b101-ccide-10-099]--[@b107-ccide-10-099] Growth modification and orthopedic treatment -------------------------------------------- Growth modification in developing Class III malocclusion is indicated in patients with skeletal discrepancy. The basic aim of this interceptive treatment for developing Class III malocclusion is to improve or correct the skeletal discrepancy to allow future treatment of such patients by orthodontic camouflage only without the need of orthognathic surgery. This approach of growth modification in Class III patients can be achieved through functional appliances, chin cup therapy, protraction facemask, and bone-anchored appliances. A brief description of each modality based on current evidence has been given below. Functional appliances --------------------- Functional appliances have been used to modify the skeletal pattern by enhancing the growth of the maxilla and restricting or redirecting the growth of the mandible. Two commonly used functional appliances to intercept Class III malocclusion are Frankel functional regulator III appliance (FR III) and reverse twin-block appliance. FR III has maxillary vestibular shields in the depth of the sulcus. These shields are placed away from the maxilla to stretch the periosteum and encourage anterior development of the maxilla. The lower part of the appliance attempts to restrict mandibular growth or redirect it posteriorly. In the reverse twin-block appliance, the blocks are positioned so that there are posterior forces on the mandible and anterior forces on the maxilla. Current research suggests that functional appliances can improve occlusal relationships, but this is principally due to dentoalveolar changes, proclining upper incisors, and retroclining the lowers incisors.[@b108-ccide-10-099]--[@b110-ccide-10-099] Evidence from a recent systematic review suggests that the FR III might restrict mandibular growth but not stimulate forward movement of the maxilla.[@b111-ccide-10-099] Current evidence suggests that functional appliances can successfully correct a developing Class III malocclusion, but they have principally dentoalveolar effects, with minimal or no effects on the underlying skeletal pattern. Functional appliances especially FR III can be challenging to wear in the mouth and are subject to breakage; hence, a simpler method like orthodontic camouflage may be used in place of functional appliances. Chin cup -------- Chin cup appliance treatment is indicated in young growing patients with mandibular prognathism. It has been found that chin cup therapy does not restrain mandibular growth but redirects the mandible growth vertically, causing a backward rotation of the mandible.[@b112-ccide-10-099] These changes in the direction of mandibular growth help to improve Class III malocclusion. Recent systematic reviews showed that there is considerable agreement between studies in that chin cup therapy may be used for interceptive treatment of growing Class III malocclusion based on short-term favorable results.[@b113-ccide-10-099],[@b114-ccide-10-099] It has been seen that these changes are not maintained in the long term and the normal growth pattern of the mandible reestablishes itself, if chin cup appliance therapy is discontinued before growth completion.[@b115-ccide-10-099] Hence, it is recommended that patients with Class III malocclusion with mandibular prognathism wear the chin cup appliance until growth is completed to maintain the treatment effects of chin cup therapy. Special care should be taken while deciding chin cup therapy in patients who present in the mixed dentition with marked mandibular prognathism, particularly if associated with increased vertical proportions, as these patients are often best treated by surgical orthognathic approach, when their growth is completed. As the long-term prognosis of chin cup therapy is unpredictable, patient caregivers should always be fully informed of this before initiating chin cup therapy. Line of force of applied force for chin cup therapy should be directed along the lines from the chin point to the condyle heads, bilaterally in the range of 400--500 g, 10--14 hours per day. While trying the chin cup appliance on patients, care should be taken to ensure that the chin cup does not impinge on the lower lips as it may cause retroclination of the lower incisors and recession of labial gingiva ([Figure 3](#f3-ccide-10-099){ref-type="fig"}). Protraction facemask -------------------- Protraction facemask also referred to as reverse headgear is one of the most commonly used interceptive tools to intercept developing skeletal Class III malocclusion.[@b116-ccide-10-099] The appliance is composed of two components: an extraoral framework (facemask) that fits on the forehead and chin, and an intraoral attachment to the maxillary dentition ([Figure 4](#f4-ccide-10-099){ref-type="fig"}). The chin and forehead part of the extraoral framework are connected by a middle bar for the connection of the elastics to the intraoral attachment to the maxillary dentition. The intraoral attachment is of various designs, including removable, banded, and acrylic-bonded versions. They all incorporate hooks bilaterally positioned near the maxillary canines. To minimize unwanted rotation of the palatal plane, Class III elastics should be attached near the maxillary canines at 30° to the occlusal plane. Bonded expansion appliance is preferred as it provides a temporary bite plane effect in hyperdivergent cases and facilitates the jumping of anterior crossbite in deep bite cases. The elastic forces are typically 400--450 g per side (14 to 16 OZ) and need to be worn 12--14 hours per day. The total treatment time is usually 6--9 months. An increased release of growth hormone and other growth promoting endocrine factors has been observed during evening and night than during the day. As a result, it is recommended to wear the appliance during evening and nighttime. One controversial area is the use of rapid maxillary expansion (RME) at the same time as the protraction facemask. Often patients with a Class III skeletal pattern have a constricted maxilla in the transverse dimension as well as the anteroposterior dimension; so this expansion is a helpful component of the treatment. It has been suggested that expansion may loosen the circummaxillary sutures and increase forward movement of the maxilla. Results of a recent randomized controlled trial (RCT)[@b117-ccide-10-099] and data from meta-analysis support the notion that facemask with and without RME treatment are both equally effective clinically in early treatment methods for skeletal Class III malocclusion. Hence, in case of no transverse discrepancy facemask without RME treatment can be undertaken. Maxillary expansion and protraction treatment for early permanent dentition is undertaken in case of transverse maxillary constriction.[@b118-ccide-10-099] The expansion appliance is activated twice daily (0.25 mm per turn) for 7--10 days. In case of severely constricted maxilla, activation of the screw is carried out for ≥2 weeks. In a randomized controlled clinical trial (CCT) comparing protraction facemask with no treatment, successful correction of the reverse overjet was observed in 70% of patients, with an average increase in overjet of 4 mm, and significant skeletal changes, principally due to forward movement of the maxilla and improvement in the ANB angle of 2.6° compared to the control at the end of treatment.[@b119-ccide-10-099] These patients were followed up 6 years later to see if favorable changes were maintained toward the end of growth and in particular to assess whether the interceptive use of a facemask in the developing dentition can help reduce the need for orthognathic surgery.[@b120-ccide-10-099] Of the patients that wore a protraction facemask, 36% needed orthognathic surgery at the age of 15, whereas 66% of patients in the control group required orthognathic surgery. Encouragingly, 68% of patients who wore the protraction facemask had a positive overjet after 6 years. To conclude, it appears that the use of a protraction facemask in the developing dentition will correct Class III malocclusion and reduce the need for orthognathic surgery in the future in the following types of cases: child under the age of 10;mild--moderate Class III;retrusive maxilla;average or reduced vertical proportions. As the mandible resumes its original downward and forward growth after Phase I (facemask) therapy, patients may have reduced corrected overjet and more Class III molar relation during follow up and the second phase of orthodontic camouflage. Hence, the use of an extraoral orthopedic appliance (eg, chin cup) is recommended after the first phase of interceptive therapy with facemask till the completion of mandibular growth. It is also recommended to take the serial posttreatment cephalogram to calculate the GTRV to make a decision on whether to opt for orthodontic camouflage or wait till the completion of growth. A recent review published in 2017, in evidence-based dentistry, by Smyth and Ryan assessed early treatment of Class III malocclusion with facemask therapy and included randomized clinical trials and controlled clinical trials in children aged 7--12 years undergoing fixed or removable orthodontic treatment with the primary outcome of correction of the reverse overjet. The authors concluded that there is a moderate amount of evidence to show that early treatment with a facemask appliance has positive improvement for both skeletal and dental effects in the short term.[@b121-ccide-10-099] Bone-anchored appliances ------------------------ Interceptive treatment of Class III malocclusions with a tooth-borne protraction appliance (eg, facemask) often poses problems of unwanted dental changes such as: buccal tilting of maxillary molars and extrusion may lead to an increase in vertical dimensions and downward and backward growth of the mandible;decreased arch length due to mesial movement of maxillary molars leading to crowding in the anterior teeth. In an attempt to overcome the limitations of tooth-borne appliances in the interceptive treatment of Class III malocclusions, bone-anchored maxillary protraction (BAMP) appliances have recently been used.[@b122-ccide-10-099] BAMP appliances typically involve the use of Class III elastics attached between miniplates placed in the infrazygomatic crest to miniplates placed in the mandibular symphysis region or attached to the extraoral facemask ([Figure 5](#f5-ccide-10-099){ref-type="fig"}). The success of these miniplates is related to the surgical technique and the thickness and quality of the bone. Particularly in the maxilla, the bone quality is often not as good until the patient is at least 11 years old; so this interceptive technique tends to be used in slightly older patients than the tooth-borne appliances. The results of an initial study on the effects of BAMP compared with growth of the untreated Class III subjects showed that the BAMP protocol induced an average increment on skeletal and soft tissue advancement of maxillary structures of about 4 mm with negligible changes in the maxillary incisor inclination and vertical skeletal pattern.[@b123-ccide-10-099] Recent research also found that a Hybrid Hyrax bone-anchored rapid palatal expansion appliance minimized the side effect encounter by tooth-borne rapid palatal expansion appliances for maxillary expansion and protraction and may serve as an alternative treatment appliance for correcting Class III patients with a hyperdivergent growth pattern.[@b124-ccide-10-099] Hence, BAMP has demonstrated promising initial results in its potential to offer greater skeletal changes, with less unwanted displacement of the dentition. However, there are unpredictable variations in individual outcomes, and high-quality research is needed to further investigate this technique. Summary ======= For accurate diagnosis and successful execution of orthodontic treatment, it is very important that children with developing Class III malocclusion must be evaluated for family history, and should undergo dental examination for molar and incisor relationships, functional assessment to evaluate CO or CR shift on mandibular closure, and cephalogram analysis to determine sagittal jaw discrepancy. Evaluation of GTRV to determine the individual growth direction and rate should always be performed in borderline cases. A summary of diagnosis and treatments of developing Class III malocclusions is presented and [Figure 1](#f1-ccide-10-099){ref-type="fig"} may be used for quick reference. Interceptive treatment of Class III malocclusions may be undertaken if it prevents damage to the oral tissues and/or prevents or significantly reduces the amount, or severity, of future orthodontic treatment.Simple anterior dental crossbites can be successfully corrected with removable or fixed appliances in mixed dentition.Treatment with chin cup or functional appliances can correct a Class III incisor relationship, but any orthopedic changes are likely to be minimal with these appliances.Optimal timing for interceptive treatment with face-mask is in the deciduous or early mixed dentition stage. Early interceptive treatment with a facemask allows for favorable sutural response for maxillary expansion and protraction; correction of any CO or CR discrepancies; and improvement in the facial profile and self-esteem in patients who are under 10, with mild to moderate Class III and a retrusive maxilla and with average or reduced vertical proportions. In case of mandibular prognathism, after facemask therapy patients should be advised to the wear the chin cup appliance until growth completion. A follow-up lateral cephalogram should be taken at 2--4 years after maxillary protraction to calculate the GTRV ratio. Information obtained from the GTRV ratio and from the cephalogram for the mandibular growth rate and direction is helpful in deciding whether Class III malocclusion can be camouflaged orthodontically or surgical intervention would be needed in the future. A case of Class III malocclusion treated successfully with facemask therapy has been presented in [Figures 6](#f6-ccide-10-099){ref-type="fig"}[](#f7-ccide-10-099){ref-type="fig"}[](#f8-ccide-10-099){ref-type="fig"}[](#f9-ccide-10-099){ref-type="fig"}[](#f10-ccide-10-099){ref-type="fig"}[](#f11-ccide-10-099){ref-type="fig"}[](#f12-ccide-10-099){ref-type="fig"}--[13](#f13-ccide-10-099){ref-type="fig"}.Bone-anchored appliances may offer the potential for more skeletal changes, but before its routine use more evidence is needed. Furthermore, bone-anchored appliances require help from surgeons. Written informed consent was obtained from the parents or legal guardians of the children, to have the images published. The authors are grateful to Professor OP Kharbanda, Chief of the Centre for Dental Education and Research, All India Institute of Medical Sciences (AIIMS), New Delhi for all clinical facilities, Dr Pradip Sangroula, Ex-Orthodontic resident, AIIMS for [Figure 5](#f5-ccide-10-099){ref-type="fig"}, and Mr Netrapal, Dental Technician, AIIMS for scanning of models. **Disclosure** The authors report no conflicts of interest in this work. ![Summary of clinical practice guidelines for developing Class III malocclusion.\ **Abbreviations:** CR, centric relation; CO, centric occlusion; GTRV, growth treatment response vector.](ccide-10-099Fig1){#f1-ccide-10-099} ![Horizontal growth changes of the maxilla and the mandible between the posttreatment and the follow-up cephalograms: **(A)** pretreatment, **(B)** post-facemask, **(C)** follow-up cephalogram, and **(D)** GTRV calculation on post-facemask and follow-up cephalogram.\ **Abbreviation:** GTRV, growth treatment response vector.](ccide-10-099Fig2){#f2-ccide-10-099} ![Chin cup appliance with line of force passing through the condyle.](ccide-10-099Fig3){#f3-ccide-10-099} ![Facemask therapy. **(A)** Occlusal view showing maxillary splint; **(B)** frontal view showing hooks for Class III elastics; and **(C)** facemask attached to maxillary splint through Class III elastics for maxillary protraction.](ccide-10-099Fig4){#f4-ccide-10-099} ![Bone-anchored maxillary protraction: **(A)** placement of miniplate in infrazygomatic crest; **(B)** miniplate in situ; **(C)** OPG showing miniplates; and **(D)** facemask attached to miniplates through Class III elastics for maxillary protraction.\ **Abbreviation:** OPG, orthopantomogram.](ccide-10-099Fig5){#f5-ccide-10-099} ![Pretreatment photographs showing Class III malocclusion.](ccide-10-099Fig6){#f6-ccide-10-099} ![Pretreatment digital study models showing Class III malocclusion.](ccide-10-099Fig7){#f7-ccide-10-099} ![Maxillary splint with Hyrax expander cemented for facemask therapy.](ccide-10-099Fig8){#f8-ccide-10-099} ![Posttreatment photographs.](ccide-10-099Fig9){#f9-ccide-10-099} ![Posttreatment digital study models.](ccide-10-099Fig10){#f10-ccide-10-099} ![Pretreatment and posttreatment cephalometric and panoramic radiographs.](ccide-10-099Fig11){#f11-ccide-10-099} ![Pretreatment and posttreatment superimposed occlusal view of maxillary and mandibular models showing treatment changes (green color showing pretreatment and rustic brown color showing posttreatment).](ccide-10-099Fig12){#f12-ccide-10-099} ![Posttreatment chin cup appliance therapy for mandibular growth redirection.](ccide-10-099Fig13){#f13-ccide-10-099} ###### Reported prevalence of Class III malocclusion globally and in different continents Continents Prevalence (%) ------------------- ---------------- Globally 0--26.7 East Asian 4--14 Southeast Asian 15.80 African 4.59 Middle Eastern 10.18 Indian 1.19 European 4.88 (2--6) Northern European 0.8--4.2 American 5 ###### Reported prevalence of Class III malocclusion among different nationality groups Nationality Prevalence (%) -------------------- ---------------- Chinese 15.69 Malaysian 16.59 Japanese 2.3--14 Korean 9--19 Taiwanese 1.65 Indian 0--4.76 Israeli Arabs 1.30 Iranians 15.20 Turkish 10.30 Egyptians 11.38 Tanzanian 1.81--19.72 Nigerian 1.22--11.79 United Kingdom 3--5 Scandinavian 3--5 Swedish 6 Brazilian 3--5 Saudi Arabian 9.40 African Americans 3--6 European Americans 0.80 Latino Americans 9.10 Mexican Americans 8.30 Italians 5 German 2.80 Belgian 6 British 2.9 Danish 4.30 Polynesian 5.50 Syrian 14.0 Lebanese 5.10 ###### Optimum timing of interventions and indicators for Class III malocclusion --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Problem Optimum timing of intervention Main indicators Treatment modality ------------------------------------------------------------ -------------------------------- ------------------------------------------------------ ---------------------------------------------- Class III malocclusion with or without constricted maxilla Pubertal growth phase Chronological age (up to 8 Y -- girls; 9 Y -- boys)\ Facemask with or without maxillary expansion Phases of dentition (up to mixed dentition)\ CVM (CS1--CS2)\ HWM (SMI 1--SMI 2) --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- **Abbreviations:** CS, cervical stage; SMI, skeletal maturation indicator; CVM, cervical vertebral maturation; HWM, hand and wrist maturation. ###### Turpin's positive and negative factors for decision making for interception of developing Class III malocclusion Positive factors Negative factors -------------------------------------------------------------- -------------------------------- Convergent facial type Divergent facial type Anteroposterior functional shift No anteroposterior shift Symmetrical condyle growth Asymmetrical growth of condyle Young subject with remaining growth Mild skeletal disharmony Growth completed Severe skeletal disharmony Good cooperation expected Poor cooperation expected No familial prognathism Familial pattern established Good facial esthetics Poor facial esthetics [^1]: These authors contributed equally to this work

{ "pile_set_name": "PubMed Central" }

1. Introduction =============== Pulmonary embolism (PE) is the 3rd cause of cardiovascular death worldwide. More than 600,000 patients suffering from PE are admitted to the hospital every year in the United States but \<5% of them die within 3 months.^\[[@R1]--[@R5]\]^ Based on these epidemiologic data, experts emphasize the importance of early stratification to identify patients with PE at high risk of death, who may benefit from more intensive monitoring or aggressive therapy.^\[[@R6],[@R7]\]^ Recent international guidelines defined "high risk" population in the presence of persistent arterial hypotension (\<90 mm Hg) or shock. However, hemodynamic parameters defining shock status in the context of PE remain unclear and their correlation to outcome has never been evaluated. In critically ill patients admitted in the intensive care unit (ICU) for septic shock or myocardial infarction-related cardiogenic shock, several groups have reported that alterations of microcirculation blood flow are strongly predictive of mortality independently of arterial blood pressure or cardiac output.^\[[@R8]--[@R10]\]^ Microcirculatory perfusion could be evaluated using computer-assisted devices but these tools are not available at bedside. Tissue perfusion could also be assessed with more simple tools including arterial lactate level or noninvasive clinical parameters such as skin mottling or urinary output. In our department, we are interested in clinical evaluation of organ perfusion and we have showed that skin mottling, reflecting skin hypoperfusion, correlated with organ failure and predicted mortality in sepsis^\[[@R11]\]^ and septic shock patients.^\[[@R12]\]^ In this study, we aimed at investigating the relationship between 28-day mortality and tissue perfusion parameters, that is, arterial lactate level, skin mottling and urinary output, in patients admitted to ICU for acute PE. 2. Materials and methods ======================== We conducted a retrospective observational study in a 16-bed ICU in a tertiary teaching hospital. All consecutive patients older than 18 years admitted for acute PE from January 1993 to December 2015 were included. Patients were identified by querying the electronic health records with the following keywords (in French): "pulmonary embolism," "acute respiratory failure," "anti-coagulant," "deep venous thrombosis," "cardiac arrest," "heparin," "vitamin K antagonists," "cava filter," or "thrombolysis." General characteristics of patients were recorded: demographic data, diagnoses, severity of illness evaluated by the sequential organ failure assessment (SOFA) score and simplified acute physiology score II (SAPS II). Global hemodynamic variables, right ventricle (RV) dilation and tissue perfusion parameters were collected at ICU admission. RV dilation was defined as a right to left ventricle diameter ratio superior or equal to 1 on echocardiography or computed tomography (CT) scan. We recorded 3 tissue perfusion parameters at admission: arterial lactate level, the presence/absence of skin mottling around the knee area, and mean urinary output (within the first 6 hours). An ethical approval was not necessary because it was a monocenter retrospective study using anonymous data. 2.1. Statistical analysis ------------------------- Data were summarized as median (25th--75th percentiles) for skewed distributions and percentages as appropriate. Association of tissue perfusion parameters with 28-day mortality was tested using Mann--Whitney nonparametric test for continuous variables, Fischer exact test, and the Chi-squared test for categorical data according to sample size. Finally, we used logistic regression for multivariable analysis. All statistical analyses were performed using the R software (v 2.12.0; [http://cran.r-project.org](http://cran.r-project.org/)). Significance was defined as 2-sided *P-*value \<.05. 3. Results ========== 3.1. Studied population ----------------------- Over a 22-year period, we identified 317 patients with PE admitted in our intensive care department by querying the electronic records. After reviewing medical charts, 122 patients were excluded because PE was not the main diagnosis for ICU admission and 87 patients were excluded because of missing data, leaving 108 patients for study (Supplemental Figure 1). PE was routinely diagnosed using nuclear planar V/Q-imaging during the 1993 to 2008 period, whereas CT pulmonary angiography became the diagnosis test of choice between 2008 and 2015. Echocardiography use in the context of acute severe PE also increased between the 2 time periods (Supplemental Figure 2). Characteristics of the studied population are summarized in Table [1](#T1){ref-type="table"}. The median age was 72 (61--81) years with no gender predominance. The median length of stay was 4 (2--7) days in the ICU and 14 (8--21) days in the hospital. At admission, median SOFA score was 2 (0--6) and median SAPS II was 29 (16--43). Thirty-nine patients (36%) required support organ therapy including vasopressor infusion (n = 30, 28%) and mechanical ventilation (n = 31, 29%). Thirty-seven patients (34%) received thrombolytic therapy. Day 28 mortality rate was 25% (n = 27). Nonsurvivor patients had higher organ failure severity and received more frequently both support therapy and thrombolytic treatment. ###### General characteristics of the studied population. ![](medi-97-e11993-g001) 3.2. Tissue perfusion parameters -------------------------------- We analyzed tissue perfusion parameters at ICU admission according to 28-day outcome. When compared to survivors, nonsurvivor patients had higher arterial lactate level (4.5 \[2.3--10.3\] mmol/L vs 1.4 \[1--2.9\] mmol/L, *P* \< .001), more frequent skin mottling around the knee area (56% vs 25%, *P* = .003) and lower mean 6-hour urinary output (0.35 \[0--1\] mL/kg/h vs 0.88 \[0.62--1.68\] mL/kg/h, *P* = .0002) (Table [2](#T2){ref-type="table"}). ###### Analysis of tissue perfusion parameters at admission according to 28-day outcome. ![](medi-97-e11993-g002) To assess the cumulative predictive value of tissue perfusion parameters, we analyzed the relationship between 28-day mortality and the number of tissue perfusion abnormalities at admission, that is, skin mottling, hyperlactatemia (\>2 mmol/L), and oliguria (\<0.5 mL/kg/h). Interestingly, 28-day mortality increased with the number of tissue hypoperfusion parameters at 8%, 21%, 28%, and 85% in the presence of 0, 1, 2, and 3 alterations, respectively (*P* \< .0001, Chi-squared for a trend test) (Fig. [1](#F1){ref-type="fig"}). ![Day 28 mortality according to the number of tissue perfusion alterations. Tissue perfusion alterations were defined as the presence of skin mottling around the knee area, arterial hyperlactatemia \>2 mmol/L and oliguria (mean 6-hour urinary output \<0.5 mL/kg/h). Chi-squared test for a trend.](medi-97-e11993-g003){#F1} In a multivariate analysis, the relationship between the number of tissue perfusion alterations and 28-day mortality was maintained after adjustment on the presence of shock and RV dilation at admission (Table [3](#T3){ref-type="table"}) ###### Multivariable logistic regression analysis of risk factors for 28-day mortality, including tissue perfusion alterations, shock, and right ventricular dilation. ![](medi-97-e11993-g004) 4. Discussion ============= In this study, we found that 3 tissue perfusion parameters measured at admission, arterial lactate level, skin mottling, and urinary output were synergistically associated with 28-day mortality in ICU patients admitted for PE, independently of global hemodynamic status and right ventricular dilation. In patients with PE, current guidelines recommend the evaluation of systolic blood pressure and RV injury to identify patients at high risk of death.^\[[@R7]\]^ However, both parameters fail to fully identify patients who would benefit from more aggressive treatment (thrombectomy or fibrinolysis) with a substantial reduction on mortality without a major incidence of hemorrhagic complications.^\[[@R13],[@R14]\]^ Recently, the pathophysiology of severe PE has been extended beyond simple RV obstruction. Several studies highlighted the role of inflammation^\[[@R15]\]^ and endothelial dysfunction^\[[@R16]\]^ in the alteration of organ perfusion at the microcirculation level in the context of acute circulatory failure related to infection^\[[@R17]\]^ or myocardial infarction.^\[[@R18]\]^ In this study, we evaluated microcirculatory hypoperfusion parameters available at bedside and found these "easy to use and easy to learn" tools are helpful to identify high-risk patients in acute PE. First, we investigated arterial lactate level the measurement of which is widely available in routine laboratories. Hyperlactatemia is a hallmark characteristic of shock states^\[[@R19]\]^ and predicts organ failure and mortality^\[[@R20]\]^ in the ICU, even more than refractory hypotension.^\[[@R21]\]^ In patients with PE, Vanni et al reported that arterial lactate level measured in the emergency department was significantly associated with in-hospital mortality independently of systolic blood pressure and right ventricular dysfunction.^\[[@R22],[@R23]\]^ Here we confirmed these results in critically ill patients admitted to the ICU. We found that nonsurvivor patients had significantly higher lactate levels when compared to survivors. Mortality rate was higher in our study when compared to Vanni\'s study probably because ICU patients had more frequent comorbidities and more severe organ failure. Next, we analyzed skin mottling, an old clinical sign, readily available at bedside. Skin mottling is characterized by locally reduced blood flow^\[[@R24]\]^ and decreased tissue oxygen content. In several prospective observational studies in sepsis patients, we have showed that mottling extension is strongly predictive of mortality.^\[[@R11],[@R12]\]^ Here, for the first time, we found that the presence of skin mottling is significantly associated with 28-day mortality in patients with PE, corroborating data from the severe infection context. Urinary output can also be routinely measured. In sepsis, urinary output monitoring is recommended to assess hemodynamic status and to guide treatment.^\[[@R25]\]^ Guidelines are based on seminal studies that reported an association between oliguria and pejorative outcome.^\[[@R26],[@R27]\]^ Here, we showed that in critically ill patients with PE, urinary output was significantly lower over the first 6 hours in nonsurvivors when compared to survivors. We have previously reported a significant relationship between these tissue perfusion parameters. Indeed, in septic shock patients, mottling extension correlates negatively with urinary output and positively with arterial lactate levels.^\[[@R12]\]^ Here, we found that these parameters could be combined as the risk of mortality increased with the number of tissue perfusion alterations. In light of all these data, we believe that combining arterial lactate level, mottling and urinary output could be helpful at the bedside to improve risk stratification of patients with PE. Our study has several limitations. First, it is a retrospective study and the predictive value of these tissue perfusion parameters has to be confirmed prospectively in a larger cohort of patients with PE. The next step would be to evaluate if clinical detection of tissue perfusion disorders could have an impact on the therapeutic management of patients with PE. Second, we did not investigate the variations of tissue perfusion parameters during resuscitation, although a reduction of lactate arterial level or a decrease in mottling skin areas over time have been both reported to be correlated with better outcome.^\[[@R28],[@R29]\]^ Third, because a large number of patients were included before 2008 international guidelines, circulating markers of RV injury, including troponin and B-type Natriuretic peptide (BNP) were not measured and not included in the risk stratification for death. 5. Conclusion ============= In this retrospective study, we found that 3 tissue perfusion parameters, arterial lactate level, mottling, and urinary output are significantly associated with 28-day mortality in ICU patients admitted for acute PE independently of shock and right ventricular dilation. These tissue perfusion parameters could be helpful to improve risk stratification of patients with PE. Acknowledgment ============== The authors thank R. Leboursicaud for her participation and help in recovering medical files. Author contributions ==================== Study concept and design, all authors. Data collection T.U., Y.N. Drafting of the manuscript, T.U., N.B., E.M., B.G. and H.A.O. Critical revision of manuscript, all authors. Statistical analysis, T.U., N.B., PYB and H.A.O. **Conceptualization:** Tomas Urbina, Hafid Ait-Oufella. **Data curation:** Pierre-Yves Boelle, Hafid Ait-Oufella. **Formal analysis:** Tomas Urbina, Pierre-Yves Boelle, Bertrand Guidet, Hafid Ait-Oufella. **Investigation:** Tomas Urbina, Naike Bige, Yann Nguyen, Vincent Dubee, Jeremie Joffre, Idriss Abdallah, Jean-Luc Baudel, Hafid Ait-Oufella. **Methodology:** Pierre-Yves Boelle, Vincent Dubee, Eric Maury, Bertrand Guidet, Hafid Ait-Oufella. **Project administration:** Hafid Ait-Oufella. **Software:** Yann Nguyen, Hafid Ait-Oufella. **Supervision:** Hafid Ait-Oufella. **Validation:** Tomas Urbina, Naike Bige, Yann Nguyen, Pierre-Yves Boelle, Hafid Ait-Oufella. **Visualization:** Jean-Luc Baudel, Hafid Ait-Oufella. **Writing -- original draft:** Tomas Urbina, Naike Bige, Hafid Ait-Oufella. **Writing -- review & editing:** Naike Bige, Vincent Dubee, Jeremie Joffre, Jean-Luc Baudel, Eric Maury, Bertrand Guidet, Hafid Ait-Oufella. Supplementary Material ====================== ###### Supplemental Digital Content Abbreviations: ICU = intensive care unit, PE = pulmonary embolism, RV = right ventricle, SAPS II = simplified acute physiologic score II, SOFA = sequential organ failure assessment, sPESI = simplified pulmonary embolism severity index. The authors have no funding and conflicts of interest to disclose. Supplemental Digital Content is available for this article.

{ "pile_set_name": "PubMed Central" }

All relevant data is in the paper and supporting information files. Introduction {#sec001} ============ Polysaccharides are an extremely diverse family of natural biopolymers, which are industrially used as thickeners, stabilizers and gelling agents in foodstuffs \[[@pone.0171469.ref001]\]. There is currently growing interest in their biological functions, such as their antioxidant and prebiotic activity. Although polysaccharides are derived from various sources such as microorganisms, algae and higher plants, the market is dominated by polysaccharides obtained from algae (such as carrageenans, alginates and agar) and higher plants (such as starch, cellulose and pectin) \[[@pone.0171469.ref002]\]. The synthesis of microbial polysaccharides has emerged as an important source of new biopolymers for industrial use \[[@pone.0171469.ref003]\]. It is an attractive alternative as microorganisms can grow under controlled conditions and produce a large variety of polysaccharides with unique properties. Production can be carried out in large quantities by biotechnological routes and from renewable and low-cost raw materials, which are easily recovered \[[@pone.0171469.ref004]\]. One of the microbial polysaccharides of industrial interest is curdlan, a neutral exopolysaccharide composed exclusively of glucose residues joined by β-(1→3) bonds, commercially produced by bacterial species of *Agrobacterium* \[[@pone.0171469.ref002],[@pone.0171469.ref005]--[@pone.0171469.ref007]\]. After being discovered by Harada in 1966, curdlan has received considerable attention in both the food and non-food industry due to its physicochemical properties, which are unique when compared to other polysaccharides commonly used, such as starch. When in aqueous suspension, curdlan may form a thermoreversible gel (low-set gel) when heated to temperatures close to 55°C with subsequent cooling. This gel has similar behavior to agar and gelatin gels. In addition, curdlan can form a thermoirreversible gel (high-set gel) when its aqueous suspension is heated to temperatures above 80°C, which is very stable. Furthermore, gels with different strengths can be formed by varying the temperature, the heating time and the concentration of curdlan \[[@pone.0171469.ref001],[@pone.0171469.ref006]\]. Curdlan was approved for use in foodstuffs in Korea, Taiwan and Japan in 1989 and registered in 1996 by the FDA (*Food and Drug Administration*) in the USA as a food additive with the following functions: a formulation and processing aid, and a stabilizer, thickener and texturizer \[[@pone.0171469.ref008]\]. It is widely used due to its ability to form the thermoirreversible gel, which exhibits great stability during the industrial processes of autoclaving, frying and freezing-thawing cycles. It has no taste, color or odor, can form gels at a wide pH range (from 2 to 10) and can mimic the palatability of foods containing fat \[[@pone.0171469.ref009]\]. It is also considered a dietary fiber, thus improving the functionality of various food products. The versatility of its applications associated with its health benefits make curdlan a valuable tool in the development of innovative food systems \[[@pone.0171469.ref001],[@pone.0171469.ref003]\]. The dispersion and gelation characteristics of curdlan in an aqueous system, as well as its mechanical properties, are associated with several factors such as concentration, temperature, heating time, dispersal method, and the presence of ions, salts and low molecular weight sugars \[[@pone.0171469.ref010]\]. Therefore, the form of recovery from the culture medium and the purification of the polysaccharide are steps which decisively influence the physical and technological properties of curdlan, which directly affect the sensory characteristics of foods in which the polysaccharide is inserted. So, it is necessary to study these properties to define the most appropriate form of the implementation of the polysaccharide in the industry \[[@pone.0171469.ref011]\]. Considering the interest in the properties of dispersion, viscosity, gelation and the health benefits of curdlan, the present study aimed to characterize the physicochemical and technological properties of curdlan produced by *Agrobacterium* sp. IFO 13140 and recovered using the pre-gelation and precipitation methods, as well as commercial curdlan subjected or otherwise to pre-gelation treatment. The methodologies employed for the recovery of the curdlans were studied. The structures of all curdlans were compared by FT-Raman spectroscopy. In addition, the application of curdlans in yogurts was evaluated. Materials and methods {#sec002} ===================== Materials {#sec003} --------- The bacterial strain *Agrobacterium* sp. IFO 13140 was purchased in lyophilized form from the Institute for Fermentation of Osaka (Japan). Commercial curdlan was acquired from Wako Pure Chemical Industries, Ltd. (Osaka, Japan). All solvents were of analytical grade. Curdlan production by *Agrobacterium* sp. IFO 13140 {#sec004} --------------------------------------------------- The culture medium used to reactivate the microorganism was proposed by the supplier (g L^-1^), pH 7: polypeptone (10), yeast extract (2), MgSO~4~.7H~2~O (1). 30 mg of the lyophilized bacteria were incubated in 100 mL of the medium at 30°C and 120 rpm for 48 h. The cells were recovered through centrifugation (6000 ×g, 10 min), washed with 9 g L^-1^ NaCl and transferred to the production medium. For curdlan production, the liquid medium described by Martinez et al. \[[@pone.0171469.ref012]\], pH 7, was used (g L^-1^): glucose (50), KH~2~PO~4~ (2.7), NH~4~Cl (1.6), MgSO~4~ (0.5) and trace elements (10 mL L^-1^). The composition of the trace elements (g L^-1^) in HCl 0.1 mol L^-1^ was: FeCl~3~.6H~2~O (1), MnCl~2~.4H~2~O (1), ZnCl (1), CaCl~2~ (1) and CaCO~3~ (0.03). The reactivated microorganisms were transferred to Erlenmeyer flasks containing 100 mL of the production medium and maintained at 30°C and 150 rpm for 5 days. Recovery of curdlan from production medium {#sec005} ------------------------------------------ Two methodologies were used to recover curdlan from the medium. In pre-gelation method \[[@pone.0171469.ref012],[@pone.0171469.ref013]\], NaOH 3 mol L^-1^ was added to the Erlenmeyer flasks containing the production medium at a ratio of 1.8:1 (NaOH:medium) for curdlan solubilization. This mixture was centrifuged (18000 ×g, 15 min, 4°C) to separate the cells. HCl 3 mol L^-1^ was added to the supernatant until pH 6--7 to obtain the curdlan gel, which was recovered by centrifugation (18000 ×g, 15 min, 4°C). Subsequently, it was washed three times with distilled water and lyophilized. The commercial curdlan was subjected to the same treatment and it was named pre-gelled commercial curdlan. The precipitation method is an adaptation of the industrial purification of curdlan. Industrially produced curdlan is purified by dissolution in a strong alkaline solution and dried in a spray-dryer, then washed with water until neutralization \[[@pone.0171469.ref014]\]. NaOH 3 mol L^-1^ was added to the Erlenmeyer flasks containing the production medium at a ratio of 1.8:1 (NaOH: medium) to solubilize the curdlan and the mixture was centrifuged at 18000 ×g for 15 min at 4°C to separate the cells, before being lyophilized. The dried mixture was resuspended in water, filtered to remove larger impurities and then ultra-filtered using the system described by Fenelon et al. \[[@pone.0171469.ref015]\] with a 30 kDa membrane and pressure of 1 kgf cm^-2^. During ultrafiltration, the material was washed with water until the pH was neutralized, causing the precipitation of the curdlan, that was thereafter lyophilized. Structural analysis of curdlans by FT-Raman {#sec006} ------------------------------------------- Samples of the commercial and commercial pre-gelled curdlans and those produced by *Agrobacterium* sp. IFO 13140 (pre-gelation and precipitation methods) were analyzed by Fourier transform Raman scattering infrared spectroscopy (FT-Raman) using a Fourier Transform infrared spectrometer (Vertex 70v model with Ram module II, Bruker, Germany) equipped with a Germanium detector cooled with liquid nitrogen. A Nd:YAG laser was used for excitation at 1064 nm. The spectra were based an average of 200 scans with a resolution of 4 cm^-1^. Morphology of curdlans {#sec007} ---------------------- The morphology of the commercial and commercial pre-gelled curdlans, and the curdlans produced by *Agrobacterium* sp. IFO 13140 (pre-gelation and precipitation methods) was analyzed in a scanning electron microscope (SS-550 model, Superscan, Shimadzu, Japan), at an accelerating voltage of 15 kV. Physicochemical characterization of curdlans {#sec008} -------------------------------------------- The commercial and commercial pre-gelled curdlans and those produced by *Agrobacterium* sp. IFO 13140 (pre-gelation and precipitation methods) were characterized for their carbohydrate, moisture and sodium content. The carbohydrate content was determined through the phenol-sulfuric method \[[@pone.0171469.ref016]\]. The moisture content was determined by the gravimetric method \[[@pone.0171469.ref017]\] and sodium was identified using atomic absorption spectrometry \[[@pone.0171469.ref018]\]. Technological properties of curdlans {#sec009} ------------------------------------ ### Water dispersion and gel formation capacities of curdlans {#sec010} To verify the dispersal capacity in water of the commercial and commercial pre-gelled curdlans and those produced by *Agrobacterium* sp. IFO 13140 (pre-gelation and precipitation methods), they were subjected to stirring in a mixer and/or in a magnetic stirrer until homogeneous dispersion was achieved. The dispersions were heated at 95°C/1 h to assess the gel formation capacity. ### Rheological characteristics and gel strength of curdlans {#sec011} Dispersions of the pre-gelled commercial curdlan and the curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) were prepared at three concentrations: 20 g L^-1^, 40 g L^-1^ and 80 g L^-1^ in water. The curdlans were dispersed in water using a mixer at room temperature for 5 min, sonicated for 10 min and then stirred in a magnetic stirrer for 12 h at 40°C. The samples were analyzed in a controlled stress rotational rheometer (HAAKE MARS II model, Thermo Fisher Scientific Inc., Newington, Germany), with steel cone/plate geometry (60 mm, gap 0.052 mm). The elastic (G\') and viscous (G\") modulus, and the apparent viscosity were measured depending on temperature (20--60°C) at a frequency of 10 Hz. For gel strength evaluation, the dispersions of both samples at the concentration of 20 g L^-1^ were kept or not in a water bath at 61°C/1 h to prepare the low-set gel and at 95°C/1 h for preparation of the high-set gel \[[@pone.0171469.ref019]\]. After heating, they were cooled at room temperature. The strength of the suspensions and gels was evaluated in triplicate in a Stable Micro Systems texturometer (TA-XT Plus model, Texture Technologies Corp., UK), using a 36 mm probe for compression analysis. ### Water Holding Capacity (WHC), Oil Holding Capacity (OHC) and Water Solubility Index (WSI) {#sec012} Samples of 0.25 g of the commercial and pre-gelled commercial curdlans and curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) were diluted in 10 mL of distilled water or soya oil at 30°C, homogenized in a magnetic stirrer for 30 min and centrifuged (1500 ×g, 10 min). The WHC was expressed as g of water absorbed per g of curdlan sample and the OHC was described as g of oil absorbed per g of curdlan sample. To determine the WSI, the supernatant of the WRI analysis was oven dried at 105°C and the ratio between the weight of the solid residue present in the supernatant after drying and the weight of the curdlan sample was calculated \[[@pone.0171469.ref011]\]. Application of curdlan in yogurt {#sec013} -------------------------------- ### Preparation of yogurts {#sec014} Eight yogurt samples were prepared: A1) without curdlan or heat treatment; A2) without curdlan and with heat treatment; B1) with commercial curdlan and without heat treatment; B2) with commercial curdlan and heat treatment; C1) with pre-gelled commercial curdlan and without heat treatment; C2) with pre-gelled commercial curdlan and heat treatment; D1) with curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) without heat treatment; D2) with curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) with heat treatment. The curdlan was mixed in UHT (ultra-high temperature) whole milk at a concentration of 15 g L^-1^ using a mixer. Subsequently, the milk of the samples subjected to heat treatment was heated at 90°C for 120 seconds and cooled to 42°C. For samples without heat treatment, the milk was directly heated to 42°C. Next, all samples were inoculated with 2 g L^-1^ of mixed culture of *Streptococcus thermophilus* and *Lactobacillus delbrueckii* subsp. *bulgaricus*, and maintained at 42°C for 6 h. The yogurts were cooled and refrigerated for 24 h for rheological, texture and syneresis analysis \[[@pone.0171469.ref003],[@pone.0171469.ref020]\]. ### Texture Profile Analysis (TPA) {#sec015} The TPA of the yogurts was carried out in a TA-XT Plus texturometer (Stable Micro Systems, Godalming, UK), equipped with Texture Expert software (Stable Micro Systems, Godalming, UK). A 10 mm cylindrical probe was used (ref. P/0.5R, Stable Micro Systems). Two cycles were applied, at a velocity of 2 mm s^-1^ and a depth of 15 mm, producing force-time curves which were used to determine the firmness, cohesiveness, adhesiveness, springiness, gumminess and chewiness of the samples. The analysis was performed in triplicate, at 8°C. ### Syneresis {#sec016} The syneresis analysis of each preparation was performed after 28 days of storage of the yogurt at 8°C in triplicate. Yogurt samples (30--40 g) were centrifuged (222 ×g, 10 min, 4°C) and the supernatant was separated and centrifuged again (222 ×g, 10 min, 4°C), weighed and the syneresis (%) was calculated from the ratio between the mass of the supernatant and the initial mass of the yogurt \[[@pone.0171469.ref020]\]. ### Rheological analysis {#sec017} The rheological properties of flow of the yogurts were analyzed in triplicate, in a controlled stress rotational rheometer (HAAKE MARS II model, Thermo Fisher Scientific Inc., Newington, Germany), with steel cone/plate geometry (60 mm, gap 0.052 mm). The measurements were performed at 8°C. The scan of the deformation rate was performed from 0 to 116 s^-1^, obtaining the outward and return data. The flow and viscosity curves were obtained by determining the stress and viscosity versus the shear rate \[[@pone.0171469.ref020]\]. The rheological parameters K (consistency index) and n (flow behavior index) were calculated using the Ostwald de Waele model, while the parameter τ~0~ (yield stress) was calculated by the Herschel-Bulkley model. The hysteresis was calculated from the area of the flow curves of the yogurts using the RheoWin 4.10.000 software program (HAAKE software, Thermo Fisher Scientific Inc., Newington, Germany). Statistical analysis {#sec018} -------------------- Data were analyzed by analysis of variance (ANOVA), and means were compared with the Tukey Test (p\<0.05) using the Statistica 8.0/2008 software package (Stat Soft, Inc., Tulsa, USA). Results and discussion {#sec019} ====================== Structural analysis of curdlans by FT-Raman {#sec020} ------------------------------------------- [Fig 1](#pone.0171469.g001){ref-type="fig"} shows FT-Raman spectra of commercial and pre-gelled commercial curdlans and those produced by *Agrobacterium* sp. IFO 13140 (pre-gelation and precipitation methods). From the data, it is verified a structural similarity between all samples, which means that the curdlan produced by the microorganism has a similar structure to commercial curdlan and it also means that the method employed to recover curdlan from the medium did not modify the structure of the polysaccharide. ![FT-Raman spectra of: (A) commercial curdlan, (B) pre-gelled commercial curdlan, (C) curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gellation method), (D) curdlan produced by *Agrobacterium* sp. IFO 13140 (precipitation method).\ The dotted lines show the characteristic peaks of the samples.](pone.0171469.g001){#pone.0171469.g001} Several authors described the chemical bonds present in the structure of curdlan molecule based in FT-IR analysis \[[@pone.0171469.ref003],[@pone.0171469.ref014],[@pone.0171469.ref021]\]. They stated that, from the FT-IR data for curdlan, the bands around 890, 1080 and 1160 cm^−1^ correspond to the *β*-(1,3)-glucan linkages. Despite the difference in band assignments for FT-IR and FT-Raman spectroscopies, the spectra of carbohydrates in both techniques show a characteristic absorption band of β-anomeric configuration at \~890 cm^−1^ (886 cm^-1^ in this work) \[[@pone.0171469.ref022]\]. The Raman bands and shoulders at 1090 and 1137 cm^−1^ are typical for β-glucans. Intense highly overlapped Raman bands between 990 and 1200 cm^−1^ are attributed to COC and CC stretching vibrations of polysaccharides. The features between 1200 and 1440 cm^-1^ are mainly assigned to in-plane ring deformation including CH and OH bending modes. Finally, the band at 1457 cm^-1^ is assigned to CH~2~ in-plane bending in CH~2~OH of the molecule \[[@pone.0171469.ref023]\]. Morphology of curdlans {#sec021} ---------------------- [Fig 2](#pone.0171469.g002){ref-type="fig"} shows scanning electron microscopy images of the commercial curdlan ([Fig 2A and 2B](#pone.0171469.g002){ref-type="fig"}), the pre-gelled commercial curdlan ([Fig 2C and 2D](#pone.0171469.g002){ref-type="fig"}) and the curdlans produced by *Agrobacterium* sp. IFO 13140 recovered by pre-gelation and precipitation methods ([Fig 2E--2H](#pone.0171469.g002){ref-type="fig"}) with a range of magnifications. ![Scanning electron microscopy of: commercial curdlan---(A) 40x and (B) 1000x magnification; pre-gelled commercial curdlan---(C) 40x and (D) 1000x magnification; curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method)---(E) 40x and (F) 1000x magnification; curdlan produced by *Agrobacterium* sp. IFO 13140 (precipitation method)---(G) 40x and (H) 1000x magnification.](pone.0171469.g002){#pone.0171469.g002} Marchessault and Deslandes \[[@pone.0171469.ref024]\] define the shape of the granules of commercial curdlan as collapsed or invaginated, which coincides with the commercial curdlan shape illustrated in [Fig 2A and 2B](#pone.0171469.g002){ref-type="fig"}. The granules have widely differing sizes, varying from around 10 to 100 μm in diameter. Industrially produced curdlan is dried by spray-drying \[[@pone.0171469.ref014]\] and the type of drying has a major influence on the particle bead structure \[[@pone.0171469.ref025]\]. The expected range of particle size of polymers dried by ordinary spray dryers is 5--150 μm. Furthermore, spherical microspheres or those with pores/concavities, as shown in [Fig 2A and 2B](#pone.0171469.g002){ref-type="fig"}, are characteristic of the drying of products by a spray-dryer. The formation of pores or concavities is associated with the rapid evaporation of the liquid particles of this process \[[@pone.0171469.ref026]\]. The pre-gelled commercial curdlan and both the curdlans produced by the microorganism, when viewed at low magnification (40x), were in flake form and displayed irregularities. These characteristics are expected for lyophilized products, since the working conditions of the lyophilizer exert great pressure on the particles to be dried, meaning that a product dried by lyophilization remains amorphous in comparison with a spray dried product \[[@pone.0171469.ref025]\]. Furthermore, at a higher magnification (1000x), it can be seen that the particles obtained for the three samples that underwent the lyophilization process ([Fig 2D, 2F and 2H](#pone.0171469.g002){ref-type="fig"}), were considerably smaller than the commercial curdlan (obtained by spray-dryer). Therefore, it can be inferred that the lyophilization process can facilitate the dispersion of curdlan molecules in water, especially the pre-gelled commercial curdlan and the one produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method), which have smaller particle sizes ([Fig 2D and 2F](#pone.0171469.g002){ref-type="fig"}). The smaller diameter increases the accessibility of water molecules to the inside of the particles, facilitating dispersion. The structure displayed by the commercial curdlan is maintained by a large amount of hydrogen bonds. But, when added to an alkaline solution, these bonds are broken due to their ionization, and the granule loses its structure. The neutralization of the curdlan suspension causes a reshaping of the hydrogen bonds that depends on the neutralizing agent used. With the addition of HCl (pre-gelation method), the reassociation causes the formation of a firm gel, with small structures of less than 1 μm in diameter ([Fig 2D and 2F](#pone.0171469.g002){ref-type="fig"}). With water (precipitation method), reassociation causes the precipitation of the curdlan into small particles with diameters greater than those obtained by neutralization with HCl, ranging from 2--10 μm in diameter ([Fig 2H](#pone.0171469.g002){ref-type="fig"}). Therefore, the difference between the sizes of the curdlan particles produced by the microorganism using the two methods is due to the different forms of reassociation of the hydrogen bonds, which are dependent on the recovery method employed, and can also influences the characteristics of water dispersion of the polysaccharide. Physicochemical characterization of curdlans {#sec022} -------------------------------------------- As an exopolysaccharide, curdlan is secreted to the extracellular medium in the form of biofilms, and has the advantage of being easy to recover. As a result, the polysaccharide should present a low degree of impurities, and consequently the carbohydrate content represents an indirect measure of the purity of the samples. The levels of carbohydrates and moisture found in the curdlan samples are shown in [Table 1](#pone.0171469.t001){ref-type="table"}. 10.1371/journal.pone.0171469.t001 ###### Carbohydrate, moisture and sodium content (%) of different samples of curdlan. Values indicate mean ± standard-deviation. ![](pone.0171469.t001){#pone.0171469.t001g} Curdlan sample Carbohydrate Moisture Sodium ----------------------------------------------------------------------- -------------- ----------- --------------- Commercial 94.5 ± 0.2 4.1 ± 0.4 1.2 ± 0.3 Commercial pre-gelled 82 ± 1 4.0 ± 0.3 4.4 ± 0.8 Microbial[\*](#t001fn001){ref-type="table-fn"} (pre-gelation method) 70.6 ± 0.6 5.4 ± 0.9 0.024 ± 0.001 Microbial[\*](#t001fn001){ref-type="table-fn"} (precipitation method) 85.4 ± 0.7 4.8 ± 0.6 \< 0.001 \*Produced by *Agrobacterium* sp. IFO 13140. The highest carbohydrate content was found for the commercial sample. The pre-gelled commercial sample had a carbohydrate content lower than the commercial curdlan due to the salt (NaCl) incorporated in the pre-gelling process. The two samples produced by *Agrobacterium* sp. IFO 13140 had different carbohydrate contents, with that produced by precipitation method having greater purity due to the successive washings performed. The impurities in the last samples consist of remnants of the production medium and the salt formed in the recovery step. The sodium content of curdlans was determined because curdlan is a glucose polymer with a low amount of inorganic salts, mainly sodium chloride \[[@pone.0171469.ref027]\]. Thus, sodium content is an indirect estimate of the salt content of the material. The sodium concentrations disposed in [Table 1](#pone.0171469.t001){ref-type="table"} directly influence the dispersion and gelling characteristics of the polysaccharide, which will be further discussed in the next section. Technological properties {#sec023} ------------------------ ### Water dispersibility and gelling capacities {#sec024} Curdlan can be used as a thickener, stabilizer and texturizer in food industry \[[@pone.0171469.ref008],[@pone.0171469.ref027]\]. However, as curdlan is insoluble in water, it is necessary to use an efficient homogenizer to obtain a homogeneous dispersion and apply a further stirring of the dispersion prior to the carrying out of analyses requiring uniformity \[[@pone.0171469.ref021]\]. The pre-gelled commercial curdlan and the curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) dispersed easily in water when subjected to stirring in a mixer for less than 5 min, and also acted as thickening agents. Both dispersions remained visibly homogeneous for days and, when heated at 95°C/1 h, formed a firm and homogeneous gel. The commercial curdlan did not easily disperse in water using the mixer. After 24 hours of stirring in a magnetic stirrer, a homogeneous dispersion formed for a brief period, displaying a phase separation after 10 min of rest. After 48 h of stirring, phase separation began to be observed at 30 min; after 72 h, at 1 hour; and after 96 h, phase separation began at 1.5 h. The gelling of commercial curdlan also occurred after heating at 95°C/1 h, but there was phase separation in the gel formed, which became more pronounced as the stirring period for formation of the dispersion was reduced. Marchessault and Deslandes \[[@pone.0171469.ref024]\] have previously identified the formation of a non-homogenous gel from curdlan in its native form. Probably, the difficulty of dispersion of commercial curdlan compared with those submitted to pre-gelation process is related to the particle size of the sample (as noted in morphological analysis), which makes difficult the accessibility of water molecules to the inside of the particles. Another important factor related to the dispersion and gelling of curdlan is the presence of salts. In the gelling of curdlan, the swelling phase of the molecule is essential and may be influenced by salts, as they affect the mobility of the water molecules during hydration \[[@pone.0171469.ref010]\]. The swelling of curdlan is facilitated with the salts present because it increases intermolecular association and consequently increases the viscosity of curdlan water dispersions. The curdlan produced by *Agrobacterium* sp. IFO 13140 (precipitation method) displayed greater difficulty in forming a homogenous dispersion in water and, in contrast to the other curdlans, did not form a gel when heated. This can be explained by its considerably lower sodium content, due to the countless washings with water used to achieve neutralization. Samples of the other curdlans were also washed in the ultrafiltration device and, following such process, no longer had the same dispersion and gelling capacities. This proves therefore, that as with other biopolymers, for curdlan, the rheological characteristics depend excessively on the methods of recovery employed following production. Due to the characteristics displayed, the technological characteristics of the curdlan produced by *Agrobacterium* sp. IFO 13140 (precipitation method) were not evaluated, and this sample was not applied in yogurt. Additionally, dispersions of commercial curdlan were not evaluated by rheological and gel strength analysis because the commercial curdlan did not act as a thickener with the methodology employed for preparation of the dispersions. ### Rheological characteristics and gel strength of curdlans {#sec025} Thermal scanning rheological measurements were made to evaluate both the apparent viscosity and the gelation behavior of curdlan dispersions. The temperature dependence of apparent viscosity and of G\' and G\" modulus for aqueous dispersions of pre-gelled commercial curdlan and of the curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) at 20, 40 and 80 g L^-1^ are shown in [Fig 3A and 3B](#pone.0171469.g003){ref-type="fig"}. ![Temperature dependency of: (A) apparent viscosity and (B) G\' (continuous line) and G\" (dotted line) modulus of the aqueous dispersions of curdlan.\ Pre-gelled commercial curdlan (empty symbol) and curdlan produced by *Agrobacterium* sp. IFO 13140 recovered by the pre-gelation method (full symbol) at (■) 20, (●) 40 and (▲) 80 g L^-1^.](pone.0171469.g003){#pone.0171469.g003} The curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) exhibited considerably higher viscosity than pre-gelled commercial curdlan ([Fig 3A](#pone.0171469.g003){ref-type="fig"}). For example, at 20°C, the apparent viscosity of curdlan produced by the microorganism was almost 100% higher than pre-gelled commercial curdlan at 20 and 80 g L^-1^. Equally, G\' and G\" modulus ([Fig 3B](#pone.0171469.g003){ref-type="fig"}) were also higher for the curdlan produced by the microorganism, revealing the highest thickening potential of this curdlan when compared to the pre-gelled commercial one. Both the apparent viscosity and the G\' module of both curdlans exhibited a different behavior as a function of temperature for each concentration employed. At 20 g L^-1^, both parameters increased with temperature for pre-gelled comercial curdlan, while for curdlan produced by the microorganism they decreased. This behavior is related to the swelling of curdlan because of the breakage of hydrogen bonds during heating. The difference between the curdlanas is because, for pre-gelled commercial curdlan, inter-molecular entanglements between the particles that were swollen formed pseudo-links, which contributed to the little increase in G\'. This phenomenon did not occur for curdlan produced by the microorganism. However, for higher concentrations (40 and 80 g L^-1^), the decrease in the viscosity and in G\' no longer occurred with temperature for any sample, because the high concentration favors the inter-molecular entanglements between the particles, which contribute to the increase of both parameters. Additionally, the increase in G\' and G\" modulus and in apparent viscosity of curdlan dispersions at about 50--55°C, especially at 80 g L^-1^, suggests the beginning of the formation of the thermo-irreversible gels due to hydrophobic interactions. Jin et al. \[[@pone.0171469.ref021]\] stated that G\' decreased until 50°C for curdlan suspensions at 20 g L^-1^, which is related to the swelling of curdlan because of the breakage of hydrogen bonds, as observed for curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) in this study. However, Funami et al. \[[@pone.0171469.ref028]\] stated only increasing in G\' parameter for dispersions of curdlans at 20, 40 and 100 g L^-1^ with increasing temperature. The authors attributed this behavior to inter-molecular entanglements between the particles that were swollen forming pseudo-links, being this result similar to the one obtained for pre-gelled comercial curdlan in all concentrations evaluated and for curdlan produced by the microorganism for concentrations from 40 to 80 g L^-1^ in this work. The strength data of suspensions and gels of the pre-gelled commercial curdlan and the curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) at 20 g L^-1^ are presented in [Table 2](#pone.0171469.t002){ref-type="table"}. According to the strength values of the suspensions and gels, the curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) has a better thickening capacity than the pre-gelled curdlan, as the suspension of the same without heat treatment was approximately 10% stronger. This result agrees with the results of rheological analysis. However, the gelling capacity of the pre-gelled commercial curdlan is considerably higher. 10.1371/journal.pone.0171469.t002 ###### Strength (×10^−3^ N) of pre-gelled commercial curdlan and the curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) samples after undergoing different heat treatments. Values indicate mean ± standard-deviation. ![](pone.0171469.t002){#pone.0171469.t002g} Heat treatment of the dispersion in a concentration of 20 g L^-1^ Strength (× 10^−3^ N) ------------------------------------------------------------------- -------------------------------------------------- ---------------------------------------------- Without heat treatment 69 ± 2[^d^](#t002fn001){ref-type="table-fn"} 76 ± 2[^c^](#t002fn001){ref-type="table-fn"} Low-set gel (61°C/1 h) 79 ± 1[^c^](#t002fn001){ref-type="table-fn"} 77 ± 2[^c^](#t002fn001){ref-type="table-fn"} High-set gel (95°C/1 h) 96.8 ± 0.8[^a^](#t002fn001){ref-type="table-fn"} 83 ± 1[^b^](#t002fn001){ref-type="table-fn"} ^a--d^ Means with different letters are significantly different (p\< 0.05). The strength of the low-set gel of the pre-gelled commercial curdlan was around 15% greater than that of the suspension without heat treatment, and statistically equal to that of the low-set gel of the curdlan produced by the microorganism. The high-set gel of the pre-gelled commercial curdlan had strength 40% greater than the suspension without heat treatment, and 17% greater than the high-set gel produced by the microorganism. When an aqueous dispersion of curdlan is heated around 55°C, the low-set gel formed is maintained by intramolecular hydrogen bonds and the curdlan chains adopt a predominantly single helix conformation. But when heated above 80°C, the high-set gel formed is maintained by intermolecular hydrophobic interactions and the curdlan chains adopt a predominantly triple helix conformation \[[@pone.0171469.ref014],[@pone.0171469.ref029]\], forming an organized and rigid gel configuration. As such, the increase in the strength of the high-set curdlan gel (95°C/1 h) is due to the greater presence of the triple helix conformation. The fact that the curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) have formed a considerably weaker gel than the pre-gelled commercial curdlan, as well as the fact that it does not exhibit the same behavior that pre-gelled commercial sample in rheological analysis at low concentrations (increase of G\' and apparent viscosity with temperature), are probably due to a difference of molecular weight/degree of polymerization of the polymers. A variety of physical properties of β-(1→3)-glucans, including gel strength (or gel-forming ability) is related to the molecular weight/degree of polymerization of the biopolymer. The higher the degree of polymerization of the polysaccharide, the greater is its gel forming ability with heating \[[@pone.0171469.ref030],[@pone.0171469.ref031]\]. Nakanishi et al. \[[@pone.0171469.ref032]\] studied the formation of a complex of curdlan with aniline blue dye and noted a relation between the rate of interaction of the polymer with the dye and the concentration, degree of polymerization, and gel-forming ability of the polymer. The variation in absorbance at 590 nm is proportional to the concentration of curdlan and to its gel-forming ability (and consequently to its degree of polymerization). [S1 Appendix](#pone.0171469.s001){ref-type="supplementary-material"} contains the methodology used by the authors described to evaluate curdlan interaction with aniline blue dye and the results of the relationship between absorbance variation and concentration of the pre-gelled commercial curdlan and curdlan produced by the microorganism. By comparing the results of this study to those obtained by Nakanishi et al. \[[@pone.0171469.ref032]\] and using the relation obtained by these authors between absorbance variation and degree of polymerization, it is concluded that the degree of polymerization of the pre-gelled commercial curdlan is about 45% higher than of the curdlan produced by the microorganism, which explains the lowest gel strength and the decrease in apparent viscosity and G\' in low concentration dispersions of the polymer. ### Water Holding Capacity (WHC), Oil Holding Capacity (OHC) and Water Solubility Index (WSI) {#sec026} Determining the technological properties of polysaccharides is of great importance when predicting their possible industrial applications. The values for the properties of water solubility index (WSI) water holding capacity (WHC) and oil holding capacity (OHC) of the curdlan samples are described in [Table 3](#pone.0171469.t003){ref-type="table"}. 10.1371/journal.pone.0171469.t003 ###### Water Holding Capacity (WHC), Oil Holding Capacity (OHC) and Water Solubility Index (WSI) (g g^-1^) of the commercial, pre-gelled commercial curdlans and those produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method). Values indicate mean ± standard-deviation. ![](pone.0171469.t003){#pone.0171469.t003g} Curdlan sample WHC OHC WSI --------------------------------- --------------------------------------------------- --------------------------------------------------- ------------------------------------------------------ Commercial 4.6 ± 0.4[^a^](#t003fn001){ref-type="table-fn"} 0.62 ± 0.08[^c^](#t003fn001){ref-type="table-fn"} 0.006 ± 0.002[^a^](#t003fn001){ref-type="table-fn"} Commercial pre-gelled 2.20 ± 0.08[^c^](#t003fn001){ref-type="table-fn"} 4.4 ± 0.2[^b^](#t003fn001){ref-type="table-fn"} 0.00923 ± 7E-5[^a^](#t003fn001){ref-type="table-fn"} Microbial (pre-gelation method) 3.6 ± 0.3[^b^](#t003fn001){ref-type="table-fn"} 8.7 ± 0.1[^a^](#t003fn001){ref-type="table-fn"} 0.0068 ± 0.003[^a^](#t003fn001){ref-type="table-fn"} ^a--c^ Means in the same column with different letters are significantly different (p\< 0.05). The three curdlans had very low and statistically equal WSIs, which was expected as curdlan is insoluble in water. Despite this, however, the curdlans presented some water absorption, with the commercial curdlan achieving the highest value, followed by the curdlan produced by the microorganism and the pre-gelled commercial curdlan. Seguchi and Kusunose \[[@pone.0171469.ref033]\] found water absorption rates for curdlan between 5.244 g g^-1^ and 7.724 g g^-1^. Compared to other polysaccharides employed in food industry as gums, curdlan has low water holding capacity; the guar and xanthan gums have a WHC of 25.77 g g^-1^ and 27.33 g g^-1^, respectively \[[@pone.0171469.ref034]\]. However, both are soluble in water. The low solubility and water holding capacity of curdlans is explained by the existence of a large amount of intra/intermolecular hydrogen bonds within the polymer. This also explains the fact that commercial curdlan has a higher water absorption index than the other types as the recovery by the pre-gelation process employed in the latter two types favors the formation of large quantities of hydrogen bonds in the polymer. Thus, the polysaccharide interacts more strongly with itself than with water \[[@pone.0171469.ref021],[@pone.0171469.ref034]\]. The three evaluated curdlans also differed with respect to oil holding capacity. With the exception of the commercial curdlan, they presented greater OHC values than WHC, indicating that the samples have a higher lipolytic than hydrophilic capacity. Oil absorption values greater than three make curdlan a potentially useful ingredient in structural interactions in foods, especially in aroma retention, improved palatability and maintenance of the stability of the product during storage \[[@pone.0171469.ref011]\]. This result justifies the use of curdlan as a fat replacer or mimicker of fat in the food industry \[[@pone.0171469.ref009],[@pone.0171469.ref035]\]. Application of curdlan in yogurt {#sec027} -------------------------------- The main characteristics that define the quality of yogurts are its texture and propensity for serum separation (syneresis). Typically, polysaccharides such as xanthan, guar, gellan, pectin, carrageenan are used to give the product a firmer texture, increase its stability and hence make it more acceptable to the consumer \[[@pone.0171469.ref003],[@pone.0171469.ref020]\]. [Table 4](#pone.0171469.t004){ref-type="table"} displays the parameters of texture and syneresis after 28 days of storage of yogurt samples with and without curdlan, submitted or not to heat treatment with the aim of gelling the curdlan in the milk prior to the fermentation process. 10.1371/journal.pone.0171469.t004 ###### Texture parameters and syneresis of yogurts with and without curdlan and submitted or not to heat treatment. Values indicate mean ± standard-deviation. ![](pone.0171469.t004){#pone.0171469.t004g} Sample Firmness (× 10^−3^ N) Cohesiveness Adhesiveness (N × mm) Springiness (mm) Gumminess (N) Chewiness (N) Syneresis (%) -------- --------------------------------------------------- ----------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------------- -------------------------------------------------- A1 77 ± 2[^c^](#t004fn001){ref-type="table-fn"} 0.77 ± 0.02[^a^](#t004fn001){ref-type="table-fn"} 0.02 ± 0.01[^c^](#t004fn001){ref-type="table-fn"} 1.13 ± 0.01[^a^](#t004fn001){ref-type="table-fn"}[^b^](#t004fn001){ref-type="table-fn"} 5.9 ± 0.3[^c^](#t004fn001){ref-type="table-fn"} 6.7 ± 0.4[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"}[^d^](#t004fn001){ref-type="table-fn"} 49.3 ± 0.8[^a^](#t004fn001){ref-type="table-fn"} A2 84 ± 8[^c^](#t004fn001){ref-type="table-fn"} 0.71 ± 0.07[^a^](#t004fn001){ref-type="table-fn"}[^b^](#t004fn001){ref-type="table-fn"} 0.04 ± 0.03[^c^](#t004fn001){ref-type="table-fn"} 1.00 ± 0.04[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 5.9 ± 0.1[^c^](#t004fn001){ref-type="table-fn"} 5.9 ± 0.3[^c^](#t004fn001){ref-type="table-fn"}[^d^](#t004fn001){ref-type="table-fn"}[^e^](#t004fn001){ref-type="table-fn"} 48.7 ± 0.4[^a^](#t004fn001){ref-type="table-fn"} B1 83 ± 8[^c^](#t004fn001){ref-type="table-fn"} 0.66 ± 0.02[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 0.08 ± 0.03[^c^](#t004fn001){ref-type="table-fn"} 0.94 ± 0.04[^c^](#t004fn001){ref-type="table-fn"} 5.5 ± 0.7[^c^](#t004fn001){ref-type="table-fn"} 5.1 ± 0.7[^e^](#t004fn001){ref-type="table-fn"} 45.6 ± 0.9[^b^](#t004fn001){ref-type="table-fn"} B2 129 ± 7[^b^](#t004fn001){ref-type="table-fn"} 0.47 ± 0.04[^d^](#t004fn001){ref-type="table-fn"} 0.4 ± 0.1[^b^](#t004fn001){ref-type="table-fn"} 0.97 ± 0.05[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 6.1 ± 0.2[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 5.9 ± 0.4[^c^](#t004fn001){ref-type="table-fn"}[^d^](#t004fn001){ref-type="table-fn"}[^e^](#t004fn001){ref-type="table-fn"} 38.2 ± 0.3[^c^](#t004fn001){ref-type="table-fn"} C1 94 ± 9[^c^](#t004fn001){ref-type="table-fn"} 0.65 ± 0.02[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 0.16 ± 0.04[^c^](#t004fn001){ref-type="table-fn"} 1.2 ± 0.1[^a^](#t004fn001){ref-type="table-fn"} 6.0 ± 0.6[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 7.1 ± 0.3[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 26.4 ± 0.8[^d^](#t004fn001){ref-type="table-fn"} C2 125.1 ± 0.6[^b^](#t004fn001){ref-type="table-fn"} 0.58 ± 0.02[^c^](#t004fn001){ref-type="table-fn"} 0.58 ± 0.02[^a^](#t004fn001){ref-type="table-fn"}[^b^](#t004fn001){ref-type="table-fn"} 1.01 ± 0.01[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 7.2 ± 0.2[^b^](#t004fn001){ref-type="table-fn"} 7.3 ± 0.2[^b^](#t004fn001){ref-type="table-fn"} 29.2 ± 0.7[^d^](#t004fn001){ref-type="table-fn"} D1 128 ± 6[^b^](#t004fn001){ref-type="table-fn"} 0.46 ± 0.02[^d^](#t004fn001){ref-type="table-fn"} 0.68 ± 0.04[^a^](#t004fn001){ref-type="table-fn"} 0.96 ± 0.03[^c^](#t004fn001){ref-type="table-fn"} 5.8 ± 0.3[^c^](#t004fn001){ref-type="table-fn"} 5.6 ± 0.5[^d^](#t004fn001){ref-type="table-fn"}[^e^](#t004fn001){ref-type="table-fn"} 28.8 ± 0.2[^d^](#t004fn001){ref-type="table-fn"} D2 160 ± 7[^a^](#t004fn001){ref-type="table-fn"} 0.59 ± 0.01[^c^](#t004fn001){ref-type="table-fn"} 0.5 ± 0.1[^a^](#t004fn001){ref-type="table-fn"}[^b^](#t004fn001){ref-type="table-fn"} 0.99 ± 0.06[^b^](#t004fn001){ref-type="table-fn"}[^c^](#t004fn001){ref-type="table-fn"} 9.5 ± 0.6[^a^](#t004fn001){ref-type="table-fn"} 9.3 ± 0.3[^a^](#t004fn001){ref-type="table-fn"} 35.8 ± 0.9[^c^](#t004fn001){ref-type="table-fn"} ^a--e^ Means within the same column with different letters are significantly different (p\< 0.05). Yogurts: A1) without curdlan or heat treatment; A2) without curdlan and with heat treatment; B1) with commercial curdlan and without heat treatment; B2) with commercial curdlan and heat treatment; C1) with pre-gelled commercial curdlan and without heat treatment; C2) with pre-gelled commercial curdlan and heat treatment; D1) with curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) without heat treatment; D2) with curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) with heat treatment. It can be seen that the use of curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) produced significant alterations in the parameters of firmness, cohesiveness and adhesiveness of the yogurt, even without being subjected to heat treatment. This is due to the thickening potential of the material. Heat treatment caused major changes in these parameters, both for the yogurts with commercial curdlan and pre-gelled commercial curdlan, although the highest values for firmness, stickiness, gumminess and chewiness were obtained for the yogurt with curdlan produced by the microorganism. Heat treatment promotes the formation of a firm gel due to the intermolecular hydrophobic interactions that structure the system, making the yogurt more consistent, resulting in greater difficulty to separate in the mouth and making it denser during chewing, which therefore takes longer. It was notable that the pre-gelled commercial curdlan formed a stronger gel in water, but had less effect on the texture parameters than that produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) when applied to the yogurt, which was not expected. Thus, it is likely that the latter has a greater ability to interact with water and the other components of milk, especially proteins, more efficiently stiffening the protein network formed after the fermentation process. As described before, the curdlan produced by the microorganism has the capacity to absorb 64% more water and 98% more oil than the pre-gelled commercial curdlan, with these parameters being important for products such as yogurts prepared with whole milk (3% fat). The use of commercial curdlan without heat treatment did not produce many changes in the texture parameters or in the syneresis of the yogurts. This is because it was not effectively homogenized in the milk as it dispersed with greater difficulty, interfering little in the formation of the protein network during fermentation. It was found that syneresis was reduced for the yogurt samples with curdlan, particularly the pre-gelled commercial variety and the curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method). The latter displayed higher syneresis after the first 28 days of storage because its gel is less stable than the pre-gelled commercial curdlan. As such, the pre-gelled commercial curdlan demonstrated a greater ability to preserve the structure of the yogurt, avoiding rearrangements in the casein network due to the time of storage and avoiding whey expulsion. Martinez et al. \[[@pone.0171469.ref003]\] produced yogurts with added curdlan and achieved a significant reduction in the syneresis of the products compared to the control (without curdlan). The authors found that the high stability obtained by the yogurt with curdlan is attributed to interactions between the curdlan molecules, either between one another or with the proteins, which cause the formation of a more compact and continuous three-dimensional protein network, which effectively entraps the protein molecules and water in its structure. Importantly, there was no difference between any of the texture parameters and the syneresis of yogurts without curdlan submitted or not to heat treatment. This was expected as the aim of treatment was to gel the curdlan and, therefore, the A2 sample was not submitted to rheological analysis. [Fig 4](#pone.0171469.g004){ref-type="fig"} shows the flow curves ([Fig 4A and 4B](#pone.0171469.g004){ref-type="fig"}) and the viscosity curves ([Fig 4C and 4D](#pone.0171469.g004){ref-type="fig"}) for the yogurts produced with curdlan with and without heat treatment of the milk prior to fermentation. ![Flow curves of yogurt samples without heat treatment (A) with heat treatment (B); viscosity curves of yogurt samples without heat treatment (C) and with heat treatment (D).\ Yogurt without curdlan (--), with commercial curdlan (Δ), with pre-gelled commercial curdlan (○) and with curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) (■). The direction of the gray arrows indicates the ascendant and descendant curves.](pone.0171469.g004){#pone.0171469.g004} All the yogurts behaved as non-Newtonian pseudo-plastic fluids, as their viscosity decreased due to the shear rate applied. It can be seen that the samples subjected to heat treatment of the milk had higher viscosity values across the entire shear rate analyzed due to the gelling of the curdlan. However, the decrease in viscosity versus the shear rate in the return data was much more significant for these samples. All the samples also exhibited thixotropic characteristics, due to the difference of tension and viscosity between the ascending and descending shear rate curves. The hysteresis observed is due to the breakdown of the gel structure formed by coagulation of the protein during fermentation, in the presence and absence of curdlan. Hysteresis is measured as the area between the ascending and descending curves, where the greater the area (when positive), the greater the thixotropic effect. Hysteresis, together with the other rheological parameters, are described in [Table 5](#pone.0171469.t005){ref-type="table"}. 10.1371/journal.pone.0171469.t005 ###### Rheological parameters of yogurt samples with and without curdlan and submitted or not to heat treatment. Hysteresis, consistency index (K), flow behavior index (n) and yield stress (τ~0~). Values indicate mean ± standard-deviation. ![](pone.0171469.t005){#pone.0171469.t005g} *Sample* *K (mPa s*^*n*^*)* *n* *τ*~*0*~ *(Pa)* *Hysteresis (Pa s*^*-1*^*)* ---------- ------------------------------------------------- ----------------------------------------------------- -------------------------------------------------- ------------------------------------------------- A 98 ± 3[^d^](#t005fn001){ref-type="table-fn"} 0.997 ± 0.006[^a^](#t005fn001){ref-type="table-fn"} 2 ± 1[^e^](#t005fn001){ref-type="table-fn"} 362 ± 43[^e^](#t005fn001){ref-type="table-fn"} B1 223 ± 26[^d^](#t005fn001){ref-type="table-fn"} 0.938 ± 0.001[^b^](#t005fn001){ref-type="table-fn"} 12 ± 1[^d^](#t005fn001){ref-type="table-fn"} 1088 ± 75[^d^](#t005fn001){ref-type="table-fn"} B2 1685 ± 18[^a^](#t005fn001){ref-type="table-fn"} 0.662 ± 0.004[^e^](#t005fn001){ref-type="table-fn"} 30 ± 4[^c^](#t005fn001){ref-type="table-fn"} 2241 ± 39[^b^](#t005fn001){ref-type="table-fn"} C1 922 ± 35[^b^](#t005fn001){ref-type="table-fn"} 0.73 ± 0.04[^d^](#t005fn001){ref-type="table-fn"} 12 ± 1[^d^](#t005fn001){ref-type="table-fn"} 1134 ± 92[^d^](#t005fn001){ref-type="table-fn"} C2 ---[\*](#t005fn002){ref-type="table-fn"} ---[\*](#t005fn002){ref-type="table-fn"} 36 ± 2[^b^](#t005fn001){ref-type="table-fn"} 2200 ± 49[^b^](#t005fn001){ref-type="table-fn"} D1 467 ± 56[^c^](#t005fn001){ref-type="table-fn"} 0.87 ± 0.02[^c^](#t005fn001){ref-type="table-fn"} 10.9 ± 0.5[^d^](#t005fn001){ref-type="table-fn"} 1623 ± 36[^c^](#t005fn001){ref-type="table-fn"} D2 ---[\*](#t005fn002){ref-type="table-fn"} ---[\*](#t005fn002){ref-type="table-fn"} 75 ± 3[^a^](#t005fn001){ref-type="table-fn"} 3461 ± 10[^a^](#t005fn001){ref-type="table-fn"} ^a--e^ Means within the same column with different letters are significantly different (p\< 0.05). \*It was not possible to calculate the parameters by the Oswald de Waele model. In [Fig 4B](#pone.0171469.g004){ref-type="fig"}, the heat-treated samples of yogurt with pre-gelled commercial curdlan and curdlan produced by the microorganism displayed a peak in shear strain at the beginning of the growth of the deformation rate. This peak is related to the lack of homogeneity of the product, because during the gelling process of curdlan small lumps are formed, which are not necessarily noticeable to the palate, but which alter rheology. With the presence of these lumps, the tension needed for a small deformation is high, and as a result, it was not possible to calculate the K and n parameters for samples C2 and D2 by the Oswald de Waele model. From the data of [Table 5](#pone.0171469.t005){ref-type="table"} it is clear that the higher viscosity samples in [Fig 4C and 4D](#pone.0171469.g004){ref-type="fig"}, which are those subjected to heat treatment, had more evident thixotropic characteristics (higher hysteresis), as they underwent a major reduction in apparent viscosity with time, in a rate constant with shearing. This is due to the breakdown of the organized yogurt structure when submitted to a determined strain. After remained stationary, the samples returned to their original state more quickly, but with lower viscosity during the return, being this recovery dependent on time. Therefore, the gel formed with heat treatment presented low stability, in particular the gel of the curdlan produced by the microorganism, corroborating with the syneresis data of the yogurt. Curiously, it was observed that of the samples without heat treatment, that which had the greater viscosity in the increasing shear rate curve did not present higher K and n values, or in other words, the sample with pre-gelled commercial curdlan was greater than the curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) in the consistency index and flow behavior. This was due to the sharp drop in viscosity in the decreasing curve of the latter, revealing again the lower stability of the gel formed. The results of the yield strength parameter (τ~0~) were consistent with the firmness of the yogurt, as the yogurt with curdlan produced by *Agrobacterium* sp. IFO 13140 with thermic treatment withstood the greatest tension before suffering deformation, followed by the yogurts with pre-gelled commercial curdlan and commercial curdlan with heat treatment. It is important to note that yield stress refers to the maximum stress that the material can bear before yielding (in the elastic deformation regime) and that the greatest τ~0~ values are those of the firmer samples. These samples displayed a minimum stress for deformation that was more difficult to break due to the increased organization and stiffening of the protein network formed, originating from the intermolecular hydrophobic interactions of the curdlan with heating. Conclusions {#sec028} =========== The characteristics of dispersion, gelation and the rheological properties of curdlan depended greatly on the recovery methods employed after its production, as the curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) and the commercial pre-gelled curdlan dispersed better in water, acted as thickeners and formed more homogeneous gels. The use of the pre-gelation method exerts a major influence on the size of the particles obtained and in the presence of NaCl, which contributes significantly to the dispersion and gelation characteristics described. However, the recovery method employed did not influence the structure of the polysaccharide. Even the pre-gelled commercial curdlans and those produced by the microorganism through the pre-gelation method had significantly different technological properties. The curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) showed a greater thickening and water and oil holding capacity than the pre-gelled commercial curdlan, while the latter demonstrated a considerably greater gelling capacity, that is related to degree of polymerization of the polysaccharide. As a consequence, although both types of curdlan increased firmness, viscosity and reduced the syneresis of yogurts, the curdlan produced by *Agrobacterium* sp. IFO 13140 (pre-gelation method) caused a greater increase in the parameters, but also showed a less stable gel formation than the pre-gelled commercial curdlan. The commercial curdlan did not influence the firmness and syneresis of yogurts tested. Nevertheless, the curdlans recovered by the pre-gelation method provided rigid gels with a stable structure, allowing improvements in the texture of a range of products, and therefore, has great potential for application in the food industry. Supporting information {#sec029} ====================== ###### Curdlan interaction with aniline blue dye. (DOC) ###### Click here for additional data file. The authors are thankful to the Brazilian Agencies CAPES, CNPq and Fundação Araucária for their financial support of this work. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: **Conceptualization:** CSM GM.**Data curation:** CSM TTS VCF LNK SBSF.**Formal analysis:** CSM VCF MLB.**Funding acquisition:** GM MLB.**Investigation:** CSM GM.**Methodology:** CSM MLB GM.**Project administration:** GM.**Resources:** MLB GM.**Supervision:** CSM GM.**Validation:** CSM GM.**Visualization:** CSM GM.**Writing -- original draft:** CSM GM.**Writing -- review & editing:** MLB GM.

{ "pile_set_name": "PubMed Central" }

The content published in Cureus is the result of clinical experience and/or research by independent individuals or organizations. Cureus is not responsible for the scientific accuracy or reliability of data or conclusions published herein. All content published within Cureus is intended only for educational, research and reference purposes. Additionally, articles published within Cureus should not be deemed a suitable substitute for the advice of a qualified health care professional. Do not disregard or avoid professional medical advice due to content published within Cureus. Introduction ============ Enchondroma is a benign cartilaginous lesion commonly found in small bones of the hands and feet \[[@REF1]\]. It is common and affects all age groups. Generally being asymptomatic, it is discovered incidentally during an unrelated radiographic examination or in case of a pathological fracture. Proximal locations such as proximal humerus and scapula are extremely rare sites \[[@REF2]\]. Radiographs show a lytic lesion with intralesional calcification. Asymptomatic lesions can be managed conservatively with follow-up and serial radiographs. If the lesion grows or if it becomes symptomatic, extended curettage usually is curative \[[@REF3]\]. We report the case of a six-year-old male child with enchondroma of the acromion, which, to the best of our knowledge, has never been published. Case presentation ================= A six-year-old male child presented to our orthopaedics department with the complaints of pain and swelling over the right shoulder for six months. Both pain and swelling were of insidious onset and had gradually progressed. Pain was of dull-aching type. This was not associated with radiation of the pain or any history of trauma, fever or night pains. On clinical examination, there was a globular, tender swelling, about 7x7 cm in size with firm consistency, present over the posterolateral corner of the right shoulder. Overlying skin was normal and there were no dilated veins, scar marks or sinuses. There was no local rise in temperature, lymphadenopathy, distal neurovascular deficit or any other swelling elsewhere in the body. Blood investigations showed normal counts and kidney and liver functions. Serum calcium, erythrocyte sedimentation rate, C-reactive protein, alkaline phosphatase and parathyroid hormone were normal. X-ray of the right shoulder showed an expansile lytic lesion in the acromion process of scapula with no periosteal reaction or soft tissue component. CT scan showed an expansile lytic lesion with incomplete septations and a cortical breach. MRI showed the lesion to be hypointense in T1 images and hyperintense in T2 images (Figure [1](#FIG1){ref-type="fig"}). A skeletal survey did not reveal any other lesion. ![Preoperative radiographic assessment\ (A) X-ray showing an expansile lytic lesion in the acromion; (B) CT scan showing cortical breach and incomplete septations; (C) MRI showing the lesion to be hypointense on T1 and hyperintense on T2.](cureus-0012-00000007630-i01){#FIG1} The differential diagnoses of simple bone cyst, aneurysmal bone cyst and chondroblastoma were kept in mind. A core biopsy from the posterolateral corner of the shoulder was done. Straw coloured fluid within the bony cavity was found. Biopsy specimen was sent for histopathological examination, which revealed lobules of hypocellular cartilage separated by fibroconnective stroma, features suggestive of enchondroma. Due to persistent pain in the subsequent follow-ups, an extended curettage with hydrogen peroxide and fibular bone grafting of the lesion was done (Figure [2](#FIG2){ref-type="fig"}). The patient was discharged on the third day after wound inspection. Sutures were removed on day 14, and passive range of motion exercises of the shoulder were started as tolerated by the patient. ![Extended curettage and fibular grafting\ (A) Posterolateral surgical approach; (B) postoperative X-ray.](cureus-0012-00000007630-i02){#FIG2} At one-year follow-up, the patient remained symptom-free, having complete and painless range of motion of the shoulder joint (Figure [3](#FIG3){ref-type="fig"}). The patient was explained about the need of a long-term regular follow-up. ![Clinical assessment at one-year follow-up\ (A) Well-healed surgical scar; (B) full elevation at the right shoulder.](cureus-0012-00000007630-i03){#FIG3} Discussion ========== Chondroma is a benign cartilaginous lesion, and is called an enchondroma when arising from the medullary canal. It is called a perisosteal or juxtacortical chondroma when it arises from the surface of the bone \[[@REF4]\]. Enchondroma is a benign tumour of the bone containing hyaline cartilage. Multiple enchondromatosis is known as Ollier's disease, and when associated with hemangioma of overlying soft tissues, it referred to as Maffuci's syndrome \[[@REF5]\]. It is the most common tumour of the hands and feet. Proximal femur, pelvis, proximal humerus and scapula are rare sites. Proximal sites are more prone to recurrence and malignant change (chondrosarcoma), and hence should be monitored regularly \[[@REF6]\]. Various lesions arising from the acromion, such as osteochondroma, simple bone cyst (SBC), aneurysmal bone cyst (ABC), giant cell tumour with secondary ABC, chondroblastoma, metastasis and rarely multiple myeloma, have been described in the literature. Our six-year-old patient with expansile lytic lesion of the acromion had differential diagnoses of SBC, ABC, and chondroblastoma. Both SBCs and ABCs present as lytic lesions, but ABCs are blood-filled cavities with septations and fluid-fluid levels on MRI scan \[[@REF7]\]. Chondroblastoma is a rare, benign tumour with predilection for proximal humerus, but has also been reported in the acromion \[[@REF8]\]. Our case was proved to be a solitary enchondroma on biopsy. It could be managed conservatively with serial radiographs, but an extended curettage and fibular bone grafting was done due to persistent pain and its proximal location which made it prone to a malignant change. Various other options of treatment for an acromial tumour such as en bloc excision with or without stabilisation have been described \[[@REF9]\]. Conclusions =========== Enchondroma represents a rare differential diagnosis of a lytic lesion of the acromion and must be kept in mind. Extended curettage and bone grafting in a symptomatic patient yields excellent clinical outcomes. A long-term and regular follow-up is required to monitor for recurrence or malignant transformation. The authors have declared that no competing interests exist. Consent was obtained by all participants in this study

{ "pile_set_name": "PubMed Central" }

INTRODUCTION ============ Standard of care for suspicious microcalcifications in the breast is stereotactic-guided percutaneous biopsy or mammography (MG)-guided wire localization and surgical excision ([@B1]). Stereotactic biopsy, however, requires a sufficient breast volume to guarantee a safe procedure limiting its universal application. Ultrasound (US)-guided biopsy has many advantages over the MG-guided procedure including no ionizing radiation, greater patient comfort, no need for breast compression, real-time confirmation of needle location, lower cost, and universally available sonography equipment ([@B1]-[@B4]). Improvements in the resolution of sonographic equipment have led to the use of US-guided biopsy for histologic assessment of suspicious microcalcifications in the breast those are apparent on sonography ([@B1], [@B5]-[@B12]). Due to its higher sensitivity and lower rate of common sampling errors, US-guided vacuum-assisted biopsy (VAB) has been preferred to 14-gauge (G) automated core needle biopsy (CNB) ([@B5], [@B6], [@B10]-[@B12]). 14-G semi-automated CNB is a coaxial system similar to VAB in terms of the single skin puncture and repeated needle passes. Using an introducer, radiologists can precisely obtain contiguous tissue specimens and reduce the procedure time. Additionally, 14-G semi-automated CNB device is less costly than the VAB device. The purpose of the present study was to evaluate the rate of successful retrieval of microcalcifications and diagnostic accuracy of US-guided 14-G semi-automated CNB for suspicious microcalcifications in the breast those were apparent on sonography. MATERIALS AND METHODS ===================== Between January 2008 and September 2011, 138 lesions containing isolated suspicious microcalcifications without an associated mass or asymmetry were detected on contact MG. Based on findings from magnified images, each lesion was categorized as suspicious (category 4) or highly suggestive of malignancy (category 5) according to the American College of Radiology Breast Imaging-Reporting and Data System (BI-RADS) ([@B13]). Category 4 lesions were subcategorized as low (A), intermediate (B), or moderate (C) suspicious lesions. To select between US-guided 14-G semi-automated CNB and MG-guided wire localization and surgical excision, targeted US was performed for all suspicious calcifications. We prospectively determined the biopsy method and excluded lesions that were not apparent on US. In addition, lesions located in the posterior portion of the mammary zone of very thin patients who had obliterated retromammary fat layer were excluded to avoid complications such as pneumothorax. Finally, 33 consecutive lesions were selected for US-guided 14-G semi-automated CNB and were enrolled into the study. Written informed consent for US-guided 14-G semi-automated CNB was obtained from all patients and institutional review board approval was obtained from our institute. Targeted US (iU22®; Philips, Bothell, WA, USA) examination was performed using a linear array transducer (L12-5) to identify a correlating lesion containing microcalcifications. Unilateral orthogonal mammographic views were obtained with a skin marker only when there was uncertainty about the mammography-sonography correlation. The biopsy was performed using a STERICUT® 14-G semi-automated CNB set (TSK Laboratory, Tochigi, Japan). One breast radiologist with 10 years of experience performed both breast US and US-guided 14-G semi-automated CNB in all cases. For biopsy, the skin was punctured using an introducer with a stylet and advanced forward near the margin of the lesion. The stylet was removed and a biopsy gun was inserted through the introducer. The biopsy needle was pushed manually into the lesion. The location of the needle within the lesion was confirmed on transverse and longitudinal scans. The biopsy gun was fired, causing it to follow the biopsy needle and remove tissue from the lesion. The biopsy gun was then withdrawn through the introducer and the needle was pushed manually. The specimen could then be observed in the notch. Insertion of the biopsy gun through the introducer was repeated after changing the angle of insertion. Several cores of more than 1 cm length were obtained for each lesion. A 1- or 2-cm cutting edge excursion could be used; we usually selected a 2-cm excursion. After the biopsy, the retrieved cores were placed on wet paper or a plastic weigh boat and specimen radiography (Senograph DS®; GE, Buc, France) was performed to identify the presence of microcalcifications. The parameters of exposure for specimen radiography were 22 kVp, 8 mAs, and 1.8 × magnification. A core containing any calcification on specimen radiography was defined as a successfully retrieved core. The total number of retrieved cores, the number of successfully retrieved cores containing any calcification, and the total number of calcifications on specimen radiography were enumerated. Concordance between the MG findings of microcalcifications and the results of US-guided 14-G semi-automated CNB was reviewed. Radiology-pathology concordance was considered when a BI-RADS category 4 or higher lesion showed a malignant result on 14-G semiautomated CNB. There were two exceptions for category 4A lesions. In low suspicious lesions, 14-G semi-automated CNB results showing a specific benign diagnosis, such as fibroadenoma or fat necrosis, were considered to be concordant. Also, nonspecific benign results, such as fibrocystic changes (FCC), were considered to be concordant in low suspicious lesions. In contrast, for category 4B or higher lesions, nonspecific benign results were considered to be discordant because of the possible sampling error. Additionally, we reviewed the accuracy of US-guided 14-G semi-automated CNB of microcalcifications by comparing the results with surgical pathology. When both results were the same, the 14-G semi-automated CNB was considered accurate. If the 14-G semi-automated CNB result of ductal carcinoma *in situ* (DCIS) yielded a diagnosis of invasive ductal carcinoma (IDC) in a subsequent surgery, it was deemed a DCIS underestimate. If the 14-G semi-automated CNB result of atypical ductal hyperplasia (ADH) or ductal intraepithelial neoplasia (DIN) 1B yielded a diagnosis of DCIS in a subsequent surgical excision, it was deemed an ADH underestimate. Mean lesion size, morphology, and distribution of microcalcifications on magnified images, and associated features of calcifications, such as mass or non-mass lesions on US, were analyzed. These factors were compared between the successful and failed retrieval groups. Biopsy-related complications including infection, hemorrhage, and pneumothorax were checked by radiology reports and chart review. The chi-square test and Fisher\'s exact test were performed for nonparametric variables and Student\'s *t* test was performed for parametric interference. Statistical analyses were performed using SPSS 17.0 software (SPSS Inc., Chicago, IL, USA). *p* values \< 0.05 were considered to indicate statistical significance. RESULTS ======= Thirty-three (23.9%, 33/138) lesions in 32 patients were subjected to US-guided 14-G semi-automated CNB. One patient had suspicious microcalcifications in both breasts. Other suspicious microcalcifications underwent MG-guided wire localization excision (76.1%, 105/138). Mean patient age was 49 years (range, 24-67 years). Thirty patients were asymptomatic and were detected on screening MG. One patient had bloody nipple discharge and the other had palpable cancer in the upper inner quadrant and incidental suspicious microcalcifications in the lower inner quadrant of the same breast. Magnified images identified four low suspicious lesions (4A; 15.2%), five intermediate suspicious lesions (4B; 12.1%), eight moderate suspicious lesions (4C; 24.2%), and 16 highly suggestive of malignancy lesions (5; 48.5%) ([Table 1](#T1){ref-type="table"}). Calcifications were successfully retrieved in 30/33 lesions (90.9%) and confirmed on specimen radiography. No complications related to 14-G semi-automated CNB were observed. Thirty cases of successfully retrieved microcalcifications were assigned to the successful retrieval group and the remaining three lesions were assigned to the failed retrieval group. Results of US-guided 14-G semi-automated CNB identified 26 (78.8%) malignant and seven (21.2%) benign lesions ([Fig. 1](#F1){ref-type="fig"}). The successful retrieval group comprised 25 malignant (15 DCIS and 10 IDC) and five benign lesions. The five benign lesions were one ADH (or DIN 1B), one columnar cell hyperplasia (CCH), one FCC, one fibroadenoma, and one fat necrosis. In the failed retrieval group, one DCIS and two FCC were identified. After 14-G semi-automated CNB, 30 lesions underwent cancer surgery (n = 26) or repeated biopsy after MG-guided wire localization (n = 4), and final pathology results were provided ([Fig. 1](#F1){ref-type="fig"}). Of the 16 DCIS on 14-G semi-automated CNB, 14 were confirmed to be DCIS on final pathology ([Fig. 2](#F2){ref-type="fig"}). The remaining two DCIS, which belonged to the successful retrieval group, were upgraded to IDC. The rate of DCIS underestimates was 12.5% (2/16) and their BIRADS assessments were category 4C and 5, respectively. All 10 IDC on 14-G semi-automated CNB were confirmed to be IDC on final pathology. Of the seven benign lesions, four underwent MG-guided wire localization and excision, and were finally diagnosed as three FCC and one DCIS. The latter was the only false-negative case of 14-G semi-automated CNB and belonged to the failed retrieval group. The other three benign lesions underwent mammographic follow-up for 2 years and no missed cancers were found. There was no ADH underestimate. According to the BI-RADS categories, three of four category 4A calcifications were successfully retrieved on US-guided 14-G semi-automated CNB. The results were all benign (one fibroadenoma, one fat necrosis, and one FCC) and concordant ([Table 1](#T1){ref-type="table"}). None of the successfully retrieved benign lesions showed a missed cancer after 2 years of mammographic follow-up. One low suspicious lesion diagnosed as DCIS on biopsy, despite failed retrieval of microcalcifications, was finally confirmed to be DCIS after surgery, confirming the 14-G semi-automated CNB results to be accurate. Four of five category 4B calcifications were successfully retrieved and 14-G semi-automated CNB results were all malignant (one DCIS and three IDC) and concordant ([Table 1](#T1){ref-type="table"}). All lesions were finally diagnosed as one DCIS and three IDC after breast cancer surgery, confirming the 14-G semi-automated CNB results to be accurate. Microcalcification retrieval failed in one intermediate suspicion lesion. Although the 14-G semi-automated CNB result was FCC and considered to be discordant, the lesion was finally diagnosed as FCC after MG-guided wire localization excision, confirming the 14-G semi-automated CNB result to be accurate. All of the category 4C lesions were successfully retrieved and the 14-G semi-automated CNB results were one ADH (or DIN 1B), four DCIS, and three IDC. The ADH case was considered discordant, but was finally confirmed to be FCC after MG-guided wire localization excision, confirming the 14-G semi-automated CNB result to be accurate. All of the moderate suspicious lesions except one were finally confirmed to be malignant after breast cancer surgery. In one case among them, however, the final pathology result was upgraded to IDC, confirming the 14-G semi-automated CNB result to be a DCIS underestimate. Of the 16 category 5 calcifications, 15 were successfully retrieved ([Table 1](#T1){ref-type="table"}). The 14-G semi-automated CNB results in the successful retrieval group were 10 DCIS, four IDC, and one CCH. Of the category 5 lesions, 14 were finally diagnosed as malignancy after surgery. In one case, however, the final pathology result was upgraded to IDC, confirming the 14-G semi-automated CNB result to be a DCIS underestimate. The CCH case was considered to be discordant, but was finally confirmed to be FCC after MG-guided wire localization excision, confirming the 14-G semi-automated CNB result to be accurate ([Fig. 3](#F3){ref-type="fig"}). Retrieval of calcification failed in one highly suspicious lesion and its 14-G semi-automated CNB result was FCC, which was considered to be discordant. The case was finally diagnosed as DCIS after MG-guided wire localization and surgical excision, confirming the 14-G semi-automated CNB result to be a false-negative. Of the 30 lesions in the successful retrieval group, two (6.7%) showed a radiology-pathology discordance, whereas two of the three (66.7%) lesions in the failed retrieval group showed discordance. The discordance rate was higher in the failed retrieval group (*p* = 0.035). Of the 30 lesions in the successful retrieval group, 28 (93.3%) lesions were accurately diagnosed, and two of the three (66.7%) lesions in the failed retrieval group were accurately diagnosed. The accuracy of 14-G semi-automated CNB was not significantly different (*p* = 0.256). The overall accuracy of US-guided 14-G semi-automated CNB was 90.9% (30/33) because of two DCIS underestimates and one false-negative case. We compared the characteristics of specimen radiography according to the retrieval of calcifications ([Table 2](#T2){ref-type="table"}). The total number of retrieved cores did not differ between groups: 6.5 (range, 4-9) in the successful retrieval group and 7.3 (range, 6-10) in the failed retrieval group (*p* = 0.61). In the successful retrieval group, the mean number of cores containing calcification was 4.8 (range, 1-8) and the mean total number of calcifications in the retrieved cores was 24 (range, 2-101). We reviewed imaging findings of microcalcifications, including the extent of the calcification, morphology, and distribution on MG, and associated features on US ([Table 3](#T3){ref-type="table"}). Based on the MG findings, the extent of calcification did not differ between two groups (16.7 mm \[range, 3-70 mm\] in the successful retrieval group and 18.0 mm \[range, 9-30 mm\] in the failed retrieval group, p = 0.093). Punctate (66.7%, 2/3) calcifications were more common in the failed retrieval group than in the successful retrieval group (*p* = 0.033). Linear/branching (33.3%, 10/30) and coarse heterogeneous (16.7%, 5/30) calcifications were more common in the successful retrieval group. The proportions of linear/segmental and clustered distributions were similar in both groups (36.7% vs. 33.3%; 60.0% vs. 66.7%; *p* = 0.095). On US, microcalcifications were associated with mass (63.3%; 19/30) or non-mass (30.0%; 9/30) lesions in the successful retrieval group, whereas they were associated with mass (100%, 3/3) in the failed retrieval group. DISCUSSION ========== For suspicious microcalcifications in the breast those are either not palpable or not associated with a mass on MG, standard of care is stereotactic-guided VAB or MG-guided wire localization and surgical excision ([@B1], [@B2], [@B9], [@B14]). Stereotactic equipment, however, is not universally applicable. Further, an adequate breast thickness is essential for safe procedures ([@B15]). As high-resolution sonography equipment is universally available, US-guided VAB can be an effective alternative to stereotactic-guided VAB for suspicious calcifications in the breast ([@B10]-[@B12]). VAB is superior to automated CNB for the retrieval of microcalcifications because the former provides larger tissue specimens and contiguous sampling ([@B1]-[@B3], [@B5], [@B6], [@B9]-[@B12], [@B14]). On the other hand, using a 14-G semi-automated CNB device, radiologists can precisely obtain multiple cores of calcifications because the introducer reduces reverberation artifacts and the possibility of missing the target ([@B16]). The cost of 14-G semi-automated CNB is lower than that of VAB. Using an introducer in the semi-automated gun does not significantly add to the cost. Thus, US-guided 14-G semi-automated CNB may be an acceptable and cost-effective procedure for suspicious microcalcifications compared with US-guided VAB. In the present study, the retrieval rate of microcalcifications using the semi-automated gun was 91%. The retrieval rates of microcalcifications in VAB are reported to be 91-100% under stereotactic guidance ([@B1], [@B2]) and 71-97% under US guidance ([@B10]-[@B12]). Thus, in the retrieval of microcalcifications, US-guided 14-G semi-automated CNB is comparable to VAB, particularly considering the smaller tissue specimens and lower cost. The retrieval rate of microcalcifications using an automated gun is 86-94% ([@B1], [@B2]) and the results of the present study are comparable to US-guided automated CNB. Interestingly, the retrieval rate in the present study was superior to that in a report of US-guided VAB (91% vs. 71%) ([@B11]). This difference may be related to the exclusion criteria used in the present study. We excluded calcifications that were located in the posterior portion of the mammary zone in very thin patients who had obliterated retromammary fat layer to avoid pneumothorax. Cho et al. ([@B11]) reported that microcalcifications located in the anterior or middle part of the breast were retrieved more frequently than those in the posterior part of the breast. The lower retrieval of the latter likely reflected the difficulty in inserting a biopsy needle to a sufficiently deep position because of fear of an iatrogenic pneumothorax. So, the retrieval rate of posteriorly located calcifications was lower than that of other calcifications. The accuracy of 14-G semi-automated CNB was 91%, due to the two DCIS underestimates and one false-negative diagnosis. The accuracies of US-guided VAB were 94-96% ([Table 4](#T4){ref-type="table"}) ([@B10]-[@B12]). These excellent results, however, came from a mean of 10-17 VAB cores, which represent larger tissue specimens compared with stereotactic-guided VAB (mean, 11-13 cores) ([@B1], [@B2]) and a mean of seven cores in the present study. Histologic underestimation by 14-G semi-automated CNB is mostly an issue in calcified lesions rather than in mass lesions, and DCIS underestimates are more common in 14-G semi-automated CNB than that of VAB due to sampling error ([@B2]). In stereotactic-guided biopsy of microcalcifications, the rate of DCIS underestimates is reported to be 16-37% with 14-G semi-automated CNB and 0-23% with VAB ([@B2], [@B9], [@B14]). In US-guided biopsy of microcalcifications, the rate of DCIS underestimates ranges from 7-19% with VAB ([Table 4](#T4){ref-type="table"}) ([@B10]-[@B12]). Hahn et al. ([@B12]) reported a rate of DCIS underestimates of 0%, but an ADH (or DIN 1B) underestimation of 33% (1/3). In the present study, the rate of DCIS underestimates of 14-G semi-automated CNB was 13% (2/16), which is comparable to that of US-guided VAB and there was no ADH underestimate. One false-negative diagnosis was identified. The MG finding in that case was segmental fine pleomorphic calcifications and was assessed as a BI-RADS category 5. On sonography, a few calcifications with subtle non-mass lesions were identified. US-guided VAB or MG-guided wire localization excision would be a more desirable procedure for the particular case. Fortunately, the discordant 14-G semi-automated CNB result led the patient to act quickly on the diagnosis of breast cancer. The false-negative rate of the present study was 3.7% (1/27), which is comparable to those of US-guided automated CNB (mean, 2.8%; range, 0.3-8.2%) and MG-guided wire localization excision (mean, 2%; range, 0-8%) ([@B1]). Breast microcalcifications are observed on sonography as distinct echogenic dots without posterior acoustic shadowing and are more readily identified when associated with a hypoechoic mass or duct-like structures than when no associated sonographic abnormality is present ([@B1], [@B7]). In the present study, 31 of 33 microcalcifications visible on sonography were associated with a mass or nonmass lesions on sonography. The remaining two lesions of isolated calcifications on sonography were retrieved successfully and accurately diagnosed. Although the number of cases was too small to deduce a conclusive inference, it could be suggested that isolated calcifications without associated lesions could be successfully retrieved using 14-G semi-automated CNB in the era of high-resolution sonography. This study has some limitations. First, the number of enrolled cases in this study was small, which may have limited a conclusive inference. Second, all procedures were performed by one experienced breast radiologist who had used 14-G semi-automated CNB for many years. The proficiency of 14-G semi-automated CNB can affect the retrieval rate of microcalcifications. Third, US-guided 14-G semi-automated CNB was not applicable to all suspicious microcalcifications; only 24% of microcalcifications were considered amenable in the current study, similar to the findings of other studies ([@B1], [@B5], [@B9]). Nevertheless, in selected cases, US-guided semi-automated 14-G CNB can provide a reliable and cost-effective diagnosis of suspicious microcalcifications. Last, we did not deploy marker clips after US-guided biopsy because they are not generally used in Korea. In conclusion, US-guided 14-G semi-automated CNB provided a retrieval rate and diagnostic accuracy of 90.9% with a DCIS underestimate rate of 13%. Thus, this could be useful procedure for suspicious isolated microcalcifications apparent on sonography. This work was supported in part by the Soonchunhyang University Research Fund. ![US-guided 14-G semi-automated core needle biopsy results and pathologic diagnosis of microcalcifications.\ ^\*^Magnification mammography, ^†^Microcalcifications. DCIS = ductal carcinoma *in situ*, IDC = invasive ductal carcinoma, US = ultrasound, ADH = atypical ductal hyperplasia, DIN = ductal intraepithelial neoplasia, CCH = columnar cell hyperplasia, FCC = fibrocystic changes](kjr-15-12-g001){#F1} ![Details from case of 54-year-old asymptomatic woman.\ **A.** Mammography shows clustered fine pleomorphic calcifications (arrows). BI-RADS assessment is category 5. **B.** Targeted US shows multiple echogenic dots (arrows) within irregular indistinct isoechoic mass. **C.** Specimen radiography after US-guided 14-G semi-automated CNB shows multiple microcalcifications (arrows) in five of six cores retrieved. Pathology results of both CNB and final surgery were DCIS. BI-RADS = Breast Imaging-Reporting and Data System, US = ultrasound, CNB = core needle biopsy, DCIS = ductal carcinoma *in situ*](kjr-15-12-g002){#F2} ![Details from case of 42-year-old asymptomatic woman.\ **A.** Magnification mammography shows compactly clustered fine linear and fine pleomorphic calcifications (arrows). BI-RADS assessment is category 5. **B.** Targeted US shows multiple microcalcifications (arrows) within oval partially indistinct hypoechoic mass. **C.** Specimen radiography shows 69 microcalcifications in five (arrows) of six cores retrieved. Pathology result of CNB was columnar cell hyperplasia with microcalcifications and that of mammography-guided localization and surgical excision was fibrocystic change with microcalcifications. BI-RADS = Breast Imaging-Reporting and Data System, US = ultrasound, CNB = core needle biopsy](kjr-15-12-g003){#F3} ###### Results of US-Guided 14-G Semi-Automated CNB According to BI-RADS Categories ![](kjr-15-12-i001) **Note.**-US = ultrasound, MG = mammography, CNB = core needle biopsy, BI-RADS = Breast Imaging-Reporting and Data System, 4A = low suspicious lesions, 4B = intermediate suspicious lesions, 4C = moderate suspicious lesions ###### Characteristics of Specimen Radiography According to Retrieval of Calcifications ![](kjr-15-12-i002) **Note.**-^\*^Microcalcifications. No = number ###### Imaging Findings of Microcalcifications According to Retrieval ![](kjr-15-12-i003) **Note.**-BI-RADS = Breast Imaging-Reporting and Data System, US = ultrasound ###### Comparison of Performance in US-Guided Biopsies for Suspicious Microcalcifications ![](kjr-15-12-i004) **Note.**-US = ultrasound, VAB = vacuum-assisted biopsy, DCIS U-E = ductal carcinoma *in situ* underestimation, NA = not applicable, A-CNB = automated core needle biopsy, SA-CNB = semi-automated core needle biopsy

{ "pile_set_name": "PubMed Central" }

Ju L, Shan L, Yin B, Song Y. δ‐Catenin regulates proliferation and apoptosis in renal cell carcinoma via promoting β‐catenin nuclear localization and activating its downstream target genes. Cancer Med. 2020;9:2201--2212. 10.1002/cam4.2857 1. INTRODUCTION {#cam42857-sec-0001} =============== Renal cell carcinoma (RCC) is the ninth most common malignancy, with 403 262 new cases diagnosed and 175 098 cancer deaths worldwide in 2018.[1](#cam42857-bib-0001){ref-type="ref"}, [2](#cam42857-bib-0002){ref-type="ref"} In the management of localized or locally advanced RCC, surgical resection including partial and radical nephrectomy is a common treatment strategy.[3](#cam42857-bib-0003){ref-type="ref"}, [4](#cam42857-bib-0004){ref-type="ref"} However, up to 40% risk of developing recurrence after nephrectomy.[4](#cam42857-bib-0004){ref-type="ref"}, [5](#cam42857-bib-0005){ref-type="ref"} For patients with metastatic RCC, cytoreductive nephrectomy followed by systemic drugs is an established treatment approach.[6](#cam42857-bib-0006){ref-type="ref"}, [7](#cam42857-bib-0007){ref-type="ref"} Nevertheless, the disease still progresses after therapy and metastasis is the major cause of their death.[1](#cam42857-bib-0001){ref-type="ref"} Therefore, treatments for the recurrence and metastasis of RCC are clearly needed to improve current therapies. δ‐Catenin, alternatively known as CTNND2 or NPRAP (neural plakophilin‐related armadillo protein), is a member of the catenin family. It is the adhesive junction‐associated protein that was originally identified to be enriched in brain tissues.[8](#cam42857-bib-0008){ref-type="ref"}, [9](#cam42857-bib-0009){ref-type="ref"}, [10](#cam42857-bib-0010){ref-type="ref"}, [11](#cam42857-bib-0011){ref-type="ref"} Recent researches have heightened that δ‐catenin is also implicated in some cancers. It is evidenced that δ‐catenin facilitates the malignant phenotype of non‐small‐cell lung cancer through binding to the juxtamembrane domain (JMD) of E‐cadherin.[12](#cam42857-bib-0012){ref-type="ref"} Beyond this, two research articles indicated that ectopic overexpression of δ‐catenin is related to prostate cancer progression. One reported δ‐catenin causes the alteration of cell cycle and survival gene profiles. The other presented the mechanism that δ‐catenin induces E‐cadherin processing and activates β‐catenin‐mediated oncogenic signals.[13](#cam42857-bib-0013){ref-type="ref"}, [14](#cam42857-bib-0014){ref-type="ref"} To date, the precise role of δ‐catenin in tumorigenesis and development of RCC is still poorly understood, neither is the correlation between its expression and clinical pathological parameters. All these deserve to be explored. β‐Catenin, a crucial member of the adherent junctions,[15](#cam42857-bib-0015){ref-type="ref"} is located at adhesion complexes at the cytoplasmic side of membrane in normal cells.[16](#cam42857-bib-0016){ref-type="ref"} β‐Catenin is phosphorylated in the cytoplasm by a multiprotein destruction complex, which leads to proteasome‐mediated degradation.[17](#cam42857-bib-0017){ref-type="ref"} In pathological conditions, due to mutations on β‐catenin sequence, blocking the activity of the destruction complex or preventing β‐catenin from being phosphorylated by GSK3‐3β, β‐catenin may stabilize and escape from its degradation fate.[18](#cam42857-bib-0018){ref-type="ref"} This results in elevation of cytoplasmic β‐catenin and its translocation to the nucleus. In the nucleus, β‐catenin interacts with transcription factors from the T‐cell factor (TCF)/Lymphoid enhancer‐binding factor (LEF) family,[19](#cam42857-bib-0019){ref-type="ref"}, [20](#cam42857-bib-0020){ref-type="ref"}, [21](#cam42857-bib-0021){ref-type="ref"} which triggers transcription of Wnt/β‐catenin target genes.[22](#cam42857-bib-0022){ref-type="ref"}, [23](#cam42857-bib-0023){ref-type="ref"} The aberrant activation of these genes involved in proliferation, apoptosis and invasion contributes to cancer progression.[24](#cam42857-bib-0024){ref-type="ref"}, [25](#cam42857-bib-0025){ref-type="ref"}, [26](#cam42857-bib-0026){ref-type="ref"} Moreover, it has been demonstrated that β‐catenin can be activated in RCC development,[27](#cam42857-bib-0027){ref-type="ref"}, [28](#cam42857-bib-0028){ref-type="ref"} suggesting a possible dysregulation of this protein. However, the regulation of β‐catenin by δ‐catenin and the corresponding mechanism have not been studied yet. In the present study, we detected the expression of δ‐catenin in RCC specimens by Western blot analysis and immunohistochemistry staining, and then analyzed the correlation between its expression and clinicopathological factors. To gain insights into the role of δ‐catenin in RCC, we explored whether it was implicated in proliferation and apoptosis of RCC cells. In addition, the regulation of δ‐catenin on β‐catenin and β‐catenin‐mediated oncogenic signals were also investigated. Moreover, we determined the influence of δ‐catenin on tumor growth in the xenograft mouse model. Our results illuminated that elevated δ‐catenin expression in RCC caused the activation of β‐catenin and its target genes, thereby affecting proliferation and apoptosis. This cognition of δ‐catenin may help to establish a novel targeted therapy and improve current therapies of RCC. 2. MATERIALS AND METHODS {#cam42857-sec-0002} ======================== 2.1. Cell lines and cell culture {#cam42857-sec-0003} -------------------------------- Normal human renal epithelial cells (HK‐2) and human renal cancer cell lines (A498, ACHN and 786‐O) were obtained from Procell Life Science & Technology (Wuhan, China) and cultured in MEM medium or RPMI‐1640 medium supplemented with 10% fetal bovine serum (BI). All the cells were maintained in a humidified incubator with 5% CO2 at 37°C. 2.2. Nuclear and cytoplasm protein extraction {#cam42857-sec-0004} --------------------------------------------- The cells were lysed in 200 μL cytoplasmic lysis buffer A supplemented with PMSF on ice for 10‐15 minutes, and then added with 10 μL cytoplasmic lysis buffer B. The mixture was vortexed for 5 seconds and centrifuged at 12 000 × *g* for 5 minutes at 4°C. The supernatant was a cytoplasmic extract. The pellet was resuspended with 50 μL nuclei lysis buffer (Beyotime, P0028) and kept in ice for 30 minutes (vortexed 15‐30 seconds at a high speed every 1‐2 minutes). Nuclear proteins were extracted by centrifugation at 12 000 × *g* for 10 minutes at 4°C. The supernatant was the nuclear extract. The protein concentration was assessed by Enhanced BCA Protein Assay Kit (Beyotime, P0009). 2.3. Western blot analysis {#cam42857-sec-0005} -------------------------- Cells were collected and lysed with RIPA lysis buffer (Beyotime, P0013B). The protein concentrations were evaluated via Enhanced BCA Protein Assay Kit (Beyotime). Protein lysates were separated by electrophoresis on a sodium dodecyl sulfate‐polyacrylamide gel (8%‐15%), transferred onto PVDF membranes, and incubated with primary antibodies at 4°C overnight followed by goat anti‐rabbit or goat anti‐mouse IgG (Proteintech) at 37°C for 40 minutes. The protein bands were visualized by enhanced chemiluminescence (ECL). Primary antibodies used in this study were as follows: anti‐δ‐Catenin antibody (Bioss, bs‐22251R), anti‐Cyclin D1 antibody (CST, \#2922), anti‐CDK6 antibody (Proteintech, 19117‐1‐AP), anti‐CDK4 antibody (CST, \#12790), anti‐c‐myc antibody (Proteintech, 10828‐1‐AP), anti‐Bcl2L1 antibody (Proteintech, 10783‐1‐AP), anti‐cleaved‐caspase‐3 antibody (CST, \#9661), anti‐survivin antibody (Proteintech, 10508‐1‐AP), anti‐β‐catenin antibody (CST, \#8480), anti‐β‐actin antibody (Proteintech, 60008‐1‐Ig), anti‐Histone H3 antibody (Proteintech, 17168‐1‐AP). 2.4. Plasmid and siRNA transfection {#cam42857-sec-0006} ----------------------------------- A498 and ACHN cells were transfected by δ‐Catenin‐specific siRNA (5ʹ‐CCCUAGCAGUACUGACCAATT‐3ʹ and 5ʹ‐UUGGUCAGUACUGCUAGGGTT‐3ʹ) or siCtrl (5ʹ‐UUCUCCGAACGUGUCACGUTT‐3ʹ and 5ʹ ‐ACGUGACACGUUCGGAGAATT‐3ʹ) encapsulated with Lipofectamine 2000 (Invitrogen, 11668‐019) following its transfection protocol. After 48 hours of transfection, subsequent assays were performed as described. ACHN cells were transfected by shRNA plasmids or shCtrl plasmids as the negative control with Lipofectamine 2000 for silencing δ‐catenin. After 24 hours, the cells were selected in media supplemented with 300 μg/mL G418 to obtain the stably transduced cells. 2.5. RNA extraction and quantitative real‐time PCR (qRT‐PCR) {#cam42857-sec-0007} ------------------------------------------------------------ Total RNA was isolated from RCC cells using total RNA rapid extraction kit (BioTeke, RP1201) according to the manufacturer\'s protocol. The obtained RNA was reverse transcribed into cDNA by M‐MLV (Takara, 2641A). The cDNA was used to perform real‐time PCR with Taq HS Perfect Mix (Takara, R300A) and SYBR Green (BioTeke, EP1602). The data were determined by Exicycler™ 96 (Bioneer, Daejeon, Korea), calculated by 2^−∆∆Ct^ parameters. Sequences of real‐time PCR primers were as follows: δ‐Catenin: forward: 5ʹ‐TCTGAGAAACCTGGTGTATGG‐3ʹ, reverse: 5ʹ‐CAGTCGTCTTGCGGAGTAA‐3ʹ, β‐actin: forward: 5ʹ‐CCACTGCCGCATCCTCTT‐3ʹ, reverse: 5ʹ‐GGTCTTTACGGATGTCAACG‐3ʹ. 2.6. CCK‐8 assay {#cam42857-sec-0008} ---------------- Cell counting kit‐8 (CCK‐8) assay was employed to detect cell viability. RCC cells were seeded into 96‐well plates with 4 × 10^3^ per well beforehand. After being adhered, the cells were transfected with siCtrl and δ‐catenin siRNA; and CCK‐8 reagent (Sigma) was added into the 96‐well plates with 10 µL per well and incubated for 1 hour. The optical density (OD) value was measured at 450 nm with a microplate reader (BioTek). The OD450 values at 0, 24, 48, 72, and 96 hours were tested, respectively. 2.7. Immunofluorescence {#cam42857-sec-0009} ----------------------- The slides of cells were fixed for 15 minutes with 4% paraformaldehyde and permeabilized with 0.1% tritonX‐100 for 30 minutes. The samples were blocked by goat serum for 15 minutes. Primary antibody (diluted to 1:200 with PBS) such us Ki67 and β‐catenin was added to the cells and incubated overnight at 4°C. After PBS washing, cells were incubated with Cy3‐labeled goat anti‐rabbit IgG (Beyotime, A0516) diluted in PBS (1:200) for 1 hours at room temperature. The nuclei were counterstained by DAPI. Finally, the samples were sealed by anti‐fade reagent and observed using the OLUMPUS fluorescence microscope (at 400 × magnification). 2.8. Flow cytometry {#cam42857-sec-0010} ------------------- Flow cytometry was performed to detect the cell cycle and apoptosis of RCC cells. After transfection for 48 hours, cells were collected, treated by Cell Cycle Analysis kit (Beyotime) following the manufacturer\'s protocol, and detected using NovoCyte flow cytometer (ACEA Biosciencs). For the apoptosis analysis, the cells were treated with Annexin V‐FITC Apoptosis Detection Kit (Beyotime) and accessed through flow cytometer. 2.9. TOP/FOP luciferase reporter assay {#cam42857-sec-0011} -------------------------------------- ACHN cells transfected with δ‐catenin siRNA or siCtrl were seeded in 6‐well plates. After incubation for 24 hours, the cells were transfected with 2 μg β‐catenin responsive firefly luciferase reporter plasmid TOPflash or negative control FOPflash and 0.5 μg constitutively active vector encoding Renilla luciferase using lipofectamine 2000. At 24 hours after transfection, the cells were lysed and then firefly and Renilla luciferase activities were determined using the dual luciferase reporter assay system (Promega) according to the manufacturer\'s instructions. 2.10. Animal studies {#cam42857-sec-0012} -------------------- All animal experiments in this study were approved by the Institutional Animal Ethics Committee of China Medical University, and were carried according to the Guideline for the Care and Use of Laboratory Animals. Male Balb/c nude mice at the age of 6 weeks were divided into two groups randomly (6 mice per group). ACHN cells (1 × 10^7^) with stable knockdown of δ‐catenin with shRNA plasmids or ACHN cells transfected with shCtrl plasmids were injected subcutaneously into the right flank of the nude mice. The tumor size was measured every four days, and the tumor volume was calculated according to the following: *V* = (width^2^ × length)/2. The mice were sacrificed 27 days after injection. Tumors were excised and the tumor weight was recorded. 2.11. Immunohistochemistry staining assay {#cam42857-sec-0013} ----------------------------------------- The slides were heated at 60°C for 2 hours, deparaffinized in xylene, and rehydrated by graded ethanol. After antigen retrieval, the tissue sections were blocked with goat serum. The sections were then exposed to antibodies against δ‐catenin or Ki67 (Abcam, ab15580) overnight at 4°C. The slides were washed with PBS and incubated with HRP‐labeled goat anti‐rabbit antibody (ThermoFisher, \#31460) for 1 hour. After washing, the slides were added with diaminobenzidine (DAB) and counterstained with hematoxylin. In the end, the sections were dehydrated by graded ethanol, sealed by neutral balsam, and visualized with microscope (at 400 × magnification). 2.12. Statistical analysis {#cam42857-sec-0014} -------------------------- The data were presented as mean ± SD in the figures. Statistical analysis was performed with two‐tailed unpaired t‐test, one‐way ANOVA or two‐way ANOVA as indicated. *P* value \< .05 was considered statistically significant. 3. RESULTS {#cam42857-sec-0015} ========== 3.1. δ‐Catenin was highly expressed in human renal cancer tissues and correlated with poor prognosis of RCC patients {#cam42857-sec-0016} -------------------------------------------------------------------------------------------------------------------- We firstly determined the expression of δ‐catenin in human renal cancer tissues. Western blot analysis was performed to examine δ‐catenin expression from 28 cases of patients with RCC. The results showed that δ‐catenin was highly expressed in tumor tissues compared with adjacent normal tissues (Figure [1](#cam42857-fig-0001){ref-type="fig"}A and Figure [S1](#cam42857-sup-0001){ref-type="supplementary-material"}A). Furthermore, we detected the expression of δ‐catenin by immunohistochemistry in 58 cases of RCC, and evaluated the correlation between δ‐catenin expression and clinicopathological factors. The representative images of low and high expression of δ‐catenin from patients were presented in Figure [1](#cam42857-fig-0001){ref-type="fig"}B. As shown in Table [1](#cam42857-tbl-0001){ref-type="table"}, high expression of δ‐catenin was correlated with pTNM stage, tumor stage, and lymph node metastasis (*P* \< .05), but not correlated with patient\'s gender, age, tumor size, and distant metastasis (*P* \> .05). These results together demonstrated that the expression of δ‐catenin was enhanced in RCC tissues, which was associated with poor prognosis in RCC patients. ![δ‐Catenin was highly expressed in human renal cancer tissues. A, The representative bands of δ‐catenin detected by Western blot analysis from human renal cancer tissues and adjacent normal tissues (16 pairs). B, The representative images of low and high expression of δ‐catenin in human renal cancer tissues visualized by immunohistochemical staining. Scale bar, 50 μm](CAM4-9-2201-g001){#cam42857-fig-0001} ###### Relationship between δ‐catenin and clinical characteristics in RCC patients Parameters Group δ‐catenin expression *P* value ------------------------ ---------- ---------------------- ----------- ------------ Gender Male 17 13 .80628 Female 14 14 Age \<60 20 14 .47797 ≥60 11 13 Tumor size (cm) \<4 cm 1 3 .50751 ≥4 cm 30 24 pTNM stage I/II 21 27 **.00378** III/IV 10 0 Tumor stage T1‐T2 23 27 **.01384** T3‐T4 8 0 Lymph nodes metastasis Negative 25 27 **.04747** Positive 6 0 Distant metastasis Negative 29 27 .53404 Positive 2 0 A *P* value of less than .05 was considered to be statistically significant (shown in bold). John Wiley & Sons, Ltd 3.2. δ‐Catenin was highly expressed in human renal cancer cells {#cam42857-sec-0017} --------------------------------------------------------------- Next, we determined the level of δ‐catenin in normal human renal epithelial cells (HK‐2) and three different RCC cell lines (A498, ACHN and 786‐O). As expected, δ‐catenin was highly expressed in RCC cells compared with HK‐2 cells (Figure [2](#cam42857-fig-0002){ref-type="fig"}A). We employed specific siRNA to knock down δ‐catenin in A498 and ACHN cells showing its relatively higher expression. qRT‐PCR measurement of A498 and ACHN cells revealed the mRNA level of δ‐catenin was reduced after δ‐catenin knockdown (Figure [2](#cam42857-fig-0002){ref-type="fig"}B,C). The remarkable reduction of δ‐catenin protein level was also observed in RCC cells after δ‐catenin silencing (Figure [2](#cam42857-fig-0002){ref-type="fig"}D,E). All the data suggested that δ‐catenin expression was elevated in RCC cells. ![δ‐Catenin was highly expressed in human renal cancer cells. A, The expression of δ‐catenin in renal cancer cell lines tested by Western blot analysis. Renal cancer cells A498 and ACHN were treated with δ‐catenin‐specific siRNA. After 48 h transfection, B, C, δ‐catenin mRNA levels were analyzed by qRT‐PCR (n = 3). D, E, The protein levels of δ‐catenin were detected by Western blot. Data were presented as mean ± SD (n = 3) with statistic significance calculated from one‐way ANOVA for multiple comparisons (\*\**P *\< .01, \*\*\**P *\< .001, \*\*\*\**P *\< .0001)](CAM4-9-2201-g002){#cam42857-fig-0002} 3.3. δ‐Catenin promoted renal cancer cell proliferation {#cam42857-sec-0018} ------------------------------------------------------- Considering the correlation between high expression of δ‐catenin and poor prognosis, we investigated the biological role of δ‐catenin in RCC development. We performed CCK‐8 assay to evaluate the effect of δ‐catenin knockdown on RCC cell proliferation. As shown in Figure [3](#cam42857-fig-0003){ref-type="fig"}A,B, δ‐catenin knockdown resulted in decreased cell viability in A498 and ACHN cells. The immunofluorescence analysis of Ki67 also illustrated reduced proliferative capacity after δ‐catenin silencing (Figure [3](#cam42857-fig-0003){ref-type="fig"}C,D). We next explored whether δ‐catenin participated in RCC cell cycle regulation. We examined cell numbers in each phase of mitosis in RCC cells following δ‐catenin knockdown by flow cytometry. The results indicated that much higher percentages (53.87% in A498, 57.53% in ACHN) of RCC cells were arrested in G1 phase upon δ‐catenin knockdown (Figure [3](#cam42857-fig-0003){ref-type="fig"}E,F). We also determined the expression of cell cycle‐related proteins in RCC cells following δ‐catenin silencing. The expression of cyclin D1, CDK4, and CDK6 was significantly reduced after downregulating endogenous δ‐catenin (Figure [3](#cam42857-fig-0003){ref-type="fig"}G,H). These results strongly indicated that δ‐catenin knockdown has an inhibitory effect on proliferation of RCC cells. ![δ‐Catenin promoted renal cancer cell proliferation. δ‐Catenin‐specific siRNA was employed to silence δ‐catenin in A498 and ACHN cells. A, B, After 24, 48, 72, and 96 h of transfection, CCK‐8 assay was performed to observe cell viability. C, D, The immunofluorescence analysis of the expression of Ki67 in renal cancer cells transfected with siRNA. Scale bar, 50 μm. E, F, Cell cycle was analyzed in A498 and ACHN cells by flow cytometry. G, H, The expression level of cell cycle‐associated markers was determined by Western blot. Data in panel A, B, E, F, G, and H were shown as the mean ± SD from three separate experiments, with two‐way ANOVA for A, B, E, F, and one‐way ANOVA for G and H](CAM4-9-2201-g003){#cam42857-fig-0003} 3.4. δ‐Catenin inhibited apoptosis of renal cancer cells {#cam42857-sec-0019} -------------------------------------------------------- Furthermore, we identified the effect of δ‐catenin on apoptosis of RCC cells. A498 and ACHN cells transfected with siRNA were stained with Annexin V‐FITC/PI and subjected to flow cytometric analyses. The percentage of apoptotic cells was increased in δ‐catenin‐knocked down RCC cells (Figure [4](#cam42857-fig-0004){ref-type="fig"}A,B). We also examined protein expression of apoptosis‐related markers by Western blot analysis. As shown in Figure [4](#cam42857-fig-0004){ref-type="fig"}C,D, silence of δ‐catenin led to the downregulation of Bcl2L1 (anti‐apoptotic marker) but upregulation of cleaved‐caspase 3. All these data substantiated δ‐catenin restrained the induction of apoptosis in RCC cells. ![δ‐Catenin inhibited apoptosis of renal cancer cells. A498 and ACHN were transfected with δ‐catenin‐specific siRNA for 48 h. A, B, The cells were analyzed by FACS using Annexin V‐FITC/PI staining, and the percentage of apoptotic cells was calculated. C, D, The protein levels of apoptosis‐related markers in renal cancer cells following δ‐catenin knockdown were detected by Western blot. The percentage of apoptotic cells and quantitative analysis for apoptosis‐related proteins were presented as mean ± SD with one‐way ANOVA with Tukey\'s test for multiple comparisons. \*\**P* \< .01, \*\*\**P* \< .001, \*\*\*\**P* \< .0001](CAM4-9-2201-g004){#cam42857-fig-0004} 3.5. δ‐Catenin affected β‐catenin nuclear localization and its downstream gene expression in renal cancer cells {#cam42857-sec-0020} --------------------------------------------------------------------------------------------------------------- Considering δ‐catenin could activate β‐catenin and its downstream tumor‐associated signaling pathways to participate in the development of prostate cancer,[14](#cam42857-bib-0014){ref-type="ref"} we investigated the effect of δ‐catenin on this pathway in RCC. Immunofluorescence analysis was employed to visualize the change in nuclear localization of β‐catenin under the condition of δ‐catenin silence. As shown in Figure [5](#cam42857-fig-0005){ref-type="fig"}A, δ‐catenin silencing in ACHN cells led to a remarkable decrease in nuclear β‐catenin, which was further confirmed by Western blot analysis (Figure [5](#cam42857-fig-0005){ref-type="fig"}C). To further investigate whether the altered β‐catenin distribution interacts with TCF/LEF to activate downstream gene expression, TCF/LEF transcriptional reporter activity was measured in ACHN cells. Knockdown of δ‐catenin exhibited a lower specific TCF/LEF transcriptional reporter activity (Figure [5](#cam42857-fig-0005){ref-type="fig"}B), implying an alteration in the transcription co‐activator activity of β‐catenin, which caused the subsequent downregulation of downstream target gene, such as c‐myc and survivin (Figure [5](#cam42857-fig-0005){ref-type="fig"}C). Together, these results indicated that δ‐catenin upregulated downstream gene expression of β‐catenin by promoting β‐catenin translocation to the nucleus in RCC cells. ![δ‐Catenin affected β‐catenin nuclear localization and its downstream gene expression in renal cancer cells. ACHN cells were treated with δ‐catenin specific siRNA for 48 h. A, The expression and localization of β‐catenin were observed by immunofluorescence analysis. Scale bar, 50 μm. B, The transcriptional activity of β‐catenin following δ‐catenin knockdown was analyzed by the TOP/FOP luciferase reporter assay. The relative luciferase activity was normalized and shown (n = 3). C, The expressions of β‐catenin in the whole cell lysate, cytoplasm, and nucleus as well as the protein levels of c‐myc and survivin were tested by Western blot. Data in panel B and C were shown as the mean ± SD based on one‐way ANOVA with Tukey\'s test (\**P *\< .05, \*\**P *\< .01, \*\*\**P *\< .001, \*\*\*\**P *\< .0001)](CAM4-9-2201-g005){#cam42857-fig-0005} 3.6. δ‐Catenin promoted tumor growth in ACHN xenograft mouse model {#cam42857-sec-0021} ------------------------------------------------------------------ To further detect the regulation of δ‐catenin on tumorigenesis in vivo, we examined the effect of δ‐catenin knockdown on tumor growth of ACHN cell‐derived xenograft in nude mice. The ACHN cells transfected with shCtrl plasmids and δ‐catenin shRNA plasmids were subcutaneously injected into BALB/c nude mice, which were divided into two groups (n = 6 for each group). Tumor volume was recorded over a period of 27 days and the solid tumors were photographed and weighed after the mice were sacrificed. As shown in Figures [6](#cam42857-fig-0006){ref-type="fig"}A‐C, knockdown of δ‐catenin markedly suppressed tumor growth in mice, compared with the shCtrl group. Moreover, immunohistochemical analysis of the tumor tissues also revealed δ‐catenin knockdown suppressed tumor growth, as evidenced by reduced Ki67 expression (Figure [6](#cam42857-fig-0006){ref-type="fig"}D). Furthermore, δ‐catenin knockdown restrained translocation of β‐catenin to the nucleus, as well as regulated the expression of c‐myc, surviving, cyclinD1, Bcl2L1, and cleaved‐caspase‐3 (Figure [6](#cam42857-fig-0006){ref-type="fig"}E). All these results validated that δ‐catenin facilitated tumor growth in vivo in RCC. ![δ‐Catenin promoted tumor growth in ACHN xenograft mouse model. A, The images of tumor harvested at 27 d after transplantation of ACHN cells. B, Tumor diameters were measured at a regular interval of 4 d for up to 27 d and the tumor volume was calculated. C, Tumor weight from each group was measured after the mice were sacrificed. D, The expression of δ‐catenin and Ki67 in tumor tissue sections was tested by immunohistochemistry staining. Scale bar, 50 μm. E, The protein expression of δ‐catenin, β‐catenin (in the whole cell lysate, cytoplasm, and nucleus), c‐myc, survivin, cyclinD1, Bcl2L1, and cleaved‐caspase‐3 in tumor tissues was detected by Western blot. Data in panel B were shown as the mean ± SD based on two‐way ANOVA (n = 6). Data in C and E were presented as the mean ± SD (n = 6, two‐tailed unpaired *t* test). \*\*\**P* \< .001, \*\*\*\**P* \< .0001](CAM4-9-2201-g006){#cam42857-fig-0006} 4. DISCUSSION {#cam42857-sec-0022} ============= It has been proven that δ‐catenin has ectopic overexpression in lung cancer and prostate cancer and promotes tumor progression,[12](#cam42857-bib-0012){ref-type="ref"}, [13](#cam42857-bib-0013){ref-type="ref"}, [14](#cam42857-bib-0014){ref-type="ref"} but little is known about its biological role and the associated signaling pathways in RCC. Our study set out to determine whether δ‐catenin was implicated in the development of RCC. Based on the results of Western blot and immunohistochemistry, we found that the level of δ‐catenin is elevated in clinical tissue samples with RCC. Therefore, we reasoned that δ‐catenin also played a role in promoting the development of RCC. In subsequent exploration, we found that δ‐catenin affected the translocation of β‐catenin, which further modulated the expression of its downstream apoptosis‐related protein Bcl2L1, cell cycle‐associated marker cyclin D1 as well as tumorigenic gene c‐myc and survivin, thereby facilitated proliferation and suppressed apoptosis in RCC. Moreover, the RCC cell xenograft mouse model revealed knockdown of δ‐catenin restrained tumor growth in vivo, which was concordant with our in vitro findings. Researches conducted by Zhang et al[12](#cam42857-bib-0012){ref-type="ref"} have assessed the clinical significance of δ‐catenin and its correlation with clinicopathological factors in a cohort of 115 specimens with non‐small‐cell lung cancer (including 65 cases with follow‐up records and 50 cases with paired lymph node metastases lesions). The results demonstrated that high expression of δ‐catenin could be observed in adenocarcinoma. Positive expression of δ‐catenin was enhanced in lymph node metastasis lesions, compared with corresponding primary tumors. The elevated δ‐catenin expression was significantly correlated with higher pTNM stage and lymph node metastasis. Furthermore, patients harbouring tumors with δ‐catenin positive expression had significantly shorter postoperative survival period than patients harboring tumors with negative expression. Besides, it has been reported that increased δ‐catenin expression in estimated 85% of prostate cancer based on tissue samples (composed of 90 cases with prostate cancer and 90 cases with benign prostate). Meanwhile, the enhanced δ‐catenin expression was concerned with increased Gleason scores.[29](#cam42857-bib-0029){ref-type="ref"} Consistent with these reports, a significant finding of our study is that RCC patients showed high expression of δ‐catenin and it is correlated with some clinicopathological factors, such as pTNM stage, tumor stage and lymph nodes metastasis. It gave evidence that positive expression of δ‐catenin was relevant to poor prognosis in RCC. δ‐Catenin might be established as a potential prognostic factor and a therapeutic target of RCC. In this study, we found δ‐catenin promoted RCC cell proliferation by altering G1 to S phase transition in cell cycle. Furthermore, δ‐catenin restrained apoptosis of RCC cells through upregulating Bcl2L1 (a long isoform of Bcl2) that protects cells from undergoing apoptosis.[30](#cam42857-bib-0030){ref-type="ref"} In the past decades, a large and growing body of literature has shown that cyclin D1 was induced by hyperactivated β‐catenin, and thereby led to G1 phase progression and proliferation of tumor cell.[31](#cam42857-bib-0031){ref-type="ref"}, [32](#cam42857-bib-0032){ref-type="ref"}, [33](#cam42857-bib-0033){ref-type="ref"}, [34](#cam42857-bib-0034){ref-type="ref"} On the other hand, Rosenbluh et al stated that β‐catenin forms a complex with YAP1 and transcription factor TBX5 which binds to the promoter of BCL2L1, thereby directly regulating BCL2L1 expression at transcriptional level.[35](#cam42857-bib-0035){ref-type="ref"} Taken together, aberrant cyclin D1 and BCL2L1 upregulation take place when the Wnt/β‐catenin pathway is activated. Combined with our investigation about the effects of δ‐catenin on proliferation and apoptosis in RCC cells, we deduced that δ‐catenin may exert these functions by regulating β‐catenin in RCC development. As expected, our data showed that δ‐catenin knockdown suppressed β‐catenin nuclear localization and subsequent alteration of Bcl2L1 and cyclin D1 level in both ACHN cells and ACHN‐derived xenograft tumor tissue. Given that δ‐catenin affected subcellular β‐catenin distribution and activated the downstream effectors of β‐catenin in prostate cancer,[14](#cam42857-bib-0014){ref-type="ref"} we explored whether δ‐catenin played the same role in RCC progression. Figure [5](#cam42857-fig-0005){ref-type="fig"} elucidated δ‐catenin realized the tumor‐promoting function in renal tumorigenesis by altering β‐catenin translocation. Notably, silencing of δ‐catenin prevented β‐catenin from translocating to the nucleus, which resulted in the repression of β‐catenin‐TCF/LEF‐mediated downstream effector activation. Knockdown of δ‐catenin in RCC cells revealed a markedly decreased expression in β‐catenin target genes such as Bcl2L1, cyclin D1, c‐myc, and survivin. In our study, we used RNA interference (RNAi) including transient siRNA and stable shRNA for loss‐of‐function studies to investigate the role of δ‐catenin in RCC development. One major drawback of these means is the limited utility, as evidenced by δ‐catenin‐silenced cells still could grow and preserve tumorigenesis capacity. In future research, the inducible system (such as the tetracycline‐controlled operator system) should be combined with RNAi, which could reveal the direct effects caused by the loss of δ‐catenin via comparing noninduced and induced cells, thereby accurately evaluating the crucial role of δ‐catenin in RCC cell growth and apoptosis. In summary, our study is the first to report δ‐catenin was overexpressed in tissues of RCC patients and its correlation with pTNM stage, tumor stage, and lymph nodes metastasis. In addition, knockdown of δ‐catenin can obviously decrease the proliferation capacity and induce apoptosis of RCC cells by affecting β‐catenin nuclear localization and its downstream target gene expression. Therefore, our results illustrated that δ‐catenin might be a prognostic marker and a promising therapeutic target in RCC. CONFLICT OF INTEREST {#cam42857-sec-0024} ==================== The authors declare that they have no competing interests. AUTHOR CONTRIBUTIONS {#cam42857-sec-0025} ==================== Yongsheng Song conceived and designed the study. Ju Lincheng performed the experiments and analyzed the data. Shan Liping wrote the paper. Yinbo obtained funding. All authors have read and approved the final manuscript. Supporting information ====================== ###### ###### Click here for additional data file. This study was supported by a grant from the Guide Project for the Key Research and Development Project of Liaoning Province (No. 2017225038) and 345 Talent Project of Shengjing Hospital of China Medical University. The pathologic sections were provided by the Biobank from Shengjing Hospital of China Medical University. DATA AVAILABILITY STATEMENT {#cam42857-sec-0027} =========================== All data generated or analyzed during this study are included in this published article.

{ "pile_set_name": "PubMed Central" }

Introduction {#s1} ============ Over the last two decades, optical traps (or optical tweezers) have become a standard tool in the physical and biological sciences, allowing the measurement of sub-nanometer displacements of optically trapped microparticles, as well as the exertion of piconewton-level, controlled forces on these particles [@pone.0057383-Svoboda1], [@pone.0057383-Bustamante1], [@pone.0057383-Ashkin1]. Taking advantage of this versatility, studies on the mechanical and biochemical properties of biomolecules at the single-molecule level using optical tweezers are now widespread. In these assays, individual molecules, attached to trapped particles, can be pulled on or stretched (or even twisted) using the laser trap, and the molecular displacement and force (or torque) responses can be measured with high spatial (∼1 nm) and temporal (∼100 kHz) resolution [@pone.0057383-Fazal1]. These experiments provide unique information on molecular mechanisms that complement traditional biochemical studies. Typically, an optical tweezers arrangement consists of a single laser beam tightly focused by a microscope objective of high numerical aperture (NA) (1.0--1.4), which traps microparticles near its focal point [@pone.0057383-Ashkin2], [@pone.0057383-Neuman1]. Position detection of the trapped particle relative to the laser beam axis is achieved through back focal plane (BFP) detection, where the trapping beam (or a secondary beam of low-power) is directed to a position-sensitive detector (PSD) located in a plane conjugate to the condenser BFP [@pone.0057383-Pralle1]. In many trapping setups, these requirements are implemented by substantial modifications of a commercial inverted light microscope in order to accommodate holders, mounts, and a stable stage platform fitted with a piezoelectric stage (for fine sample movement). Alternatively, an optical trap can be built entirely from individual optical components, providing increased flexibility in the design, often reducing costs, and facilitating the choice of parts. This last option is the one we have followed in this work. Successful development of an optical tweezers apparatus often requires appropriate means to visualize small objects (∼20--200 nm). Indeed, a variety of cellular and single-molecule assays require the localization and subsequent tracking of nanoparticles [@pone.0057383-Sako1], or the imaging of individual slender polymer filaments such as microtubules (MTs, with diameters of only ∼25 nm) [@pone.0057383-Block1]. Because small objects scatter light weakly, their visualization has traditionally not been performed using bright field microscopy. Instead, localization is achieved by employing a number of specialized imaging techniques, in the case of MTs: fluorescence [@pone.0057383-Sammak1], dark field [@pone.0057383-Kuriyama1], polarization [@pone.0057383-Oldenbourg1], phase contrast [@pone.0057383-Kuriyama1], or Nomarski differential interference-contrast (DIC) microscopy [@pone.0057383-Schnapp1]. However, some of these visualization methods exhibit inherent limitations in the context of optical trapping. For example, it is well-known that the Wollaston prisms required in DIC optics can introduce significant asymmetries in the optical trap [@pone.0057383-Lang1]. Likewise, because the Wollaston prisms generally produce or recombine two displaced beams with orthogonal polarizations, DIC microscopy prohibits the operation of an optical tweezers setup where polarization of the trapping beam requires high purity and independent adjustment (as in an "optical torque wrench" [@pone.0057383-GutirrezMedina1]). Introducing optical elements such as prisms, apertures, phase masks, or polarizers in the laser beam path unavoidably constrains the possibilities for optical trapping, especially when control of phase or polarization effects is necessary, as in holographic [@pone.0057383-Grier1] or interferometric [@pone.0057383-MacDonald1] optical tweezers. Recently, it has been shown that unlabeled, individual MTs can be visualized by computer-enhanced bright field microscopy (CEBFM) [@pone.0057383-GutirrezMedina2], a method that does not require the addition of specific optical elements or processing instrumentation to effectively remove the background, reduce noise, and enhance contrast of images. CEBFM provides an alternative of remarkable simplicity to visualize small objects, and is therefore ideally suited to complement optical trapping. In bright field microscopy, ample access to optical paths in the setup is ensured, and it is possible to independently optimize Koehler illumination and the BFP detection stage of the optical trap, allowing excellent imaging and trapping. By using CEBFM, a basic microscope could be fit to perform standard single-molecule "gliding filament" or optical trapping "bead" assays without resorting to advanced imaging modalities. Furthermore, CEBFM may prove useful in emerging microscopy techniques that use image contrast to perform biophysical measurements and that incorporate optical tweezers as a tool for microparticle manipulation, such as "defocusing microscopy" [@pone.0057383-Agero1] (see Section "Digital image processing"). In this work, we describe the design, construction, and testing of an elementary optical tweezers instrument that allows trapping and visualization of weakly scattering objects, integrating a straightforward approach for the study of biomaterials and individual biomolecules. Our design is based on CEBFM together with a single laser beam coupled into the microscope for optical trapping and BFP detection. This arrangement keeps the number of mechanical and optical components to a minimum, resulting in ease of construction and cost-effectiveness that do not compromise final stability or resolution. To test the capabilities of our system, we performed two single-molecule assays and compared our results with previous reports. We have measured the persistence length of individual double-stranded DNA (dsDNA) molecules at relatively high ionic strength (\[Na^+^\] = 150 mM), and have observed the stepping of single kinesin motor proteins as they advance along clearly imaged, individual MTs. To our knowledge, this is the first time that the classic cytoskeletal motor protein bead-assay has been demonstrated using plain bright field illumination to localize filament substrates. Results and Discussion {#s2} ====================== Optical and Mechanical Layout {#s2a} ----------------------------- Our optical tweezers system (shown in [Figure 1](#pone-0057383-g001){ref-type="fig"}) was built entirely from individual components, mounted on a vibration-isolation optical table (I-2000, Newport). Trapping light is provided by a CW, Yb-doped, fiber laser (YLR-10-1064-LP, 10 W, λ = 1064 nm, IPG Photonics) featuring excellent pointing stability (deviations \<1 µrad/s), and good power stability (fluctuations ∼2%). A 10 m-long optical fiber coupled to the laser allows installation of the laser head (together with the noisy power electronics) in a separate room. The laser is nominally operated at ∼4.2 W of output power, above the minimum ∼3.0 W, that according to the manufacturer accounts for the stability figures cited above. ![Schematic of the optical tweezers instrument.\ A fiber-coupled laser provides the trapping light (red lines), the intensity of which is controlled with a combination of a λ/2-waveplate and a polarizing beam splitter (PBS). An acousto-optic modulator (AOM) is part of a feedback loop that stabilizes laser intensity. The trapping beam is coupled into and out of a custom-built inverted microscope (black dotted line box) using two dichroic mirrors (DM). Illumination (blue lines) for the inverted microscope is provided by a blue LED. The green and blue arrows indicate the positions of the two sets of optically conjugated planes that satisfy the conditions for Koehler illumination and back focal plane detection. See text for details. BS: beam sampler, PD: photodiode, M: mirror, L: lens, D: diaphragm iris, PSD: position-sensitive detector.](pone.0057383.g001){#pone-0057383-g001} We now describe our setup in detail. The end terminal of the laser fiber is secured onto the optical table, and a combination of a half-wave plate and a polarizing cube beamsplitter (PBS) control the laser power going to the optical trap. To provide means of automatically steering the laser trap on the sample plane, an *x*-*y* acousto-optic modulator (AOM) (DTD-274HD6M, IntraAction) is placed immediately after the PBS, and an iris diaphragm (D1) selects the (+1,+1) diffracted beam that emerges from the AOM. Next, two lenses (L1 and L2), forming a Keplerian telescope, expand the beam to a final waist size, *w,* of ∼3 mm, corresponding to a beam diameter that roughly matches the diameter of the objective back aperture (∼6 mm). Two additional lenses (L3 and L4) then form a 1∶1 telescope, with L3 (see [Figure 1](#pone-0057383-g001){ref-type="fig"}) mounted to an *x-y*-*z* translation stage to provide fine manual control to steer and collimate the laser beam going into the objective. A stage-mounted lens is a useful feature during the initial alignment of the trap. To transform beam deflection by either the AOM or lens L3 into beam translation in the specimen plane and minimize beam clipping, the AOM and the lens L3 are optically conjugate to the back aperture of the objective [@pone.0057383-Neuman1] (see [Figure 1](#pone-0057383-g001){ref-type="fig"}). The AOM is useful not only to steer the trap but also to control the laser power going into the objective. Our instrument is placed in a room that does not have temperature control, and we have found that the output laser power may drift during the course of the day by as much as ∼5%, presumably due to room temperature changes. To correct for this problem, a laser power feedback loop was implemented using the AOM. A beam sampler placed after the AOM picks up part of the main beam and sends it to a photodiode, whose signal is amplified and fed into a data-acquisition board (U12, LabJack) connected to a computer. A proportional-integral-derivative (PID) feedback routine is implemented in software, using LabView 8.5 (National Instruments), and its corrected amplitude signal is communicated to the AOM driver (AFG3022B, Tektronix) via a USB port. Finally, the *x* and *y* outputs of the AOM driver are each amplified by an RF amplifier (ZHL-3A, Minicircuits), and connected to the AOM, closing the feedback loop. The resulting feedback control has a characteristic time response of ∼1 s, and maintains the power constant (to a level of ∼0.2%) for hours. Alternatives to this feedback scheme using an AOM driver are available [@pone.0057383-Mack1]. To afford imaging of the specimen, trapping chamber manipulation, and laser focusing by the objective, an inverted microscope was built. Because one of the main purposes of this work is to attain visualization of weak light-scattering objects by means of bright field microscopy (without the aid of additional optical parts or special types of illumination), the components of the optical train of our microscope are arranged to satisfy Koehler illumination. Among other advantages of Koehler illumination, the field is homogeneously bright, image contrast is optimum, and maximal lateral and axial resolutions are achieved [@pone.0057383-Inou1]. In our microscopy system (see [Figure 1](#pone-0057383-g001){ref-type="fig"}), the illuminator is composed of a high-power light-emitting diode (LED) (LXHL-LR5C, 700 mW, peak λ = 455 nm, Luxeon) as the light source, together with a collector lens (L6) and a field iris diaphragm (D3) placed immediately adjacent to L6. The LED is driven by an inexpensive analog current controller, originally designed for laser diodes, that provides extremely stable current (∼10^−4^ rms variations) up to 500 mA (design available at: <http://george.ph.utexas.edu/~meyrath/informal/laser%20diode.pdf>). This intensity stability is necessary to perform appropriate background subtraction during image processing (as described in the next section). The collector lens collimates incoming light from the LED, and an additional lens (L7) focuses the image of the light source onto the condenser iris diaphragm (D4). To maximize contrast during imaging, we close this aperture almost entirely (∼85%) to effectively produce a point-like illumination source. Relay lens (L8) images the condenser iris onto the condenser BFP. The condenser lens (1.4NA, Zeiss) is mounted on an *x*-*y*-*z* translation stage (461-XYZ-M, Newport), and aligned to focus an image of the field diaphragm onto the sample plane. This ensures that the specimen and the images of the light source are in different (reciprocally related) planes, and that the specimen is illuminated with a set of plane waves. Following the optical train, the objective lens captures the illuminating plane waves and creates an image of the light source onto the rear focal plane of the objective (exit pupil). We use an infinity-corrected, oil-immersion objective of high numerical aperture (100X/1.3NA/PlanFluorite, part RMS100X-PFO, Olympus). Finally, a tube lens (L9) captures illuminating and scattered light from the specimen and images the sample onto an inexpensive CCD camera (STC-TB33USB-B, Sentech) that is connected to the computer via a USB interface. With the condition for Kohler illumination satisfied in this optical arrangement, two mutually reciprocal sets of optically conjugated planes are formed. The first set includes the field diaphragm, the sample plane, and the camera sensor plane. The second set includes the light source, the condenser iris, the rear focal plane of the condenser, and the back focal plane of the objective. Coupling of the trapping laser into and out of the microscope is achieved using two dichroic mirrors (Z1064RDC-SP, Chroma). To perform BFP detection, a single lens (L5) images the rear focal plane of the condenser onto a PSD (DL100-7-PCBA2, Pacific Silicon Sensor). A diaphragm (D2) placed directly in front of the PSD modulates the NA of the detection optics, optimizing the PSD detection signal [@pone.0057383-Rohrbach1]. The output voltage signals from the PSD (fitted with a built-in preamplifier) are further amplified using a custom-made electronic circuit (200 kHz bandwidth) and sent to a data-acquisition board (PCI-6052E, National Instruments) fitted in a computer, where data are processed by custom software written in LabView. A laser power of 90 mW applied to the trap (measured before entering the objective rear pupil) was typically used for our experiments. To block the trapping laser whenever necessary, an inexpensive, custom-made mechanical shutter was used (design available at: <http://george.ph.utexas.edu/~meyrath/informal/shutter.pdf>). An advantage of our optical trapping and microscopy setup over previous designs is that BFP detection and Koehler illumination are effectively decoupled and can be optimized independently. In typical microscopes working under Koehler illumination, the condenser iris is located at the back focal plane of the condenser lens (entrance pupil). This arrangement is not ideal for an optical tweezers arrangement, as any aperture placed in the proximity of the condenser compromises BFP detection. Conversely, although standard optical trapping systems often do not use the condenser iris, here it is necessary to maximize contrast during visualization of weak light-scattering specimens. We resolved this potential conflict by using relay lens L8 to image the condenser iris onto the condenser BFP, leaving the condenser lens entrance pupil unobstructed. Lens L8 thus allows us to adjust the condenser iris aperture without affecting the NA of the BFP detection optics. The mechanical design of the microscope is completed with a piezoelectric stage (Nano-LP100, Mad City Labs) that is used for fine sample movement, and a supporting crossed-roller-bearing mechanical stage (Manual MicroStage, Mad City Labs) that provides coarse positioning. The *x*-*y*-*z* piezoelectric stage has nanometer-level step resolution over 100×100×100 µm. The stacked stages are supported by four 8″-tall, 1.5″-diameter stainless steel posts that provide mechanical stability to the stage platform. All of the condenser, detection, and illumination optics are mounted on a breadboard that is suspended vertically on a structure formed by two large vertical construction rails (XT95, Thorlabs), cross-linked at the top by a third rail, and further supported by one additional beam joining the top rail to the optical table. Digital Image Processing {#s2b} ------------------------ Images are digitally processed to remove background, perform frame averaging, and enhance contrast. We implemented computer-based image processing essentially as described in Ref. [@pone.0057383-GutirrezMedina2], with appropriate improvements for our work. All of the procedures described here are implemented in software using LabView 8.5 (add-on package Vision, National Instruments). First, using the piezoelectric stage, the surface of the coverslip is brought into focus and subsequently defocused by a few micrometers. At this point, the sample is oscillated in the *x* or *y* direction (about 5 micrometers in amplitude, 2 seconds in period) using the piezoelectric stage, while background images (typically 250) are simultaneously acquired. An average of the background images provides a reference image that is subsequently subtracted from all incoming frames. As background frames are acquired this way, images belonging to any relatively strong scattering objects that may be stuck to the coverslip surface and present in the field of view are defocused and blurred in the final reference background image, thereby minimizing their influence in subsequent processing. This allows for changes in the field of view without the need to acquire a new background reference. After completing this operation, focus is readjusted to visualize the sample. Next, background-free images are averaged using a simple exponential-averaging (or exponential-smoothing) procedure, during which incoming time series data are averaged using exponentially decreasing weights [@pone.0057383-Salmon1]. One advantage of exponential over boxcar averaging is that it is easier to implement, as only two data (the last, smoothed datum and the incoming datum) need to remain in memory. In addition, the weight assigned to recent or older data can be readily adjusted. Thus, let *F~i~* represent the *i* ^th^ original, background-free image acquired sequentially in time and *S~i~* the corresponding final, smoothed image. Exponential averaging is implemented according to the following rule [@pone.0057383-NISTSEMATECH1]:where parameter α is the smoothing constant (see [Figure 2A](#pone-0057383-g002){ref-type="fig"}). Full or zero weight is assigned to the most recent datum in the time series by setting α = 1 or α = 0, respectively. We typically use α = 0.3 in our experiments. ![Individual microtubules are visualized using bright field microscopy together with digital image processing.\ (A) Flow diagram of digital image processing, depicting operations for background acquisition (dashed area) and subtraction, exponential averaging, and contrast enhancement. (B-E) Frames show MTs immobilized on coverslips, as imaged during the various stages of processing marked in (A): (B) raw image; (C) the same as (B) with the background removed; (D) the same as (C) after multiple-frame exponential averaging; (E) the same as (D) after a convolution routine (shown with the symbol in the diagram) that enhances contrast in the vertical direction, and a fast-Fourier-transform routine (shown with the symbol in the diagram) that smoothes out rapidly-varying spatial features. Field of view is 24 µm × 21 µm.](pone.0057383.g002){#pone-0057383-g002} Background-free, exponentially-averaged images can be further processed using two additional routines that enhance contrast and reduce noise: (*i*) a spatial convolution routine, which involves a point-by-point multiplication of the pixel values of the original image multiplied by the corresponding pixel values of a filtering convolution mask [@pone.0057383-GutirrezMedina2], and (*ii*) a fast Fourier transform smoothing routine, which removes spatial features within a given size range. To test our system, we visualized MTs, polymer filaments that are involved in cellular structure and organization [@pone.0057383-WatermanStorer1], previously immobilized on coverslips (see [Materials and Methods](#s3){ref-type="sec"}). [Figures 2B--2E](#pone-0057383-g002){ref-type="fig"} show that individual MTs are clearly distinguished, with an apparent thickness determined by the diffraction limit of the microscope (∼250 nm). We noticed that the contrast of MT images changes as the coverslip surface is brought in and out of focus, reaching an optimum when the coverslip surface is slightly (∼300 nm) defocused (see [Figure 3](#pone-0057383-g003){ref-type="fig"}). This fact is consistent with the early observation by Fritz Zernike that when a bright field microscope is focused precisely, the image of a thin, transparent specimen disappears due to interference of light diffracted by the specimen and of background illumination, and contrast is maximized only by slightly defocusing the microscope [@pone.0057383-Inou1]. Recently, this effect has been used to develop "defocusing microscopy," a quantitative visualization method where image contrast (shown proportional to the amount of defocusing and to the curvature of the specimen) is measured, from where it is possible to learn about the material properties of the sample [@pone.0057383-Agero1]. In our experiment, the MT filaments are transparent under visible light and present enough shape curvature, therefore they act as effective phase objects despite being hollow and having a diameter of only λ/20. ![Microtubule image contrast is optimized by microscope defocusing.\ Images of a single MT (top row) displaying the same field of view when the focus is placed 300 nm below (left), precisely at (center), and 300 nm above (right) the coverslip surface. The bottom row shows the corresponding pixel count profiles, obtained for each image by averaging all the pixel count values along a given pixel column. Images were digitally-processed as described in the text to remove background and perform exponential averaging (α = 0.3). No convolution or fast-Fourier-transform routines were applied in this case. Field of view is 3.4 µm × 4.8 µm.](pone.0057383.g003){#pone-0057383-g003} The previous MT imaging result demonstrates that our setup achieves excellent sample visualization at standard video rates (∼30 Hz), offering several advantages over a recent system that also imaged MTs using bright field microscopy [@pone.0057383-GutirrezMedina2], among which are: improved background subtraction (by using a light source with excellent intensity stability and by moving the piezoelectric stage during background acquisition), real-time specimen visualization (by implementing exponential averaging of images), and the possibility of using the full NA of the condenser (by using relay lens L8 to image the condenser iris onto the condenser BFP). Most importantly, the instrument presented here has the full capabilities of optical trapping for measuring molecular-level displacements and forces. Calibration of the Optical Trap {#s2c} ------------------------------- We perform position and force calibrations using well-established procedures for trapping spherical beads [@pone.0057383-Neuman1]. First, the relationship between the PSD voltage in a given direction, *V~x~*, and the displacement of the bead from the equilibrium position, *x*, is determined by scanning a bead affixed to the coverslip surface across the laser beam. As shown in [Figure 4A](#pone-0057383-g004){ref-type="fig"}, the resulting profile *V*(*x*) is well-described by the derivative of a Gaussian function, and the slope of the central, linear part of the profile is the conversion factor from nanometers to volts, ξ*~x~* (units of V/nm), which is subsequently used to convert all PSD signals along *x* from volts to nanometers. Next, the stiffness of the optical trap, κ, along the lateral directions is established by using three separate methods, which are based on measurements of particle position variance, power spectrum, and Stokes' drag, respectively. The stiffness along the axial direction is established using the particle position variance and power spectrum methods only. ![Calibration of the optical trap.\ (A) Response of the position-sensitive detector (PSD) as a bead immobilized on the coverslip is scanned across the laser beam along the *x* axis (red circles). The data are well-fit by the derivative of a Gaussian (red line), from which the nanometer to volt conversion factor is determined, ξ*~x~* = 0.013 V·nm^−1^. (B) The position of a trapped bead along the *x* axis is monitored and the respective histogram is well-fit by a Gaussian (thick red line). The position variance is = 51 nm^2^ (or, the raw PSD signal variance is = 0.0086 V^2^). Using the principle of equipartition of energy, the trap stiffness is found: κ*~x~* = 0.080 pN·nm^−1^. (C) The power spectrum of the position signal along the *x* (red line), *y* (blue line), and *z* (green line, inset) directions is shown, together with fittings to a Lorentzian function for *x*, *y*, and *z* (red, blue, and green thin lines, respectively). Measurement of the roll-off frequency of the fit yields the trap stiffness: κ*~x~* = 0.063 pN·nm^−1^ and κ*~y~* = 0.055 pN·nm^−1^. (D) In the Stokes drag calibration method, a trapped bead is subject to fluid drag by moving the stage at velocity *v*, resulting in the bead moving to a new equilibrium position that is measured (inset). Bead displacement vs. stage velocity data are shown for the *x* (red squares) and *y* (blue circles) axes. Linear fits to the data (red and blue lines) yield the slopes: *s~x~* = 0.090 nm·µm^−1^·s, and *s~y~* = 0.099 nm·µm^−1^·s (or the corresponding raw-signal slopes *s~Vx~* = 0.0012 V·µm^−1^·s, and *s~Vy~* = 0.0013 V·µm^−1^·s), which yield trap stiffness along each direction: κ*~x~* = 0.076 pN·nm^−1^, and κ*~y~* = 0.063 pN·nm^−1^. As discussed in the text, ξ and κ can alternatively be found using: ξ = (γ/*s~V~*), and κ = γξ/*s~V~*. This method yields ξ*~x~* = 0.012 V·nm^−1^, and κ*~x~ = *0.07 pN·nm^−1^ for the example presented here. The average value of κ*~x~* using all four of these methods provides the best estimate of trap stiffness along *x*: κ*~x~* = 0.072±0.007 (mean ± SD). Calibrations were performed using 0.54-µm diameter polystyrene beads.](pone.0057383.g004){#pone-0057383-g004} Briefly, for the first method, the variance of the position signal of a trapped bead () is measured and, using the principle of equipartition of energy, the stiffness κ*~x~* is given by κ*~x~* = , where *k* ~B~ is Boltzmann's constant and *T* is temperature (see [Figure 4B](#pone-0057383-g004){ref-type="fig"}). For the other two procedures, the drag coefficient (γ) of a bead of radius *r* immersed in a fluid of viscosity η, γ = 6πη*r*, is assumed to be known. In the second method, the power spectrum of the position signal of a trapped bead is computed and fitted by a Lorentzian profile (see [Figure 4C](#pone-0057383-g004){ref-type="fig"}). From a measurement of the roll-off frequency of the fit, *f* ~0~, the stiffness can be derived as κ = 2πγ*f* ~0~. Finally, in the third method, a steady drag force is exerted on a trapped bead by moving the sample with the piezoelectric stage at constant speed (*v~x~*). The resultant viscous drag places the bead at a new equilibrium position, *x* = γ*v~x~/*κ*~x~* (from Stokes' law), that is measured. This procedure is then repeated for various *v~x~*, and the stiffness is inversely proportional to the slope of the linear fit of the *x* vs. *v~x~* data (see [Figure 4D](#pone-0057383-g004){ref-type="fig"}). Alternatively, the variance and the Stokes' drag methods can be combined to provide simultaneous position (ξ) and force (κ) calibrations along the lateral directions, using the raw PSD signal, *V~x,y~*. The advantage of this approach over the previous independent methods is that it does not require an explicit position calibration (based on coverslip-stuck beads), and can therefore be applied to free beads. In this case, we measure the variance of the position signal, = *k* ~B~ *T*ξ*~x,y~* ^2^/κ*~x,y~* (units of V^2^), and the slope of the *V~x,y~* vs. *v~x,y~* linear fit in the Stokes' drag method, *s~x,y~* = γξ*~x,y~*/κ*~x,y~* (units of V·µm^−1^·s). These two relationships are combined to yield ξ*~x,y~* = (γ/*s~x,y~*), and κ*~x,y~* = γξ*~x,y~*/*s~x,y~*. Like the power spectrum and Stokes' drag methods, with the previous procedure it is assumed that the drag coefficient of the bead is known. For a sphere near a surface, the relationship γ = 6πη*r* for the hydrodynamic drag no longer holds, due to surface proximity effects [@pone.0057383-Neuman1]. Accordingly, to determine the drag coefficient, we take into account Faxen's law in all our measurements. Very good agreement is obtained between all calibration methods presented here, resulting in variations of ∼10% in values of κ*~x,y~*. The Persistence Length of dsDNA {#s2d} ------------------------------- A number of studies have measured the persistence length of dsDNA under a variety of buffer conditions [@pone.0057383-Smith1], [@pone.0057383-Wang1], [@pone.0057383-Seol1], [@pone.0057383-Baumann1]. We tested our optical trapping instrument by pulling on individual dsDNA molecules, using a standard surface-based assay [@pone.0057383-Seol1], at a relatively high ionic strength (\[Na^+^\] = 150 mM). Following established protocols, a ∼3.1 kb dsDNA template was attached by the 5′ end of one of its strands to a 0.7 µm-diameter polystyrene bead, and affixed by the 5′ end of the complementary strand to the coverslip surface (see [Figure 5A](#pone-0057383-g005){ref-type="fig"} and [Materials and Methods](#s3){ref-type="sec"}). The procedure to measure force-extension curves for dsDNA is described in Ref. [@pone.0057383-Wang1]. ![Stretching single dsDNA molecules.\ (A) Schematic of the surface-based assay, not to scale. (B) A typical force-extension record (black circles) together with a worm-like chain (WLC) polymer model fit (black line). (C) Histogram of contour length values. Fit to a Gaussian (black line) yields: 1116±2 nm (μ ± σ~μ~). (D) Histogram of persistence length values. Fit to a Gaussian (black line) yields: 47.0±0.2 nm (μ ± σ~μ~). Legends display data mean ± SEM.](pone.0057383.g005){#pone-0057383-g005} Briefly, surface-tethered beads were held by the laser tweezers and centered with respect to the surface attachment point, along the *x* and *y* directions, by adjusting the piezoelectric stage position such that stage motions in the ±*x* or ±*y* directions produced symmetric bead displacement responses (as measured by the PSD). Next, beads were placed at a specified distance (20 nm) above the coverslip surface, taking into account the effective focal shift (∼0.8) of the optical trap [@pone.0057383-Neuman1]. Then, the DNA molecule was stretched by moving the coverslip horizontally (using the piezoelectric stage) in steps of 10 nm in distance, 10 ms in duration. For each step, *x*,*y*,*z* bead displacements were recorded at a rate of 10 kHz and averaged in each direction. Finally, the extension of the elongated molecule (*L*) was determined from the known stage movement and the measured bead position with respect to the trap, and the force (*F*) acting on the molecule along the stretching direction was determined from the position of the bead with respect to the trap and the trap stiffness [@pone.0057383-Wang1]. Because the method used here to determine bead-surface distances is precise to 20 nm [@pone.0057383-Lang1], the distance between the bead and the surface was set to 0 nm to obtain force-extension curves. All DNA molecules were stretched in the ±*x* and ±*y* directions. To discriminate single from multiple tethers, only ±*x* and ±*y* symmetrical force-extension curves were considered for analysis. We report results from stretching along the *y* direction only. For each molecule, data corresponding to the +*y* scan direction were superimposed on the --*y* scan direction, and treated as a single *y* record. Care was taken to perform measurements within the respective linear regions of position and force calibrations. [Figure 5B](#pone-0057383-g005){ref-type="fig"} shows a typical single-molecule force-extension record. Maximal stretching forces of 4 pN were applied to all molecules. In this low-force regime, the elasticity of DNA is essentially entropic, and the corresponding *F* vs. *L* relationship is well-described by the worm-like chain (WLC) polymer model [@pone.0057383-Bouchiat1]:where *L* ~0~ and *L* ~p~ are the contour length and the persistence length of the polymer, respectively. Force-extension records were well-fit using the WLC model (see [Figure 5B](#pone-0057383-g005){ref-type="fig"}), from which values for *L* ~0~ and *L* ~p~ were inferred. Statistical analysis of data for *L* ~0~ and *L* ~p~ (see [Figs. 5C and 5D](#pone-0057383-g005){ref-type="fig"}) yielded measured values: nm and nm (mean ± SEM), which are in excellent agreement with the expected value for the contour length *L* ~0~ ∼1053 nm (see [Materials and Methods](#s3){ref-type="sec"}) and with previous reports of *L* ~p~ at comparable ionic strengths [@pone.0057383-Baumann1]. The Stepping of Kinesin along MTs {#s2e} --------------------------------- Kinesin is a homodimeric motor protein that is involved in intracellular transport, using the energy of ATP hydrolysis to ferry cargo along MTs [@pone.0057383-Asbury1]. Previous studies using optical trapping, among other single-molecule techniques, have determined that kinesin advances in 8.2-nm increments and can sustain retarding loads as large as ∼7 pN before stalling [@pone.0057383-Block2]. In traditional optical-tweezers based bead assays of kinesin, DIC microscopy is used to localize MTs [@pone.0057383-Block1]. Here, we demonstrate the versatility of our instrument by observing kinesin motors stepping along MTs, where we simultaneously use both optical trapping and the excellent visualization features available from our experimental setup. To record kinesin motion, a standard motility bead assay was carried out (see [Figure 6A](#pone-0057383-g006){ref-type="fig"}) [@pone.0057383-Svoboda2]. Single recombinant kinesin motors were attached to 0.54-µm polystyrene beads (see [Materials and Methods](#s3){ref-type="sec"}). Kinesin-carrying beads were optically trapped and placed directly above individual MTs that were immobilized on the coverslip surface, previously identified using the microscope's digitally-enhanced imaging (see [Figure 6B](#pone-0057383-g006){ref-type="fig"}). When kinesin binds to the MT in the presence of ATP in the buffer solution, it advances unidirectionally towards the 'plus' end of the MT. In our experiment, the (fixed) optical trap was used both to measure the stepping of kinesin, and to exert a retarding force on motors that increases as beads are transported away from the center of the laser beam. For MTs aligned along the *x* direction, the retarding force on kinesin is determined through Hookés law (*F~x~* = -κ*~x~*·*x*). [Figure 6C](#pone-0057383-g006){ref-type="fig"} presents examples of kinesin stepping records obtained in this work, showing that expected steps of 8.2 nm are clearly resolved. Additionally, we observed maximal forces sustained by motors before stalling or detachment from the microtubule of ∼7 pN, consistent with previous, detailed studies of kinesin motility [@pone.0057383-Block2]. ![Kinesin stepping along MTs.\ (A) Schematic of the kinesin bead assay, not to scale. (B) A kinesin-carrying trapped bead is placed directly above an individual MT, visualized using digital image processing. Field of view is 27 µm × 6 µm. (B) Representative records of kinesin-driven bead motion (gray line: unfiltered trace acquired at 20 kHz, black line: trace after a 20-point median filtering), display steps at 8.2-nm intervals (dashed horizontal lines) and the development of forces up to ∼7 pN. \[ATP\] = 100 µM.](pone.0057383.g006){#pone-0057383-g006} Conclusions {#s2f} ----------- An optical trapping system has been developed with exceptional visualization features at minimal cost and complexity. The technique of CEBFM provides an alternative to elaborate microscopy techniques for high-contrast visualization, avoiding the inclusion of optical elements in the instrument (such as prisms, apertures or polarizers) that could compromise operation of optical trapping. Using our instrument, it is possible to achieve independent control of BFP detection (in optical trapping) and Koehler illumination (in sample visualization). The approach presented here is uniquely suited for single-molecule experiments and studies of biomaterials at the micro- and nano-scales. Unconventional optical trapping techniques, such as an optical torque wrench (for rotating microparticles), or holographic and interferometric optical tweezers (for creating multiple-trap arrangements) can be incorporated immediately without sacrificing imaging capabilities. The setup can also be readily expanded to incorporate additional features, such as a second trapping beam (to perform off the surface, multiple-bead assays) [@pone.0057383-Abbondanzieri1], or strategies to maintain a constant force during experiments (a 'force clamp') [@pone.0057383-Lang1], [@pone.0057383-Greenleaf1]. Materials and Methods {#s3} ===================== Unless specified, all reagents were purchased from Sigma. Microscope slides and coverslips were cleaned prior to use for 5 min in a plasma cleaner (Harrick Plasma) at 1 Torr (ambient air). Flow channels were made using double-sided tape as described [@pone.0057383-Appleyard1]. All buffers were filtered before use through 0.22 µm filters (Millipore). Preparation of dsDNA {#s3a} -------------------- A 3039-bp dsDNA coding for a fragment of the *blr-1* gene of *Trichoderma atroviride* [@pone.0057383-CasasFlores1] was amplified by a polymerase chain reaction (PCR) (Applied Biosystems). To allow attachment of the dsDNA to a polystyrene bead and to the coverslip, two functionalized PCR primers (Integrated DNA Technologies) were used: a forward primer (biotin-5′-GGGCTTCTACCAGACAAACCA-3′), and a reverse primer (digoxigenin-5′-CGCTCTTCTCGTATTGAAGCC-3′). The reaction tube contained: 5.0 µL of the 5XPCR buffer (Promega), 2.5 µL of 25-mM MgCl~2~ (Promega), 0.5 µL of 100-mM dNTPs, 0.5 µL of 10-µM reverse primer, 0.5 µL of 10-µM forward primer, 1.0 µL of cDNA from the *blr-1* gene (generous gift of Sergio Casas Flores, IPICYT, Mexico), 0.5 µL of Taq Polimerase (GoTaq, Promega), diluted in 15 µL of Milli-Q water. The PCR ran for 25 cycles with an alignment temperature of 58°C and 1 min for extension. The amplified fragment was purified using a QIAQuick PCR purification kit (Qiagen). The expected length of the dsDNA molecule is ∼1053 nm, computed by taking into account a 3039-bp chain with 0.34 nm rise per bp [@pone.0057383-Watson1], together with ∼20 nm for the biotin-streptavidin and digoxigenin-anti-digoxigenin linkages. Recombinant Kinesin {#s3b} ------------------- We expressed the homodimeric, recombinant kinesin construct DmK401, a His-tagged derivative of *Drosophila melanogaster* kinesin heavy chain, that includes the first 401 N-terminal residues (previously described [@pone.0057383-Asbury2]). Briefly, BL21(DE3) cells transformed with plasmid pCA1 (generous gift of Steven Block, Stanford University) were grown to logarithmic phase in Luria Broth medium (10 g/L tryotone, 5 g/L yeast extract, 10 g/L NaCl) supplemented with 0.1 mg/mL ampicillin (GIBCO). Kinesin expression was chemically induced with 1 mM IPTG (Invitrogen) at 27°C during 12 h. Cells were lysed by sonication in extraction buffer (200 mM Na~2~HPO~4~, 50 mM NaCl, 2 mM imidazole, 20 uM ATP, 2 mM MgCl~2~, 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride (PMSF), and protease inhibitor cocktail (P8465, Sigma); 1% Tween was added after sonication). Cellular lysate was clarified by centrifugation (30 min, 15,000 rpm, 4°C). Clarified lysate was stored at −20°C in 10% glycerol, and used in motility assays. Single-molecule Assays {#s3c} ---------------------- ### dsDNA flexibility {#s3c1} A solution of 0.05 mg/mL of antidigoxigenin (3210--0488, Spherotech) in phosphate buffer saline (137 mM NaCl, 2.7 mM KCl, 10 mM Na~2~HPO~4~, 2 mM KH~2~PO~4~) was introduced in a flow channel, and incubated for 10 min. After washing with 200 µL of washing buffer (5 mg/mL BSA, 77.4 mM Na~2~HPO~4~, 0.1% Tween), the channel was filled with a sample of 0.25 nM dsDNA diluted in phosphate buffer (77.4 mM Na~2~HPO~4~), and incubated for 10 min. Unbound DNA was removed by flowing 200 µL of washing buffer through the channel. Finally, 30 µL of 730-nm diameter, avidin-coated beads (generous gift of Steven Block, Stanford University), diluted in phosphate buffer to a final concentration of ∼1 pM, were introduced into the channel, and the flow cell was sealed using nail polish. ### Kinesin motility assay {#s3c2} The motility, bead assay follows previous experiments. Briefly, 10 µL of the stock solution of 540-nm diameter, streptavidin-coated beads (Spherotech, SVP-05--10) were diluted in 70 µL of PEMBSA buffer (4 mg/mL BSA, 80 mM PIPES, 1 mM EDTA, 4 mM MgCl~2~, pH 6.9), and sonicated for 10 min., after which 20 µL of penta-His biotin conjugate antibody (34440, Qiagen) were added. After incubating for 1 hour at room temperature, beads were washed 5 times by centrifugation, and stored at 4°C. To bind kinesin to the beads, antibody-coated beads were diluted in assay buffer (3 mg/mL BSA, 0.05 M potassium acetate, 100 µM ATP, 1 µM DTT, 80 mM PIPES, 1 mM EDTA, 4 mM MgCl~2~, 0.02 mM Taxol (TXD01, Cytoskeleton), pH 6.9), sonicated, and incubated for 12 h at 4°C with clarified lysate diluted in assay buffer at various concentrations. To immobilize MTs on coverslips, flow channels were prepared using coverslips coated with poly-L-lysine. A rack of plasma-cleaned coverslips was submerged in a solution of 600 µL of poly-L-lysine diluted with 300 mL of ethanol, incubated for 15 min, dried in an oven at 40°C, and stored in a closed container. Tubulin (TL238-C, Cytoskeleton) was polymerized to produce MTs as described [@pone.0057383-GutirrezMedina2]. Stabilized MTs were diluted in PEMTAX buffer (0.02 mM Taxol, 80 mM PIPES, 1 mM EDTA, 4 mM MgCl~2~, pH 6.9), introduced into the flow channel and incubated for 10 min. Unbound MTs were removed by washing the channel with 40 µL of PEMTAX. This procedure was followed to obtain samples used to tests the imaging quality of our microscope. To complete the kinesin motility assay, the channel was washed first with 60 µL of 20-mg/mL BSA diluted in PEMTAX (to minimize the sticking of beads to the coverslip surface), and then with 100 µL of assay buffer. Finally, 40 µL of kinesin-bead complexes were introduced into the channel and the flow cell was sealed. To minimize the presence of reactive oxygen species, final samples were protected with an oxygen scavenger system (0.25 mg/mL glucose oxidase, 0.03 mg/mL catalase, 4.7 mg/mL beta-D-glucose (MP Biomedicals)). To ensure motility in the single-molecule regime, kinesin dilutions were used in which only ∼50% of the tested beads in a given sample displayed movement. Data processing and analysis were performed using Igor Pro 5.0 (Wavemetrics). We thank J.S. Casas Flores, M.G. Cervantes Badillo, and C. García-García for continuous support and advice; F.J. Cárdenas Flores for a generous loan of the AOM driver; and P.E. Riley, C. García-García, N.R. Guydosh, and anonymous reviewers for comments on the manuscript. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: CNHC STM BGM. Performed the experiments: CNHC STM BGM. Analyzed the data: CNHC STM BGM. Contributed reagents/materials/analysis tools: BGM. Wrote the paper: CNHC BGM. [^3]: Current address: Program in Biophysics, University of California, Berkeley, California, United States of America

{ "pile_set_name": "PubMed Central" }

1.. Introduction {#sec1} ================== Recent progress in direct electron detector technology, the provision of intense and coherent electron beams using field emission guns (FEGs and X-FEGs) in cryo-electron microscopes and advances in image-processing algorithms and sample preparation have led to the so-called 'resolution revolution' (Kühlbrandt, 2014[@bb15]; Mitra, 2019[@bb17]) in cryo-electron microscopy (cryo-EM). Indeed, two thirds of the more than 6000 entries currently in the Electron Microscopy Data Bank (EMDB; Patwardhan, 2017[@bb20]) have been deposited within the last five years. However, an average resolution for single-particle cryo-EM structures of between 5 and 6 Å for the 3DEM density maps deposited suggests that there is still some way to go before cryo-EM becomes a routine technique for the production of high-resolution structural information. Nevertheless, with this aim in mind, and coupled with significant advances in sample preparation and data processing, many national and international centres for cryo-EM have opened during the last few years (see, for example, Alewijnse *et al.*, 2017[@bb2], and references therein; Stuart *et al.*, 2016[@bb27]; Clare *et al.*, 2017[@bb6]). In 2015, the European Synchrotron Radiation Facility (ESRF) took the decision to complement its portfolio of cutting-edge X-ray-based facilities for structural biology (Pernot *et al.*, 2013[@bb21]; von Stetten *et al.*, 2015[@bb26]; Linden *et al.*, 2014[@bb16]; Mueller-Dieckmann *et al.*, 2015[@bb19]) by acquiring, installing and commissioning a cryo-EM facility (CM01; <http://www.esrf.eu/home/UsersAndScience/Experiments/MX/About_our_beamlines/CM01.html>) based around a Titan Krios microscope and making it available to its structural biology user community. CM01 is managed as is any other public ESRF beamline, with user access via either peer-reviewed applications or a paid-for proprietary mechanism. CM01 forms part of the Partnership for Structural Biology (PSB; <http://www.psb-grenoble.eu>) and its operation and development is assured by staff from a consortium of institutes co-located on the European Photon and Neutron Campus (EPN) in Grenoble. Experiments at CM01 are currently restricted to pre-characterized projects for which the sample-freezing conditions have been optimized and initial cryo-EM analyses indicate a high potential for obtaining information to the secondary-structure level. In the intermediate future it is planned that this restriction will be removed and that access to CM01 will be based on scientific merit alone. Here, preliminary sample screening will be carried out using the PSB cryo-EM platform (<http://www.ibs.fr/research/research-groups/methods-and-electron-microscopy-group/electron-microscopy-platform/>) microscopes located at the Grenoble Institut de Biologie Structurale (IBS). Once this mode of operation is fully functional, it will provide a unique opportunity for scientists of the ESRF's international community with limited or no access to EM facilities to access the technique for scientifically important projects. 2.. General setup and infrastructure {#sec2} ====================================== CM01 is sited on the 'golden slab' of the Belledonne extension of the ESRF experimental hall (Fig. 1[▸](#fig1){ref-type="fig"} *a*), which was constructed as part of Phase 1 of the ESRF Upgrade. The location was chosen following preliminary measurements confirming that it would provide a stable, vibration-free environment coupled with minimal variations in magnetic field (Table 1[▸](#table1){ref-type="table"}) that might otherwise be associated with installing a cryo-electron microscope in the experimental hall of a synchrotron source. The infrastructure (Figs. 1[▸](#fig1){ref-type="fig"} *b*--1[▸](#fig1){ref-type="fig"} *e*) in which CM01 is hosted comprises a grey room (ISO8 level) air-conditioned with a temporal thermal stability of 20 ± 0.1°C within 30 min and ±0.8°C per 24 h, a technical cabinet and an experimental control cabin. This room also contains the high-voltage generator, electronics and fluidics cabinets as well as the computer and processor units for the detector. These are screened from the microscope itself (Fig. 1[▸](#fig1){ref-type="fig"} *c*). All walls and the ceiling are acoustically damped to avoid any external disturbances, while the water-chilling unit that is needed to cool the electro-magnetic constant power lenses of the microscope is located in a technical cabinet (Fig. 1[▸](#fig1){ref-type="fig"} *d*) outside the experimental hutch, again for acoustic and vibrational insulation reasons. The microscope around which CM01 is built is a Titan Krios G3 (ThermoFisher Scientific; <https://www.fei.com>) currently operated at its maximal voltage of 300 kV and equipped with a GIF Quantum LS energy filter coupled to a Gatan K2 summit direct electron-counting camera. The microscope also features a Volta phase plate (VPP), which increases the contrast and as such enables the high-resolution imaging of smaller macromolecules of less than 150 kDa (Danev & Baumeister, 2016[@bb8]). CM01's spacious control cabin is equipped with the computing infrastructure required to control both the microscope and the detector (Fig. 1[▸](#fig1){ref-type="fig"} *e*) and also includes infrastructure for data processing and backup. Before each experiment, the microscope is set up and aligned to reduce the astigmatism and coma values to less than 0.5 µm (*AutoCTF*) by the facility personnel, who also clip the grids if necessary, load them into the cassette and then into the microscope. The microscope is operated in EFTEM mode with a nanoprobe for data acquisition. Currently, automatic data collection for single-particle reconstruction is carried out using the manufacturer's control software package *EPU* v.1.11 (ThermoFisher). A typical user session involves 4 h for grid screening (up to six grids) followed by the selection of the best grid squares and hole positions for image acquisition. Many experiments are carried out at a nominal magnification of 130 000 (pixel size of 1.056 Å) using a detector flux density of 5 e^−^ per pixel per second (usually denoted the 'dose rate' in EM) and exposure times of 8 s per movie (40 frames per movie), resulting in an accumulated fluence of 40 e^−^ Å^−2^ (usually denoted as the 'dose' in EM). Although the setting up and launching of image acquisition is carried out by the CM01 staff, the data-collection strategies are discussed directly with the users, who are either present on-site or access the facility remotely (see below). Using the current setup, the throughput is about 75--80 movies per hour for collection in counting mode for grids with a 2 µm hole size (five movies per hole). 3.. Computing infrastructure, data pre-processing and presentation of data using EXI/ISPyB {#sec3} ============================================================================================ The cryo-EM facility greatly benefits from the computing and network infrastructure available at the ESRF for experiments on its X-ray beamlines (Fig. 2[▸](#fig2){ref-type="fig"} *a*). Images or movies recorded by the detector are directly transferred to the ESRF central storage system (NICE, Networked Interactive Computing Environment; <https://www.esrf.eu/Infrastructure/Computing/NICE>) using a dedicated fibre-optic connection (10 Gb s^−1^). This facilitates not only fast data transfer but also allows both on-the-fly data backup onto an external hard drive and image pre-processing. In general, raw data are maintained on disk for up to 50 days from the experiment date and are then archived on tape and stored without time limit. Data backup to external hard disk drives is usually carried out by the users using the same easy-to-use interface that is available on all ESRF beamlines for structural biology. Alternatively, users can also download and transfer data to their home storage system via an rsync command. CM01 implements the ESRF data policy (<http://www.esrf.fr/datapolicy>). Parallelized image pre-processing (Fig. 2[▸](#fig2){ref-type="fig"} *b*) using several GPU-equipped computing clusters is automatically launched on the collected movies using the *SCIPION* wrapper (Conesa Mingo *et al.*, 2018[@bb7]; Gómez-Blanco *et al.*, 2018[@bb11]). The imported movies are drift-corrected using *MotionCor*2 (Zheng *et al.*, 2017[@bb30]) and contrast-transfer function (CTF) parameters are estimated using *Gctf* (Zhang, 2016[@bb29]). With the current GPU infrastructure, which is subject to constant upgrades, and for an incoming rate of up to 80 movies per hour, pre-processing results are calculated with almost no delay. The output of *MotionCor*2 and *Gctf* are uploaded into the ISPyB database (Delagenière *et al.*, 2011[@bb9]) and are displayed, along with other metadata, as snapshots in EXI (Fig. 3[▸](#fig3){ref-type="fig"}; <https://exi.esrf.fr>). Several relevant parameters, including maximum resolution and astigmatism, are constantly uploaded and plotted both in EXI and on a separate screen inside the control cabin, allowing on-the-fly quality control of incoming data. In the imminent future, extension of the existing data-processing pipeline to include automated or semi-automated particle picking and the generation of 2D classes are foreseen. 4.. Sample-preparation laboratory and storage facility {#sec4} ======================================================== As noted above, experiments at CM01 are currently restricted to pre-characterized samples for which frozen grids have been prepared. However, and in preparation for the operation phase of CM01, in which preliminary sample screening and sample optimization will be carried out using the PSB cryo-EM platform, a fully equipped sample-preparation laboratory is associated with the facility (Fig. 4[▸](#fig4){ref-type="fig"}). The equipment in this laboratory comprises a ThermoFisher Scientific Mark IV Vitrobot and all of the tools necessary for grid preparation, including a glow-discharge apparatus (Ted Pella easyGlow). Users currently applying for microscope time to analyse grids based on pre-characterization can ask for access to this laboratory during the application procedure (see below). External users of CM01 have the option of either coming to the ESRF to oversee their experiments or sending samples for remote and mail-in data collection. In both cases, cryo-EM grids can be sent to the ESRF prior to the experiment date and stored using individually barcoded grid-boxes provided by the ESRF (<https://www.mitegen.com/product/mitegen-cryo-em-pucks-generation-2-0/>) in a dedicated storage area. As for experiments on the ESRF's X-ray-based structural biology beamlines, samples must be registered using the ISPyB sample-tracking module (Delagenière *et al.*, 2011[@bb9]). In the case that more data need to be collected from a set of grids, samples can be stored for up to six months, a period that can be prolonged upon request. Again, as for the ESRF's X-ray-based structural biology beamlines, if users do not come to the ESRF for their experiment then the costs of sample transport to and from Grenoble are met by the ESRF (see <https://www.esrf.eu/UsersAndScience/Experiments/MX/How_to_use_our_beamlines/Prepare_Your_Experiment/Dewar_sending>). 5.. How to apply for access {#sec5} ============================= Currently, CM01 delivers ∼600 8 h shifts per year for external user experiments. Experiments are scheduled on a three-day, two-day or one-day basis according to the beamtime demanded in the proposal, with the vast majority of experiments using three days. 10% of the beamtime given is dedicated to in-house research. One day per week is reserved for microscope maintenance and the remaining time is used for commissioning (including service interventions and upgrades) and developments on the microscope. As for all of the ESRF's facilities for structural biology, access to CM01 can be on either a public or a proprietary basis. For the former, applications, peer-reviewed by the ESRF's MX Beam-time Allocation Panel (MX-BTAP) comprising several cryo-EM specialists, can currently be submitted at any time via the ESRF's Rolling Access Mechanism (<http://www.esrf.eu/UsersAndScience/UserGuide/Applying/MXApplications>). For accepted proposals, the ESRF provides financial support for either user travel to Grenoble (including accommodation and subsistence) or for sample transport to/from Grenoble. Aside from this core mission of providing user service, hands-on workshops on sample preparation (see, for example, <http://www.esrf.eu/cryo-em2019-1-workshop.html>) as well as on theoretical principles of cryo-EM are held on a regular basis. The normal waiting time from proposal submission until experiment is currently around two months. Proprietary access to CM01 is via the ESRF's Business Development Office (BDO; <http://www.esrf.fr/Industry/contact-industrial-services>). 6.. Microscope benchmarking {#sec6} ============================= In order to benchmark the microscope installed on CM01, we used Tobacco mosaic virus (TMV) as a test sample for single-particle reconstruction. TMV grids were prepared using the Vitrobot by pipetting 3.5 µl of 3 mg ml^−1^ TMV onto Quantifoil 1.2/1.3 400 mesh grids (<https://www.quantifoil.com/>); excess liquid was blotted off for 2 s. The K2 detector was used in counting mode for automatic data collection using the *EPU* software. Movies were collected as gain-normalized and unpacked MRC files at a magnification of 130 000, yielding a calibrated pixel size of 1.067 Å. Drift correction was carried out using *MotionCor*2 (Zheng *et al.*, 2017[@bb30]) and CTF estimation was performed by *Gctf* (Zhang, 2016[@bb29]). Further image-processing steps were performed using *RELION* 2.1 (Kimanius *et al.*, 2016[@bb14]; Scheres, 2012[@bb24]). Segments were automatically picked and extracted with about 90% overlap between segments. 2D reference-free classification was performed and class averages showing high-resolution features were selected and included for 3D refinement. As a final step, local refinement of the CTF parameters was performed using *Gctf* (Zhang, 2016[@bb29]). The map was further post-processed by automatically applying a *B* factor of −83 Å^2^, yielding a gold-standard FSC (Henderson *et al.*, 2012[@bb13]) resolution of 2.3 Å (Fig. 5[▸](#fig5){ref-type="fig"}), one of the highest resolutions for the TMV structure to be published (Weis *et al.*, 2019[@bb31]). Even though the resolution of the TMV structure presented here is not the highest to be reported (Song *et al.*, 2019[@bb25]), the result obtained clearly shows the very high quality of data that can be obtained using CM01. The final helical parameters are 1.408 Å for the helical rise and 22.03° for the helical twist. One TMV protomer (PDB entry [4udv](http://scripts.iucr.org/cgi-bin/cr.cgi?rm=pdb&pdbId=4udv); Fromm *et al.*, 2015[@bb10]) was rigid-body fitted into the density map using *UCSF Chimera* (Pettersen *et al.*, 2004[@bb22]) and the atomic positions were further refined using *PHENIX* real-space refinement (Afonine *et al.*, 2018[@bb1]). The final model fits well into the density with very good statistics (Table 2[▸](#table2){ref-type="table"}). The TMV protomer modelled consists of 153 residues (residues 65 and 98--102 were not visible). Most of the side-chain densities are well defined (Fig. 5[▸](#fig5){ref-type="fig"}) and only 17% of residues exhibit poorly defined side-chain density. Of these, about 40% are negatively charged amino acids such as aspartic and glutamic acid residues, which are prone to radiation damage (Hattne *et al.*, 2018[@bb12]). 7.. The first year of operation {#sec7} ================================= The vast majority of users are from academic institutions located in the 13 ESRF member states (Fig. 6[▸](#fig6){ref-type="fig"}). At the time of writing, seven publications based, in whole or in part, on data collected at CM01 have already been published. These include different states of the 5HT3 receptor, which allow a description of its activation cycle upon binding to its serotonin substrate (Polovinkin *et al.*, 2018[@bb23]), the structure of the membrane complex (MC) of a type VI secretion system (T6SS), providing insights into the mechanism of action of the MC complex during T6SS assembly (Cherrak *et al.*, 2018[@bb5]), the natural tetrameric structure of human butyrylcholinesterase (BChE; Boyko *et al.*, 2019[@bb4]), the structure of Hantaan virus (Arragain *et al.*, 2019[@bb3]), the structure of the human ferritin--transferrin receptor 1 complex (Montemiglio *et al.*, 2019[@bb18]) and the structure of adenovirus type 3 fibre with its receptor desmoglein 2 (Vassal-Stermann *et al.*, 2019[@bb28]). However, given that most of the experiments carried out are rather challenging, no reliable statistics on publications per year can yet be given. 8.. Perspectives {#sec8} ================== With the rapid growth of the cryo-EM community in the last few years, the demand for data collection at high-end microscopes such as the Titan Krios installed at CM01 has skyrocketed. Since its installation, CM01 has proved to be a highly successful facility, acting in concert with the ESRF's X-ray-based beamlines for structural biology as a source of primary or complementary structural information in a wide range of projects. One of the major driving forces for the creation of CM01 was to provide a facility that can be accessed by all of the ESRF's international structural biology user community. However, a lack of infrastructure for sample preparation and screening remains a bottleneck, particularly for researchers in smaller laboratories. To address this and to broaden the accessibility of cryo-EM to researchers working in ESRF member states, the PSB Cryo-EM platform will, for projects approved by the ESRF MX-BTAP, soon provide additional services to users based on one of the following models: (i) the screening of frozen grids and a preliminary data collection if negative-stain characterization has already been carried out and (ii) sample characterization using negative-staining methods followed by cryo-EM sample preparation and screening. If the above steps are successful, data collection using the Titan Krios at CM01 will follow. The ESRF will also continue its training workshops for external users on both theoretical principles and the practical aspects of sample preparation. We would like to thank all of the ESRF support groups, such as the technical infrastructure division, safety, communication and data-analysis units and computing and network services, for their invaluable support during the construction, setup and commissioning of the microscope. The support of Francesco Sette (Director General of the ESRF), Harald Reichert (Director of Research of the ESRF), Winfried Weissenhorn (Director of the IBS) and Stephen Cusack (Head of EMBL Outstation Grenoble) was crucial for this project. We would also like to mention FEI, Gatan and particularly their application support teams for continuous training and support, specifically Pierre-Yves Mille, Remi Boucher, Bruno Barlassina, Ana Alarcon, Wim Woorhut, Eric Bathelt, Felix de Haas, Thomas Hoffmann and Evgeniya Pechnikova. ![CM01 infrastructure. (*a*) Location of CM01 in the Belledonne extension of the experimental hall of the ESRF. (*b*) Schematic top view of CM01 showing the experimental hutch (EH) in blue, the technical gallery in yellow and the control cabin (CC) in violet. (*c*) The CM01 experimental hutch (grey room) showing the Titan Krios microscope (1), the duct sock porous mesh for uniform air flow (2), SF6 extraction (3) and the acoustic damping sheets (4) behind which the high-voltage generator is located. (*d*) Technical gallery, where the chiller unit (indicated) is placed. (*e*) The large CM01 control cabin.](d-75-00528-fig1){#fig1} ![(*a*) Schematic of the IT infrastructure at CM01 for data collection and transfer connections between computers (pc TF, pc GATAN and the microscope are connected via ethernet). (*b*) The single-particle cryo-EM image pre-processing workflow currently carried out at CM01. The collected movies are motion-corrected and CTF values are estimated. The obtained metadata are then accessed and uploaded to the EXI web interface.](d-75-00528-fig2){#fig2} ![The EXI web interface (<https://exi.esrf.fr>) provides a summary of the data-collection parameters and pre-processing results on the fly. (*a*) Snapshot of the pre-processing results showing the unaligned average (first column), motion-corrected average (second column) and CTF estimation (third column) images and associated values. (*b*) Overview of data quality. The first three rows show the image series plots and histograms for the information limit in the micrographs, the astigmatism angle and defocus *U* and *V* (astigmatism in the middle frame), respectively. The last plot indicates the average motion per frame in micrometres from the alignment of the acquired movies.](d-75-00528-fig3){#fig3} ![Laboratory access is provided to users for cryo-EM sample preparation. The laboratory includes a Vitrobot positioned in the fume hood, as can be seen on the left side.](d-75-00528-fig4){#fig4} ![3D reconstruction of TMV particles. (*a*) 3D reconstruction of TMV. Alternate subunits are coloured in yellow and green. The scale bar represents 5 nm. (*b*) Masked-corrected Fourier shell correlation (FSC) curve for resolution. (*c*) An asymmetric unit modelled in the electron density. (*d*) Some examples representing the quality of the EM density and the model.](d-75-00528-fig5){#fig5} ![Breakdown of the shifts delivered in 2018 by proposal country of origin for experiments at CM01.](d-75-00528-fig6){#fig6} ###### Vibrational and electromagnetic measurements inside the experimental hutch of CM01 The mechanical vibrations are measured in units of *g* (the acceleration owing to gravity in m s^−2^) in three directions approximately at the position of the column at the floor level. Both the AC and DC electromagnetic fields (EMF) are measured in nanotesla p/p in three directions to the column position for three different heights \[the GIF approximately 0.5 m from the floor, the compustage approximately 1.5 m from the floor and the gun (source) approximately 2.5 m from the floor\]. Front to back Left to right Vertical ---------------------- --------------------- --------------------- -------------------- Mechanical vibration 1.1 (1.56 µm s^−1^) 1.1 (1.56 µm s^−1^) 2.3 (3.2 µm s^−1^) EMF AC GIF level 42 (50 nT) 21 (50 nT) 29 (75 nT) Stage level 43 (80 nT) 20 (80 nT) 33 (80 nT) Source level 42 (80 nT) 21 (80 nT) 39 (80 nT) Filtered GIF level 19 (50 nT) 18 (20 nT) 23 (75 nT) Stage level --- --- --- Source level --- --- --- Near DC GIF level 14 (50 nT) 7 (50 nT) 19 (75 nT) Stage level 6 (80 nT) 11 (80 nT) 19 (80 nT) Source level 3 (80 nT) 12 (80 nT) 58 (80 nT) ###### Data-collection and refinement statistics for the single-particle reconstruction of TMV -------------------------------- -------------- Data collection Accelerating voltage (kV) 300 Nominal magnification 130 000 Calibrated pixel size (Å) 1.067 Total No. of movies collected 357 Total fluence (e^−^ Å^−2^) 40 No. of frames 40 Defocus range (µm) −1.0 to −3.0 Data processing Final segments 109763 Helical parameters Twist (°) 22.04 Rise (Å) 1.41 Gold-standard resolution (Å) 2.3 *MolProbity* statistics All-atom clashscore 2.41 Ramachandran plot Outliers (%) 0 Allowed (%) 3.97 Favoured (%) 96.03 Rotamer outliers (%) 0 C^β^ deviations (%) 0 Deviations from ideal values Bond lengths (µm) 0.007 Bond angles (°) 1.030 Chirality (Å^3^) 0.060 Planarity (°) 0.009 Dihedral (°) 8.534 Map--model CC Main chain (760 atoms) 0.68 Side chain (513 atoms) 0.63 EMDB ID EMD-4928 -------------------------------- --------------

{ "pile_set_name": "PubMed Central" }

1. Introduction =============== Suture anchors are increasingly used to fix soft tissue to bone.^\[[@R1]\]^ The modified Brostrom operation (MBO) is the most widely used procedure for the correction of ankle instability.^\[[@R2]\]^ Favorable outcomes of the MBO using suture anchors have been reported.^\[[@R1],[@R3]\]^ The infection rate after such surgery is low: 0% to 4.5%.^\[[@R1],[@R2],[@R4]\]^ However, a suture anchor is not free of complications because a suture anchor acts as an implant. After shoulder surgery, suture anchor complications include pull-out,^\[[@R5],[@R6]\]^ arthritis,^\[[@R5]\]^ and osteolysis.^\[[@R7]\]^ However, complications of suture anchors used in ankle surgery are rare. We encountered a chronic infection of the sinus tract after the MBO using a suture anchor. The wound healed completely after removal of the sinus tract and infected suture anchor. 2. Case description =================== This case report was approved by the Institutional Review Board of Eulji Medical Center, and the patient gave written informed consent. The patient was a healthy 18-year-old male who underwent anterior talofibular ligament (ATFL) and calcaneofibular ligament (CFL) repair using a suture anchor at a local clinic 1 month after being injured while playing soccer. There were no other injuries. The ATFL and CFL were repaired using a JuggerKnot 1.45-mm LongSoft Anchor (Zimmer Biomet, Warsaw, IN). The ATFL was repaired at the talar side and the CFL at the fibular side. According to the medical record of the first hospital, an infection featuring a purulent exudate developed in the surgical site 1 week after the operation. Irrigation and debridement were performed in the operating room every day for 10 days. The wound discharge and swelling decreased, and a simple suture was placed. There was no evidence of wound infection, and the suture was removed and the patient discharged. However, abrupt wound discharge and dehiscence developed after suture removal. He did not visit that hospital again, for personal reasons, and personally dressed the chronic wound at home. He visited our outpatient clinic 1 year later. A clinical examination revealed a 1.5 × 2-cm-sized ulcerative wound accompanied by a red-black discharge on the anterior border of the right lateral malleolus (Fig. [1](#F1){ref-type="fig"}). We could approximate the wound to the joint capsule level by applying forceps at the wound site. A simple standing anteroposterior radiograph of the right ankle revealed mild lateral soft-tissue swelling. Magnetic resonance imaging revealed a sinus tract running from the wound to the talar body across the anterior joint capsule, and bone marrow edema surrounding the talus (Fig. [2](#F2){ref-type="fig"}). The American Orthopedic Foot and Ankle Society (AOFAS) Ankle-Hind Foot score was 45. Intravenous vancomycin was commenced on the day of surgery. ![A 1.5 × 2-cm-sized ulcerative wound accompanied by a red-black discharge on the anterior border of the right lateral malleolus.](medi-97-e13738-g001){#F1} ![A T2-weighted axial image (A) and a proton density coronal image (B) showing the sinus tract (arrow) running from the wound to the talar body (crossing the anterior joint capsule), and bone marrow edema (arrowhead) surrounding the talus.](medi-97-e13738-g002){#F2} In the operation theater, we excised the sinus tract, sent material for culture, and performed irrigation and debridement; our surgical approach was via the chronic ulcerative wound. Surgery was performed with the patient right side up in the lateral position after placement of a pneumatic tourniquet under general anesthesia. A rectangular skin incision encompassing the ulcerative wound was created, followed by dissection along the chronically infected sinus tract (Fig. [3](#F3){ref-type="fig"}). Intraoperatively, the sinus tract was seen to lie on top of the suture anchor (Fig. [4](#F4){ref-type="fig"}). The suture anchor was loose and infected and was removed (Fig. [5](#F5){ref-type="fig"}). The chronically infected sinus tract was excised (Fig. [6](#F6){ref-type="fig"}) and a pathological examination was performed (Fig. [7](#F7){ref-type="fig"}). A hole through the sinus tarsi was evident near the wound; we performed wound culture (Fig. [8](#F8){ref-type="fig"}). As the surgical site was clean and no discharge was evident, we decided to perform immediate repair. As a rectangular wound cannot be easily sutured, we created rectangular superior and inferior skin flaps (Fig. [9](#F9){ref-type="fig"}). The anterior flap was lowered and the posterior flap raised to create a zigzag-shaped wound (Fig. [10](#F10){ref-type="fig"}). After simple suturing, a short leg cast was applied and the wound dressed through the cast window (Fig. [11](#F11){ref-type="fig"}). The patient received intravenous vancomycin during 14 days of hospitalization. He was discharged on postoperative day 14 with a 1-week prescription of oral amoxicillin/clavulanic acid. At 4 weeks postoperatively, we removed the suture and cast; no wound dehiscence was evident. No bacterium grew on wound culture; the pathological diagnosis was chronic inflammation and abscess formation. He visited our outpatient clinic for scheduled follow-up at 12 months after surgery. The wound had completely healed (Fig. [12](#F12){ref-type="fig"}). He reported no symptoms of infection and regained full function of the right ankle. The anterior drawer test was negative. The AOFAS Ankle-Hind Foot score improved to 90. ![A rectangular skin incisions encompassing the ulcerative wound was created, followed by dissection along the chronically infected sinus tract.](medi-97-e13738-g003){#F3} ![Intraoperatively, the sinus tract was found to lie over the suture anchor. The anchor was loose and infected, and was removed.](medi-97-e13738-g004){#F4} ![The removed suture anchor.](medi-97-e13738-g005){#F5} ![The chronically infected sinus tract was excised.](medi-97-e13738-g006){#F6} ![The sinus tract; pathological examination.](medi-97-e13738-g007){#F7} ![A hole through the sinus tarsi was evident at the wound site.](medi-97-e13738-g008){#F8} ![Rectangular skin flaps were created superior and inferior to the wound.](medi-97-e13738-g009){#F9} ![The anterior flap was lowered and the posterior flap raised to create a zigzag-shaped wound.](medi-97-e13738-g010){#F10} ![After simple suturing, a short leg cast was applied and the wound dressed through the cast window.](medi-97-e13738-g011){#F11} ![The wound exhibited no complication or recurrence of infection at the 1-year follow-up exam.](medi-97-e13738-g012){#F12} 3. Discussion ============= Chronic lateral ankle instability (CLAI) is defined as an instability more than 6 months in duration, manifesting as repeated episodes of "giving out", persistent pain, and an inability to resume the pre-injury activity level.^\[[@R1]\]^ About 50 surgical procedures are available to treat CLAI. Anatomical repair of the ATFL and CFL can restore normal ankle and subtalar joint mechanics. The MBO is the most popular technique for anatomical reconstruction of lateral ankle ligaments; the success rate is 85% to 90%.^\[[@R8]--[@R10]\]^ Suture anchors were developed to facilitate soft tissue fixation to bone, and have been placed during both open and arthroscopic MBOs.^\[[@R1]--[@R3],[@R11],[@R12]\]^ Such placement is technically simple and allows accurate ligament re-attachment to the normal anatomical origins.^\[[@R1]\]^ However, suture anchors, being implants, can cause problems.^\[[@R5],[@R13],[@R14]\]^ Reported complications associated with suture anchor used in shoulder surgery include pull-out,^\[[@R5],[@R15]\]^ arthritis,^\[[@R5]--[@R7]\]^ and osteolysis.^\[[@R7],[@R16]\]^ To the best of our knowledge, complications associated with suture anchor use during ankle surgery have been rarely reported. Suture anchors implanted during orthopedic surgery can cause infections, both perioperatively and postoperatively. In our present case, the exact cause of infection was unknown, but given that a chronically infected sinus tract lay over the suture anchor, the anchor is likely to have caused the infection. The removal of orthopedic implants is important to control infection. We removed the infected suture anchor and the chronically infected sinus tract; the wound healed after repair without infection recurrence. A few reports have described the incidence and management of culture-negative infections developing after orthopedic surgery. Such infections arise after prior antibiotic treatment to counter infections with unusual organisms such as fungi, mycobacteria, or bacteria that form biofilms. Two case reports have appeared.^\[[@R17],[@R18]\]^ Bacterial cultures were negative, but *Mycobacterium* was found in the wounds. This pathogen must be considered if, postoperatively, wounds present with chronically discharging sinuses, even in healthy non-immunocompromised patients. In our case, the sinus tract was chronically infected, but Gram staining and bacterial cultures were negative. We did not stain for acid-fast bacilli or seek to culture *Mycobacterium*; these tests should be considered in cases featuring the chronic infection of a sinus tract. In conclusion, the MBO with placement of a suture anchor affords good results, but the anchor may become a source of infection. Removal of the chronically infected sinus tract and the infected suture anchor are required to heal the wound. Acknowledgments =============== The authors would like to thank the Soonchunhyang University Research Fund for support. Author contributions ==================== **Conceptualization:** Woo Jong Kim. **Investigation:** Ki Won Young. **Software:** Sung Hun Won, Kyu Hwan Bae. **Supervision:** Hong Seop Lee. **Writing -- original draft:** Hong Seop Lee. **Writing -- review & editing:** Chang Hyun Kim, Hong Seop Lee. Hong Seop Lee orcid: 0000-0002-3589-314X. Abbreviations: AOFAS = American Orthopedic Foot and Ankle Society, ATFL = anterior talofibular ligament, CFL = calcaneofibular ligament, CLAI = chronic lateral ankle instability, MBO = modified Brostrom operation. WJK and KWY contributed equally to this study and should be considered co-first authors. No benefit in any form has been or will be received from a commercial party related directly or indirectly to the subject of this article. The authors have no potential conflicts of interest to disclose.

{ "pile_set_name": "PubMed Central" }

The face is perhaps the most powerful communication channel for human social interaction, enabling the observer to quickly infer information about the sender[@b1][@b2]. Faces allow not only recognizing a unique identity, but also communicating basic emotional states. Dating back to Darwin[@b3], faces and facial expressions have been extensively studied in psychology and neuroscience, resulting in elaborate cognitive[@b4] and neuroscientific theories (e.g. see refs [@b5] and [@b6]). Some theories also had a big impact on other disciplines. Ekman's theory of universal basic emotions[@b7][@b8] and the resulting Facial Action Coding System (FACS) became very popular in computer graphics and are a fundamental concept in current facial animation systems. By blending between different FACS expressions a complex facial expression is created for a virtual character. In this regard, advancements in face perception research are also of high interest for many commercial productions in the game or visual effects industry. Convincing characters highly contribute to the story and visual quality of a movie binding a lot of time and money at production time. In the movie "Maleficent" for example, expensive hardware, custom software, and months of work by highly skilled professionals and researchers was required to create the digital faces of the flower pixies[@b9]. However, little is known about how stylized faces are processed by the brain and about which facial details are important in particular. Even less is known about how a presented emotional expression interacts with stylization of a face. Concerns about the influence of stylization have recently also become an issue for face perception research in psychology, where computer generated characters start to replace photographs of real people due to higher flexibility and better experimental stimulus control[@b10]. Regarding the effect of character realism on human perception, so far, the uncanny valley theory[@b11] dominates the discourse in research and practice. According to Mori, humanoids are in general perceived as more appealing if they are more realistic. However, humanoids that are more realistic than cartoon characters, but do not fully achieve human-likeness, may cause an "eerie feeling" and fall into the "uncanny valley". This theory has been developed on the basis of Mori's personal observation in the 1970s and has been subject of some empirical research in the last decade. Initial research used morphs of different characters[@b12] or videos of different robots at different stylization levels[@b13]. Also image editing algorithms[@b14] or different rendering algorithms[@b15] have been applied to create facial stimuli at different stylization levels. Other work investigated the threshold beyond which characters are perceived as real, by morphing between photographs and puppet faces[@b16] or photographs and virtual faces[@b17]. Despite its intuitive appeal, the uncanny valley concept still lacks empirical support[@b18] and no final agreement exists about the measurement scales[@b19] or potentially underlying perceptual mechanisms. However, facial features such as skin impurities[@b20] or strong deviations of facial proportions like too big eyes[@b21] have been identified as reliable characteristics of unappealing characters. Event-related potentials (ERPs) are an excellent way to analyse face perception, as the measured brain responses can distinguish between highly automatic and more controlled stages of processing and do not require overt responses that may be distorted by experimental demand characteristics. Neuroscientific research has shown that face processing strongly activates dedicated areas in the visual cortex, in particular the occipital face area[@b22][@b23] and the fusiform face area[@b24]. The fusiform face area is also involved in generating the face-sensitive N170 component in ERP studies[@b25]. Previous work showed that the N170 component peaks selectively for faces[@b26]. Real faces elicit a stronger N170 compared to abstract sketches of faces, but the difference is not statistically significant compared to schematic faces[@b27], suggesting a gradual increase. Larger N170 amplitudes are found for human faces compared to those of other species[@b28]. Attractiveness also seems to induce small modulations in the N170 for characters of the same stylization level[@b29][@b30]. Further, it has been found that baby faces cause a higher N170 amplitude than adult faces, most likely due to neotenic features, like proportionally larger eyes[@b31]. Finally, a recent meta-analysis indicated that larger N170 amplitudes are observed for angry, fearful, and happy than for neutral facial expressions[@b32]. Moreover, emotion effects at the N170 can be found for faces of medium stylization[@b33] and even for robots with rather schematic faces[@b34]. Enhanced processing of emotional faces is also reflected in an enhanced Early Posterior Negativity (EPN), as well a larger Late Positive Potential (LPP)[@b35][@b36]. The EPN indicates early attention mechanisms, whereas LPP is viewed as reflecting higher-order evaluation and episodic memory encoding (for an overview see ref. [@b37]). Both EPN and LPP amplitudes are sensitive to voluntary attention deployment to a stimulus and can be increased by explicit attention instructions[@b38][@b39][@b40]. Both ERP components are also sensitive to emotional content of various stimulus types, including faces[@b41][@b42][@b43][@b44][@b45]. This emotion sensitivity is thought to reflect intrinsically motivated attention to emotional stimuli and often co-varies with stimulus intensity or arousal[@b35][@b46][@b47][@b48]. As, in the absence of relevant social context, participants typically report less subjective arousal for picture of emotional faces than for pictures of emotional scenes, differences in arousal-level might account for generally smaller ERP emotion effects in response to faces compared to scenes[@b49]. Apart from emotional expressions, larger EPN[@b50] and LPP amplitudes[@b30][@b50] are reported for attractive faces compared to unattractive faces, also applying to moderate stylizations[@b29]. Interestingly, in a recent study contrasting real neutral faces with neutral faces of puppets, no differences at the N170 level were observed, while from 400 ms onwards a larger LPP was found for real faces[@b51]. This was attributed to the salience and unique identity of a real face and to mentalizing about the depicted individual[@b51], as generally computer generated faces are harder to remember[@b52][@b53]. In this vein, manipulating perceived uniqueness or distinctiveness via shape or reflectance manipulations of initially non-distinctive real faces has been found to result in a larger late positivity as well as a better memory performance[@b54][@b55]. Further, biographical information facilitates the processing of real faces at late ERP components[@b56][@b57], which has recently also been shown for realistic computer generated faces[@b58]. The combined evidence suggests that a unique identity, either pre-existing or ascribed, enhances processing particularly at late stages. Against the above background, the present research systematically addresses how brain responses known to reflect distinct stages of face processing vary across different levels of character stylization and for different emotional expressions. While previous work, in general, compared a real photograph of a person to a single stylization level, no previous study systematically manipulated face stylization of the same identity and studied how this influences brain responses towards emotional and neutral expressions. Our stimulus set exhibits six levels of stylization for two identities, depicting happy, neutral and angry expressions[@b20]. These cover a wide range of stylization, were professionally created, and match quality and style of current movies. All stimuli are rendered under the same conditions, thus reducing inconsistencies to a minimum. We expected to find strongest responses for real faces at the N170, EPN and LPP[@b26]. For very abstract faces the proportionally larger eyes and uniformly coloured textures might also cause a strong N170[@b31]. For angry and happy faces, we expected to find larger N170, EPN and LPP amplitudes. Interactions between level of stylization and emotional expression were explored on all components. Moreover, source analyses were employed to uncover the cortical generators of stylization effects, addressing the possibility of differential recruitment of visual areas such as fusiform or occipital face areas. Results ======= ERP results ----------- ### N170 For the N170 component significant modulations of stylization (*F*~(5,160)~ = 15.93, *p* \< 0.001, partial η~P~^2^ = 0.332; see [Fig. 1](#f1){ref-type="fig"}), emotion (*F*~(2,64)~ = 50.33, *p* \< 0.001, partial η~P~^2^ = 0.611; see [Fig. 2](#f2){ref-type="fig"}), as well as an interaction of stylization with emotion (*F*~(10,320)~ = 2.44, *p =* 0.008, partial η~P~^2^ = 0.071) were found over the two large symmetrical occipital sensor groups (see [Fig. 1](#f1){ref-type="fig"}). For stylization levels, we tested a linear (*F*~(1,32)~ = 0.09, *p =* 0.765, partial η~P~^2^ = 0.003), compared to a u-shaped, quadratic, contrast (*F*~(1,32)~ = 68.50, *p* \< 0.001, partial η~P~^2^ = 0.682), and found a u-shaped form for the face-specific N170 component. Although these u-shaped forms are visible and highly significant for all emotions, we identified with increased realism less intense reactions towards happy expressions and more intense reactions towards angry faces (see [Fig. 1](#f1){ref-type="fig"}, upper panel). Moreover, emotional modulations resulted in a stronger quadratic (*F*~(1,32)~ = 59.81, *p* \< 0.001, partial η~P~^2^ = 0.651), compared to a linear contrast (*F*~(1,32)~ = 41.64, *p* \< 0.001, partial η~P~^2^ = 0.565), showing the most pronounced N170 for angry faces, smallest for neutral faces and in between happy expressions (all *ps* \< 0.001). Additionally, there was a trend for a main effect of channel group (*F*~(1,32)~ = 3.37, *p =* 0.075, partial η~P~^2^ = 0.095) and an interaction of stylization with channel group (*F*~(5,160)~ = 3.30, *p =* 0.025, partial η~P~^2^ = 0.087), showing in tendency a u-shaped distribution with two maxima over the left and a u-shaped distribution with one maximum over the right sensor cluster. The interaction of emotion with channel group (*F*~(2,64)~ = 0.58, *p =* 0.532, partial η~P~^2^ = 0.018) and triple interaction were both insignificant (*F*~(10,320)~ = 1.03, *p =* 0.471, partial η~P~^2^ = 0.031). ### EPN In the EPN time range, over the same cluster, main effects of emotional content (*F*~(2,64)~ = 19.89, *p* \< 0.001, partial η~P~^2^ = 0.383) and laterality were observed (*F*~(1,32)~ = 33.83, *p* \< 0.001, partial η~P~^2^ = 0.514; see [Fig. 2](#f2){ref-type="fig"}). The EPN effect was somewhat better explained by a u-shaped form (*F*~(1,32)~ = 22.73, *p* \< 0.001, partial η~P~^2^ = 0.415), compared to a linear contrast (*F*~(1,32)~ = 17.59, *p* \< 0.001, partial η~P~^2^ = 0.355), where the strongest responses were found for angry faces, least responses to neutral faces and happy faces in between. The effect of laterality showed a larger EPN on the right compared to the left sensor group. There was no effect of stylization (*F*~(5,160)~ = 1.56, *p =* 0.215, partial η~P~^2^ = 0.046), and no interaction of stylization with emotion (*F*~(10,320)~ = 1.04, *p =* 0.412, partial η~P~^2^ = 0.031), or channel group (*F*~(5,160)~ = 0.80, *p =* 0.555, partial η~P~^2^ = 0.024). Further, there was no interaction of emotion with channel group (*F*~(2,64)~ = 0.76, *p =* 0.474, partial η~P~^2^ = 0.023) and no triple interaction (*F*~(10,320)~ = 1.59, *p =* 0.108, partial η~P~^2^ = 0.047). ### LPP In the LPP time window, over a large parietal sensor group, we found only a main effect of stylization (*F*~(5,160)~ = 12.62, *p* \< 0.001, partial η~P~^2^ = 0.283). Here, a linear increase in the LPP amplitude was observed (see [Fig. 3](#f3){ref-type="fig"}). The linear contrast (*F*~(1,32)~ = 31.28, *p* \< 0.001, partial η~P~^2^ = 0.494) accounted for substantially more variance explained than did a u-shaped contrast (*F*~(1,32)~ = 11.39, *p =* 0.002, partial η~P~^2^ = 0.262). There was no effect of emotion (*F*~(2,64)~ = 0.04, *p =* 0.965, partial η~P~^2^ = 0.001), and no interaction of stylization with emotion (*F*~(10,320)~ = 1.00, *p =* 0.444, partial η~P~^2^ = 0.030). Source reconstruction --------------------- Source reconstructions were calculated for significant effects of stylization level. All presented faces elicited strong visual responses (see [Figs 4](#f4){ref-type="fig"} and [5](#f5){ref-type="fig"} upper panel). However, the extreme poles (photographs compared to cartoon faces) showed a distinct processing already in the N170: While real faces led to larger inferior and middle occipital activations, highly stylized faces caused stronger responses in the right cuneus/lingual gyrus (see [Fig. 4](#f4){ref-type="fig"} and [Table 1](#t1){ref-type="table"}). Thus, despite similar N170 peaks for the extreme poles, the cortical generators differ. For neither of the extreme poles did we find statistical differences in source localization compared to moderately stylized characters. Later, in the LPP stronger superior occipital and superior parietal activations are observed for real compared to cartoon faces (see [Fig. 5](#f5){ref-type="fig"} and [Table 2](#t2){ref-type="table"}). These source estimations mirror the linearly increasing LPP for higher realism. With increasing realism of the faces, the differences to real faces become smaller and finally disappear in middle occipital areas. For the reverse contrasts, no differences were found. Discussion ========== This study investigated the cerebral processing of stylized faces across six levels of realism and three levels of emotional expression (happy, angry, and neutral), as reflected by the N170, EPN and LPP components. The results demonstrate that character stylization affects both the N170 and the LPP amplitudes, albeit in a qualitatively different manner. For the N170, a u-shaped modulation was observed, while continuous amplitude increases with increasing realism occurred for the LPP. For the N170 both highly stylized faces and real faces elicited strongest responses. Further, at the N170 level, a differential effect of face stylization on emotional expression was found: For cartoon characters, happy expressions caused similar N170 amplitudes as did angry faces, while for realistic faces (levels 5 and 6) only angry expressions were selectively processed. The EPN component was modulated solely by emotional expression, angry faces eliciting largest amplitudes. The results are striking in that they reveal a dissociation of stylization effects on the N170 and LPP: Taking into account that (i) the N170 amplitude is larger for faces than for objects[@b24][@b25][@b31], (ii) larger for real compared to schematic faces[@b27] and (iii) cuteness and baby-like features have been associated with a larger N170[@b31], we suggest neoteny and perceived face realism to drive the u-shaped N170 modulation. Analysis of stimulus properties (see [Fig. 6](#f6){ref-type="fig"}) indicates that neotenic features, such as eye size, decrease non-linearly and very quickly for stylization levels 1 to 3. On the other hand, perceived realism increases linearly for the tested characters. N170 generators were found to differ between very abstract and realistic faces: Although all faces activated extended visual regions, including the right fusiform gyrus, cartoon faces elicited stronger early visual cortex activations (cuneus, lingual gyrus, inferior occipital gyrus), while for real faces, stronger activations were found in middle occipital regions. Results suggest that processing of highly stylized faces relies more on structural analysis, associated with the so-called occipital face area, whereas realistic faces activate to a greater extent holistic processing, associated with the fusiform face area[@b6]. Fusiform responses have been found for a variety of face stimuli[@b59]. However, within computer generated characters, stronger fusiform responses were found when these looked and acted naturally and meaningfully[@b60]. The present U-shaped modulation could result from an interaction of perceived realism activating fusiform-dependent holistic processing and neoteny features activating feature-based processing in more occipital face areas. Emotion effects on the N170 and EPN are in line with previous work[@b32][@b33][@b35]. They indicate that emotional expressions modulate the N170 and EPN responses across stylization levels, while, also in line with the literature, in real faces, angry expressions had the largest impact on ERPs[@b32]. It is remarkable that stylization and emotional expressions interact on the N170, indicating an early interplay of structural analysis and emotional classification, rather than dual processing routes for identity and expression. The present data suggest that with increasing realism more resources are captured by cues signalling threat[@b61]. Accordingly, more realistic angry faces were rated more intense compared to happy faces see [Figs 6](#f6){ref-type="fig"} and [7](#f7){ref-type="fig"} and[@b20]. On the other hand, for very stylized faces, a relatively stronger processing of happy expressions was observed. Neotenic features may selectively enhance the processing of happy expressions. With the exception of the forehead, all neotenic features decrease or remain nearly unchanged with increasing realism ([Fig. 6b](#f6){ref-type="fig"}). As big eyes and a small nose contribute highly to a cute perception of the character, possibly amplifying processing of positive expressions. Unlike the N170 amplitude, LPP amplitude increase parallels perceived face realism across the tested stylization levels. This might be due to the uniqueness of a real face, prompting a multitude of ad hoc social inferences[@b62]. Bruce and Young's influential model suggests that after initial perceptual and structural analyses, the observed stimuli are compared with face representations stored in memory and if there is a match, person-specific knowledge is retrieved[@b63]. Indeed, person-related semantic information enhances LPP amplitudes[@b56]. Recently, it has been further shown that biographical information can increase the LPP to computer-generated faces[@b58]. Although we did not provide participants with explicit biographical information, as a result of social inferences, the more realistic faces might be perceived as having a unique biographical identity. Behavioural evidence showed that computer generated faces are harder to remember, possibly because they are not encoded as a unique person[@b52][@b53]. The noticeable discontinuity between levels 1--4 and 5--6 could also imply a categorical change between realistic and non-realistic characters as shown by classification tasks at a similar stylization level[@b16][@b17]. Typically, distinctiveness is achieved by exaggerating certain spatial differences between an individual and an average face[@b64]. Distinctiveness by shape or reflectance manipulations has been found to result in a larger EPN and LPP as well as a steeper learning curve and better memory trace for initially non-distinctive faces[@b54][@b55][@b65]. However, naturally distinctive faces lead to the largest LPP and are remembered even better[@b55]. Regarding the creation of the currently used face stimuli, spatial differences were not overexaggerated in comparison to an average face. However, rated face-realism and distinctiveness might not be uncorrelated, as more realistic faces for example exhibit a more detailed texture. Other studies have related the enhanced LPP for real compared to doll faces to the unique identity of the real face, generating an impression of personal social presence[@b51]. Future studies should aim to disentangle effects of face-distinctiveness from face-realism. Other factors can also modulate the LPP, but are unlikely to play a role in the current experiment: For instance, LPP responses increase with higher perceived familiarity[@b66]. However, Zell *et al*.[@b20] report equal familiarity across the present stylization levels. Similarly, facial attractiveness enhances the LPP[@b30][@b50]. For our stimuli, appeal, which is conceptually similar to attractiveness, was rated highest for medium-stylized faces (see [Fig. 6](#f6){ref-type="fig"} and Zell *et al*.[@b20]). Therefore, it is unlikely that attractiveness is responsible for LPP modulations in our experiment. In source space, the linear modulation of the LPP was reflected in larger and broader activations in occipito-parietal areas. The localization of this increase is in line with both enhanced perceptual processing of more realistic faces and, in particular, also memory-related processes. Although higher LPP amplitudes have been reported for emotional than for neutral stimuli e.g. refs [@b36], [@b41] and [@b44], we found no differences between emotional and neutral expressions on this component. In general, during passive viewing, emotion effects are smaller for faces compared to complex scenes and participants typically report less subjective arousal for faces[@b49]. Nevertheless, large emotion effects were present for the N170 and EPN time window. Similarly, Thom and colleagues[@b49] found emotion effects for the N170, while for the EPN emotion effects were only descriptively visible and no differences were found for the LPP. This suggests that without an explicit task a highly automatic response towards emotional facial expressions modulates early components without affecting late stages of processing. In this vein, it has been shown that at late stages emotion effects benefit more from attention to the emotional category than do early responses[@b39][@b40]. Overall, we demonstrated that stylized characters elicit neural effects that are different from the ones elicited by real faces. Importantly, the pattern changes qualitatively across different processing stages, although the measured facial features changed continuously across similar stylization levels. For face perception experiments, which use computer generated stimuli, this means that, unless a high level of realism is achieved, results cannot be transferred directly to real humans. Thus, computer generated stimuli may be suitable to test initial hypotheses, but require final validation with real photographs. So far, it is unclear why or when exactly realism is beneficial in practical applications like games or perceptual studies, but the present study, in demonstrating that realism affects different processing stages in a distinct manner, may offer some clues: Both highly stylized faces, with their neotenic features and very realistic faces influence early stage processing and are equally efficient in transient attention capture. On the other hand, only more realistic faces induce the kind of post-processing necessary to build an individual identity representation and likely facilitating identification with the character. Beyond gaming, these findings have implications for the design of virtual reality therapy settings, for instance of social phobia. They underscore that depending on the overall goal optimal character design will differ. If so, the uncanny valley phenomenon may also arise from a perceived mismatch between situational expectations resulting from a given virtual scenario and character appearance. In character design, the main problem for artists is that, for adult characters, neotenic and realistic features often exclude each other. For instance, skin smoothness is a neotenic feature, but detailed pores and skin-impurities are required to achieve full realism for a virtual character. Similarly, big eyes are considered as cute, but realistic characters that have unnaturally big eyes are perceived as creepy[@b21], inverting the intended effect. This dichotomy between realistic characters on one side and rather cute characters on the other, could also explain the plausibility of the "uncanny valley" concept and the present data indeed reveal a neural dissociation that might support it. In practice, considering two independent scales--one for realism and one for neotenic features--seems to be a promising future direction to predict whether a stylized character will be perceived positively or negatively when used in game or as an interactive agent. Moreover, these parameters can be controlled more easily than appeal or attractiveness, which depend on many aspects and are more subjective. As a limitation of the present work, it has to be noted that only two different identities were used. Therefore, our results might not generalize across all conceivable characters. In order to increase validity of the obtained results, our characters have been stylized based on popular 3D characters. Furthermore, the stimuli have been designed to match the quality of current animation movies as much as possible. To our best knowledge, our results are unique in that they provide many different stylization levels for the same characters. Conclusion ========== We measured EEG responses elicited by carefully manipulated faces with six different degrees of stylization and three emotional expressions. We tested a stimulus set for which a linear modulation of realism was achieved for the same identity. Our results indicate that face realism has a strong, but qualitatively different, influence on the N170 as well as the LPP component. While perceived realism influenced the N170 component in a u-shaped manner and interacted with emotional expression, the LPP component was only influenced by perceived realism, increasing continuously with face realism. For the N170, main generators differed between highly stylized and very realistic faces, suggesting to distinct contributing processes. The increased LPP was based on enhanced activity in broad occipito-parietal areas, in line with enhanced perceptual processing and memory encoding of more realistic faces. Methods ======= Participants ------------ Thirty-three participants were recruited at Bielefeld University. They gave written informed consent and received 11 Euros or course credit for participation. The study was approved by the Bielefeld University ethics committee (EUB number 2016-112). All methods were performed in accordance with the guidelines and regulation at Bielefeld University. The participants (22 females) were 23.30 years on average (*SD* = 3.68), all of them right-handed and had normal or corrected-to normal vision. Upon structured interview, no participant reported a current neurological or psychiatric disorder or relevant medication intake. Stimuli ------- Two face characters (one male, one female) varying across six stylization levels, with three emotional expressions (happy, angry, neutral) per stylization level, were used as stimuli. Creating continuous stylization of the same character is still an unsolved problem. While certain stylization effects can be achieved by image editing, using non-photorealistic rendering algorithms or character generators, all these approaches either do not modify the shape at all or address only a single stylization level. So far no automatic approach exists that accomplishes character stylization comparable to trained artists. Professional stylization is in general very time consuming and expensive. To circumvent this problem, most previous work used unrelated characters at different stylization levels. In contrast, Zell *et al*.[@b20] published a set of stimuli that contains six stylization levels of the same person matching the state of the art in computer graphics. For completeness, we shortly report the stimuli creation process. Level 6 are real photographs. Models for level 5 have been created from high resolution 3D scans of the photographed people. These 3D scans have been post-processed to remove visible artefacts or to add hair. The remaining stylization levels have been created by professional artists targeting popular looks of animation movies. All emotions have been created so that specific features remain consistent (e.g., teeth shown consistently for happy and angry expressions). Levels 2, 3 and 5 were created first and evaluated in a study according to perceived realism and familiarity. All characters had very similar familiarity ratings as none of the characters is known to the public. In addition, for the initial characters, discontinuities were detected for rated face-realism. In order to achieve a stepwise increase in rated realism and sample the stylization scale more uniformly, level 1 and 4 have been created and added afterwards. The whole stimulus set was finally rated according perceived realism, appeal and expression intensity ([Fig. 6](#f6){ref-type="fig"}). For the real photographs, only the neutral expression was rated in perceived realism and appeal. In addition, facial neotenic cues have been computed by measuring the relative size of facial parts[@b67]. We refer the reader to Zell *et al*.[@b20] for a full description of the stimuli creation process, including all technical details and initial evaluation. For the current study, camera view and aspect ratio have been adjusted, such that all faces are nearly of the same size and the eyes are located at similar positions. Background planes with a 50% grey were inserted in the 3D scenes before rendering. For stimuli examples see the depicted faces in [Figs 6](#f6){ref-type="fig"} and [7](#f7){ref-type="fig"}. Procedure --------- The faces were randomly presented for 600 ms, followed by a fixation cross with variable latencies (randomly between 400 and 500 ms). All faces were repeated fifteen times, for a total number of 540 presented faces, while it was prevented that the same stimulus was presented twice in a row. Faces were presented on a 15.4-inch screen (Dell Latitude D830) with a 1600 × 1200 pixel resolution (image width: 800; height: 1142). The background colour was 50% grey (RGB color values: 128; 128; 128). The stimulus presentation lasted for about 10 minutes, while the whole session took approximately 50 minutes. Participants had no task but were instructed to attend to the presented faces, while moving as little as possible. Participants were encouraged to reduce their eye-movements by focusing on the fixation cross. EEG recording and analyses -------------------------- EEG was recorded from 128 BioSemi active electrodes ([www.biosemi.com](http://www.biosemi.com)). Recorded sampling rate was 2048 Hz. During recording Cz was used as reference electrode. Biosemi uses two separate electrodes as ground electrodes: A Common Mode Sense active electrode (CMS) and a Driven Right Leg passive electrode (DLR). The two electrodes form a feedback-loop which enables to measure the average potential close to the reference in the AD-box (see <http://www.biosemi.com/faq/cms&drl.htm>, where also information about extra functions of the CMS/DRL loop can be retrieved). Four additional electrodes (EOG) measured horizontal and vertical eye-movement. These were placed at the outer canthi of the eyes and below the eyes. Pre-processing and statistical analyses were done using BESA^®^ ([www.besa.de](http://www.besa.de)), EMEGS[@b68] and SPM8 for EEG data (<http://www.fil.ion.ucl.ac.uk/spm/>). Offline, data were re-referenced to the average reference and then filtered with a forward 0.16 Hz high-pass and a zero-phase 30 Hz low-pass filter. Filtered data were segmented from 100 ms before stimulus onset until 600 ms after stimulus presentation. The 100 ms before stimulus onset were used for baseline correction. Eye-movements were corrected using the automatic eye-artefact correction method implemented in BESA[@b69]. Additionally, trials exceeding a threshold of 120 μV were rejected. Overall, 4.04 percent of all electrode measurements were interpolated. Per participant, on average 5.41 percent of all trials were rejected, leaving 28.38 trials per cell, leading to 85 trials per realism condition and 170 trials per emotion condition. Cortical source reconstructions of significant ERP differences were generated and statistically assessed with SPM8 for EEG[@b70], following recommended procedures. First, a realistic boundary element head model (BEM) was derived from SPM's template head model based on the Montreal Neurological Institute (MNI) brain. Electrode positions were then transformed to match the template head, which is thought to generate reasonable results even when individual subjects' heads differ from the template[@b71]. Average electrode positions as provided by BioSemi were co-registered with the cortical mesh template for source reconstruction. This cortical mesh was used to calculate the forward solution. For the inverse solution, the group inversion algorithm was used[@b70] and the solution was calculated from 100 ms pre-baseline to 600 ms after stimulus onset. Statistical analyses -------------------- EEG scalp-data were statistically analyzed with EMEGS. Six (stylization: level 1, level 2, level 3, level 4, level 5, level 6) by three (emotional display: angry, neutral, happy) repeated measure ANOVAs were set-up to investigate main effects of the communicative sender, emotion and their interaction in time windows and electrode clusters of interest. Partial eta-squared (η~P~^2^) was estimated to describe effect sizes, where η~P~^2^ = 0.02 describes a small, η~P~^2^ = 0.13 a medium and η~P~^2^ = 0.26 a large effect[@b72]. When Mauchly's Test detected a violation of sphericity, degrees of freedom were corrected according to Greenhouse-Geisser. For readability, the original degrees of freedom but corrected p-values and effect sizes are reported. For significant main effects linear compared to U-shaped contrasts were calculated. Time windows were segmented from 150 to 190 ms to investigate the N170, from 250 to 400 ms to investigate the EPN, and from 400 to 600 ms to the LPP component, after collapsing all conditions and visual inspection of the ERP components. For the N170 and EPN time windows, two large symmetrical temporo-occipital clusters of thirteen electrodes each were examined (left: I1, OI1, O1, POO3, PO9, PO9h, PO7, PO7h, P9, P9h, P7, TP9h, TP7; right: I2, OI2, PO10, POO4, PO10, PO10h, PO8, PO8h, P10, P10h, P8, TP10h, TP8). For the LPP time windows a large parietal cluster was investigated (twenty-six electrodes: CCPz, CP5, CP5h, CP3, CP1, CPz, CP2, CP4, CP6, CPP5h, CPP3, CPPz, CPP4, P3, P1, Pz, P2, P4, PPO3, PPO1, PPOz, PPO2, PPO4, PO3, POz, PO4). Results did not change qualitatively, when selecting different literature-based electrode clusters for the N170, EPN or LPP[@b34][@b36][@b54]. Source reconstructions were performed for the main effects of face-stylization. For each analyzed time window in scalp space, three-dimensional source reconstructions were generated as NIFTI images (voxel size = 2 mm · 2 mm · 2 mm). These images were smoothed with a Gaussian kernel using an 8 mm full-width half-maximum. The statistical comparisons used in source space were based on significant differences on the scalp. In line with previous studies[@b39][@b42][@b73], we describe statistical differences in source activity of voxels differing at least at an uncorrected threshold of *p* \< 0.005 and a minimum of twenty-five significant voxels per cluster. Additionally, in all tables results are shown applying a family-wise error corrected threshold of *p* \< 0.05. The identification of activated brain regions was performed using the LONI atlas[@b74]. Additional Information ====================== **How to cite this article:** Schindler, S. *et al*. Differential effects of face-realism and emotion on event-related brain potentials and their implications for the uncanny valley theory. *Sci. Rep.* **7**, 45003; doi: 10.1038/srep45003 (2017). **Publisher\'s note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. We thank Liane Stritz, Dominik Neu and Maria Vogt for help with data acquisition and all participants contributing to this study. All authors were supported by the Cluster of Excellence Cognitive Interaction Technology "CITEC" (EXC 277) at Bielefeld University, funded by the German Research Foundation (DFG). We acknowledge support of the publication fee by Deutsche Forschungsgemeinschaft and the Open Access Publication Funds of Bielefeld University. The authors declare no competing financial interests. **Author Contributions** E.Z. and S.S. contributed to the study design. S.S. measured participants and analysed the data under supervision of J.K. E.Z. and S.S. drafted and revised the manuscript under supervision of M.B. and J.K. All authors read and approved the final version. ![N170 ERP results.\ The upper panel displays the N170 at electrode PO8. From this panel, the interaction between emotion and realism can be observed. The lower panel shows the mean N170 over the occipital sensor cluster. Error bars are +/− one standard error of the mean. Note that, while negative-going, the N170 peak is still in the positive range (see top panel). Therefore, smaller bars represent higher N170 amplitudes.](srep45003-f1){#f1} ![Emotion main effects for the N170 and the EPN.\ Left: Displayed are the difference topographies for the main effects of emotion across all stylization levels. Blue colours indicate a relatively larger negativity and red colours a larger positivity. Right: Displayed is the time course at electrode PO8.](srep45003-f2){#f2} ![LPP ERP results.\ The upper panel shows the time course for electrode CPPz. Here an increase of LPP amplitudes can be seen for increasing realism. The lower panel displays the mean LPP over the parietal sensor cluster. Error bars are +/− one standard error of the mean.](srep45003-f3){#f3} ![Source estimations for main effects of stylization level for the N170.\ Upper panel: Displayed are the family-wise error corrected (*p* \< 0.05) changes in source activity for each stylization level. For all faces, significant generators can be observed for the N170 in bilateral superior, middle and inferior occipital gyri, as well as in the bilateral fusiform gyri. Lower panel: Displayed are the differences in source activity between stylization levels (post-hoc contrasts, uncorrected *p* \< 0.005). In the N170 the real faces lead to larger middle and inferior occipital activations, while the most stylized faces are processed more intensely in the right inferior occipital gyrus/cuneus/lingual gyrus.](srep45003-f4){#f4} ![Source estimations for main effects of stylization for the LPP time window.\ Upper panel: Displayed are the family-wise error corrected (*p* \< 0.05) changes in source activity for each realism level. For all faces, significant generators can be observed for the LPP in bilateral superior, middle and inferior occipital gyri, as well as in the bilateral fusiform gyri and bilateral superior parietal areas. Lower panel: Displayed are the differences between realism levels (post-hoc contrasts, uncorrected *p* \< 0.005). In the LPP, real faces are processed more intensely in bilateral middle and superior occipital and superior parietal areas. However, with increased realism, these differences become smaller and finally disappear in middle occipital regions.](srep45003-f5){#f5} ![(**a**) Realism, appeal and intensity ratings from Zell *et al*.[@b20], which tested the same characters but from slightly different point of view. Icons along the x-axis represent the stylization level. (**b**) Length and width of different facial parts, proportionally to the length and width of the face--averaged across both characters. For more stylized characters, the bigger eyes and smaller nose follow neotenic characteristics. However, this is not the case for vertical eye placement or forehead height. Error Bars denote 95% confidence levels.](srep45003-f6){#f6} ![Example for emotional and neutral expressions for all six realism levels.\ The stimulus set consists of a male and female character. Here, the male character is shown for illustration purpose. Please refer to [Fig. 6](#f6){ref-type="fig"} for illustration of the female character.](srep45003-f7){#f7} ###### Source estimations for stylization level main effects for the N170 (150--190 ms). Cluster-level peak-level MNI coordinates LONI ---------------------- ------------ ----------------- ------ ----- ----- ------------------- *Level 6 \> Level 1* 408 3.35 \<0.001 28 −90 2 Mid Occipital G R 86 2.70 \<0.005 −34 −90 8 Mid Occipital G L *Level 1 \> Level 6* 169 3.31 \<0.001 18 −98 −14 Inf Occipital G R Results show differences only between very stylized faces (Level 1) and real faces (Level 6). Real faces elicited more activity in middle occipital regions, while cartoon faces elicited more activity in the right inferior frontal gyrus/cuneus/lingual gyrus. Notes. No. of sig. voxel = the number of voxel which differ significantly between both conditions. Peak *p*-unc = uncorrected *p* value. For each significant peak, respective coordinates (x, y and z) are displayed in MNI space. If a cluster exhibited more than one peak, only the largest peak is reported. Area = peak-level brain region as identified by the LONI atlas. R/L = laterality right or left. G = Gyrus; Mid = middle. ###### Source estimations for stylization level main effects for the LPP (400--600 ms). Cluster-level Peak-level MNI coordinates LONI ---------------------- ------------ ----------------- ------ ----- ---- ------------------- *Level 6 \> Level 1* 1331 (707^a^) 7.47 \<0.001 −36 −86 18 Mid Occipital G L 700 (483^a^) 7.09 \<0.001 28 −90 2 Mid Occipital G R 673 (209^a^) 4.74 \<0.001 20 −82 30 Sup Occipital G R *Level 6 \> Level 2* 1192 (639^a^) 6.51 \<0.001 -38 −84 20 Mid Occipital G L 617 (356^a^) 5.51 \<0.001 34 −88 16 Mid Occipital G R 685 (301^a^) 4.94 \<0.001 20 −82 30 Sup Occipital G R *Level 6 \> Level 3* 1215 (610^a^) 6.06 \<0.001 −36 −86 20 Mid Occipital G L 684 (345^a^) 5.13 \<0.001 20 −82 30 Sup Occipital G R 749 (53^a^) 4.98 \<0.001 34 −92 2 Mid Occipital G R *Level 6 \> Level 4* 1167 (348^a^) 5.63 \<0.001 −34 −88 20 Mid Occipital G L 705 (369^a^) 5.24 \<0.001 20 −82 30 Sup Occipital G R 565 4.10 \<0.001 36 −86 0 Mid Occipital G R *Level 6 \> Level 5* 243 3.18 \<0.001 −38 −82 22 Mid Occipital G L 329 2.91 \<0.005 20 −82 30 Sup Occipital G R Results show enhanced activity for real faces compared to stylized faces. Real faces (Level 6) elicited more activity in middle and superior occipital regions. Differences become smaller with increasing realism of the stylized faces. Notes. ^a^Resulting cluster size with FWE-corrected threshold of *p* \< 0.05 (≥25 significant voxels). No. of sig. voxel = the number of voxel which differ significantly between both conditions. Peak *p*-unc = uncorrected *p* Value. For each significant peak, respective coordinates (x, y and z) are displayed in MNI space. If a cluster exhibited more than one peak, only the largest peak is reported. Area = peak-level brain region as identified by the LONI atlas. R/L = laterality right or left. G = Gyrus; Inf = inferior, Mid = middle, Sup = superior. [^1]: These authors contributed equally to this work.

{ "pile_set_name": "PubMed Central" }

Introduction {#s1} ============ Genome size, defined as the haploid DNA content of a cell, has increased as eukaryotes evolved from single-cell species to more complex, multicellular organisms. Within the same evolutionary timeframe, nuclear volume has also increased but at a slower rate than genome size expansion ([@bib45]; [@bib49]). While the ratio of nuclear to cell size has remained essentially constant in eukaryotes ([@bib8]), the disproportional increase in genome size relative to the nuclear volume has required organisms with larger genomes to compact their chromatin to greater extents than organisms with small sized genomes. Indeed there is a positive correlation between genome size and native chromatin compaction as measured by dye incorporation into chromatin ([@bib66]). In most eukaryotes, the genome is organized into chromatin by the repeating nucleosomal structure ([@bib42]). The nucleosomes stack and fold into higher order structures, serving to systematically compact the genome ([@bib38]; [@bib17]) and to regulate molecular processes that are based on DNA ([@bib9]; [@bib67]; [@bib24]; [@bib35]; [@bib26]). The surface of the histone octamer has 14 DNA interaction sites. Each interaction is mediated by an arginine residue that intercalates into the minor groove of the DNA to stabilize the nucleosomal structure ([@bib42]; [@bib71]). Arginine is the most commonly used amino acid for interaction with DNA due to its positive charge and the lower energetic cost compared to lysine for displacing water when intercalating into the minor groove ([@bib54]). Nucleosomes mediate chromatin compaction through a variety of mechanisms. For instance, nucleosomes form higher order structures through inter-nucleosomal contacts between the histone H4 N-terminal domain (NTD) and the acidic patch of H2A and between two H2B C-terminal domains (CTD) ([@bib42]; [@bib15], [@bib16]; [@bib27]; [@bib56]). Histone variants, such as H2A.Z or H2A.Bbd, as well as post-translational modifications of histones, such as H4K16ac, can further regulate the degree of compaction ([@bib63]; [@bib1]; [@bib61]; [@bib73]; [@bib10]; [@bib34]; [@bib23]). Polycomb complexes compact large domains of chromatin ([@bib21]) and are important for proper development. Histones of the H1 family promote additional compaction by binding between nucleosomes to linker DNA near the DNA entry/exit site on nucleosomes and stabilize the intrinsic ability of nucleosomal arrays to fold in vitro ([@bib6]; [@bib53]; [@bib64]). Linker histones may affect chromatin compaction globally ([@bib22]), at specific stages of the cell cycle such as mitosis ([@bib44]) or at specific regions of the genome ([@bib37]). In contrast to canonical histones, the linker histones are much less conserved ([@bib7]), and ectopic expression of human linker histones in the budding yeast even at low levels is lethal for the cell ([@bib46]). Finally, structural proteins such as condensin also contribute to chromatin condensation ([@bib65]). Many of these modulatory mechanisms are dynamic in nature ([@bib41]) and may help explain why multicellular organisms can compact chromatin to different degrees in different cell types. However, despite the existence of these mechanisms for genome compaction in higher eukaryotes, it has not been known whether the canonical histones themselves have evolved sequence features that also contribute to the generally increased chromatin compaction observed in organisms with larger genomes. In this study, we provide evidence from analysis of 160 fully-sequenced eukaryotic genomes that arginine (R) residues at specific positions in the N-terminal tail of histone H2A---which protrudes from the nucleosome on the opposite side of DNA entry/exit site---have co-evolved with increasing genome size with a concomitant decrease in serines/threonines. Although increases in genome size are associated with phylogenetic evolution from protozoa to fungi to more complex plants and animals, we present genetic and molecular evidence from the budding yeast and human cells as well as in vitro biochemical data to demonstrate that the evolutionary changes in H2A directly regulate chromatin compaction in vivo and in vitro with consequences for the nuclear volume. The evolutionary changes in H2A regulate chromatin compaction in yeast and human cells, revealing a surprising flexibility in the dynamics of the chromatin fiber that has been conserved across distant eukaryotes. This previously unrecognized structural feature of the nucleosome has evolved to enable greater chromatin compaction when genome size is disproportionately larger than the nuclear volume. Our findings also suggest that the reported mutations in the histone H2A NTD may contribute to the altered chromatin compaction that is commonly observed in cancer cells ([@bib74]). Results {#s2} ======= H2A acquires specifically positioned arginines as genome size increases {#s2-1} ----------------------------------------------------------------------- To determine whether specific residues in the four core histones have co-evolved with increasing genome size, we performed residue composition analysis of canonical histone protein sequences from 160 fully sequenced eukaryotes with genome sizes ranging from 8 to 5600 Mbp encompassing protozoa, fungi, plants, and animals. The canonical histone proteins for each organism were defined based on at least 90% overlap and 35% identity with the histone fold domain of the corresponding human sequence ('Materials and methods'). Each organism was categorized as having a small (\<100 Mbp), medium (100--1000 Mbp), or large (\>1000 Mbp) genome ([Figure 1---figure supplement 1A](#fig1s1){ref-type="fig"}). Of the canonical histones, the H2A NTD showed the most statistically significant variability in amino acid residues, where the number of arginines increased with increasing genome size ([Figure 1A](#fig1){ref-type="fig"}), while the number of serines (S) and threonines (T) decreased ([Figure 1B](#fig1){ref-type="fig"}). Other amino acid residues in the H2A NTD, including lysines (K), did not correlate with genome size ([Figure 1---figure supplement 1B](#fig1s1){ref-type="fig"}).10.7554/eLife.02792.003Figure 1.Histone H2A N-terminal sequence has co-evolved with genome size.Violin plots of the number of (**A**) arginines or (**B**) serines/threonines in the H2A NTD for species with small, medium, and large genomes. Plot widths correspond to species frequency within each group. (**C**) H2A NTD sequences for *S. cerevisiae* and *H. sapiens*. (**D**) Heat map of H2A NTD residue composition at the indicated positions ordered by genome size. Example species are shown with kingdom and genome size information. (**E**) Protein sequence motifs surrounding the four H2A NTD arginine residues. (**F**) Positioning of evolutionarily variable residues relative to the H2A N-terminus (left) or histone fold (right). See also [Figure 1---figure supplements 1 and 2](#fig1s1 fig1s2){ref-type="fig"}.**DOI:** [http://dx.doi.org/10.7554/eLife.02792.003](10.7554/eLife.02792.003)10.7554/eLife.02792.004Figure 1---source data 1.H2A multiple sequence alignments, heat map data, and canonical H2A isoforms.**DOI:** [http://dx.doi.org/10.7554/eLife.02792.004](10.7554/eLife.02792.004)10.7554/eLife.02792.005Figure 1---figure supplement 1.Phylogenetic distribution of species analyzed in this paper.(**A**) The bar graph indicates the number and proportion of organisms in our data set that belong to the indicated phylogenetic kingdoms for each genome size category. (**B**) Violin plots of the number of lysines in the H2A NTD grouped by genome size as in [Figure 1](#fig1){ref-type="fig"}. (**C**) Boxplot of genome sizes of the organisms which have an H2A without (left) or with (right) the indicated residue. p-values (Mann--Whitney U test) of the difference in means between the absence or presence of the indicated residue are indicated.**DOI:** [http://dx.doi.org/10.7554/eLife.02792.005](10.7554/eLife.02792.005)10.7554/eLife.02792.006Figure 1---figure supplement 2.H2A arginines 3 and 11 are situated adjacent to DNA within the nucleosome.Structure of a di-nucleosome obtained from the crystal lattice of a mono-nucleosome structure is shown from (**A**) the side or (**B**) close up highlighting potential intra- and inter-nucleosomal interactions between arginines and DNA backbones. Green is H2A, yellow is H2B, cyan is H3, and salmon is H4. The red and blue spheres are R3 and R11, respectively, both of which are in chain C ([@bib12]).**DOI:** [http://dx.doi.org/10.7554/eLife.02792.006](10.7554/eLife.02792.006) The acquisition of arginines and loss of serines/threonines in the H2A NTD with increasing genome size occur at specific positions and in sequential order. For instance, the human H2A NTD contains arginine residues at positions 3 and 11 that are absent in yeast and at position 20 which is correspondingly a lysine in yeast ([Figure 1C](#fig1){ref-type="fig"}). In contrast, the human sequence lacks S10 and S15 that are observed in the yeast H2A NTD ([Figure 1C](#fig1){ref-type="fig"}). Alignment of all H2A NTD sequences also revealed similar trends across all eukaryotes studied here. The heat map in [Figure 1D](#fig1){ref-type="fig"} shows the occurrence of arginines and serines/threonines in the H2A NTD as a function of genome size (see [Figure 1---source data 1](#SD1-data){ref-type="supplementary-material"} for raw data and [Figure 1---figure supplement 1C](#fig1s1){ref-type="fig"} for statistical analysis). At position 3, an arginine (R3) is predominantly present in medium and large species but is lacking in small species. At position 11, a lysine (K11) is observed in species with medium genomes that evolves to an arginine (R11) mainly in organisms with large genomes. R17 is present in most organisms examined, suggesting a very conserved function for this residue ([@bib72]). At position 20, small genomes contain predominantly a lysine residue, which converts to arginine in medium and large genomes. In contrast, serines/threonines at positions 10 and 15 are found primarily in organisms with small genomes and much less so in organisms with medium and large genomes ([Figure 1D](#fig1){ref-type="fig"}). Additionally, each of the four H2A NTD arginines is surrounded by a conserved motif ([Figure 1E](#fig1){ref-type="fig"}). The residues surrounding R3 and R17 are mainly glycine and serine, respectively. At position 11, the motif varies based on genome size. Species with medium-sized genomes contain VKG and those with large genomes contain ARA. The same is true of position 20, where AKA is present in organisms with small genomes and (S/T)RA in larger genome species ([Figure 1E](#fig1){ref-type="fig"}). Interestingly, except for R3, the positions of all the other evolutionarily varying residues in the H2A NTD are strongly conserved relative to the histone fold domain and not the N-terminus ([Figure 1F](#fig1){ref-type="fig"}). When counting conventionally from the N-terminus, amino acids R11, R17, and R20---which are numbered based on the human sequence---were not observed consistently at the same positions in other organisms. However, these residues are respectively 12, 6, and 3 amino acids away from the histone fold in most species (note the vertical axes in [Figure 1F](#fig1){ref-type="fig"}). S10 and S15---which are numbered based on the yeast sequence---also show more uniform positioning when counted from the histone fold. Altogether, as genome size increases, arginines appear in conserved positions within the H2A NTD relative to the histone fold, and serines and threonines are lost. Arginines 3 and 11 in the H2A NTD increase chromatin compaction {#s2-2} --------------------------------------------------------------- To determine whether arginines and serines/threonines of the H2A NTD affect chromatin compaction in vivo*,* we took advantage of a strain of *Saccharomyces cerevisiae*, that has both chromosomal copies of H2A deleted and carries a single copy of H2A on a plasmid (TSY107), to construct mutant strains containing single or multiple insertions of arginines into their conserved motifs, deletions of serines, or combinations thereof (see [Table 1](#tbl1){ref-type="table"} for specific amino acid changes and [Supplementary file 1A](#SD2-data){ref-type="supplementary-material"} for a description of the mutant strains). Two mutants, R3(ΔGS10)R11 and R11ΔS15, were also designed such that the spacing between R3 and R11 or R11 and the histone fold, respectively, is the same as in the H2A NTD of organisms with large genomes ([Figure 1C,F](#fig1){ref-type="fig"}). As a control for positive charge, mutant strains with lysines inserted in the same positions as arginines were also generated.10.7554/eLife.02792.007Table 1.List of H2A mutations, sequence changes and their effects on chromatin compaction and nuclear volume**DOI:** [http://dx.doi.org/10.7554/eLife.02792.007](10.7554/eLife.02792.007)FISHNuclear volumeH2A mutantH2A NTD Protein sequence% Changep-value% Changep-valueYeast WTSG--GKG--GKAGSA--AKASQSRSAKAG--1.0E+00--1.0E+00 R3SG**[R]{.ul}**GKG--GKAGSA--AKASQSRSAKAG**−18**9.5E−04−54.1E−01 R11SG--GKG--GKAGSA**[R]{.ul}**AKASQSRSAKAG**−15**8.6E−04**−20**5.9E−05 R3R11SG**[R]{.ul}**GKG--GKAGSA**[R]{.ul}**AKASQSRSAKAG**−22**8.2E−06**−16**3.0E−03 R3(ΔGS10)R11SG**[R]{.ul}**GKG--GKA··A**[R]{.ul}**AKASQSRSAKAG**−30**2.1E−06+63.7E−01 R11ΔS15SG--GKG--GKAGSA**[R]{.ul}**AKA·QSRSAKAG**−41**3.9E−08**−9**[\*](#tblfn1){ref-type="table-fn"}4.7E−04 K3SG**[K]{.ul}**GKG--GKAGSA--AKASQSRSAKAG+98.6E−01**+13**9.4E−03 K11SG--GKG--GKAGSA**[K]{.ul}**AKASQSRSAKAG+163.1E−01+32.6E−01 K3K11SG**[K]{.ul}**GKG--GKAGSA**[K]{.ul}**AKASQSRSAKAG+68.3E−01**+31**5.4E−08 K11ΔS15SG--GKG--GKAGSA**[K]{.ul}**AKA·QSRSAKAG−79.2E−02+26.6E−01 ΔGS10SG--GKG--GKA··A--AKASQSRSAKAG−63.2E−02+103.0E−02 ΔS15SG--GKG--GKAGSA--AKA·QSRSAKAG+39.4E−02**+9**9.7E−03 R6SG--GKG**[R]{.ul}**GKAGSA--AKASQSRSAKAG−55.6E−02**+10**1.0E−03 K20RSG--GKG--GKAGSA--AKASQSRSA**[R]{.ul}**AG−33.0E−01+71.5E−02 R17KSG--GKG--GKAGSA--AKASQS**[K]{.ul}**SAKAG−17.9E−0102.7E−01Human---HA Tag WTSGRGKQGGKTRAKAKSRSSRAG--1.0E+00--1.0E+00 ΔR3SG·GKQGGKTRAKAKSRSSRAG**+39**8.3E−03**+42**1.3E−08 R11KSGRGKQGGKT**[K]{.ul}**AKAKSRSSRAG**+20**2.3E−02**+14**1.2E−03 R11ASGRGKQGGKT**[A]{.ul}**AKAKSRSSRAG**+43**1.0E−05**+21**5.7E−07 ΔR3R11ASG·GKQGGKT**[A]{.ul}**AKAKSRSSRAG**+35**3.5E−03**+18**5.9E−04Human---FLAG Tag WTSGRGKQGGKARAKAKSRSSRAG--1.0E+00--1.0E+00 Δ1--12············KAKSRSSRAG**+47**4.9E−03**+18**2.8E−04[^2][^3][^4] To test the effects of H2A NTD changes on chromatin compaction, the physical distance between two probes on chromosome XVI spaced 275 kb apart was assessed in each of the H2A mutants using fluorescent in situ hybridization (FISH) ([Figure 2A](#fig2){ref-type="fig"}; [@bib30]; [@bib4]). The probes were differentially labeled and visualized by confocal microscopy. The distance between the probes was measured in a single plane in which both probes were present within each nucleus ([@bib4]). When compared to the isogenic wild type (WT), addition of a single arginine at position 3 (R3) or 11 (R11) to the H2A NTD was sufficient to significantly decrease the average interprobe distance by 18% and 15%, respectively ([Figure 2B,C](#fig2){ref-type="fig"}; [Table 1](#tbl1){ref-type="table"}). The average interprobe distance was further decreased by 22% when both arginines were present (R3R11) and even more so (30%) in R3(ΔGS10)R11. Deleting G9S10 (ΔGS10) alone caused slightly increased compaction with low statistical significance ([Table 1](#tbl1){ref-type="table"}). The largest decrease in interprobe distance (41%) was observed in the R11ΔS15 mutant, which places R11 12 amino acids from the histone fold, the same position as in organisms with large genomes. Removal of S15 (ΔS15) alone had no effect. The effect of arginines was not simply due to increasing the positive charge of the H2A NTD, as insertions of lysines at positions 3 and 11 did not significantly affect the interprobe distances ([Figure 2C](#fig2){ref-type="fig"} and [Supplementary file 2](#SD3-data){ref-type="supplementary-material"}). Although lysines are found at these positions in certain species ([Figure 1D](#fig1){ref-type="fig"}), the lack of potential compaction by lysines may be due to the absence of other evolutionary changes in yeast histones (see [Figure 1E](#fig1){ref-type="fig"}). Additionally, R17K or K20R mutations did not affect compaction, nor did a randomly inserted arginine at position 6 (R6) ([Figure 2C](#fig2){ref-type="fig"} and [Supplementary file 2](#SD3-data){ref-type="supplementary-material"}), suggesting that not every arginine in the H2A NTD contributes to chromatin compaction.10.7554/eLife.02792.008Figure 2.Ectopic expression of H2A NTD arginines causes compaction in yeast.(**A**) Schematic position of probes on chromosome XVI that were used for FISH. The letters correspond to the probe sets. (**B**) FISH images and (**C**) boxplot of the distributions of interprobe distances for probe set A in the indicated strains. (**D**) The mean interprobe distances for the indicated yeast strains for probe sets A, B, C, and D are plotted as a function of genomic distance. Solid lines are best fit equations. (**E**) Boxplot of the distributions of interprobe distances for probe set A in the indicated strains. Dashed lines mark the median value for the WT strain. The boxplot whiskers contain 90% of the data. All scale bars are 1 µm. Boxes are colored if the mean of the indicated strain is significantly different from WT. Red stars denote level of significance: \*p\<0.01; \*\*p\<0.001; \*\*\*p\<0.0001 (For exact values, see [Supplementary file 2](#SD3-data){ref-type="supplementary-material"}). (**F**) Agarose gel electrophoresis of MNase-digested chromatin in the indicated strains including the densitometric profiles comparing the WT to each of the mutant H2A strains for a given amount of enzyme. See also [Figure 2---figure supplement 1](#fig2s1){ref-type="fig"}.**DOI:** [http://dx.doi.org/10.7554/eLife.02792.008](10.7554/eLife.02792.008)10.7554/eLife.02792.009Figure 2---figure supplement 1.Ectopic expression of H2A NTD arginines causes compaction in yeast.(**A**--**C**). Boxplot of the distributions of interprobe distances in the indicated H2A mutant strains for probe sets B, C, and D as shown in [Figure 3A](#fig3){ref-type="fig"}. Boxes are colored if the mean of the indicated strain is significantly different from WT. Red stars denote level of significance: \*p\<0.01; \*\*\*p\<0.0001 ([Supplementary file 2](#SD3-data){ref-type="supplementary-material"}). Cell cycle analysis of yeast strains in the TSY107 (**D**) or FY406 (**E**) background. 1C and 2C refer to G1 and G2 DNA content, respectively. Note that the WT strains (TSY107, FY406) carry one copy of the H2A gene on a plasmid with the two chromosomal copies deleted ([@bib59]; [@bib33]). Dosage alterations of the H2A protein cause G2/M arrest ([@bib62]) which is evident in our WT strains. Nonetheless, all the mutants display highly similar cell cycle profiles. (**F**) Agarose gel electrophoresis of MNase-digested chromatin in the indicated strains. The amount of enzyme used to digest chromatin is indicated. Also shown are the densitometric profiles of the agarose gel that compares WT to each of the indicated mutant H2A strains for a given amount of enzyme.**DOI:** [http://dx.doi.org/10.7554/eLife.02792.009](10.7554/eLife.02792.009) We further confirmed the effects of R11 on chromatin compaction using three additional probe sets ([Figure 2A](#fig2){ref-type="fig"}). The level of compaction seen in our WT strain is similar to what has been previously reported in yeast using a different strain background ([@bib4]). The interprobe distances for all probe sets were significantly decreased in R11 compared to WT and even more so in R11ΔS15 but not ΔS15 alone ([Figure 2---figure supplement 1A--C](#fig2s1){ref-type="fig"} and [Supplementary file 2](#SD3-data){ref-type="supplementary-material"}). Plotting the physical vs genomic distances for all probe sets revealed uniform compaction across large genomic distances ([Figure 2D](#fig2){ref-type="fig"}). The effect of R11 on chromatin compaction was not strain-specific as H2A R11 and R11ΔS15, but not ΔS15, caused chromatin compaction in a different strain background ([Figure 2E](#fig2){ref-type="fig"}; [Supplementary file 2](#SD3-data){ref-type="supplementary-material"}). We therefore conclude that H2A arginines at positions 3 and 11, especially when R11 is placed at the evolutionarily-conserved position relative to the histone fold, increase the degree of chromatin compaction. Chromatin is differentially compacted at different cell cycle stages ([@bib30]). Cell cycle profile analysis showed little difference between the strain harboring WT H2A and any of the mutant strains ([Figure 2---figure supplement 1D--E](#fig2s1){ref-type="fig"}), indicating that the observed differences in chromatin compaction are not due to altered cell cycle profiles. Chromatin compaction may also be influenced by nucleosomal spacing; indeed the linker DNA length is larger in human cells than in yeast ([@bib28]). We find that there are essentially no differences in nucleosomal density in H2A arginine mutants using Micrococcal nuclease (MNase) digestion ([Figure 2F](#fig2){ref-type="fig"}, [Figure 2---figure supplement 1F](#fig2s1){ref-type="fig"}), indicating that the average nucleosomal spacing is not affected by these mutations. But the more compact mutants displayed decreased accessibility to MNase as indicated by the delayed appearance of the nucleosomal digestion pattern ([Figure 2F](#fig2){ref-type="fig"}, [Figure 2---figure supplement 1F](#fig2s1){ref-type="fig"}). H2A arginines and serines affect nuclear volume in yeast {#s2-3} -------------------------------------------------------- Since chromatin structure may influence the volume of the nucleus ([@bib8]), we asked whether nuclear volume was affected by H2A-mediated chromatin compaction. We tagged a nuclear pore protein, Nup49, in its chromosomal locus with GFP to visualize the nuclear membrane and used confocal microscopy to capture three-dimensional images of the nucleus to quantify volumes of ≥150 cells per H2A mutant ([Figure 3A,B](#fig3){ref-type="fig"}; [Table 1](#tbl1){ref-type="table"}; [Supplementary file 3](#SD4-data){ref-type="supplementary-material"}, see 'Materials and methods' for volume calculations). As compared to WT cells, H2A mutants containing R11 or R3R11, both of which contain more compact chromatin, displayed significantly decreased nuclear volumes. The average nuclear volume in the R3 mutant was also less than WT but did not reach statistical significance. Interestingly, H2A mutants from which serines 10 and 15 were removed displayed larger nuclear volumes. Simultaneous insertions of arginines into these strains (R3(ΔGS10)R11 and R11ΔS15) decreased their nuclear volume (R11ΔS15 p\<0.001 compared to ΔS15), restoring them to levels similar to WT. The control strains with either lysines or R6 had nuclear volumes similar to or larger than WT. Neither arginines nor serines had any effect on total cell size as measured by concanavalin A staining ([Figure 3---figure supplement 1A,B](#fig3s1){ref-type="fig"}; [Supplementary file 3](#SD4-data){ref-type="supplementary-material"}). In the FY406 strain background, ΔS15 did not cause an increase in nuclear volume; and thus both R11 and R11ΔS15 strains exhibited smaller nuclear volumes than isogenic WT ([Figure 3C](#fig3){ref-type="fig"}; [Supplementary file 3](#SD4-data){ref-type="supplementary-material"}). These data suggest that modulation of chromatin compaction through the H2A NTD, especially in the presence of R11, affects the nuclear volume but this effect may be indirect (see human data below).10.7554/eLife.02792.010Figure 3.Ectopic expression of H2A NTD arginines decreases nuclear volume in yeast.(**A**) Images of the nuclear envelope, as visualized by Nup49p-GFP, and boxplot of the distributions of nuclear volumes in the indicated strains in the TSY107 background (**B**) or the FY406 background (**C**). Dashed lines mark the median value for the WT strain. The boxplot whiskers contain 90% of the data. All scale bars are 1 µm. Boxes are colored if the mean of the indicated strain is significantly different from WT. Red stars denote level of significance: \*p\<0.01; \*\*p\<0.001; \*\*\*p\<0.0001 ([Supplementary file 3](#SD4-data){ref-type="supplementary-material"}). Red dagger (†) indicates that mean nuclear volume of R11ΔS15 is significantly smaller than its isogenic WT strain (ΔS15; p\<0.001). See also [Figure 3---figure supplement 1](#fig3s1){ref-type="fig"}.**DOI:** [http://dx.doi.org/10.7554/eLife.02792.010](10.7554/eLife.02792.010)10.7554/eLife.02792.011Figure 3---figure supplement 1.H2A arginines do not affect cell size.(**A**) Images of the cell wall, as visualized by concanavalin A staining, and (**B**) boxplot of the distributions of cellular volumes in the indicated yeast strains. The scale bar is 5 µm.**DOI:** [http://dx.doi.org/10.7554/eLife.02792.011](10.7554/eLife.02792.011) Loss of H2A arginines causes de-compaction of chromatin in human cells {#s2-4} ---------------------------------------------------------------------- Since the H2A NTD in large genomes contains both R3 and R11, we expected that their removal would cause de-compaction of chromatin. To test this prediction, we ectopically expressed WT or mutant H2A in several human cell lines and measured the distances between probes 0.49 Mbp apart on chromosome 1 by FISH, as well as the largest nuclear cross-sectional areas ('Materials and methods'). The H2A gene was HA-tagged and mutated to remove R3 (ΔR3), to replace R11 with alanine (R11A) or lysine (R11K), or to combine two mutations (ΔR3R11A). The H2A constructs were overexpressed using the strong CMV promoter in the normal human IMR90 fibroblasts, the breast cancer cell line MDA-MB-453, or the HEK293 cells. Cells overexpressing ΔR3, R11A, or ΔR3R11A H2A mutants had increased interprobe distances, indicating de-compaction of chromatin. Expression of H2A R11K had modest effects on chromatin de-compaction with marginal statistical significance ([Figure 4A,B](#fig4){ref-type="fig"}, [Figure 4---figure supplement 1A,B](#fig4s1){ref-type="fig"}; [Supplementary file 4](#SD5-data){ref-type="supplementary-material"}). Cells expressing any of the H2A mutants displayed larger nuclear areas, suggesting that nuclear size is increased ([Figure 4C,D](#fig4){ref-type="fig"}, [Figure 4---figure supplement 1C--F](#fig4s1){ref-type="fig"}; [Supplementary file 5](#SD6-data){ref-type="supplementary-material"}). Equal degrees of overexpression were confirmed by immunofluorescence analysis with an anti-HA antibody and detection of HA-H2A by western blotting ([Figure 4D](#fig4){ref-type="fig"}, [Figure 4---figure supplement 1G](#fig4s1){ref-type="fig"}). Ectopic expression of a C-terminally FLAG-tagged H2A mutant missing residues 1--12 (Δ1--12) also caused significant de-compaction of chromatin and increased nuclear area despite being expressed at a lower level than WT ([Figure 4E,F](#fig4){ref-type="fig"}). These data demonstrate that, consistent with our predictions, the H2A NTD, especially arginines 3 and 11, function to compact chromatin in human cells.10.7554/eLife.02792.012Figure 4.Loss of H2A NTD arginines decreases chromatin compaction in human cells.(**A**) FISH images of probes on chromosome 1 in normal primary IMR90 fibroblasts with HA-tagged WT or mutant H2A overexpressed as indicated. (**B**) Boxplot of the distributions of inter-probe distances. Note that R11K was only marginally significant at p=0.023. (**C**) Immunofluorescence images of IMR90 cells overexpressing HA-tagged WT or mutant H2A as indicated. (**D**) Top: boxplot of the distributions of largest nuclear cross-sectional areas in the indicated H2A overexpressing cells. Bottom: boxplot of the distributions of α-HA fluorescence intensities. (**E**) Left: FISH images, as in (**A**), of IMR90 cells expressing a C-terminal FLAG-tagged WT or tailless (Δ1--12) H2A. Right: boxplot of the distributions of inter-probe distances. (**F**) Top: immunofluorescence images of IMR90 cells overexpressing FLAG-tagged WT or tailless H2A. Bottom: boxplot of nuclear areas and fluorescence intensities, as indicated. Dashed lines mark the median value for the WT strain. All scale bars are 10 µm. Boxes are colored if the mean of the indicated strain is significantly different from WT. Red stars denote level of significance: \*p\<0.01; \*\*p\<0.001; \*\*\*p\<0.0001 ([Supplementary files 4 and 5](#SD5-data SD6-data){ref-type="supplementary-material"}). See also [Figure 4---figure supplement 1](#fig4s1){ref-type="fig"}.**DOI:** [http://dx.doi.org/10.7554/eLife.02792.012](10.7554/eLife.02792.012)10.7554/eLife.02792.013Figure 4---figure supplement 1.Loss of H2A NTD arginines decreases chromatin compaction in human cells.(**A**) FISH images of probes on chromosome 1 in MDA-MB-453 cells with either WT or mutant HA-tagged H2A overexpressed. (**B**) Boxplot of the distributions of interprobe distances. Immunofluorescence images of (**C**) MDA-MB-453 or (**E**) HEK293 cells overexpressing WT or mutant HA-tagged H2A. Boxplot of the distributions of the largest nuclear cross-sectional areas in (**D**) MDA-MB-453 or (**F**) HEK293 for the indicated H2A over-expressing cells. (**G**) Western blot of lysates from HEK293 cells overexpressing the indicated WT or mutant HA-tagged H2A. All scale bars are 10 µm. Dashed lines mark the median value for the WT strain. Boxes are colored if the mean of the indicated strain is significantly different from WT. Red stars denote level of significance: \*p\<0.01; \*\*\*p\<0.0001.**DOI:** [http://dx.doi.org/10.7554/eLife.02792.013](10.7554/eLife.02792.013) H2A R11 regulates compaction of nucleosomal arrays in vitro {#s2-5} ----------------------------------------------------------- Because R11 compacts chromatin in vivo, we investigated whether this effect is directly on the chromatin fiber. We used step-wise salt dialysis to assemble nucleosomal arrays with a DNA template containing 12 copies of the 177 bp '601' nucleosome positioning sequence (601-177-12) and recombinant *Xenopus laevis* histone octamers that contain either WT H2A or one with R11 deleted (ΔR11). We assembled nucleosomal arrays at different octamer-to-template ratios (0.9, 1, and 1.1 octamer to 1 template) and monitored the quality of the arrays by MgCl~2~ precipitation and restriction digest analysis using ScaI. We found that a 1:1 octamer-to-template ratio gave the best results as the ScaI digest demonstrated well-assembled arrays compared to the 5% free DNA loaded as a comparison ([Figure 5A](#fig5){ref-type="fig"}). We used analytical ultracentrifugation to determine the sedimentation velocity combined with van Holde--Weischet analysis ([@bib70]) to ascertain the distribution of sedimentation coefficients (S) for each nucleosomal array in the absence or presence of 0.8 mM MgCl~2~, a concentration of the divalent cation that promotes intra-molecular folding of nucleosomal arrays ([@bib60]). In the absence of Mg^2+^, arrays containing WT H2A sedimented with a coefficient of 33.1, which is a value that has been previously shown for similar arrays ([@bib15]; [@bib61]; [@bib73]). In contrast, arrays missing R11 adopted a more extended conformation with a smaller sedimentation coefficient of 31.0 ([Figure 5B](#fig5){ref-type="fig"}). Addition of Mg^2+^ increased compaction of both arrays and shifted the sedimentation coefficients to 39.3 and 37.4 for WT and ΔR11 H2A, respectively ([Figure 5B](#fig5){ref-type="fig"}). A second independent chromatin assembly and ultracentrifuge analysis confirmed these results ([Figure 5---figure supplement 1](#fig5s1){ref-type="fig"}). Thus, in the absence of R11 in the H2A NTD, nucleosomal arrays adopt a less compact conformation even in the presence of divalent cations, showing that R11 directly increases chromatin compaction.10.7554/eLife.02792.014Figure 5.H2A NTD R11 directly modulates the compaction of chromatin fibers in vitro.(**A**) Polyacrylamide gel electrophoresis (PAGE) of ScaI-digested 601-177-12 DNA template assembled with octamers containing recombinant WT or ΔR11 H2A. As a control, 5% of the 601-177-12 DNA without octamers was also digested. (**B**) The distribution of sedimentation coefficients determined by van Holde-Weischet analysis plotted against the percent boundary fraction in the absence or presence of 0.8 mM MgCl~2~ as indicated. S~20°C,W~ is the sedimentation coefficient corrected to water at 20°C. See also [Figure 5---figure supplement 1](#fig5s1){ref-type="fig"}.**DOI:** [http://dx.doi.org/10.7554/eLife.02792.014](10.7554/eLife.02792.014)10.7554/eLife.02792.015Figure 5---figure supplement 1.H2A NTD R11 directly modulates the compaction of chromatin fibers in vitro.The distribution of sedimentation coefficients determined by van Holde-Weischet analysis plotted against the percent boundary fraction in the absence or presence of 0.6 mM MgCl~2~ as indicated. S~20°C,W~ is the sedimentation coefficient corrected to water at 20°C.**DOI:** [http://dx.doi.org/10.7554/eLife.02792.015](10.7554/eLife.02792.015) Compaction of chromatin by H2A NTD arginines does not alter global gene expression in yeast {#s2-6} ------------------------------------------------------------------------------------------- To determine whether chromatin compaction through H2A arginines interferes with transcription regulation, we examined gene expression patterns in the H2A yeast mutants. Remarkably, there was a high level of correlation (≥0.99) between all strains examined ([Figure 6A](#fig6){ref-type="fig"}), and no specific gene ontology was found among the genes that were differentially expressed by twofold or more. The expression levels of the histone genes were similar, indicating that altered levels of histone genes expression do not account for the changes in chromatin compaction. These data indicate that compaction of chromatin by H2A does not significantly alter global gene expression in exponentially growing cells.10.7554/eLife.02792.016Figure 6.Mutations to H2A NTD decrease the fitness of yeast.(**A**) Pearson correlations between the global gene expressions of the indicated strains grown in YPD. Correlations are calculated from an average of at least two experiments. (**B**) Growth curves of the indicated H2A yeast strains over 10 hr in YPD. (**C**) Spot tests with 10-fold serial dilutions for the indicated strains in the presence of different drugs. (**D**) The proportion of yeast cells in a co-culture of WT and the indicated mutant H2A carrying Pgk1 gene fusion to GFP (green) or RFP (red) as indicated by color.**DOI:** [http://dx.doi.org/10.7554/eLife.02792.016](10.7554/eLife.02792.016) All strains also showed similar growth rates in rich media ([Figure 6B](#fig6){ref-type="fig"}) and no significant differences in sensitivity to hydroxyurea, methyl methanesulfonate (MMS), bleomycin, 4-nitroquinoline 1-oxide (4NQO), cycloheximide, and rapamycin, indicating no major defects with DNA replication or repair, protein synthesis, or the TOR signaling pathways ([Figure 6C](#fig6){ref-type="fig"}). But in competition growth assays in which equal amounts of WT and H2A mutant cells harboring the PGK1 gene fused to either GFP or RFP were co-cultured, the H2A mutants regardless of any effect on chromatin compaction, were outcompeted ([Figure 6D](#fig6){ref-type="fig"}). This suggests that changes in the H2A NTD sequence can affect the overall fitness of the cell. H2A NTD arginines and their surrounding residues are mutated in cancer and affect chromatin compaction {#s2-7} ------------------------------------------------------------------------------------------------------ Deregulated chromatin compaction is often a pathological hallmark of cancer cells ([@bib18]), although the underlying mechanisms are not well-understood. A survey of the COSMIC database ([@bib25]), as of the time of writing, revealed 41 documented missense mutations within the H2A NTD with 29 (71%) affecting a residue within one of the four arginine motifs ([Figure 7A](#fig7){ref-type="fig"}). R11, which had the strongest effect of any single arginine residue on chromatin compaction, is the most commonly mutated residue in the H2A NTD. We tested the effects of three of these mutations, R11C, H, and P and found that ectopic expression of each in normal human fibroblasts decreases chromatin compaction significantly with R11P having the strongest effect ([Figure 7B,C](#fig7){ref-type="fig"}). These cancer mutations have little effect on increasing nuclear area, however ([Figure 7D,E](#fig7){ref-type="fig"}), in contrast to R11A ([Figure 4D](#fig4){ref-type="fig"}). It is unclear to what extent the H2A mutants have to be expressed in cancer cells relative to the 17 canonical H2A genes in the human genome to affect chromatin compaction. But our data suggest that over-expression of an H2A mutant has the potential to disrupt chromatin compaction in cancer.10.7554/eLife.02792.017Figure 7.Mutations of H2A NTD found in cancers decreases chromatin compaction in human cells.(**A**) Schematic of the H2A NTD showing only the mutations within the arginine motifs found in various cancers as indicated by the colored shapes ([@bib25]). The letter within each shape represents the mutated amino acid. (**B**) FISH images of probes on chromosome 1 in normal primary IMR90 fibroblasts with HA-tagged WT or mutant H2A overexpressed as indicated. (**C**) Boxplot of the distributions of inter-probe distances. (**D**) Immunofluorescence images of IMR90 cells overexpressing HA-tagged WT or mutant H2A as indicated. Anti-HA primary and Alexa Fluor 647-conjugated secondary antibodies were used to determine expression in FISH images and for measurement of nuclear areas. (**E**) Top: boxplot of the distributions of largest nuclear cross-sectional areas in the indicated H2A overexpressing cells. Bottom: boxplot of the distributions of α-HA fluorescence intensities. Dashed lines mark the median value for the WT strain. All scale bars are 10 µm. Boxes are colored if the mean of the indicated strain is significantly different from WT. Red stars denote level of significance: \*p\<0.01; \*\*\*p\<0.0001 ([Supplementary files 4 and 5](#SD5-data SD6-data){ref-type="supplementary-material"}).**DOI:** [http://dx.doi.org/10.7554/eLife.02792.017](10.7554/eLife.02792.017) Discussion {#s3} ========== In this study, we describe evolutionary adaptations of the histone H2A whereby single arginines in the NTD function to dramatically affect the degree of genome compaction. This mechanism is distinct from several other known chromatin compaction mechanisms in higher eukaryotes ([@bib2]; [@bib15]; [@bib61]; [@bib73]; [@bib23]) in that it involves the histone proteins themselves. The H2A-mediated chromatin compaction thus provides a novel but potentially complementary mechanism for genome compaction. Organisms with small genomes but also very small cell size may face similar physical constraints as those with larger genomes, and may therefore use arginine-containing H2A as a means for chromatin compaction. For instance, *Ostreococcus tauri* which possesses an R3-containing H2A, is a free-living unicellular algae that has a very small genome of 12.6 Mbp but a cell diameter of 0.8 µm ([@bib50]). *S. cerevisiae*, which does not contain an H2A with R3, has a similarly sized genome but has a cell diameter that is approximately five times larger. Furthermore, certain organisms such as *Oikopleura dioica*, which has one of the smallest genomes in animals, have distinctive life cycles and possess H2A genes with and without arginines, which may enable them to dynamically regulate genome compaction at different stages of their life cycles ([@bib47]) (for species with H2A isoforms, see attached spreadsheet). So, the H2A arginines may have evolved in circumstances when the genome size became disproportionately large compared to nuclear volume. Interestingly, the toad, *Bufo gargarizans*, which has a genome size that is twice as large as the human genome, possesses an H2A gene with not only R3 and R11 but also glutamine 6 replaced with an arginine, suggesting that additional arginines in the H2A tail may enable further compaction in organisms with even larger genomes. To better understand the three dimensional positions of the H2A NTD arginines, we examined a crystal structure of the mono-nucleosome in which R3, R11, R17, and R20 were all simultaneously crystalized, and visualized interactions between nucleosomes within the crystal lattice ([@bib12]). Interestingly, while R17 and R20 are more buried within the octamer, R3 and R11 are situated close to the DNA backbone. R3 is at 2.87 Å from the DNA and could potentially bind the DNA gyre as the DNA wraps around the histone octamer. R11 forms close contacts with the DNA phosphate backbone of self and neighboring nucleosomes (4.09, 2.90 Å, respectively). Although these interactions may have helped form the crystal lattice, they also suggest a possible mechanism for tighter nucleosomal stacking in vivo through shielding of the DNA negative charge ([Figure 1---figure supplement 2A,B](#fig1s2){ref-type="fig"}; [@bib12]). Thus, the evolutionary appearance of arginines in the H2A NTD sequence at positions 3 and 11 corresponds to strategic positioning of R3 and R11 within the nucleosome structure that may enable interactions with the DNA, leading to more compact chromatin. Our in vivo data in yeast cells demonstrate that interprobe distances shorten in the presence of the H2A arginines R3 and R11. While the mechanism of this shortening is still unknown, the two most likely explanations are due to linear chromosomal compaction or increased chromatin looping ([@bib3]). However, our data are more consistent with increased linear compaction due to several reasons. First, our analysis of multiple probes along Chr XVI in yeast demonstrates a uniform compaction between all probe pairs examined. Second, our in vivo data with human cell lines shows de-compaction of chromatin in the absence of R11. If chromatin looping was the mechanism, loops would have to be disassembled in human nuclei independent of factors such as CTCF and condensin. Third, our in vitro data show that R11 alone affects chromatin compaction even in absence of divalent cations. Because the in vitro experiments were performed with unmodified histones in arrays with equal linker lengths, this strongly points to a direct effect of H2A NTD arginines on chromatin compaction that occurs in short arrays. Both R3 and R11 are at contact distances from the DNA, and R11 may also contact the DNA backbone of the neighboring nucleosome ([Figure 1---figure supplement 2](#fig1s2){ref-type="fig"}). These intra- and inter-nucleosomal interactions with the arginines and the DNA may serve to neutralize the negative charge of the DNA backbone, leading to enhanced stacking of nucleosomes and hence increased compaction. Consistent with this model, the other two arginine residues in the H2A NTD, R17 and R20, which are more buried from the surface, do not affect compaction by themselves. However, their functions may be to modulate the effects of the surrounding residues. Although all canonical H2A genes contain R3 and R11 in humans, the cell may still be able to dynamically regulate chromatin compaction by these arginines. For instance, arginines may be subject to posttranslational modifications, such a methylation which makes the arginine residue bulkier, or citrullination which removes the positive charge ([@bib69]; [@bib31]; [@bib14]; [@bib68]). Interestingly, the H2A NTD is situated in close proximity to the H2B CTD which when ubiquitylated, disrupts chromatin compaction in vitro ([@bib23]), lending support to the ability of this region of the nucleosome to modulate chromatin compaction. The inability of lysines, especially at position 11, to increase chromatin compaction suggests exquisite structural constraints for H2A-mediated chromatin compaction. Although lysines and arginines both are positively charged, the positive charge of arginine is due to the presence of a guanidinium group that is structurally different from the positive charge of an amino group of a lysine residue. In this regard, it is interesting to note that only arginines contact DNA as it wraps around the nucleosome core ([@bib42]); and arginines preferentially bind the minor groove of DNA compared to lysines ([@bib54]). The context in which lysines appear in evolution may be important as well. We did not observe a lysine at position 3 in our list of organisms, and K11 was present in organisms with medium-sized genomes and surrounded mainly by the motif VKG ([Figure 1D,E](#fig1){ref-type="fig"}). When tested in our *S. cerevisiae* strains, K11 was in the context of AKA. So, it is conceivable that additional amino acid changes would be required for lysines in the H2A NTD to increase genome compaction. The nucleoskeletal theory proposes that chromatin structure influences the shape of the nucleus, and thus is a major determinant of nuclear volume ([@bib8]), although the amount of DNA per se does not affect nuclear volume ([@bib48]). Non-chromatin components such as nuclear import factors from the cytoplasm may also modulate nuclear size ([@bib36]). Our data suggest that in particular cases, the effects of H2A NTD mutations on chromatin compaction are linked to nuclear volume, although not in a straightforward relationship. While arginines at positions 3 and 11 increase chromatin compaction and reduce nuclear volume, lysines at the same positions have no effect on chromatin compaction yet increase nuclear volume. Removal of serines at positions 10 or 15 has little effect on compaction but also increase nuclear volume. In human cells, expression of all R11 mutants (R11A, C, H, and P) decreased compaction but only R11A also affected nuclear area. Although we do not observe a clear-cut relationship between nuclear volume and chromatin compaction, our data identify a region of the nucleosome that is directly or indirectly linked to nuclear volume control mechanisms. Since alterations in chromatin structure often cause changes in transcription ([@bib51]), we were surprised that mutant H2A-containing yeast had very similar gene expression profiles as WT cells, grew at similar rates, did not have altered cell cycle profiles, and were not sensitive to DNA damaging drugs or environmental challenges. These data raise the possibility that H2A-mediated compaction of chromatin may have evolved as a mechanism to enable regulation of chromatin compaction without having to make compensatory changes to all other processes that are also based on DNA such as transcription. Nevertheless, it remains to be determined what molecular or cellular processes govern the optimal level of chromatin compaction and nuclear volume for an organism. The stable alterations of chromatin compaction in eukaryotic model organisms through genetic manipulation of H2A should facilitate further experiments to uncover these processes. Materials and methods {#s4} ===================== Strains and media {#s4-1} ----------------- The yeast strains used in this study are listed in [Supplementary file 1A](#SD2-data){ref-type="supplementary-material"}. Yeast cells were grown in YPD at 30°C unless otherwise noted. C-terminal tagging of yeast proteins was performed as described previously ([@bib39]). Mammalian cell lines were maintained at 37°C and 5% CO~2~ and cultured with 10% fetal bovine serum and DMEM (Life Technologies, Grand Island, NY). Histone sequence database construction and analysis {#s4-2} --------------------------------------------------- Sequences were initially extracted from the Entrez database using a keyword search for 'histone', and removing non-histone sequences by using keyword searches such as 'histone-like', 'ubiquitin', and 'acetyl', yielding 54,646 results. Blast 2.0 ([@bib5]) was used to align the sequences against the highly conserved histone fold region of the four core histones from *Homo sapiens*. Thresholds for true hits were set at \>35% identity match and \>90% overlap match with the histone fold globular domain region. All duplicate sequences were removed, and further sequence comparisons were made for histone H3 and H2A sequences to filter variants within them. The canonical sequence data sets comprised 672 sequences for histone H3, 357 sequences for histone H4, 518 sequences for histone H2B, and 435 sequences for histone H2A. To further select one canonical sequence for a species among isotypes and variants when annotation was missing, the sequences were compared to the canonical *H. sapiens* and *S. cerevisiae* sequence, and the sequence with the highest similarity was selected. Using only completely sequenced species, the final histone sequence data set included canonical sequences for 160 species from plants, fungi, protozoa, and animals, with genome sizes ranging from 8 to 5600 Mbp. Sequences for the four core histones were subsequently split into the N-terminal tail, globular domain, and C-terminal tail (in the case of H2A and H2B) sub-sequences. For discovery of patterns of residue changes according to genome size, each of the sub-sequences was further sub-grouped into small (\<100 Mbp), medium (100--1000 Mbp), and large (\>1000 Mbp) genome sizes. The frequency of the amino acid residues in each sequence in the sub-groups was determined, and a p-value for the comparison between sub groups was obtained using a Mann--Whitney U Test. Multiple sequence alignment profiles were created using the Muscle sequence comparison tool from Embl-EBI ([@bib19], [@bib20]). Weblogo3 ([@bib58]; [@bib11]) was used for motif discovery. Heat maps for residue positions were constructed using Cluster 3.0 ([@bib13]) and Java Treeview ([@bib55]). Yeast H2A mutagenesis {#s4-3} --------------------- Site directed mutagenesis was performed using the QuickChange Lightning kit (Agilent Technologies, Santa Clara, CA) on the pFL142 plasmid. [Supplementary file 1B](#SD2-data){ref-type="supplementary-material"} contains all the plasmids that were used and constructed in this study. The sequences of primers are listed in [Supplementary file 1C](#SD2-data){ref-type="supplementary-material"}. The correct mutation was verified by sequencing. Measurement of yeast nuclear volume {#s4-4} ----------------------------------- Yeast strains were generated that contained a C-terminally tagged Nup49p-GFP fusion. Cells were grown in a rich medium to 0.6--0.8 × 10^7^ cells/ml, fixed in a growth medium with 4% paraformaldehyde for 15 min at room temperature, washed twice in PBS, and mounted on a poly-L-lysine-coated slide with mounting medium (Vector Laboratories, Burlingame, CA). Z-stacks were obtained as described in the microscopy imaging section, and GFP excitation was achieved at 488 nm. Resulting z-stack images were de-convolved using a constrained iterative algorithm from SlideBook 5.0 software and nuclear volumes were measured by masking the inside of each nucleus, which were delineated by the GFP signal. The resulting mask was used to calculate volumes through the SlideBook software. Statistical analysis was performed using the Student\'s *t* test. Measurement of yeast cellular volume {#s4-5} ------------------------------------ Yeast strains were grown in rich medium to 0.6--0.8 × 10^7^ cells/ml, fixed in growth medium with 4% paraformaldehyde for 15 min at room temperature, washed twice in PBS, and stained with a 1:50 dilution of concanavalin A conjugated with tetramethylrhodamine (Life Technologies) for 15 min at room temperature. Cells were washed twice in PBS, once in water, and mounted on a poly-L-lysine-coated slide with mounting medium. Z-stacks were obtained as described in the microscopy imaging section with mRFP excitation. Cell volume was measured by masking the inside of the RFP signal as described in the measurement of yeast nuclear volume. FISH probes {#s4-6} ----------- For yeast FISH analysis, DNA templates for probes 1, 3, and 4 came from cosmids 71042, 70912, and 70982 (American Type Culture Collection, Manassas, VA) as described elsewhere ([@bib29]). DNA templates for Probe 2 were obtained by PCR amplification of a 10-kb region starting at coordinate 364647 of chromosome 16 using three primer pairs (Probe2_P1, Probe2_P2, Probe2_P3, [Supplementary file 1C](#SD2-data){ref-type="supplementary-material"}). All DNA templates were digested to smaller fragments using Sau3a (New England BioLabs, Ipswich, MA). Fragments were directly labeled using BioPrime labeling kit (Life Technologies) with either ChromaTide Alexa Fluor 488-5-dUTP or ChromaTide Alexa Fluor 568-5-dUTP (Life Technologies). For human cell FISH analysis, DNA templates for probes came from BACS RP11-252L24 and RP11-195J4 spaced 0.488 Mb apart on chromosome 1. Each BAC was digested into smaller fragments using Sau3a and fragments were directly labeled using BioPrime labeling kit with either ChromaTide Alexa Fluor 488-5-dUTP or ChromaTide Alexa Fluor 568-5-dUTP, as described above. Fluorescent in situ hybridization analysis in yeast {#s4-7} --------------------------------------------------- Yeast strains were grown in rich medium to 0.6--0.8 × 10^7^ cells/ml and fixed in a growth medium with 4% paraformaldehyde for 15 min at room temperature. Cells were then washed twice in the growth medium and re-suspended in 2 ml of EDTA-KOH (0.1 M, pH 8.0) and 10 mM DTT and incubated for 10 min with shaking at 30°C. Cells were spun down and re-suspended in 2 ml of YPD + 1.2 M sorbitol with 50 µg/ml of Zymolyase 100-T (Sunrise Science Products, San Diego, CA) and 400 U/ml of lyticase (Sigma-Aldrich, St. Louis, MO) and incubated at 30°C for 16 min with shaking. Spheroplasts were then washed twice in YPD + 1.2 M sorbitol and transferred to a poly-L-lysine-coated slide. After settling for 5 min, excess liquid was aspirated away and the slides were allowed to air dry for 5 additional min. The slides were washed in methanol for 10 min and then acetone for 30 s before air drying. Cells were then dehydrated in a series of cold ethanol washes (70%, 80%, 90%, 100%, 1 min each) and allowed to air dry. Denaturing solution (70% deionized formamide, 2 × SSC) was added to the slide, and cells were denatured at 75°C for 7--10 min. The slides were immediately put through another cold ethanol dehydration series and allowed to air dry. Hybridization solution (50% deionized formamide, 2 × SSC, 10% dextran sulfate, 100 ng/µl salmon sperm DNA) containing fresh probes was added to the slide, and the probes were hybridized for 40--48 hr at 37°C. The slides were then washed in two 5 min washes in 0.05 × SSC at 48°C and washed twice in BT Buffer (0.15 M NaHCO3 pH 7.5, 0.1% Tween) for 5 min at room temperature. Mounting medium containing DAPI (Vector Laboratories) was added to the slides, and a coverslip was sealed with nail polish. Inter-probe distances were measured in single projections as described elsewhere ([@bib4]) by finding the pixel distance between weighted centers of the green signal and red signal and converted to nm by the appropriate factor. Microscopy imaging {#s4-8} ------------------ A 3i Marianas SDC confocal microscope equipped with a Zeiss AxioObserver Z1 with a 100 × /1.45 NA objective and Yokogawa CSU-22 confocal head was used. Images were captured by a Hamamatsu EMCCD C9100-13 camera controlled by Slidebook 5.0/5.5 (Intelligent Imaging Innovations, Denver, CO). DAPI, GFP, mRFP, and Far-red images were acquired by excitation at 360 nm, 488 nm, 561 nm, and 640 nm from a high-speed AOTF laser launch line. A step size of 0.3 (yeast) or 0.5 (human) µm was used for z-stack acquisition. Micrococcal nuclease digestion {#s4-9} ------------------------------ Micrococcal nuclease (MNase) digestions were performed on exponentially growing yeast cells as described previously, except that the enzyme was obtained from Sigma-Aldrich (Sigma-Aldrich) ([@bib52]). RNA expression analysis {#s4-10} ----------------------- RNA was extracted from exponentially growing yeast as described previously ([@bib57]). PolyA-RNA was prepared, labeled, and hybridized to Affymetrix Gene ChIP Yeast Genome 2.0 array by the UCLA clinical microarray core facility and data normalized according to manufacturer\'s indications. The data are accessible at Gene Expression Omnibus with accession number GSE50440. DNA template and histone preparation for in vitro studies {#s4-11} --------------------------------------------------------- A plasmid containing 12 tandem 177 bp repeats of the high affinity 601 sequence was obtained from Craig L Peterson\'s laboratory ([@bib61]). DNA arrays were prepared as described previously ([@bib43]). After excision with EcoRV, the arrays were gel purified. QuikChange Lightning Site-Directed Mutagenesis (Agilent Technologies) was used to create H2A ΔR11 using primers as listed in [Supplementary file 1C](#SD2-data){ref-type="supplementary-material"}. Recombinant *X. laevis* histones were expressed in bacteria and purified as described previously ([@bib43]). Equimolar amounts of all histones were co-folded to form octamers. Intact octamers were purified from aggregates and free H2A-H2B dimers using Pharmacia Superdex 200 gel filtration column (GE Healthcare Bio-Sciences, Pittsburgh, PA). Nucleosome array assembly {#s4-12} ------------------------- Recombinant histone octamers and the 601-177-12 DNA template ([@bib40]) were combined in stoichiometric amounts where 1.0 equivalent of histone octamers and 1.0 equivalents of DNA template were mixed in 2.0 M NaCl. Nucleosome arrays were assembled by step-wise salt dialysis in decreasing NaCl concentration: 1.6 M, 1.2 M, 1.0 M, 0.6 M, 0.4 M, 0.1 M, and 0.025 M (in 10 mM Tris pH 8.0, 0.25 mM EDTA), followed by exchanges with 2.5 mM NaCl and 10 mM Tris pH 8.0 without EDTA. Each dialysis step was performed at 4°C for 4 hr to overnight. Partially assembled chromatin was eliminated by precipitation in 4.0 mM MgCl~2~ ([@bib15]). The extent of array saturation was assessed by ScaI digestion (200 ng total DNA/chromatin, 3 units ScaI, 50 mM NaCl, 50 mM Tris pH 7.4, 0.5 mM MgCl~2~), performed for 16 hr at room temperature followed by 1 hr at 37°C, and subsequent analysis using a 5% native polyacrylamide gel ([@bib43]). Analytical ultracentrifugation {#s4-13} ------------------------------ Nucleosome arrays were allowed to equilibrate at room temperature in buffer (2.5 mM NaCl, 10 mM Tris--HCl pH 8.0) containing either 0.1 mM EDTA or 0.6 and 0.8 mM MgCl~2~. Samples were centrifuged at 20,000 RPM on a Beckman Optima XL-I analytical ultracentrifuge using an An60 Ti rotor after a 1 hr equilibration at 20°C under vacuum. Time-dependent sedimentation was monitored at 260 nm. Boundaries were analyzed by the method of van Holde and Weischet ([@bib70]; [@bib32]). Combined immunofluorescence and fluorescent in situ hybridization in human cells {#s4-14} -------------------------------------------------------------------------------- N-terminally HA-tagged WT H2A of *X. laevis* was cloned by PCR into mammalian expression vector pCMV-HA (Clontech Laboratories, Mountain View, CA) between EcoRI and NotI sites. C-terminally Myc-FLAG-tagged human H2A, in a mammalian expression vector, was obtained from OriGene (RC200688, Origene Technologies, Rockville, MD). Site directed mutagenesis was performed using the QuickChange Lightning kit (Agilent Technologies) on these expression plasmids. Human cells (HEK293, IMR90 and MDA-MB-453) were grown on glass coverslips in 24-well plates in DMEM containing 10% fetal bovine serum and transfected with the indicated H2A expression plasmids using BioT transfection reagent (Bioland Scientific, Paramount, CA) or Lipofectamine LTX with Plus reagent (Life Technologies). Cells were grown for 48 hr post-transfection. For immunofluorescence only, transfected cells were fixed with ice-cold methanol for 15 min at −20°C followed by washing with PBS-T. For combined immunofluorescence and FISH, transfected cells were fixed with 4% paraformaldehyde in PBS for 10 min at room temperature followed by washing with PBS. Cells were then permeabilized in 0.5% Triton X-100 in PBS for 10 min at room temperature followed by washing with PBS. Cells were blocked in 5% BSA and incubated with anti-HA antibody (ab9110; 1:250 dilution, Abcam, Cambridge, MA) or anti-FLAG antibody (F1804; 1:1000 dilution, Sigma-Aldrich). Cells were washed and incubated with secondary antibody (A11008; 1:500 Alexa Fluor 488 goat anti-rabbit, A21245; 1:250 Alexa Fluor 647 goat anti-rabbit, A11001; 1:500 Alexa Fluor 488 goat anti-mouse, or A21235; 1:100 Alexa Fluor 647 goat anti-mouse, Life Technologies). For immunofluorescence, cells were washed and then incubated with Hoechst stain (0.001 mg/ml in PBS). After final washes, cover slips were mounted and imaged. Fluorescence was visualized as above except with the use of 63X magnification. For FISH, cells were washed, following secondary antibody incubation, in CSK buffer (100 mM NaCl, 300 mM sucrose, 3 mM MgCl~2~, 10 mM PIPES pH 6.8) and permeabilized in CSKT buffer (CSK+0.5% Triton X-100) before being fixed for 10 min in 4% paraformaldehyde in PBS at room temperature. Cells were immediately put through a cold ethanol dehydration series (5 min each at 85%, 95%, and 100%) and allowed to air dry. Cells were rehydrated in 2 × SSC for 5 min and then RNase-treated for 30 min at 37°C in a humid chamber. Cells were washed with 2 × SSC and denatured at 80°C for 15--20 min with 70% deionized formamide and 2 × SSC. They were immediately cooled with cold 2 × SSC and put through another cold ethanol dehydration series. Probes were added to cells and allowed to hybridize for 48 hr. After hybridization, cells were washed with 50% formamide in 2 × SSC, 2 × SSC, and 1 × SSC containing DAPI. Slides were mounted, imaged, and analyzed as described above. Nuclear staining, in H2A-expressing cells, was used to measure lengths of the long and short orthogonal nuclear axes. Estimated nuclear cross-sectional area was calculated using the following formula: Area = (D~1~/2) \* (D~2~/2)\*π, where D~1~ and D~2~ are long and short axis lengths, respectively. Competition assays {#s4-15} ------------------ Two sets of yeast strains were generated in which Pgk1p was C-terminally fused with either GFP or RFP ([Supplementary file 1A](#SD2-data){ref-type="supplementary-material"}). GFP-labeled WT H2A strains were co-cultured with RFP-labeled mutant H2A strains at a 1:1 ratio and at an optical density of ∼0.4. Corresponding co-cultures with switched fluorescent labels were also made. Cultures were incubated at 30°C for 72 hr and were diluted every 6--12 hr to maintain cells in exponential growth phase. Samples were collected every 12 hr for analysis by flow cytometry. Collected cells were fixed in 70% ethanol, washed, and re-suspended in 50 mM sodium citrate, pH 7.0, and mildly sonicated to disrupt aggregates. GFP- and RFP- labeled cells were counted using a Becton Dickinson FACScan cytometer, and the proportion of each in the population was calculated. Cell cycle analysis {#s4-16} ------------------- Cell cycle analysis of exponentially growing cells was performed essentially as described previously ([@bib75]), except that cells were stained with 1 μM SYTOX Green (Life Technologies). Spot tests {#s4-17} ---------- Approximately 1.0 × 10^7^ exponentially growing yeast cells were collected and re-suspended in 100 μl of H~2~O and 10-fold serially diluted. Subsequently, 5 μl was spotted on agar plates containing media and drugs as indicated in the figures and incubated at 30°C for 2--6 days. Funding Information =================== This paper was supported by the following grants: - National Institutes of Health [FundRef identification ID:](http://www.crossref.org/fundref/)http://dx.doi.org/10.13039/100000002 Director\'s Innovator Award, 1 DP2 OD006515 to Siavash K Kurdistani. - Howard Hughes Medical Institute [FundRef identification ID:](http://www.crossref.org/fundref/)http://dx.doi.org/10.13039/100000011 Medical Student Fellowship to Benjamin R Macadangdang. - University of California [FundRef identification ID:](http://www.crossref.org/fundref/)http://dx.doi.org/10.13039/100005595 Philip Whitcome Pre-doctoral Training Program to Benjamin R Macadangdang. - University of California [FundRef identification ID:](http://www.crossref.org/fundref/)http://dx.doi.org/10.13039/100005595 UCLA Cancer Cell Biology Postdoctoral Training Program to Tanya Spektor. We acknowledge the support of Martin Phillips and the UCLA-Department of Energy Biochemistry Instrumentation Facility for analytical ultracentrifuge experiments, the Jonsson Comprehensive Cancer Center flow cytometry core facility (supported by grants P30 CA016042 and 5P30 AI028697) and the UCLA Broad Stem Cell Center Sequencing Core. We thank Michael Grunstein and Fred Winston for providing yeast strains TSY107 and FY406, respectively. BM was supported partially by a Howard Hughes Medical Institute Medical Student Fellowship and a Philip Whitcome Pre-doctoral Training Program grant. TS was supported partially by a UCLA Cancer Cell Biology Postdoctoral Training grant. This work was funded by an NIH Director\'s Innovator Award to SKK. Additional information ====================== The authors declare that no competing interests exist. BRM, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. AO, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. TS, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. OAC, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. MV, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. FS, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article. MFC, Conception and design, Analysis and interpretation of data, Drafting or revising the article. SKK, Conception and design, Analysis and interpretation of data, Drafting or revising the article. Additional files ================ 10.7554/eLife.02792.018 ###### Tables of yeast strains, plasmids, and primers. **DOI:** [http://dx.doi.org/10.7554/eLife.02792.018](10.7554/eLife.02792.018) 10.7554/eLife.02792.019 ###### Table of yeast FISH results. **DOI:** [http://dx.doi.org/10.7554/eLife.02792.019](10.7554/eLife.02792.019) 10.7554/eLife.02792.020 ###### Table of yeast nuclear and cellular volumes. **DOI:** [http://dx.doi.org/10.7554/eLife.02792.020](10.7554/eLife.02792.020) 10.7554/eLife.02792.021 ###### Table of human FISH results. **DOI:** [http://dx.doi.org/10.7554/eLife.02792.021](10.7554/eLife.02792.021) 10.7554/eLife.02792.022 ###### Table of human nuclear area results. **DOI:** [http://dx.doi.org/10.7554/eLife.02792.022](10.7554/eLife.02792.022) Major dataset ------------- The following dataset was generated: SiavashKurdistani, 2014,Expression data from Saccharomyces cerevisiae histone H2A mutants,<http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE50440>,Publicly Available at NCBI Gene Expression Omnibus. 10.7554/eLife.02792.023 Decision letter Kadonaga James T Reviewing editor University of California, San Diego , United States eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see [review process](http://elifesciences.org/review-process)). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers. Thank you for sending your work entitled "Evolution of Histone 2A for Chromatin Compaction in Eukaryotes" for consideration at *eLife*. Your article has been favorably evaluated by James Manley (Senior editor) and 3 reviewers, one of whom is a member of our Board of Reviewing Editors. The Reviewing editor and the other reviewers discussed their comments before we reached this decision, and the Reviewing editor has assembled the following comments to help you prepare a revised submission. This paper explores a novel and interesting hypothesis, which is that the inclusion of extra arginine residues at position 3 and 11 in the N-terminal tail of histone H2A allows a greater degree of compaction of the chromatin. The authors found a strong correlation between the number of arginine residues in the N-terminal domain (NTD) of histone H2A and genome size. This observation led them to postulate that the NTD arginines promote chromatin compaction and a consequent reduction in the nuclear volume. A variety of substitution, insertion, and deletion mutations were constructed in the NTDs of yeast and human H2A. A good correlation between NTD arginines and chromatin compaction was observed. There was not, however, a strong correlation between chromatin compaction and nuclear volume. This study was based on an innovative idea and contains information that will be of interest to chromatin biologists. This work could be appropriate for publication in *eLife* if the following points are suitably addressed. 1\) [Figure 2](#fig2){ref-type="fig"} is flawed and should be deleted as a main figure. [Figure 2](#fig2){ref-type="fig"} shows the locations of R3 and R11 on an image of two nucleosomes stacked against each other, to infer that these amino acids are involved in interactions between nucleosomes to promote chromatin compaction. However, the structure used (Davey et al., 2002) was a mononucleosome. This paper itself did not show anything more than a mononucleosome, and any potential nucleosome-nucleosome interactions that have been inferred from the coordinates of the mononucleosome may be crystal-packing artifacts. Furthermore, this modeling should be done using instead Tim Richmond\'s tetranucleosome structure, albeit that the nucleosome interactions in that structure have not been validated experimentally. As such, [Figure 2](#fig2){ref-type="fig"} does not stand on its own and should be deleted. The models should be remade with two nucleosomes from the tetranucleosome structure and presented as a supplement to [Figure 1](#fig1){ref-type="fig"}. If the authors wish to keep the Davey et al. models, the text also needs to acknowledge that the structure that was used to generate the models was a mononucleosome and that the interactions shown have not been validated and may be crystal packing artifacts. 2\) The authors should consider and discuss the possibility that the changes in inter-locus distances (i.e., apparent chromatin compaction) may be due to chromatin looping rather than chromatin compaction. This chromatin looping hypothesis could be tested with a 3C-like method, such as Hi-C, but we do not request that such extensive experiments be performed for this paper. This issue will, however, ultimately need to be addressed by the authors. 3\) Changes in nuclear volume do not always correlate with chromatin compaction (as summarized in Table 1). Hence, there is not a straightforward relationship between chromatin compaction and nuclear volume. This point needs to be made clearly in the text. 4\) There are two issues with [Figure 6](#fig6){ref-type="fig"}: a\) The authors reconstituted oligonucleosome arrays with either wild-type H2A or deltaR11. The sedimentation coefficients of the mutant arrays were decreased, suggesting loss of inter-fiber compaction. It should be noted, however that results from the lab of Jeff Hansen (Gordon et al., 2005) showed that reconstituted nucleosome arrays missing only the H2A N-terminal region have a sedimentation profile that is indistinguishable from that of wild-type arrays. Hence, the results from the Hansen lab argue against an essential role of the H2A N-terminal region in chromatin compaction and appear to be inconsistent with the data in [Figure 6](#fig6){ref-type="fig"}. b\) The biochemical experiments provide an excellent opportunity to examine the effect of charge on chromatin compaction in the absence of potential histone acetylation. Thus, it could be informative to test the relative effect of R vs. K in these experiments. However, given the ambiguity in the sedimentation data from different labs (as discussed in point 4a, above), it is not certain how useful such experiments would be. 5\) [Figure 7E](#fig7){ref-type="fig"} shows some mutations of H2A that have been found in cancer. The authors make the argument that the cancer mutations in a single copy of an H2A gene have the potential to affect chromatin compaction, as supported by their overexpression analysis in tissue culture cells. However, in their overexpression study, they show that the mutant histones are expressed to the same level as the endogenous histones. This is the equivalent of a heterozygous mutation in a single copy gene. However, there are 17 canonical H2A genes in humans; hence, the authors are not making a fair argument, as a mutant histone in humans would only be present at 1/34th of the level of the non-mutant histones (assuming equal expression of all histone H2A genes). This fact needs to be made clear in the Discussion. 10.7554/eLife.02792.024 Author response *1)* [*Figure 2*](#fig2){ref-type="fig"} *is flawed and should be deleted as a main figure.* [*Figure 2*](#fig2){ref-type="fig"} *shows the locations of R3 and R11 on an image of two nucleosomes stacked against each other, to infer that these amino acids are involved in interactions between nucleosomes to promote chromatin compaction. However, the structure used (Davey et al., 2002) was a mononucleosome. This paper itself did not show anything more than a mononucleosome, and any potential nucleosome-nucleosome interactions that have been inferred from the coordinates of the mononucleosome may be crystal-packing artifacts. Furthermore, this modeling should be done using instead Tim Richmond\'s tetranucleosome structure, albeit that the nucleosome interactions in that structure have not been validated experimentally. As such,* [*Figure 2*](#fig2){ref-type="fig"} *does not stand on its own and should be deleted. The models should be remade with two nucleosomes from the tetranucleosome structure and presented as a supplement to* [*Figure 1*](#fig1){ref-type="fig"}*. If the authors wish to keep the Davey et al. models, the text also needs to acknowledge that the structure that was used to generate the models was a mononucleosome and that the interactions shown have not been validated and may be crystal packing artifacts*. We agree with the comments and have moved [Figure 2](#fig2){ref-type="fig"} to the supplementary material (as a supplement to [Figure 1](#fig1){ref-type="fig"}). We have also edited the text to reflect that the internucleosomal contacts are indeed inferred from crystal lattice packing. The reason we performed this analysis in the first place was because the interaction between H4 K16 and the H2A/H2B acidic patch was described initially from crystal lattice packing (Luger et al., 1997). *2) The authors should consider and discuss the possibility that the changes in inter-locus distances (i.e., apparent chromatin compaction) may be due to chromatin looping rather than chromatin compaction. This chromatin looping hypothesis could be tested with a 3C-like method, such as Hi-C, but we do not request that such extensive experiments be performed for this paper. This issue will, however, ultimately need to be addressed by the authors*. This is a valid point that we have now explicitly discussed in the paper (see Discussion section). However, we think that this is a less likely possibility because examination of inter-probe distances for four probe-pairs on the same chromosome suggests linear compaction ([Figure 2D](#fig2){ref-type="fig"}). *3) Changes in nuclear volume do not always correlate with chromatin compaction (as summarized in Table 1). Hence, there is not a straightforward relationship between chromatin compaction and nuclear volume. This point needs to be made clearly in the text*. We agree with the reviewers and in fact had attempted to make this point in the discussion of the original submission. We have now further amended our discussion on nuclear volume changes to better emphasize this point. *4) There are two issues with* [*Figure 6*](#fig6){ref-type="fig"}*:* *a) The authors reconstituted oligonucleosome arrays with either wild-type H2A or deltaR11. The sedimentation coefficients of the mutant arrays were decreased, suggesting loss of inter-fiber compaction. It should be noted, however that results from the lab of Jeff Hansen (Gordon et al. 2005) showed that reconstituted nucleosome arrays missing only the H2A N-terminal region have a sedimentation profile that is indistinguishable from that of wild-type arrays. Hence, the results from the Hansen lab argue against an essential role of the H2A N-terminal region in chromatin compaction and appear to be inconsistent with the data in* [*Figure 6*](#fig6){ref-type="fig"}. Although we do not have a satisfactory explanation for this discrepancy, we have two counterpoints. 1\) We have performed additional in vivo experiments in which we ectopically expressed either WT H2A or an H2A gene missing amino acids 1-12 ( H2A-NTD) in IMR90 cells. Both constructs were FLAG tagged at the C-terminus. As would be predicted from our data, expression of H2A-NTD which is missing R3 and R11 decreased chromatin compaction (increased inter-probe distances) ([Figure 4E](#fig4){ref-type="fig"}) and also caused an increase in nuclear area ([Figure 4F](#fig4){ref-type="fig"}). We therefore conclude that deletion of the H2A NTD does indeed affect chromatin compaction in vivo. 2\) In Gordon et al., nucleosomal arrays missing the H4 NTD sedimented at a value almost identical to WT octamers: 23.1 S for ΔH4 NTD vs 23.2 S for WT. (Arrays missing H2A sedimented slightly slower than WT arrays with a sedimentation coefficient of 22.7 S). However, other studies have demonstrated that H4 NTD or acetylation of H4K16 do affect chromatin compaction in vitro (Dorigo et al., 2003; Shogren-Knaak et al., 2006). Some of these discrepancies may be attributable to the nature of assembled chromatin (DNA sequence, method of assembly, etc.) and inherent intricacy of such experiments. But we note that none of the referenced studies have validated their findings *in vivo*. Our *in vitro* data are supported by additional *in vivo* evidence for both ectopic addition of the H2A arginines as well as their removal in yeast and human cell line model systems, respectively. *b) The biochemical experiments provide an excellent opportunity to examine the effect of charge on chromatin compaction in the absence of potential histone acetylation. Thus, it could be informative to test the relative effect of R vs. K in these experiments. However, given the ambiguity in the sedimentation data from different labs (as discussed in point 4a, above), it is not certain how useful such experiments would be*. We share the same point of view and in fact did attempt to assemble chromatin in vitro with H2A R11K mutation. But for unknown reasons we were not able to obtain satisfactorily assembled chromatin for sedimentation velocity analysis. However, in addition to the original data on K11 insertion in yeast H2A ([Figure 2B-C](#fig2){ref-type="fig"}), we ectopically expressed an H2A gene with R11K mutation in normal human fibroblasts and found no statistically significant effects on chromatin compaction ([Figure 4A-B](#fig4){ref-type="fig"}), albeit the average values were slightly larger than WT and trending toward significance (p value=0.02 for FISH distances). Also, R11K did increase the average nuclear area. These data suggest that R11K may have subtle effects on chromatin compaction that would be more apparent if more than one region of the genome was analyzed. Together with the findings from yeast, the data suggest that R11 may be somewhat better than K11 in compacting chromatin and has a greater effect on nuclear area, explaining the evolution of R11 in large genome species. *5)* [*Figure 7E*](#fig7){ref-type="fig"} *shows some mutations of H2A that have been found in cancer. The authors make the argument that the cancer mutations in a single copy of an H2A gene have the potential to affect chromatin compaction, as supported by their overexpression analysis in tissue culture cells. However, in their overexpression study, they show that the mutant histones are expressed to the same level as the endogenous histones. This is the equivalent of a heterozygous mutation in a single copy gene. However, there are 17 canonical H2A genes in humans; hence, the authors are not making a fair argument, as a mutant histone in humans would only be present at 1/34th of the level of the non-mutant histones (assuming equal expression of all histone H2A genes). This fact needs to be made clear in the Discussion*. This is a valid observation and we have now modified the text to better explain this point. We have also performed additional experiments showing that ectopic expression of H2A that bears cancer mutations also affect chromatin compaction ([Figure 7](#fig7){ref-type="fig"}). [^1]: These authors contributed equally to this work. [^2]: The -marks indicate spacing for sequence alignment purposes. The inserted residues are bold typed and underlined. Deletions are indicated by ·. [^3]: Percent (%) change refers to the difference in median values relative to WT unless otherwise indicated; the statistically significant differences are bold typed. p-values were calculated using the *t*-test (yeast) and Mann--Whitney U test (human). [^4]: percent change was calculated relative to isogenic WT control (ΔS15).

{ "pile_set_name": "PubMed Central" }

INTRODUCTION ============ Alternative splicing (AS) is considered as the major mechanism that contributes to the increased proteomic diversity in multicellular eukaryotes (Blencowe, [@b7]; Maniatis & Tasic, [@b27]; Nilsen & Graveley, [@b33]). Through AS, one pre-mRNA could produce multiple mRNA isoforms that might be under different post-transcriptional regulation and/or encode proteins with different functions. Recent transcriptome analysis by massive parallel RNA sequencing (RNA-seq) indicated that more than 90% human genes underwent AS (Pan et al, [@b35]; Wang et al, [@b42]). Very often, the pattern of AS was tissue and developmental stage specific, thought to be under precise regulation modulated by cooperative interplays between *trans*-acting RNA binding proteins (RBPs) and *cis*-regulatory elements in nascent transcripts (Barash et al, [@b3]; Black, [@b6]; Chen & Manley, [@b10]; Witten & Ule, [@b48]). Mutations in splicing regulators (Padgett, [@b34]; Yoshida et al, [@b51]) and abnormal splicing of RNA targets have been associated with many human diseases (Cooper et al, [@b11]; Garcia-Blanco et al, [@b12]; Wang & Cooper, [@b43]). Nevertheless, the exact molecular mechanisms controlling the AS process in physiological and pathological conditions are not well-understood to date. Several splicing regulating RBPs have recently been found to modulate hundreds even thousands of functional targets (Lebedeva et al, [@b24]; Licatalosi et al, [@b25]; Mukherjee et al, [@b31]; Ule et al, [@b40]; Wang et al, [@b41]; Xue et al, [@b49]; Yeo et al, [@b50]). Therefore, elucidating the regulatory roles of splicing related RBPs requires comprehensive identification of the RBP--RNA interactions and global quantification of the splicing outcomes induced by RBPs. *RBM10* encodes a 930 amino acid protein containing two RNA recognition motifs (RRM), two zinc fingers and one G patch motif. These motifs were often found in RNA-binding proteins involved in pre-mRNA splicing, such as heterogeneous nuclear ribonucleoproteins (hnRNPs) and protein components of small nuclear ribonucleoproteins (snRNPs; Glisovic et al, [@b13]; Keene, [@b22]). Through mass spectrometric analysis, RBM10 has been reported to associate with purified splicing complex (Rappsilber et al, [@b37]), and was further identified as a component of U2 snRNPs (Makarov et al, [@b26]), spliceosomal A (or prespliceosomal) (Agafonov et al, [@b1]; Behzadnia et al, [@b4]) and B complexes (Agafonov et al, [@b1]; Bessonov et al, [@b5]). Most recently, based on yeast two hybridization method, a study on interactions between more than 200 proteins previously known to be present in spliceosome could demonstrate the physical interaction between RBM10 and multiple spliceosomal components (Hegele et al, [@b17]). Moreover, its closest paralogue RBM5, a putative tumour suppressor of lung and other cancers (Sutherland et al, [@b39]), has been shown to regulate AS of apoptosis related genes, *Fas* receptor and c-*FLIP*, resulting in isoforms with antagonistic functions in controlling programmed cell death (Bonnal et al, [@b8]). Although all these observations would suggest the potential role of RBM10 in pre-mRNA splicing regulation, it remains unclear whether and how RBM10 could regulate splicing. Nonsense and frame shift mutations in *RBM10* have been identified to be causative for TARP syndrome (Talipes equinovarus, atrial septal defect, Robin sequence and persistent left superior vena cava, MIM \#311900), an X-linked inherited disorder leading to multiple organ malformation in affected males (Gripp et al, [@b14]; Johnston et al, [@b21]). More recently, multiple truncating and missense somatic mutations were detected in lung adenocarcinomas (Imielinski et al, [@b20]). These findings implicated the important role of RBM10, but whether its potential function in splicing regulation is involved in these different pathological contexts has not been explored. In this study, we explored the AS regulated by RBM10. Here, we combined photoactivatable-ribonucleoside-enhanced crosslinking and immunoprecipitation (PAR-CLIP) with massive parallel sequencing to identify RNA binding sites for RBM10 in human embryonic kidney (HEK) 293 cells, which turned out to be significantly enriched in the vicinity of both 5′ and 3′ splice sites. Using RNA-seq, we identified 304 and 244 significant exon splicing changes following RBM10 depletion or overexpression (OE) in HEK293 cells, respectively. Among these changes, more than 74% were RBM10 enhanced exon skipping events and they were correlated with strong RBM10 binding near 5′ and 3′ splicing sites of both upstream and downstream introns. Furthermore, in a patient suffering from TARP syndrome, we identified an *in-frame* deletion in *RBM10* and demonstrated that the splicing defects in the lymphoblastoid cells derived from the patient were largely due to the loss of nuclear function of RBM10. Overall, our data provides direct experimental evidence supporting the role of RBM10 in splicing regulation. Our transcriptome-wide analysis of binding pattern and RBM10 splicing profile allows the illustration of the molecular mechanism underlying RBM10 regulated AS. RESULTS ======= Transcriptome-wide binding sites of RBM10 in HEK293 cells --------------------------------------------------------- To identify *in vivo* binding sites of RBM10, we performed PAR-CLIP sequencing (Hafner et al, [@b15]; Lebedeva et al, [@b24]) in HEK293 cells that expressed epitope (FLAG/HA)-tagged RBM10 (Materials and Methods Section). 4-Thiouridine (4SU) labelled and crosslinked cells were immunoprecipitated with monoclonal anti-FLAG antibody. The bound RNAs was then partially digested and radioactively labelled. Protein--RNA complexes were resolved on a denaturing gel. The band corresponding to RBM10--RNA complexes was excised (Supporting Information [Fig S1](#SD1){ref-type="supplementary-material"}A). The RNA was recovered, converted into cDNA and sequenced on an Illumina platform. In total, we performed two biological replicate experiments. The sequencing reads were processed and clustered as described in Materials and Methods Section. A total of 20.6 million sequencing reads could be mapped to the human genome with at most one mismatch (Supporting Information [Table S1](#SD1){ref-type="supplementary-material"}). Compared with all other mutations in the mappable sequence reads, T to C transitions were significantly enriched (Supporting Information [Fig S1](#SD1){ref-type="supplementary-material"}B), manifesting efficient crosslinking of 4SU labelled RNA (Hafner et al, [@b15]). We identified 240,712 and 218,281 RBM10 sequence clusters (putative binding sites) in the two replicates, respectively (Supporting Information [Table S1](#SD1){ref-type="supplementary-material"}, [Fig S1](#SD1){ref-type="supplementary-material"}C and D for the length distribution of binding clusters as well as the number of PAR-CLIP reads within each cluster). Of these, 87,957 sequence clusters had their preferred crosslinking sites, *i.e*. the position with the highest number of T to C transitions within a site, to be within the binding site identified in the other replicate. We defined these clusters as consensus binding clusters. Comparison of the binding scores of these consensus binding clusters between the two replicates revealed a high correlation (*R*^2^ = 0.619) (Supporting Information [Fig S1](#SD1){ref-type="supplementary-material"}E). RBM10 binding in the vicinity of intronic splicing sites -------------------------------------------------------- Ninety-one percent of the consensus binding sites could be assigned to 6396 RBM10 target genes. According to Refseq annotation, 39 and 52% of them fell into exonic and intronic regions, whereas 9% mapped to intergenic regions, which might harbour previously unannotated transcripts ([Fig 1](#fig01){ref-type="fig"}A). Given the possible involvement of RBM10 in splicing process, we examined the distribution of binding sites relative to splice sites. Intriguingly, we found that they were significantly enriched in exons and in the vicinity of both 5′ and 3′ splice sites of the introns ([Fig 1](#fig01){ref-type="fig"}B). Notably, the binding sites were more enriched at the vicinity of (∼70 nt upstream) of 3′ splice site than at 5′ splice site. Interestingly, we also observed the specific binding of RBM10 at U2 snRNA, which is known to pair with 3′ branch site (Supporting Information). Together, these binding patterns were consistent with the previous findings of RBM10 in pre-spliceosomal A and B complex (Agafonov et al, [@b1]; Behzadnia et al, [@b4]; Bessonov et al, [@b5]) and indicated that RBM10 very likely involves in splice site recognition and/or pairing, as well as further intron removal processes via coordinated interactions with snRNPs and the pre-mRNA substrates. ![RBM10 RNA binding sites identified by PAR-CLIP\ The genomic distribution of 87,957 RBM10 consensus binding clusters.Distribution of RBM10 consensus binding clusters (red line), PAR-CLIP reads (green line) and PAR-CLIP reads containing at least one T to C change (converted reads, blue line) along exon--intron and intron--exon boundaries. The density of RBM10 converted PAR-CLIP reads in the vicinity of intronic splice sites were significantly higher than that of Ago2 converted PAR-CLIP reads (grey line).](emmm0005-1431-f1){#fig01} Alternative splicing regulated by RBM10 --------------------------------------- The RNA binding patterns presented above suggested that RBM10 might function as a splicing regulator. To test this possibility, we performed RNA-seq and quantified changes in gene expression as well as AS in HEK293 cells upon RBM10 knockdown (KD) or overexpression (OE) respectively (Materials and Methods Section). The efficiency of KD and OE were assessed both at mRNA level by qPCR and at protein level by Western blot (Supporting Information [Fig S2](#SD1){ref-type="supplementary-material"}A and B). In total, we performed two and four biological replicate experiments for OE and KD, respectively. Sixty-one to 185 millions 100 nt sequencing reads were generated for each sample, of which 92--96% could be mapped to the genome reference (UCSC genome browser hg19) or a reference set of exon--exon junction sequences (see Materials and Methods Section and Supporting Information [Table S2](#SD1){ref-type="supplementary-material"}). The gene expression level was estimated based on RPKM value (reads per kilobase of exon per million mapped sequence reads, (Mortazavi et al, [@b30]), Materials and Methods Section). At false discovery rate (fdr) \<0.05, 171 and 105 genes were found to be significantly upregulated and downregulated by at least 1.5-fold upon RBM10 KD (Supporting Information [Fig S2](#SD1){ref-type="supplementary-material"}C and [Table S3](#SD1){ref-type="supplementary-material"}), whereas 19 and 49 genes were upregulated and downregulated to the same level (fdr \< 0.05, fold change ≥1.5) in response to RBM10 OE, respectively (Supporting Information [Fig S2](#SD1){ref-type="supplementary-material"}D and [Table S3](#SD1){ref-type="supplementary-material"}). Overall, the expression changes induced by KD and OE were not inversely correlated (Supporting Information [Fig S2](#SD1){ref-type="supplementary-material"}E). We then sought to characterize the splicing changes induced by RBM10 OE/KD. Based on RNA-seq data, we defined the inclusion ratio (PSI: percentage splicing in) of each exon in Refseq transcripts as the number of reads exclusively supporting inclusion divided by total number of reads supporting inclusion and exclusion of the specific exon (Supporting Information [Fig S2](#SD1){ref-type="supplementary-material"}F; Polymenidou et al, [@b36]; Wang et al, [@b42]). We then compared the inclusion ratio between KD and control, OE and control, respectively. The changes were transformed into *Z*-value (Supporting Information [Fig S2](#SD1){ref-type="supplementary-material"}H) and the results from replicate experiments were combined to evaluate statistical significance using the rank product method (Materials and Methods Section). At a stringent cutoff (fdr \< 0.05, \|ΔPSI\| ≥ 10%), we identified 256 induced cassette exon inclusion and 48 exclusion events upon RBM10 KD ([Fig 2](#fig02){ref-type="fig"}A and Supporting Information [Table S4](#SD1){ref-type="supplementary-material"}). In comparison, 27 exon inclusion and 217 exclusion events were observed upon RBM10 OE ([Fig 2](#fig02){ref-type="fig"}A and Supporting Information [Table S4](#SD1){ref-type="supplementary-material"}). ![Exon splicing changes induced by RBM10 perturbation and their association with RBM10 binding profile\ Exon splicing changes (ΔPSI, percentage splicing in) induced by RBM10 OE (*X* axis) were plotted against those induced by RBM10 KD (*Y* axis). A total of 412 cassette exons were found to be differentially spliced after RBM10 OE or KD (FDR ≤ 5%, \|ΔPSI\| ≥ 0.1).The splicing changes (left panel, ΔPSO, percentage splicing out; right panel, ΔPSI) of 21 exons (orange dots in A) quantified by RNA-seq (*X* axis) are highly correlated with those measure by qPCR (*Y* axis).Representative examples of RBM10 dependent exon skipping events. The density of RNA-seq reads obtained in control, RBM10 OE and KD experiments together with RBM10 binding sites were shown for gene *UBN1* and *PUF60*. Reads supporting the inclusion or exclusion of the cassette exon, or spanning exon--intron junctions were shown in blue, orange or green, respectively. The remaining reads were shown in grey. The size of the triangle marking RBM10 binding sites reflected the number of PAR-CLIP reads.The density of RBM10 binding clusters close to 5′ and 3′ splicing sites of introns flanking all cassette exons (black) is significantly lower than those flanking the exons more excluded after RBM10 OE (orange, *Z*~PSI~ ≤−2 and green, *Z*~PSI~ ≤−1).Cumulative distribution functions of splicing changes upon RBM10 OE for different groups of cassette exons with RBM10 binding close to none or one of the four splicing sites (upper panel), or to different number of the four splicing sites (lower panel). The numbers of exons within different groups were printed in parenthesis.](emmm0005-1431-f2){#fig02} We then selected 21 candidate transcripts for which we had detected splicing changes with different *Z* values for validation by qPCR using junction specific primers ([Fig 2](#fig02){ref-type="fig"}A and Supporting Information [Fig S3](#SD1){ref-type="supplementary-material"}). The abundance of transcript isoforms including or excluding the cassette exons was normalized based on that of constitutive exons. We could validate splicing changes in all the 21 cases (Supporting Information [Fig S3](#SD1){ref-type="supplementary-material"}). The splicing changes detected by qPCR were quantitatively correlated with that determined by RNA-seq ([Fig 2](#fig02){ref-type="fig"}B). Comparison of splicing changes induced by OE and that by KD revealed a clear inverse correlation ([Fig 2](#fig02){ref-type="fig"}A). The majority (74%) of the splicing changes observed upon RBM10 OE and KD are RBM10-enhanced exon exclusion events ([Fig 2](#fig02){ref-type="fig"}A), indicating that RBM10 primarily mediated the skipping of cassette exons. An RNA splicing map integrating RBM10 binding profile and induced splicing changes ---------------------------------------------------------------------------------- We then took a close look at functional annotations of the genes that changed the expression level and/or the splicing pattern as a result of RBM10 perturbation. Interestingly, 14 RNA-binding proteins and five known splicing regulators were found with significant expression changes (Supporting Information [Table S3](#SD1){ref-type="supplementary-material"}), and even more genes with splicing changes (22 and eight) were found to be involved in RNA-binding or splicing regulation (Supporting Information [Table S4](#SD1){ref-type="supplementary-material"}). Therefore, the overall splicing changes described above represented not only the direct RBM10-targeted splice events, but also the secondary effects resulting from the expression and/or splicing changes in those splicing regulators. In order to understand the mechanisms for AS directly under RBM10 regulation, we correlated the RBM10 RNA binding pattern with the splicing changes upon RBM10 OE and KD. Among the RBM10-enhanced exon skipping events, we often observed RBM10 binding(s) close to 5′ and/or 3′ splicing sites at upstream and/or downstream introns. Two representative examples were depicted in [Fig 2](#fig02){ref-type="fig"}C. To examine general mechanism for RBM10 enhanced exon skipping, we integrated the PAR-CLIP data and splicing profiles into an RNA splicing map. As shown in [Fig 2](#fig02){ref-type="fig"}D, the map revealed increased density of RBM10 binding clusters close to the splice sites of both introns flanking the skipped cassette exons, with the most prominent enrichment at 3′ splice site of downstream intron. The RNA splicing map of RBM10 suggested that RBM10 binding close to the splice sites of neighbouring introns are enriched for skipped exons. To assess whether such binding pattern could predict exon exclusion events, we searched our PAR-CLIP data for non-constitutive exons with RBM10 binding close (*i.e*. up to 150 nt) to the splice sites of adjacent introns. In total, 5262 such exons were found. Among these exons, 147 showed significant splicing changes (fdr \< 0.05, \|ΔPSI\| ≥ 10%) upon RBM10 OE and/or KD, accounting for 30.8% (147/412) of all the exons with significant splicing changes upon RBM10 perturbation. As shown in [Fig 2](#fig02){ref-type="fig"}E, the exons with RBM10 binding close to one of the four splice sites were more likely excluded upon RBM10 OE, and those with binding close to 3′ splice sites of upstream introns exhibiting the weakest skipping propensity. Intriguingly, exons with binding close to more of the four splice sites showed progressively stronger skipping tendency upon RBM10 OE ([Fig 2](#fig02){ref-type="fig"}E). Similarly, exons with binding close to the same four splicing junctions showed progressively stronger inclusion tendency upon RBM10 KD (Supporting Information [Fig S4](#SD1){ref-type="supplementary-material"}). Mechanistic study of RBM10 enhanced exon skipping using minigenes ----------------------------------------------------------------- We demonstrated that RBM10 binding near splice sites of flanking introns would enhance the skipping of cassette exons. In order to test the direct effect of the RBM10 binding on pre-mRNA splicing, we fused RBM10 with a modified pumilio domain, PUF3-2, which specifically recognizes an eight nucleotide sequence 'UGUAUGUA' with high affinity ([Fig 3](#fig03){ref-type="fig"}A; Wang et al, [@b44]). By co-transfecting the RBM10-PUF fusion protein with the splicing reporter pZW2C-A6G that contains PUF cognate sequence, we could tether RBM10 to an intronic region 18-nt downstream of the middle exon ([Fig 3](#fig03){ref-type="fig"}A; Wang et al, [@b44], [@b45]). Whereas the expression of PUF domain alone hardly induced any splicing changes, the expression of RBM10-PUF fusion protein exhibited strong exon skipping effects ([Fig 3](#fig03){ref-type="fig"}B and Supporting Information [Fig S5](#SD1){ref-type="supplementary-material"}), providing unequivocal support to our hypothesis that RBM10 intronic binding in the vicinity of splicing sites would facilitate the skipping of cassette exons. ![Validation of RBM10-enhanced exon exclusion with minigene experiments\ Schematic representation of the minigene experiment, in which RBM10 was tethered to an intronic region 18-nt downstream of the middle cassette exon via fusing to the PUF domain that recognizes its cognate sequence (blue rectangle).Splicing changes of the cassette exons upon the OE of RBM10-PUF fusion protein or PUF alone, detected in three biological replicates. RBM10 tethered in the vicinity of intronic splice site could significantly promote exon skipping (two-tailed paired *t*-test *p*-value: 0.001832).](emmm0005-1431-f3){#fig03} Mechanistic model underlying RBM10 regulated alternative splicing ----------------------------------------------------------------- RBM10 binding close to splice sites might interfere with splicing sites recognition and/or splice sites pairing. Our observations strongly supported the model that RBM10 binding in the vicinity of splicing sites might repress the splicing of introns and delay the splicing choice; thereby facilitate the skipping of cassette exons flanked with relatively weaker splicing sites ([Fig 4](#fig04){ref-type="fig"}). Indeed, we found several lines of evidences supporting such a working model. First, a clear positive correlation between exon skipping and the retention of flanking introns could be observed based on RNA-seq results ([Fig 5](#fig05){ref-type="fig"}A). Second, the intron retention appeared to be also enhanced by RBM10 binding in the vicinity of its splice sites, especially the binding near 5′ splice sites ([Fig 5](#fig05){ref-type="fig"}B). Third, we proposed that the exon skipping was largely due to the weaker flanking splice sites. Once the splicing of flanking intron was repressed, the use of stronger distal splicing sites would be enhanced. As shown in [Fig 5](#fig05){ref-type="fig"}C, the strength of splicing sites distal to the cassette exons was generally stronger than that of those immediately flanking the exons. Finally, RBM10 binding close to downstream 3′ splice sites might also facilitate their paring with upstream 5′ splicing sites at later steps of splicing process, an effect proposed previously by Bonnal et al for RBM5 (Bonnal et al, [@b8]). Although it would be difficult to formally disentangle such effect from its repression of intron splicing, the observation that RBM10 binding at downstream 3′ splicing sites exerts in general stronger effects than that at the immediately flanking ones ([Fig 2](#fig02){ref-type="fig"}E) implicated a possible dual function of RBM10 binding at 3′ splicing sites, *i.e*. repression of intron splicing and facilitation of the splicing site pairing. ![Mechanistic model for RBM10-mediated splicing regulation\ RBM10 binding in the vicinity of splicing sites might repress the splicing of flanking introns and delay the splicing choice, thereby facilitate the skipping of cassette exons flanked with relatively weaker splice sites.](emmm0005-1431-f4){#fig04} ![Correlation between RBM10 binding, exon splicing changes, intron splicing changes and splicing site strengths\ Cumulative distribution functions (CDF) of splicing change of introns (*Z*~PIR~, percentage intron retention) upstream (left) or downstream (right) to the cassette exons that were differentially spliced upon RBM10 OE. Exon exclusion upon RBM10 OE (orange: *Z*~PSI~ ≤−2; green: *Z*~PSI~ ≤−1, PSI: percentage splicing in) is associated with higher retention of both flanking introns than background (black) and exon inclusion (blue: *Z*~PSI~ ≥1). The numbers of exons were indicated in parenthesis.Cumulative distribution functions of *Z*~PIR~ for different groups of introns with or without RBM10 binding at the 5′ and/or 3′ splice sites. The numbers of introns within different groups were printed in parenthesis.Box plots of the strengths of splicing splice sites at upstream and downstream introns flanking all cassette exons (white) or those with higher exclusion after upon RBM10 OE (orange: *Z*~PSI~ ≤−2; green: *Z*~PSI~ ≤−1).](emmm0005-1431-f5){#fig05} Functional investigation of an *in-frame* deletion of *RBM10* identified in a patient with TARP syndrome -------------------------------------------------------------------------------------------------------- Nonsense and frame-shift mutations in *RBM10* have been identified to be causative for TARP syndrome (Johnston et al, [@b21]). More recently, during a screen of \>400 index patients from families with X-linked intellectual disability (ID; Kalscheuer et al., manuscript in preparation), a deletion of 1292 nt (ChrX: 46929367--46930658 bp, UCSC genome browser hg18) in RBM10 were found in a German family ([Fig 6](#fig06){ref-type="fig"}A and Supporting Information [Fig S6](#SD1){ref-type="supplementary-material"}A). Apart from severe ID, the two patients also suffered from multiple congenital malformations and their leading pathologic phenotypes overlapped with TARP syndrome (see Supporting Information and [Table S7](#SD1){ref-type="supplementary-material"} for detailed reports of clinical findings). Based on the annotated gene structure, the deletion spanned six exons, but appeared not to disrupt the open reading frame. We performed RT-PCR and Western blot on the lymphoblast cells deriving from the patient, and demonstrated that the in-frame deletion did not induce nonsense mediated decay, which would otherwise degrade the truncated RNA and reduce protein levels (Supporting Information [Fig S6](#SD1){ref-type="supplementary-material"}B and C). The deletion removed 239 amino acids (651--889 amino acid at NP_005667), including the second zinc finger domain and a portion of the G patch domain ([Fig 6](#fig06){ref-type="fig"}B). ![In-frame deletion of RBM10 identified in a family afflicted with multi-organ malformation\ Pedigree of the family.RBM10 protein domain structure, the deletion removed the second zinc finger domain, a portion of the G patch domain and a nuclear localization signal (NLS).Splicing difference (ΔPSI) between lymphoblastoid cell lines (LCLs) derived from the patient and those from healthy controls were correlated with splicing changes induced by RBM10 KD in HEK293.Subcellular localization of wild type and mutant RBM10 detected by antibody against RBM10 (a-RBM10) or that against HA tag. Co-staining with DAPI showed that in contrast to the nuclear localization of wild type, RBM10 mutant predominantly localized in the cytoplasm. Scale bar: 10 µm.Splicing changes of the two cassette exons in HEK293 cells upon overexpression (OE) of RBM10 wild type or mutant were measured by qPCR. The exclusion and inclusion levels were normalized based on constitutive exon expression.](emmm0005-1431-f6){#fig06} To understand how this *in-frame* deletion in *RBM10* could contribute to disease phenotype, we examined in more detail the sequence of the deleted fragment and identified a potential nuclear localization signal (NLS) within the deleted sequence (821--837 aa; [Fig 6](#fig06){ref-type="fig"}B and see also Supporting Information Supplementary Methods). To determine whether this finding was functionally relevant, we compared the subcellular localization of wild type RBM10 with that of the mutant. In contrast to the nuclear localization of wild type, RBM10 mutant predominantly localized in the cytoplasm ([Fig 6](#fig06){ref-type="fig"}D), consistent with the loss of NLS. This finding suggests that the deletion might result in loss of nuclear functions of RBM10. To investigate the impact of this deletion on gene expression especially splicing pattern, we performed RNA-seq on lymphoblastoid cell lines (LCLs), derived from the patient carrying the mutation and from healthy controls, respectively. We determined the changes in gene expression and splicing pattern using the same strategy applied on HEK 293 cells (Supporting Information [Table S3](#SD1){ref-type="supplementary-material"} and S4). In total, we identified 206 and 102 exons showing enhanced inclusion and exclusion (\|*Z*~PSI~\| ≥ 3, \|ΔPSI\| ≥ 10%) in the patient derived LCLs, respectively. Intriguingly, the splicing changes observed here correlated well with changes induced by RBM10 KD in HEK293 ([Fig 6](#fig06){ref-type="fig"}C). This observation indicated that RBM10 mutant lost its function in splicing regulation, in accordance with the observed change in subcellular localization. To further validate the functional impact of the mutant, we overexpressed RBM10 in HEK 293 cells. As shown in [Fig 6](#fig06){ref-type="fig"}E, for the two cassette exons, the OE of the mutant could not induce the same splicing changes as that of the wild type. DISCUSSION ========== RBM10 has been characterized *in vitro* as an RNA-binding protein and identified as a component of spliceosome complex. However, its putative role in splicing regulation has not been established. In this study, our transcriptome-wide analysis of RBM10 binding profile as well as changes in splicing pattern induced by RBM10 perturbation provided experimental evidence supporting its role as a novel splicing regulator. Using PAR-CLIP, we identified thousands of consensus binding sites, 51% of which are located in the introns, with a significant enrichment in the vicinity of splicing sites. Using RNA-seq, we identified hundreds of exons, the splicing pattern of which was significant changed upon increasing or decreasing cellular RBM10 abundance. An RNA splicing map that associated RBM10-binding profiles with those observed splicing changes yielded a mechanistic model underlying RBM10 mediated splicing regulation. Nonsense and frame-shift mutations in RBM10 have been identified to cause TARP syndrome (Johnston et al, [@b21]). In this study, in the two male cousins with congenital multi-organ malformation, we identified an *in-frame* deletion in RBM10, which removed the NLS of the protein and thereby largely disrupted its nuclear function. Whole-mount *in situ* expression analysis of the murine Rbm10 has shown that the gene was expressed during embryonic development in a pattern consistent with the human malformations observed in TARP syndrome (Johnston et al, [@b21]). Therefore, due to the loss of function of RBM10, our patients as well as the previously reported TARP patients would suffer from molecular defects in those tissues expressing critical amount of RBM10 during development. Given that RBM10 regulates many genes and possibly different sets of target genes in different tissues, it is likely that multiple targets would contribute to the phenotype. Indeed, among the genes with splicing pattern regulated by RBM10, some have been implicated in the TARP syndrome associated anomalies (Supporting Information [Table S4](#SD1){ref-type="supplementary-material"}). For example, *DNML1* has found to be mutated in the patients with microcephaly and optic atrophy, overlapping features of TARP syndrome (Waterham et al, [@b46]). Mutations in *CEP290* could cause Joubert syndrome, a heterogenous ciliopathy characterized by cerebellar vermis hypoplasia and severe ID, two clinical findings also typical for TARP syndrome (Sayer et al, [@b38]). Another interesting gene, *CASK*, once mutated could lead to brain anomalies similar to the patients reported here (Najm et al, [@b32]). Finally, great phenotypic overlap is also noted to individuals with mutations in the *PIGN* gene, which lead to a syndromic entity characterized by hypotonia, seizures, neonatal hypotonia, lack of psychomotor development and dysmorphic features, associated with cardiac, urinary and gastrointestinal malformations (Maydan et al, [@b28]). Among the differentially expressed genes upon RBM10 perturbation, there are also candidate genes causing entities with overlapping features. For example, *ECEL1* is a gene responsible for distal arthrogryposis type 5D with similar limb anomalies observed in our patients (McMillin et al, [@b29]). Mutation in *LHB* gene could causes hypogonadism, which is also manifested in individuals with TARP syndrome (Weiss et al, [@b47]). On one hand, the leading pathologic phenotypes observed in our patients largely overlapped with TARP syndrome (see Supporting Information for the discussion of clinical findings), indicating many, if not most of the TARP associated malformations resulted from loss of RBM10 nuclear function, *i.e*. regulation of exon skipping. On the other hand, comparing with typical TARP patients, our patients are relatively milder affected. Indeed they are the eldest patients reported so far. Given that only nonsense and frame-shift mutations in RBM10 have been reported in TARP patients, it is tempting to speculate that the mutant RBM10 in our patients might retain either some residue nuclear function or other unknown functions of the protein. Notably, a large number of RBM10-RNA interactions, especially those in exons, appeared not to be directly associated with splicing regulation. Whether such interactions hold other regulatory roles awaits further investigation. More recently, in a large sequence analysis of lung adenocarcinomas, RBM10 was found to be frequently mutated and subject to recurrent nonsense, frame-shift or splice-site mutations (Imielinski et al, [@b20]). Interestingly, among the genes differentially spliced upon RBM10 perturbation in HEK293 cells, several were known factors associated with cancers (Supporting Information [Table S4](#SD1){ref-type="supplementary-material"}). Given the frequent observation of splicing deregulation in different types of cancers, it is plausible that RBM10 might also play an important role in cancers other than lung adenocarcinomas. Taken together, our study established RBM10 as an important regulator of AS, yielded a mechanistic model for RBM10-mediated splicing regulation and provided a starting point for the future functional characterization of RBM10 in different biological systems. The paper explained ------------------- PROBLEM: TARP syndrome is an X-linked inherited disorder leading to multiple organ malformation in affected males. Nonsense and frame shift mutations in *RBM10*, a gene encoding an RBPs, have been identified to cause TARP syndrome. Although the protein has been reported to associate with spliceosome complex, the exact molecular function of RBM10 is not clear. RESULTS: We combined photoactivatable-ribonucleoside-enhanced crosslinking and immunoprecipitation (PAR-CLIP) with massive parallel sequencing to identify RNA binding sites for RBM10 and observed significant RBM10-RNA interactions in the vicinity of splice sites. Using RNA-seq, we identified hundreds of splicing changes following perturbation of cellular RBM10 abundance. Integrative analyses of binding sites as well as splicing profile suggested a mechanistic model underlying RBM10-mediated splicing regulation, which could be subsequently validated by minigene experiments. Furthermore, we demonstrated the splicing defects in a TARP patient carrying an *in-frame* deletion in *RBM10*, which could be explained by disrupted function of RBM10 in splicing regulation. IMPACT: Our study for the first time established RBM10 as an important regulator of AS, presented a mechanistic model for RBM10-mediated splicing regulation and provided a molecular link to understanding a human congenital disorder. MATERIALS AND METHODS ===================== Cell lines ---------- Stable HEK 293 T-REx Flp-In cell lines inducibly expressing FLAG/HA-tagged wild type and mutant RBM10, respectively were generated and maintained as described previously (Landthaler et al, [@b23]) with minor modifications (Supporting Information Supplementary Methods). Expression of FLAG-HA-tagged RBM10 was induced with 10 ng/ml doxycycline for 16 h. PAR-CLIP -------- The cells were labelled with 100 µM 4-thiouridine (4SU) and induced with 10 ng/ml doxycycline for 16 h. PAR-CLIP was performed as described previously (Hafner et al, [@b15]) with the following modifications. Cells were lysed in high salt lysis buffer (50 mM Tris--HCl pH 7.2, 500 mM NaCl, 1% NP40, 1 mM DTT, complete protease inhibitor (Roche)). For the second RNase T1 digestion, 10 U/µl RNase T1 and 5 min incubation was used. PAR-CLIP libraries were sequenced 1 × 50 cycles using Illumina HiSeq following the standard protocol. Detailed procedures see Supporting Information Supplementary Methods. RBM10 knockdown --------------- siRNA (Applied Biosystems, s15747) against RBM10 was reverse transfected at a final concentration 20 nM with lipofectamine RNAiMAX (Invitrogen) in HEK293 T-REx Flp-In cells. Controls were treated with only transfection reagents. Cells were harvested 48 h after transfection, respectively. Total RNA was extracted using Trizol (Invitrogen) and the quality was assessed by Agilent Bioanalyser according to the manufacturer\'s instructions. The KD efficiency was assessed by qPCR and Western blot. RBM10 overexpression -------------------- Stable HEK293 T-REx Flp-In cells inducibly expressing FLAG-HA-tagged RBM10 was induced with 10 ng/ml doxycycline for 16 h. Control was treated with equal amount of medium. Total RNA was extracted using Trizol (Invitrogen) and the quality was assessed by Agilent Bioanalyser according to the manufacturer\'s instructions. The OE efficiency was assessed by qPCR and Western blot. mRNA sequencing --------------- mRNA sequencing was performed using 1 µg total RNA. Briefly, poly (A) RNA was isolated by two rounds of oligo (dT)~25~ Dynabeads (Invitrogen) purification. Purified poly (A) RNA was fragmented at 94°C for 3.5 min using 5× fragmentation buffer (200 mM Tris--acetate, pH 8.1, 500 mM KOAc, 150 mM MgOA). The fragmented RNA was precipitated and converted to first strand cDNA using random hexmer primer and Supescript II (Invitrogen), followed by second strand cDNA synthesis with *Eschericcia coli* DNA pol I (Invitrogen) and RNAse H (Invitrogen). Then the paired-end sequencing library was prepared and sequenced on Illumina HiSeq for 2 × 100 cycles following the standard protocol. qRT-PCR ------- Total RNA was treated with TURBO DNase (Ambion) following the manufacturer\'s protocol. Reverse transcription was performed using 1 µg of DNase treated total RNA, random hexamer and Superscript III reverse transcriptase (Invitrogen) according to manufacturer\'s protocol. First stranded cDNA was diluted 1:20 and 2 µl was used as template in a 20 µl qPCR reaction system. qPCR was carried out using SYBRGreen Masrermix I (Roche) on LightCycler 480 (Roche) according to manufacturer\'s instructions. All assays were performed in triplicates. For expression quantification, the average fold change was calculated by normalization to *GAPDH*. For exon inclusion or exclusion quantification, the relative ratios were calculated by normalization to corresponding constitutive exons. The sequences of all PCR primers were listed in Supporting Information [Table S5](#SD1){ref-type="supplementary-material"}. Western blot ------------ Western blot was performed as described in Supporting Information Supplementary Methods. The following antibodies were used: rabbit polyclonal anti-RBM10 (Abcam, ab26046, 1:2000), mouse monoclonal anti-HA (Covance, MMS-101P, 1:4000), mouse monoclonal anti-FLAG (Sigma, F1804, 1:4000) and rabbit polyclonal anti-GAPDH (Santa Cruz, sc-25778, 1:2000). Secondary HRP-conjugated goat anti-mouse or human IgG (Santa Cruz) was detected with SuperSignal Kit (Thermo). Minigene experiments -------------------- To validate the direct effects of RBM10 intronic binding on exon skipping, RBM10 protein was fused upstream of a modified PUF domain (PUF3-2) with specific and high affinity RNA recognition sequence (UGUAUGUA, *i.e*. A6G) as previously reported (Wang et al, [@b44]). The plasmid expressing the RBM10-PUF fusion protein was generated by overlapping PCR as previously described (Heckman & Pease, [@b16]) and then inserted into pcDNA3.1D/V5-His-TOPO expression vector (Invitrogen). Five hundred nanograms of splicing reporter (pZW2C-A6G) containing the PUF3-2 recognition sequence (Wang et al, [@b44]) was transfected alone, or cotransfected with 100 ng of RBM10-PUF or PUF expression vector, respectively into HEK293T cells in 12-well plate. After RT-PCR, the expression level of the two isoforms including or excluding the cassette exons was measured by Bioanalyser DNA 1000 chip (Agilent). All the PCR primers were listed in Supporting Information [Table S5](#SD1){ref-type="supplementary-material"}. PAR-CLIP sequencing data analysis --------------------------------- The PAR-CLIP reads were processed as described before (Lebedeva et al, [@b24]). In brief, the reads were aligned to the human genome (UCSC Genome Browser, hg19) allowing at most one mismatch, or indel of one nucleotide. Uniquely mapped reads were overlapped to define binding clusters. For each cluster, the preferred crosslinking position was defined as the site with the highest number of T to C conversions. Based on the binding clusters identified in the two biological replicates, a consensus binding cluster was defined as a pair of clusters from the two replicates, if preferred crosslinking site of one cluster from one replicate was located within the other cluster from the other replicate and vice versa. The preferred crosslinking site in the first replicate was used as that of the consensus cluster. RNA-seq data analysis --------------------- The RNA-seq reads were mapped with at most two mismatches to the human genome reference (UCSC genome browser hg19) and a set of sequences consisting of all possible junctions between the exons of each Refseq gene. The expression level of a gene was calculated as RPKM values by dividing the number of reads which could be mapped to the exons or exon--exon junctions of this gene by its cumulative exon length (in kb) and the total number of mappable reads (in million). For each of the internal exon *E*, we computed percent splicing in value, PSI = *e*~in~/(*e*~in~ + *e*~out~), where *e*~in~ denotes the number of reads which could only be mapped to *E* or exon junctions containing *E* with an overlap of at least 6 bp, and *e*~out~ denotes the number of reads which could be mapped only to exon junctions skipping *E* and overlap with both exons by at least 6 bp. The internal exons with its PSI value between 0.02 and 0.98 in at least one data set were defined as cassette exons. For each intron *I*, we computed percent intron retention value, PIR = *i*~in~/(*i*~in~ + *i*~out~), where *i*~in~ denotes the number of reads which could only be mapped to junctions between *I* and adjacent exons, and overlap with *I* by at least 6 bp, and *i*~out~ denotes the number of reads which could be mapped only to exon--exon junctions skipping *I* and overlap with both exons by at least 6 bp. To estimate the significance of the change in gene expression level, PSI or PIR upon RBM10 KD and OE, we applied a *Z*-value transformation, *i.e*. divided Δlog~2~ RPKM/ΔPSI/ΔPIR by a local standard deviation which we computed using a sliding window approach as following. After sorting the exons according to the total number of reads used for computing the RPKM/PSI/PIR values (*e.g*. in the case of PSI, it is the sum of *e*~in~ and *e*~out~ from both OE/KD and control), we calculated for each data point the standard deviation of the respective values inside a window consisting 1% genes/exons/introns. The local standard deviations were then smoothed using loess regression before we used them for calculating *Z* values. We then estimated false discovery rates using the rank product method (Breitling et al, [@b9]). For each independent replicate, the genes/exons/introns were ranked according to the respective *Z* values and the ranks obtained in the replicates were multiplied for each gene/exon/intron. The number of genes/exons/introns expected to have a given rank product by chance was estimated using random permutations of the rank lists. The enrichment and annotation of functional categories in the set of genes with significant splicing changes or differentially expressed upon RBM10 perturbation was computed using the Database for Annotation, Visualization and Integrated Discovery (Huang et al, [@b18], [@b19]). This work was supported by the Federal Ministry for Education and Research (BMBF) and the Senate of Berlin by funds aimed at establishing the Berlin Institute of Medical Systems Biology (BIMSB) (315362A, 315362C). YW was funded by China Scholarship Council (CSC). HH and VMK were funded by project GENCODYS (241995) in the European Union Framework Programme 7. We thank Mirjam Feldkamp, Claudia Langnick from BIMSB, Melanie Bienek and Susanne Freier from MPIMG for their excellent technical assistance. We thank Gabriele Gillessen-Kaesbach from University of Luebeck for providing clinical information of the patient. We thank Zefeng Wang from UNC for help with minigene constructs. Supporting Information is available at EMBO Molecular Medicine online. The authors declare that they have no conflict of interest. Author contributions ==================== YH and WC conceived and supervised the project. YW performed the experiments. AGD analysed the data. HH and VK performed mutation screening. SF, JM, MJ performed PAR-CLIP data analysis. YM, CQ and YaW helped with minigene experiments. YaM and ML help with PAR-CLIP experiments. DW provided the clinical expertise. YW, AGD, DW, YH contributed to part of the manuscript. WC wrote the paper. For more information ==================== The raw sequencing data has been deposited in NCBI GEO database (Accession number: GSE44976). The PAR-CLIP data could also be accessed at doRiNA database (Anders et al, [@b2]): <http://dorina.mdc-berlin.de/rbp_browser/dorina.html> Supplementary material ====================== As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. [^1]: These authors contributed equally to this work.

{ "pile_set_name": "PubMed Central" }

The authors confirm that all data underlying the findings are fully available without restriction. All relevant data are within the paper and its Supporting Information files. Introduction {#s1} ============ The thymus plays a crucial role in the development and maturation of T lymphocytes. Inside the organ, thymocytes (T cells in the thymus) are provided with a wealthy microenvironment consisting of cytokines, chemokines and cell-cell interactions that ultimately leads to proliferation, T-cell receptor gene rearrangements and thymocyte differentiation into mature T cells [@pone.0110739-Gameiro1]. However, the thymus undergoes weight and cellular loss caused by ageing and environmental disturbances, such as infections and malnutrition [@pone.0110739-Savino1]. We have previously shown that infection with *Plasmodium berghei* NK65 infection, the non-cerebral malaria pathological agent, renders the thymus atrophic through the enhanced thymocyte death by apoptosis and premature egress of CD4^+^CD8^+^ (Double-positive, DP) T cells to the periphery [@pone.0110739-Andrade1]--[@pone.0110739-Gameiro2]. It is already known that some viral and bacterial infections can promote the development of autoimmunity by inducing the breakdown of T cell tolerance and development of effector T cells reactive with the self-antigens or by the phenomenon called molecular mimicry, where a foreign antigen shares sequence or structural similarities with self-antigens [@pone.0110739-Gilden1], [@pone.0110739-Herrmann1]. For instance, acute rheumatic fever, where antibodies attack the heart, can occur after the body makes immune responses against Group A β-hemolytic streptococci [@pone.0110739-Kawakita1], [@pone.0110739-Read1]. In addition, it has been proposed that the prematurely egressed DP-T cells observed during *Trypanosoma cruzi* infection play an important role in the autoimmune cardio-inflammation [@pone.0110739-MendesdaCruz1]. Experimental Autoimmune Encephalomyelitis is a T cell-driven inflammation of the Central Nervous System (CNS) that presents similar characteristics to human Multiple Sclerosis [@pone.0110739-Mix1]. In this model, following an inflammatory stimulus containing neuro-peptides T cells migrate from the peripheral immune system towards the CNS where they promote inflammation through the release of inflammatory mediators such as cytokines and chemokines [@pone.0110739-Dittel1], [@pone.0110739-Fletcher1]. Cells from the Th1 and Th17 subsets are important for disease establishment, as evidenced by previous reports [@pone.0110739-Chu1]--[@pone.0110739-Peron1]. Both in the human and animal diseases T cells play a major role. Therefore, changes in the subpopulations of T cells influence the outcome and susceptibility to autoimmune development. In this context, we aimed to evaluate whether the previous infection with *Plasmodium berghei* NK65 would interfere with the clinical course of Experimental Autoimmune Encephalomyelitis, a mouse model for human Multiple Sclerosis (MS). We observed that EAE-susceptible mice cured from malaria developed an aggravated form of EAE, with increased infiltration of DP-T cells in the Central Nervous System (CNS). Further analyses showed that thymic-prematurely egressed DP-T cells were important for the enhanced clinical manifestation of the disease. To our knowledge, this is the first study to demonstrate the possible integration between malaria and EAE through the contribution of the thymus. Materials and Methods {#s2} ===================== Animals {#s2a} ------- Six- to eight-week-old female C57BL/6 mice from the Multidisciplinary Center for Biological Research, University of Campinas, were used in this study. Mice were kept in specific-pathogen free conditions, in a controlled temperature and photoperiod environment, with free access to autoclaved food and water throughout the experiment. All protocols involving laboratory animals were approved and performed in accordance with the guidelines of the Institutional Committee on the Use and Care of Animals (CEUA, \#2687--1). Infection and treatment {#s2b} ----------------------- For these experiments, we used the NK65 strain of *Plasmodium berghei*, because this parasite do not accumulate in the CNS of mice to cause cerebral malaria. Mice (n = 6 mice/group) were intra-peritoneally injected with 10^6^ *Plasmodium berghei*-infected red blood cells (iRBCs) obtained from a source mouse. The frequency of iRBCs was assessed daily by examination with Giemsa-stained thin blood smear. Ten days after infection, animals of each group started the treatment with chloroquine (5 mg.kg^−1^, via i.p. for five consecutive days). Three days after the last dose of the drug, mice were immunized for the induction of EAE. EAE induction and evaluation {#s2c} ---------------------------- EAE was induced and evaluated in mice according to a previous published paper [@pone.0110739-Thome1]. Briefly, each mouse was subcutaneously injected with 100 µg MOG~35--55~ (MEVGWYRSPFSRVVHLYRNGK, Genemed Syn, USA) emulsified with Complete Freunds Adjuvant (CFA, Sigma-Aldrich, USA). 200 ηg Pertussis toxin (Ptx, Sigma-Aldrich, USA) was administrated via i.p. at 0 and 48 h after MOG~35--55~ inoculation. Clinical signs were followed and graded daily according to a score method, where 0: no sign, 1: flaccid tail, 2: hind limbs weakness, 3: hind limbs paralysis, 4: hind paralysis and fore limbs weakness, 5: full paralysis/dead. Disease severity was evaluated for thirty days. At the end of the experiment, results were analyzed by linear regression, thinner lines indicate 95% confidence interval. Determination of the thymic index {#s2d} --------------------------------- The calculations to determine the thymic index were previously described [@pone.0110739-Andrade1]. Briefly, at the indicated time-points, the gross weight of each mouse was recorded and then the thymuses were collected and weighed. The thymic index was calculated using the formula: organ weight (g)/body weight (g) ×100. Cell sorting and flow cytometry {#s2e} ------------------------------- Spleen cells were obtained from *P.berghei*-infected mice at ten days of infection. Total splenic T lymphocytes were isolated using Dynabeads following the manufactureŕs instructions (Mouse Pan T cell isolation kit, Life Technologies, USA). For DP-T cell isolation, spleen-derived single cell suspension was incubated with FITC-conjugated anti-CD4 (clone H129.19) and PE-conjugated anti-CD8a (clone 53--6.7) for 20 minutes. CD4^+^, CD8^+^ and CD4^+^CD8^+^ (double-positive) T lymphocytes from malaria-bearing mice at ten days of infection were sorted in FACSAria II cell sorter (BD Biosciences, USA). The purity of isolations was assessed and accounted for 98% pure subpopulations. For adoptive transfer experiments, 1,5×10^5^ cells were adoptively transferred (via i.v.) to each group of mice at the onset of EAE (around the tenth day after immunization). For activation experiments, sorted cells were seeded in U-bottom 96 well/plates (2×10^4^ cells/well) and incubated for 72h with lipopolysaccharide (1 ng/mL, from *E. coli*, Sigma-Aldrich, USA), *Plasmodium berghei* extracts (PbX, 50 µg/mL) or MOG~35--55~ peptide (20 µg/mL, Genscript, USA). At the end of culture period, the supernatants were collected and assayed for detection of mouse IFN-γ and IL-17 by Cytometric Bead Assay (CBA, BD Biosciences, USA). In some set of experiments, total T cells isolated from naïve and malaria-bearing mice were transferred (1x10^6^ cells/mouse) at the onset of EAE. Preparation of Plasmodium berghei extracts {#s2f} ------------------------------------------ The production of extracts from *P. berghei*-infected RBCs (iRBC) followed a previously published recommendation [@pone.0110739-Thome2]. Briefly, iRBC-enriched suspension was submitted to 20 cycles of freeze/thawing in liquid nitrogen and warm bath (37°C). The protein concentration was determined using the Bradford Protein Assay following the manufactureŕs instructions (Sigma-Aldrich, St Louis, MO, USA). Analysis of cellular infiltration in the CNS {#s2g} -------------------------------------------- Fourteen days after EAE induction, mice were anesthetized, perfused with ice cold PBS and half of the spinal cords were removed and stored at −80°C until use for RT-PCR assays; the remaining tissue was prepared for the enrichment of infiltrating leukocytes according to a previously described methodology and analyzed by flow cytometry [@pone.0110739-McCall1]. RT-PCR assays {#s2h} ------------- Mice were killed at the indicated time-points and frozen tissues were used for RNA extraction (Trizol reagent, Life Technologies, USA) and cDNA synthesis according to the manufactureŕs recommendations (High Capacity RNA-to-cDNA converter kit, Life Techcnologies, USA). RNA extraction from DP-T cells was carried our immediately after sorting, using an extraction kit according to the manufactureŕs instruction (RNeasy Micro kit, Qiagen USA). Expression of AIRE (Mm00477461_m1), IL-7 (Mm01295803_m1), IL-6 (Mm00446190_m1), IL-17 (Mm00439618_m1), IFNg (Mm01168134_m1), FOXP3 (Mm00475162_m1) and RAR-related orphan receptor C (RORc) (Mm01261022_m1) were analyzed in comparison to GAPDH (Mm99999915_g1, housekeeping gene) levels. RT-PCR reactions were performed using Taqman reagents according to manufactureŕs recommendations (Applied Biosystems, USA). Expression levels of genes were represented as a relative copy numbers by using the method of delta threshold (2^−ΔΔCt^). Histopathology and immunofluorescence {#s2i} ------------------------------------- At the indicated time-points mice were killed, spinal cords were removed and snap frozen; 12 µm thin slices were made in cryostat and stained with haematoxylin and eosin (H&E) for histopathological analysis. For the characterization of the infiltrating cells, the reactive sites of the slices were blocked with Phosphate-Buffer (0,1M, pH7,2)-BSA 3%. The cells were stained with FITC-conjugated rat anti-mouse CD4 (clone H129.19, at a 1:100 dilution, BD Biosciences, USA) and PE-conjugated rat anti-mouse CD8a (clone 53--6.7, at a 1:100 dilution, BD Biosciences, USA). In some set of experiments, the lumbar spinal cords were incubated with purified anti-mouse GFAP, COX-2, IL-6, NF-κB, phosphorylated iKB and iNOS. Later, secondary Cy3- or TRITC-conjugated antibodies were added. The reaction was analyzed under epifluorescence microscope (Leica, GER). *In vitro* re-stimulation and cytokine dosage {#s2j} --------------------------------------------- Splenic cells were aseptically collected from mice after 10 days of MOG~35--55~ immunization. Single cell suspensions were stained with Carboxyfluorescein succinimidyl ester (CFSE, 2,5 µM, Sigma-Aldrich, USA) following the manufactureŕs instructions. Cells (5×10^5^/well) were diluted in RPMI 1640 media supplemented with Fetal Calf Serum (FCS; 10% vol/vol), guaramicine (50 µg/mL), 2-Mercaptoethanol (2 mM) and myelin oligodendrocyte glycoprotein peptide (MOG~35--55~; 20 µg/mL), plated in U-bottom plates and incubated for 96 h. After the incubation period, cells were stained with PercPCy5-conjugated anti-CD3e, PE-conjugated anti-CD8a and PECy7-conjugated anti-CD4 antibodies and fixed in 1% paraformaldehyde prior to flow cytometer analysis. CFSE^low^ cells inside each population were considered proliferating T cells. Culture supernatants were collected and assayed for cytokines (IL-4, IL-6, IL-10, IL-17, IFN-γ and TNF-α) secretion using the Cytometric Bead Array (CBA, BD Biosciences, USA) according to manufactureŕs instructions. T cell co-cultures {#s2k} ------------------ For the co-cultures, total spleen T cells were isolated from malaria-bearing and EAE-inflicted mice after ten days of immunization. T cells derived from EAE mice were stained with CFSE according to the manufactureŕs instructions (2,5 µM, Sigma-Aldrich, USA) and seeded in U-bottom 96-well plates (5×10^5^ cells/well). T cells derived from naïve or *P.berghei*-infected mice were added afterwards at a 1∶1 proportion. MOG~35--55~ peptide was added to a final concentration of 20 µg/mL. The cultures were incubated for 96 h at 37°C. At the end of culture time, the cells were collected and stained with PercP-conjugated anti-CD3e antibodies and fixed in 1% paraformaldehyde prior to flow cytometer analysis. CFSE^low^CD3^+^ cells were considered proliferating responder T cells. Statistical analysis {#s2l} -------------------- Clinical score comparisons between control and experimental groups were done by Two-Way ANOVA and post-tested with Bonferroni. Other analyses among two and three (or more) groups were carried out with Students t test and One-Way ANOVA, respectively. Results are expressed as mean ± standard error mean (SEM) and p\<0,05 value were defined as significant. Results {#s3} ======= Exacerbation of Experimental Autoimmune Encephalomyelitis in mice cured from *P.berghei* infection {#s3a} -------------------------------------------------------------------------------------------------- The present study aimed to evaluate whether the previous infection with *P.berghei* NK65 would interfere with the clinical outcome of Experimental Autoimmune Encephalomyelitis. For that purpose, C57BL/6 mice were infected with parasitized erythrocytes and at the tenth day of infection, parasites were eliminated following chloroquine (CQ) treatment. Three days after the last dose of CQ, mice were immunized with MOG~35--55~ peptide for the induction of EAE (NK+CQ+EAE group). As a control, besides EAE-bearing mice (EAE group), we used naïve mice injected with non-parasitized erythrocytes and treated with chloroquine before EAE induction (CQ+EAE group). The results showed that mice from the NK+CQ+EAE group developed an aggravated disease course with higher clinical score when compared to EAE group ([Figure 1A](#pone-0110739-g001){ref-type="fig"}). Also noteworthy is the mild clinical scores developed by mice from CQ+EAE group. As recently published by our group, CQ reduces the severity of EAE probably by induction of regulatory T cells [@pone.0110739-Thome1]. Thus, our results also show that *P.berghei* infection was able to overcome the suppressive effect of chloroquine. ![Aggravation of EAE in mice cured from malaria correlates with increased cellular immune response towards myelin.\ C57BL/6 mice (n = 6 mice/group) were intraperitoneally (i.p.) infected with 1×10^6^ *P.berghei*-infected Red Blood Cells and treated with chloroquine (CQ, 5 mg/Kg) for five consecutive days starting at the 10^th^ day after infection. Three days after the last dose of CQ, mice were immunized with 100 µg of MOG~35--55~ peptide and Pertussis toxin was administrated (via i.p.) at 0 and 48h after peptide immunization for EAE induction. A) The clinical course of EAE was then monitored. Linear regression analyses are exposed in the side panels, thinner lines indicate 95% confidence interval. B) At the 10^th^ day after MOG-immunization, the spleens of mice were collected and dissociated. Total leukocytes (5×10^5^/well) were CFSE-stained (2,5 µM) and cultured in the presence of MOG~35--55~ (10µg/mL) peptide for 96h. At the end of culture period, the cells were surface stained with anti-CD3/CD4/CD8 antibody cocktail and events were acquired in a flow cytometer. The proliferation was analyzed inside each T cell population. C) The culture supernatants were assayed for the secreted cytokines IL-10, IL-4, IL-6, IL-17, TNF-α and IFN-γ. Data was analyzed by One-Way Anova and post-tested with Bonferroni. In all analyses, \*: p\<0,05; ns: not significant. Representative data of three independent experiments.](pone.0110739.g001){#pone-0110739-g001} The aggravation of EAE correlates with DP-T lymphocytes reactivity to neuro-peptides {#s3b} ------------------------------------------------------------------------------------ Since EAE model is highly dependent on the cellular response against neuro-antigens, the cellular immune response against MOG peptide was evaluated. Spleen cells from EAE-bearing mice were harvested ten days after neuro-peptide immunization and cultivated in the presence of MOG~35--55~. The data obtained showed that NK+CQ+EAE-derived cells proliferated significantly more when compared to cells from EAE control group ([Figure 1B](#pone-0110739-g001){ref-type="fig"}). Among the proliferating cells, the supopulation of DP-T cells had the most active proliferation, while CQ+EAE mice presented no DP-T cells in the periphery. Cells from CQ+EAE group did not present significant proliferation or the presence of DP-T cells in the spleens ([Figure 1B](#pone-0110739-g001){ref-type="fig"}). The cytokine secretion was altered as well. Cell culture supernatants from NK+CQ+EAE mice presented significant higher levels of pro-inflammatory cytokines (TNF-α, IL-17 and IFN-γ) compared with the EAE group, whereas the levels of IL-10, IL-6 and IL-4 remained similar between the two groups ([Figure 1C](#pone-0110739-g001){ref-type="fig"}). Interestingly, cell cultures from CQ+EAE mice showed significantly higher levels of IL-10 and IL-4, and lower levels of IL-17, IL-6 and TNF-α than the other two groups ([Figure 1C](#pone-0110739-g001){ref-type="fig"}). Plasmodium berghei NK65 infection promotes thymic alterations and the premature egress of CD4^+^CD8^+^ double-positive lymphocytes to the periphery {#s3c} --------------------------------------------------------------------------------------------------------------------------------------------------- The results presented so far show that the exacerbated clinical score from NK+CQ+EAE mice correlates with an enhanced cellular immune response towards neuro-antigens orchestrated, partly, by DP-T cells. The presence of DP-T cells in the spleens of NK+CQ+EAE mice is interesting. It was previously shown that during *P.berghei* infection, the thymus of BALB/c mice undergoes structural and phenotypical changes that together with the uncontrolled parasitemia culminates in death of the host fourteen days after infection [@pone.0110739-Andrade1]. To investigate whether the infection with *P.berghei* would promote similar alterations in the spleens and thymuses of a different mouse strain, C57BL/6 (B6) mice were injected with *Plasmodium berghei*-infected Red Blood Cells. In this strain, the infection is controlled when chloroquine (5 mg.kg^−1^) is administrated for five consecutive days starting at day 10 after infection ([Figure 2A](#pone-0110739-g002){ref-type="fig"}). Similarly as in the BALB/c model, the thymus of B6 mice is reduced with loss of the thymic index, starting at the third day of infection ([Figure 2B](#pone-0110739-g002){ref-type="fig"}). Interestingly, the expression of the Autoimmune Regulator (AIRE) gene was found up-regulated compared with control mice as well as the expression of IL-7 and IL-6 ([Figure 2C](#pone-0110739-g002){ref-type="fig"}). The gene expression within the thymus of B6 mice infected with other *Plasmodium* species, such as *P.yoelli* and *P.chabaudi* was also evaluated and the results showed that the gene expression of cytokines and AIRE differs between groups ([Figure S1](#pone.0110739.s001){ref-type="supplementary-material"}). ![*P. berghei* infection provokes thymic alterations and the premature egress of double-positive T lymphocytes in the spleens.\ C57BL/6 mice (n = 6 mice/group) were intraperitoneally (i.p.) infected with 1x10^6^ *P. berghei*-infected Red Blood Cells and treated with chloroquine (CQ, 5mg/Kg) for five consecutive days starting at the 10^th^ day after infection. A) Mice treated with CQ showed decrease in the parasite burden. B) The thymuses were collected at different time-points and the relative body weight was determined. C) The gene expression of AIRE, IL-7 and IL-6 was assessed in the thymuses of *P. berghei* NK65-infected mice at the fifteenth day of infection. Flow cytometry analysis of the frequency (in D) and absolute numbers (in E) of T lymphocytes subpopulations (CD4^+^ and CD8^+^) in the thymuses and spleens. Data was analyzed by Student\'s t test. In all analyses, \*: p\<0,05 and \*\*: p\<0,01. Ns: not significant. \#: p\<0,05 in comparison with untreated mice. Representative data of four independent experiments.](pone.0110739.g002){#pone-0110739-g002} It is already clear nowadays that disarranged thymic structure associated with an inflamed microenvironment result in altered maturation of thymocytes [@pone.0110739-Savino2]. Accordingly, the thymocyte subpopulations were altered in the thymus of *P.berghei*-infected mice, with reduced numbers of CD4^+^CD8^+^ (Double-positive, DP) and increased frequency of CD4^+^ and CD8^+^ thymocytes ([Figure 2D](#pone-0110739-g002){ref-type="fig"}). In the spleen, results showed an increased frequency of DP-T lymphocytes as well as CD4^+^ and CD8^+^ T cells. Naïve mice treated with chloroquine showed no alterations regarding the frequency of DP-T cells in the spleen and in thymic T cell subpopulations in comparison to mice without treatment (*data not shown*). Thus, it is possible to suppose that the responding DP-T cells observed in the NK+CQ+EAE mice was elicited during *P.berghei* NK65 infection and that these cells are related to the exacerbated EAE outcome. Infiltration of inflammatory cells in the central nervous system is characterized by the presence of DP-T cells and the local production of inflammatory cytokines {#s3d} ------------------------------------------------------------------------------------------------------------------------------------------------------------------ To investigate the possible association of DP-T cells with the exacerbation of EAE, the frequency of these cells in the Central Nervous System was evaluated. Fourteen days after EAE induction, mice were killed and spinal cords were prepared for histological analyses. The phenotype of the infiltrating T cells was determined by direct immunofluorescence labeling technique using FITC-conjugated anti-CD4 and PE-conjugated anti-CD8 antibodies. Results showed that in EAE group, among the infiltrating cells, a high frequency of CD4^+^ T cells was observed, while no infiltration was observed in the CNS of CQ+EAE mice ([Figure 3](#pone-0110739-g003){ref-type="fig"}). As expected, an elevated frequency of CD4^+^CD8^+^ (DP) T cells among the infiltrating cells in the CNS of NK+CQ+EAE mice was identified ([Figure 3](#pone-0110739-g003){ref-type="fig"}). These data indicate that the double-positive T cells that prematurely migrated from the thymus during *P.berghei* infection are antigen-responsive ([Figure 1B](#pone-0110739-g001){ref-type="fig"}) and capable to migrate to the CNS during neuro-inflammation ([Figure 3](#pone-0110739-g003){ref-type="fig"}). In addition, the CNS of NK+CQ+EAE mice showed an increased production of IL-6, NF-κB, phosphorylated iκB (iκBα-p), iNOS, COX-2 and GFAP compared with EAE group ([Figure S2](#pone.0110739.s002){ref-type="supplementary-material"}). ![Central Nervous System of malaria-cured EAE mice show increased cellular infiltration of DP-T cells.\ C57BL/6 mice (n = 6 mice/group) were intraperitoneally (i.p.) infected with 1×10^6^ *P.berghei*-infected Red Blood Cells and treated with chloroquine (CQ, 5 mg/Kg) for five consecutive days starting at the 10^th^ day after infection. Three days after the last dose of CQ, EAE was induced. As controls, naïve mice were treated with CQ or vehicle before EAE induction. The spinal cords of EAE-inflicted mice were collected fourteen days after MOG-immunization. Frozen thin sections (12 µm) were made and fixed in formalin. Cells were stained with FITC-conjugated anti-CD4 and PE-conjugated anti-CD8 and analyzed in epifluorescence microscope. Figures are representative of three independent experiments. Magnification: 200X.](pone.0110739.g003){#pone-0110739-g003} In the CNS, an increased gene expression of IL-17 in tissue from NK+CQ+EAE mice was observed in comparison with mice from EAE group ([Figure 4A](#pone-0110739-g004){ref-type="fig"}). In contrast, the expression of Foxp3 and IL-10 was not significantly altered in comparison with the naïve group ([Figure 4A](#pone-0110739-g004){ref-type="fig"}). IFN-γ expression was not significantly altered. To investigate whether DP-T cells found in the CNS tissue were able to produce inflammatory mediators, mice were killed at the fourteenth day of EAE and the infiltrating cells were enriched from the CNS. The intracellular staining results showed that NK+CQ+EAE mice had higher frequency of IFN-γ- and IL-17-producing DP-T cells than EAE mice. ([Figure 4B](#pone-0110739-g004){ref-type="fig"}). ![Inflammation in the CNS of NK+CQ+EAE mice correlates with an increased production of inflammatory cytokines by DP-T cells.\ Groups of mice (n = 6 mice/group) subjected to infection and EAE induction. A) At the 10^th^ day after MOG-immunization, mice were killed and spinal cords were removed to analyze the gene expression of IL-17, IFN-γ, Foxp3 and IL-10 in the lumbar spinal cords of mice. Data was analyzed by One-Way Anova and post-tested with Bonferroni. B) The infiltrating cells of the CNS were enriched and stimulated by Phorbol Myristate Acetate and Ionomycin in the presence of Brefeldin A for 4 h. The frequency of IFN-γ- and IL-17-producing cells inside CD4^+^CD8^+^ T cell gate was analyzed. In all analyses, \*: p\<0,05. ns: not significant. Representative data of three independent experiments.](pone.0110739.g004){#pone-0110739-g004} Transfer of lymphocytes from *P.berghei*-infected mice increases the severity of EAE {#s3e} ------------------------------------------------------------------------------------ In order to better characterize the relationship among malaria-elicited DP-T lymphocytes and the exacerbation of EAE, a series of experiments to demonstrate that the mechanism of disease aggravation is T-cell dependent was conducted. First, splenic T cells isolated from *P.berghei*-infected or naïve mice were co-cultured with splenic T cells from EAE-bearing mice in the presence of MOG~35--55~ peptide for four days. Results showed that the cultures conducted in the presence of T cells from malaria-infected mice proliferated significantly more than cells from co-cultures with naïve T cells ([Figure 5A](#pone-0110739-g005){ref-type="fig"}). ![Exacerbation of EAE in mice transferred with total T from naïve mouse or DP-T cells from malaria-cured mouse.\ A) C57BL/6 mice (n = 6) were intraperitoneally (i.p.) infected with 1×10^6^ *P.berghei*-infected Red Blood Cells. At the 10^th^ day of infection, splenocytes were collected and total T cells were isolated using dynabeads (Pan T cell isolation kit). Naïve-derived splenic T cells were used as well. EAE-inflicted mice (n = 6) were killed at the 10^th^ day after immunization and the total splenic T cell were isolated with the same methodology and CFSE-stained (2,5 µM). As controls, T cells from EAE-inflicted mice were cultured without the presence of other cells. B) Total T cells isolated from naïve and malaria-bearing mice were transferred (1×10^6^ cells/mouse) at the onset of EAE (n = 6 mice/group), and clinical course of the disease was evaluated daily. Data was analyzed by Two-Way ANOVA and post-tested with Bonferroni. C) Sorted splenic CD4^+^, CD8^+^ and CD4^+^CD8^+^ T cells from malaria-bearing mice were adoptively transferred (1,5×10^5^ cells/mouse) to EAE-inflicted mice (n = 6 mice/group) at disease onset. The clinical score of the disease was evaluated daily. The side panel contains the linear regression lines with the 95% confidence interval (thinner lines). D) Malaria-bearing mice were killed at 10 days after infection and the DP-T cells were sorted and incubated with LPS (1 ng/mL), *P. berghei* extracts (50 µg/mL) and MOG~35-55~ peptide (20 µg/mL) for 72 h, at the end of culture period the supernatants were removed and assayed for the detection of IFN-γ and IL-17. Analyses were conducted using Student\'s t test. \*: p\<0,05. \*\*: p\<0,01. Representative data of two independent experiments with similar results.](pone.0110739.g005){#pone-0110739-g005} Then, it was investigated whether the adoptive transfer of splenic T cells from malaria mice would alter the course of EAE. For that purpose, B6 mice were infected with *Plasmodium berghei* NK65 and the total splenic T cells were isolated ten days after infection. These cells were adoptively transferred to EAE mice at the beginning of the clinical signs (around the 10^th^ day after immunization). As expected, the clinical course of EAE was significantly aggravated in malaria T cells-transferred mice compared to naïve T cell-recipient mice ([Figure 5B](#pone-0110739-g005){ref-type="fig"}). Lastly, since the infiltrating T cells in the CNS of NK+CQ+EAE mice comprised mostly of DP-T cells, the next goal was to evaluate whether these cells play a role in the aggravation of EAE. Therefore, sorted splenic CD4^+^, CD8^+^ and DP-T cells from malaria-infected mice were intravenously transferred to EAE mice as soon as the clinical signs of EAE started to appear (around the 10^th^ day after MOG~35--55~ immunization). The results showed that mice that received DP-T cells developed a more severe disease than mice that received CD4^+^ and CD8^+^ T cells ([Figure 5C](#pone-0110739-g005){ref-type="fig"}). In addition, we aimed to evaluate whether the cellular response elicited by DP-T cells were antigen-specific or were raised in a by-stander fashion. For that purpose, we sorted DP-T cells from the spleens of naïve and *P. berghei*-infected mice and cultivated the cells in the presence of LPS, *P. berghei* extracts and MOG peptide. At the end of the culture time, the supernatants were assayed for the presence of inflammatory cytokines. As depicted in [figure 5D](#pone-0110739-g005){ref-type="fig"}, cultures conducted in the presence of LPS and PbX produced elevated levels of IFN-γ and IL-17 in comparison with DP-T cells from naïve mice. Interestingly, the presence of MOG was able to stimulate the production of IL-17, but not IFN-γ, in DP-T cell cultures. This data shows that the prematurely egressed DP-T cells from *P. berghei*-infected mice are able to produce inflammatory cytokines in the presence of PAMPs (LPS), specific stimulus (PbX) and auto-antigens (MOG). Discussion {#s4} ========== In this study, we show that, after infection with *Plasmodium berghei* NK 65, MOG-immunized mice develop an aggravated form of Experimental Autoimmune Encephalomyelitis. Further analysis demonstrated that the thymic prematurely egressed CD4^+^CD8^+^ (double-positive) T cells play an important role in the enhanced severity of the disease. The thymus is a key organ for the development of T lymphocytes and plays a major role in the elimination of auto-reactive T cells [@pone.0110739-Asano1], [@pone.0110739-Starr1]. T cell precursors enter the thymus through the high endothelial venules found in the cortico-medullary junction and lack the expression of antigen receptor and co-receptors CD4 and CD8 [@pone.0110739-Lind1]. The maturation process consists of both migration of thymocytes through the cortex towards the medulla and expression of T cell receptor and co-receptors [@pone.0110739-Anderson1]. During this process, immature T cells interact with a myriad of cytokines, chemokines and self-peptides presented in association with MHC molecules on thymic epithelial cells (TECs) and cells that recognize antigens with high avidity are deleted [@pone.0110739-Savino2]. We have previously shown that the lethal *Plasmodium berghei* NK65 infection promotes thymic atrophy and that this phenomenon is dependent on thymocyte death and its premature egress to the periphery [@pone.0110739-Andrade1]--[@pone.0110739-Gameiro2]. The lymphocytes subpopulations was altered in the thymus as well, with increased frequency of simple positive CD4^+^ and CD8^+^ T cells. However, due to the loss of thymic mass, the absolute numbers of cells were reduced. In the spleens, the augmentation in the numbers of CD4 and CD8 T cells may be explained by the clonal expansion observed in the course of malaria [@pone.0110739-Falanga1]--[@pone.0110739-Sardinha1]. In addition, we observed an increased expression of AIRE, IL-6 and IL-7 in thymuses from *P.berghei*-infected mice, which indicate that an altered repertoire selection might be occurring. An impaired T cell maturation and selection in the thymus may lead to the release of self-reactive T cells to the periphery and the sub-sequential development of autoimmune disorders [@pone.0110739-Capalbo1]--[@pone.0110739-Itoh1]. Indeed, we observed that mice, which were cured from *P.berghei* infection, developed a more severe form of EAE compared with uninfected animals. When we analyzed the inflammatory infiltration in the CNS, we found that most of the infiltrating cells were double-positive T cells. These DP-T cells were functional as they produced high levels of IL-17, an inflammatory cytokine related to EAE worsening [@pone.0110739-Chu1], [@pone.0110739-OConnor1], [@pone.0110739-Willenborg1]. As expected, when we adoptively transferred total T cells or sorted DP-T cells derived from malaria-bearing mice towards EAE-inflicted mice, the clinical score arose significantly. Taken together, these set of experiments demonstrated that the DP-T cells elicited in *P.berghei*-infected mice are inflammatory and exacerbate autoimmune neuro-inflammation. Co-cultures showed that malaria-elicited T cells stimulated the proliferation of T cells from MOG-immunized mice. Together with the proliferative response observed in DP-T cells from NK+CQ+EAE, these data suggest that DP-T cells, originated in the course of malaria, exacerbate EAE by, at least, two different mechanisms: (i) DP-T cells self-react and proliferate against neuro-antigens and (ii) DP-T cells stimulate the proliferation of encephalitogenic T cells. Of note, when cultured in the presence of different stimuli, DP-T cells in the spleens produced inflammatory cytokines IFN-γ and IL-17 at high levels. These observations indicate that DP-T cells elicited in *P. berghei* infection respond to Pathogen-Associated Molecular Patterns and specific antigens as well as to unrelated auto-antigens. Interestingly, infection overcame the suppressive effect of chloroquine. We previously observed that CQ treatment reduces the severity of Experimental Autoimmune Encephalomyelitis through the expansion of regulatory T cells [@pone.0110739-Thome1]. Later, our results showed that CQ is able to modulate dendritic cells *in vitro* directing them towards a tolerogenic profile that can be used to treat EAE [@pone.0110739-Thome3]. The mechanisms by which *P. berghei* surpasses the CQ suppressive effect remains to be elucidated. Still, the recently described effect of *Plasmodium* infection in Toll-like receptors (TLR) priming may explain, at least in part, the aggravation of inflammation [@pone.0110739-Hartgers1], [@pone.0110739-McCall1]. The data presented here might be concerning as malaria remains the worlds most prevalent infectious disease, with over 250 million people infected [@pone.0110739-Bhaumik1]. Several reports have correlated autoimmunity prevalence in countries that have eradicated the *Plasmodium* infection [@pone.0110739-Sotgiu1]. Although the hygiene hypothesis supports the idea of infection-induced autoimmune resistance, this theory may not fully apply to malaria. There is an increasing amount of evidence supporting or rejecting the role of malaria in autoimmune disease both in human and animal models [@pone.0110739-Butcher1]--[@pone.0110739-Adu1]. Similarly, it was previously proposed that in *Trypanosoma cruzi* infection, the inflammatory profile of prematurely egressed DP-T cells from the thymus may be involved in the autoimmune process observed in murine and human Chagas\' disease [@pone.0110739-MendesdaCruz1]. It was demonstrated that *Plasmodium chabaudi* infection increases the frequency of regulatory T cells in spleens, and these cells suppressed the development of EAE in a by-stander fashion [@pone.0110739-Farias1]. However, *P. berghei* infection was shown to induce the secretion of anti-DNA and anti-nuclear antibodies in a T cell-dependent manner [@pone.0110739-Poels1]. These observations indicate that different *Plasmodium* species induce distinct types of immune responses. Indeed, we observed that the gene expression of inflammatory cytokines and AIRE within the thymus vary depending on the type of *Plasmodium* species. Nonetheless, infected Red Blood Cells from *P.falciparum*, *P.yoelli*, *P.berghei* and *P.vivax* induce distinct maturation status of dendritic cells (DCs) [@pone.0110739-Thome2], [@pone.0110739-Bettiol1]--[@pone.0110739-Wykes1]. Indeed, we have shown that DCs modulated with plasmodium extracts are able to reduce the severity of EAE through the suppression of inflammatory responses in the Central Nervous System [@pone.0110739-Thome2]. Interestingly, it was recently observed that *Plasmodium chabaudi*-infected mice presented exacerbated response to bacterial challenge [@pone.0110739-Ataide1]. In this study, the authors showed that following infection, mice were hyper-responsive to *E. coli*-derived LPS in a mechanism that is dependent on NLRP12/NLRP3 inflammasome activation. The results presented by the authors are in line with others [@pone.0110739-Hartgers1], [@pone.0110739-McCall1], [@pone.0110739-Franklin1]. In our model, although we did not investigate the role of inflammasomes, we found that the T cell compartment is severely altered in *P. berghei*-infected mice resulting in an exacerbation of autoimmune responses. Still, further studies must be conducted in order to ascertain the possible interplay between innate cells-derived inflammasome and T cells in the hyper-responsive inflammation elicited by *P. berghei* infection. Based on our observations, we hypothesized that during malaria infection, thymocytes prematurely egress from the thymus towards the peripheral immune system. In the periphery, due to a trigger event (which can be genetic susceptibility, infection, or radiation, for instance) these cells proliferate and migrate to the target organ, where they produce high amounts of inflammatory cytokines leading to exacerbated inflammation ([Figure 6](#pone-0110739-g006){ref-type="fig"}). ![Hypothesis model for the exacerbation of autoimmune neuro-inflammation.\ Based in our observations, we propose a model for EAE exacerbation. Malaria infection promotes thymic atrophy and the premature egress of DP-T cells to the peripheral immune system. After an inflammatory trigger, which can be infection, genetic susceptibility or chronic inflammation, these cells proliferate and migrate to the target organ where they stimulate CNS inflammation by secreting cytokines. However, there is still much to be explored, as for example, whether these DP-T cells are able to induce leukocyte recruitment, microglia and astrocyte activation, and, Blood-Brain Barrier (BBB) destruction.](pone.0110739.g006){#pone-0110739-g006} Further studies must be conducted in order to evaluate how these DP-T cells are activated and whether they are antigen-specific or not. Although an increasing amount of evidences show the presence of extra-thymic DP-T cells in the control of inflammation [@pone.0110739-Bang1]--[@pone.0110739-Tutaj1], our data clearly shows that in malaria infection, the thymic-derived DP-T cells are inflammatory. To our knowledge, this is the first study that shows the interplay between malaria infection and autoimmune neuro-inflammation through the contribution of the thymus-derived double-positive T cells. Supporting Information {#s5} ====================== ###### **Analysis of the gene expression in the thymus of mice infected with distinct** ***Plasmodium*** **species.** C57BL/6 mice were intraperitoneally (i.p.) infected with 1×10^6^ infected Red Blood Cells. Mice were infected with *P.berghei* NK65, *P.chabaudi* AS and *P.yoelli* CL. The gene expression of AIRE, FOXP3, Rorγt, IL-17, IL-7 and IL-6 was evaluated. Results showed that the different *Plasmodium* species triggers distinct gene expression. Data was analyzed by One-Way ANOVA and post-tested with Bonferroni, where \*: p\<0,05. Representative data of four independent experiments. (TIF) ###### Click here for additional data file. ###### **Analysis of the inflammation in the CNS of malaria-cured EAE mice.** C57BL/6 mice were intraperitoneally (i.p.) infected with 1×10^6^ *P.berghei*-infected Red Blood Cells and treated with chloroquine (CQ, 5 mg/Kg) for five consecutive days starting at the 10^th^ day after infection. Three days after the last dose of CQ, mice were immunized with 100 µg of MOG~35--55~ peptide and Pertussis toxin was administrated (via i.p.) at 0 and 48 h after peptide immunization for EAE induction. The spinal cords of EAE mice were collected fourteen days after immunization. Frozen thin sections (12 µm) were made and fixed in formalin. Cells were stained with purified anti-mouse iNOS (in A), IL-6 (in B), COX-2 (in C), NF-κB (in D), phosphorylated iκBα (in E) and GFAP (in F). DAPI was added to stain DNA (blue color). The slices were analyzed in epifluorescence microscope. Figures are representative of three independent experiments. Magnification: 400X. (TIF) ###### Click here for additional data file. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: RT FTMC MACH ALRO LV. Performed the experiments: RT ALB LKI CR SCPL TAC RDG ITF ALFL. Analyzed the data: RT MACH FTMC LV. Contributed reagents/materials/analysis tools: LV FTMC MACH ALRO. Wrote the paper: RT LV.

{ "pile_set_name": "PubMed Central" }

All relevant data are within the paper and its Supporting Information files. Introduction {#sec001} ============ A high body mass index is a major risk factor for the incidence of numerous non-communicable diseases (NCD), such as cardiovascular and kidney diseases, diabetes and some cancers \[[@pone.0182333.ref001]--[@pone.0182333.ref006]\]. Indeed, obesity has been recognised as a major public health challenge for the 21^st^ century \[[@pone.0182333.ref007]\], with concerns regarding the health and economic burden, which have led to the identification of global targets to halt the rise in obesity prevalence by 2025 \[[@pone.0182333.ref008], [@pone.0182333.ref009]\]. However, recent figures from the NCD Risk Factor Collaboration, which analysed the aggregated data of 19.2 million participants from 200 countries, suggest that if current, post-2000, trends continue, the probability of meeting these targets is almost zero. These findings highlight the critical need to develop and implement effective interventions for the prevention and treatment of excess body mass \[[@pone.0182333.ref010]\]. A fundamental principle in the development of such interventions is the assessment of the energy expenditure associated with activities of daily living; weight management strategies are most effective when individuals can accurately determine how much energy they have expended \[[@pone.0182333.ref011]\]. Indeed, a failure to assess energy expenditure appropriately may, at least in part, explain the inconsistent evidence regarding intervention effectiveness and sustainability. Walking represents a popular, convenient and relatively safe form of activity that can easily be incorporated into weight management programmes \[[@pone.0182333.ref012]--[@pone.0182333.ref014]\]. The energy expenditure associated with walking are reported to be either linearly or slightly exponentially related to speed \[[@pone.0182333.ref015]\]. However, the applicability of these findings is based on walking in a straight line which does not align with everyday activities. In particular, recent studies have suggested that the process of changing direction is associated with significant additional energy expenditure \[[@pone.0182333.ref016]--[@pone.0182333.ref018]\]. Wilson et al. \[[@pone.0182333.ref018]\] reported that the energy expenditure of turning is linearly related to the degree of turning angle at 6 km∙hr^-1^ while Hatamoto et al. found quadratic \[[@pone.0182333.ref017]\] or linear \[[@pone.0182333.ref016]\] functions best represented the relationship between running speed and the energy expenditure of a 180° turn. A failure to account for this additional energy expenditure of turning has significant implications for both sporting and health contexts. For example, most English Premier league footballers make more than 700 turns per match \[[@pone.0182333.ref019]\] while medical treatment effectiveness is often assessed using a six-minute walking test. While the latter is intended to be conducted over a standardised 30 m straight line distance \[[@pone.0182333.ref020]\], some studies have used 20 m or 50 m straights \[[@pone.0182333.ref021], [@pone.0182333.ref022]\], significantly influencing the number of turns completed and thus potentially confounding inter-study comparisons. Indeed, the difference in the number of turns completed, and thus overall energy expenditure, may explain studies, which utilised shorter straights reporting significantly shorter distances covered \[[@pone.0182333.ref023]--[@pone.0182333.ref025]\]. The purpose of the present study was therefore to investigate the influence of walking speed, angle, and their interaction, on energy expenditure. We hypothesised that 1) as walking speed increased, so would energy expenditure; 2) as angle increased, so would energy expenditure and that 3) angle and speed would demonstrate a synergistic effect on energy expenditure while walking. Materials and methods {#sec002} ===================== Participants {#sec003} ------------ In total, 20 healthy adults (7 female, 13 male; 28 ± 7 yrs; 20.5 ± 4.1 kg∙m^2^) were recruited for the study. The participants were all recreationally active but none were highly trained. Prior to testing, participants were informed of the protocol and risks and provided written consent. All procedures were approved by a Swansea University ethics committee and were conducted in accordance with the Declaration of Helsinki. Participants were asked to arrive at the laboratory in a rested state, at least two hours postprandial and to avoid strenuous exercise in the 24 hours preceding each testing session. Participants were also asked to refrain from caffeine and alcohol for 6 and 24 h before each test, respectively. All tests were performed at the same time of day (± 2 h). Experimental design {#sec004} ------------------- Participants were required to visit the laboratory/indoor track on three occasions, separated by at least 24 hours recovery. Participants initially completed an incremental treadmill test for determination of peak oxygen uptake ($\overset{˙}{V}\text{O}_{2}{}_{\text{peak}}$) and the gas exchange threshold (GET). On each of the two subsequent visits, participants completed the turning protocol. Incremental treadmill test {#sec005} -------------------------- Following a three-minute warm-up at 6 km·hr^-1^, the treadmill speed increased by 1km·hr^-1^ every minute until participants reached volitional exhaustion. The treadmill gradient was set at 1% throughout the test \[[@pone.0182333.ref026]\], until participants reached their maximal running speed at which point it subsequently increased by 1% every minute until volitional exhaustion. Participants were given strong verbal encouragement throughout the test. Turning protocol {#sec006} ---------------- On subsequent visits to the indoor track, each participant was asked to complete repeated three-minute bouts of walking interspersed by three minutes of rest. In a randomised order, each participant walked at four different walking velocities (2.5, 3.5, 4.5, 5.5 km·hr^-1^) in combination with four different angles (0, 45, 90, 180°). Specifically, each of the sixteen conditions involved 5 m of straight walking interspaced by prescribed turns with the speed dictated by a digital, auditory metronome, which sounded once half-way between turns and once on the turns. Each condition incorporated an equal number of left- and right-handed turns, as illustrated in [Fig 1](#pone.0182333.g001){ref-type="fig"}. ![Schematic of the turning protocol showing 5m straights interspersed by prescribed turns of a) 0° b) 45° c) 90° and d) 180° with equal left and right hand turns.](pone.0182333.g001){#pone.0182333.g001} Measurements {#sec007} ------------ Throughout all the tests, gas exchange variables (MetaMax Cortex 3B, CORTEX Biophysik GmbH, Germany) were measured on a breath-by-breath basis and displayed online. Prior to each test, the gas analysers were calibrated using gases of known concentration and the turbine volume transducer was calibrated using a 3-litre syringe (Hans Rudolph, Kansas City, MO). The delay in the capillary gas transit and analyser rise time were accounted for relative to the volume signal, thereby time-aligning the concentration and volume signals. Additionally, two combined tri-axial accelerometers and tri-axial magnetometers (SLAM Tracker, Wildbyte Technologies Ltd, Swansea, UK), measuring at 60 Hz, were worn by participants; one set was worn on the right mid-axilla line at the level of the iliac crest and one set at the middle of the lower back. Data analysis {#sec008} ------------- The peak $\overset{˙}{V}\text{O}_{2}$ was defined as the highest 10 s stationary average during the incremental treadmill test. The GET was determined by the V-slope method \[[@pone.0182333.ref027]\] as the point at which carbon dioxide production began to increase disproportionately to $\overset{˙}{V}\text{O}_{2}$, as identified using purpose written software developed using LabVIEW (National Instruments, Newbury, UK). The mean $\overset{˙}{V}\text{O}_{2}$ during each condition was taken as the first 45 seconds of the final minute of that bout. Subsequent analyses were based on the premise that the energy expenditure of turning was superimposed on the baseline energy expenditure of straight line travel. Thus, the difference in $\overset{˙}{V}\text{O}_{2}$ during straight line walking (0°) at each speed compared to the $\overset{˙}{V}\text{O}_{2}$ engendered during walking with 45, 90 or 180° turns, was attributed to the additional energy expenditure of turning. This $\overset{˙}{V}\text{O}_{2}$ was converted to gross energy expenditure in kJ using a conversion factor of 20.1 J per ml of oxygen and subsequently divided by the total number of turns per condition to provide an estimate of the energy expenditure of each angle and speed combination. The raw accelerometer data were converted to dynamic body acceleration (DBA) by first smoothing each channel to derive the static acceleration using a running mean over 2 s \[[@pone.0182333.ref028]\] and then subtracting this static acceleration from the raw data \[[@pone.0182333.ref029]\]. The resulting values for dynamic acceleration were all then converted to positive values. These values for DBA were then vectorially summed to give: $$VeDBA = \ \sqrt{\left( {A_{x}^{2} + A_{y}^{2} + A_{z}^{2}} \right)}$$ Where VeDBA is the vectorial dynamic body acceleration, A~x~, A~y~, and A~z~ are the derived dynamic accelerations at any point in time corresponding to the three orthogonal axes of the accelerometer \[[@pone.0182333.ref030]\]. Mean and summed VeDBA were derived for each individual turn and straight during the middle minute of the each condition and for the overall three minute bout. Individual turns and straights were determined using custom designed C++ software (DDMT Wildbyte Technologies Ltd) written specifically for the SLAM Tracker devices, to visualise the accelerometry and magnetometry traces and identify the point at which each trace significantly deviated from the local mean. Statistics {#sec009} ---------- Gaussian distributions in data were confirmed by Shapiro-Wilks tests. To account for the repeated and correlated nature of the data, linear mixed-effects models were used to determine the influence of, and interaction between, walking speed and turn angle on energy expenditure and *VeDBA* ([S1 File](#pone.0182333.s001){ref-type="supplementary-material"}). All condition combinations were placed in one model with covariates (sex, stature, peak $\overset{˙}{V}\text{O}_{2}$ and/or turning $\overset{˙}{V}\text{O}_{2}$ for *VeDBA*) added to subsequent adjusted models to determine their modulatory effect. Pearson product moment correlation coefficients were used to analyse the degree of association between key variables. Statistical analyses were conducted using PASW Statistics 21 (SPSS, Chicago, IL). All data are presented as means ± SD. Statistical significance was accepted when P≤0.05. Results {#sec010} ======= Descriptive characteristics of the sample population are shown in [Table 1](#pone.0182333.t001){ref-type="table"}. The male participants were significantly taller and demonstrated a higher peak $\overset{˙}{V}\text{O}_{2}$, both in absolute and relative terms (normalised per kg body mass). 10.1371/journal.pone.0182333.t001 ###### Participant characteristics. ![](pone.0182333.t001){#pone.0182333.t001g} Total Male Female -------------------------------------------------------------------------------- ------------- ------------- -------------------------------------------------- **n** 10 13 7 **Age, yrs** 28.0 ± 6.7 28.5 ± 7.5 27.2 ± 5.3 **Stature, m** 1.74 ± 0.09 1.78 ± 0.08 1.67 ± 0.02[\*](#t001fn002){ref-type="table-fn"} **Body mass, kg** 72.1 ± 16.1 74.5 ± 16.1 67.6 ± 16.4 **BMI, kg∙m**^**2**^ 20.6 ± 4.1 20.8 ± 3.9 20.3 ± 4.9 **Peak $\overset{˙}{V}\mathbf{O}_{2}$, l∙min**^**-1**^ 3.54 ± 0.75 3.92 ± 0.64 2.85 ± 0.33[\*](#t001fn002){ref-type="table-fn"} **Relative peak $\overset{˙}{V}\mathbf{O}_{2}$, l∙kg**^**-1**^**∙min**^**-1**^ 49.8 ± 7.9 53.2 ± 5.5 43.5 ± 8.1[\*](#t001fn002){ref-type="table-fn"} **GET, l∙min**^**-1**^ 2.31 ± 0.66 2.52 ± 0.71 1.92 ± 0.34[\*](#t001fn002){ref-type="table-fn"} **GET, %** 65 ± 9 64 ± 10 67 ± 7 Mean ± SD. BMI, body mass index; $\overset{˙}{V}\text{O}_{2}$, oxygen uptake, GET, gas exchange threshold. \* Significant difference between males and females P \< 0.05 As shown in [Table 2](#pone.0182333.t002){ref-type="table"}, there was a significant main effect for speed (F = 121.609, *P* \< 0.001) and turn angle (F = 19.186, *P* \< 0.001) on $\overset{˙}{V}\text{O}_{2}$ and a significant interaction between these parameters (F = 4.433, *P* \< 0.001). Specifically, as speed increased, $\overset{˙}{V}\text{O}_{2}$ increased, but significant increases in $\overset{˙}{V}\text{O}_{2}$ relative to straight line walking were only observed for 90° and 180° turns at the two highest speeds (4.5 and 5.5 km∙hr^-1^; [Table 2](#pone.0182333.t002){ref-type="table"}). Males demonstrated a significantly greater $\overset{˙}{V}\text{O}_{2}$ across all conditions (F = 25.322, *P* \< 0.001), although this difference was reversed when stature was included in the model (Sex: F = 16.77, *P* \< 0.001; Stature: F = 152.493, *P* \< 0.001). $\overset{˙}{V}\text{O}_{2}$ during the turning protocol was dependent on peak $\overset{˙}{V}\text{O}_{2}$ (F = 100.970, *P* \< 0.001) but once scaled to account for differences in body size, this relationship was no longer significant (F = 0.708, *P* \> 0.05). 10.1371/journal.pone.0182333.t002 ###### Mean energy expenditure and *VeDBA* during each combination of walking speed and angle. ![](pone.0182333.t002){#pone.0182333.t002g} $\overset{˙}{V}O_{2}$ (l∙min^-1^) Mean VeDBA (g) ----------------------- -------------------------------------------------- -------------------------------------------------- ----------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------- -------------------------------------------------- -------------------------------------------------- -------------------------------------------------- -------------------------------------------------- **2.5 km∙hr**^**-1**^ 0.66 ± 0.20 0.63 ± 0.19 0.63 ± 0.20 0.67 ± 0.22 0.19 ± 0.03 0.19 ± 0.02 0.19 ± 0.04 0.20 ± 0.03 **3.5 km∙hr**^**-1**^ 0.73 ± 0.20[\*](#t002fn002){ref-type="table-fn"} 0.74 ± 0.22[\*](#t002fn002){ref-type="table-fn"} 0.74 ± 0.21[\*](#t002fn002){ref-type="table-fn"} 0.82 ± 0.24[\*](#t002fn002){ref-type="table-fn"} 0.24 ± 0.04[\*](#t002fn002){ref-type="table-fn"} 0.25 ± 0.05[\*](#t002fn002){ref-type="table-fn"} 0.25 ± 0.04[\*](#t002fn002){ref-type="table-fn"} 0.27 ± 0.04[\*](#t002fn002){ref-type="table-fn"} **4.5 km∙hr**^**-1**^ 0.86 ± 0.26[\*](#t002fn002){ref-type="table-fn"} 0.88 ± 0.26[\*](#t002fn002){ref-type="table-fn"} 0.93 ± 0.24[\*](#t002fn002){ref-type="table-fn"}[^\#^](#t002fn003){ref-type="table-fn"} 1.10 ± 0.33[\*](#t002fn002){ref-type="table-fn"}[^\#^](#t002fn003){ref-type="table-fn"} 0.32 ± 0.03[\*](#t002fn002){ref-type="table-fn"} 0.32 ± 0.04[\*](#t002fn002){ref-type="table-fn"} 0.32 ± 0.05[\*](#t002fn002){ref-type="table-fn"} 0.37 ± 0.05[\*](#t002fn002){ref-type="table-fn"} **5.5 km∙hr**^**-1**^ 1.00 ± 0.26[\*](#t002fn002){ref-type="table-fn"} 1.00 ± 0.33[\*](#t002fn002){ref-type="table-fn"} 1.14 ± 0.30[\*](#t002fn002){ref-type="table-fn"}[^\#^](#t002fn003){ref-type="table-fn"} 1.54 ± 0.36[\*](#t002fn002){ref-type="table-fn"}[^\#^](#t002fn003){ref-type="table-fn"} 0.44 ± 0.05[\*](#t002fn002){ref-type="table-fn"} 0.44 ± 0.05[\*](#t002fn002){ref-type="table-fn"} 0.46 ± 0.06[\*](#t002fn002){ref-type="table-fn"} 0.53 ± 0.08[\*](#t002fn002){ref-type="table-fn"} Mean ± SD. $\overset{˙}{V}\text{O}_{2}$, net oxygen uptake; *VeDBA*, vectorial dynamic body acceleration. \* indicates significant difference to 2.5 km∙hr^-1^ within angle (*P*\<0.05) ^\#^ indicates significant difference to straight walking within speed (*P*\<0.05) The estimated energy expenditure associated with an individual turn is represented in [Fig 2](#pone.0182333.g002){ref-type="fig"}, showing the synergistic interaction between speed and turn angle in determining the energy expenditure. ![Estimated energetic cost of individual turns at each combination of walking speed and turning angle.\ \* indicates a significant difference in the energy expenditure of turning compared to 45°. Values shown are the difference in energy expenditure between each condition and straight walking.](pone.0182333.g002){#pone.0182333.g002} In accord with $\overset{˙}{V}\text{O}_{2}$, there was a significant main effect for speed and angle on mean *VeDBA* during the straights (speed: F = 1058.734, *P* \< 0.001; angle: F = 43.416, *P* \< 0.001) and corners (speed: F = 996.671, *P* \< 0.001; angle: F = 12.534, *P* \< 0.001), with a significant interaction effect between speed and angle (straight: F = 4.497, *P* \< 0.001; corners: F = 3.053, *P* \< 0.01). Specifically, while both straight and corner VeDBA increased with speed, the effect of angle was only manifest at 4.5 and 5.5 km∙hr^-1^ ([Table 3](#pone.0182333.t003){ref-type="table"}). For straight sections, VeDBA was dependent on sex (F = 15.781, *P* \< 0.001) and peak $\overset{˙}{V}\text{O}_{2}$ (F = 7.437, *P* \< 0.01) whilst stature was not a significant contributor (F = 0.038, *P* \> 0.5). In contrast, corner VeDBA was significantly influenced by sex (F = 6.723, *P* \< 0.01) and stature (F = 4.507, *P* \< 0.05), but not peak $\overset{˙}{V}\text{O}_{2}$ (F = 0.243, *P* \> 0.05). 10.1371/journal.pone.0182333.t003 ###### Mean straight and corner *VeDBA* during each combination of walking speed and angle. ![](pone.0182333.t003){#pone.0182333.t003g} Straight Mean *VeDBA (g)* Corner Mean *VeDBA (g)* ----------------------- -------------------------------------------------- -------------------------------------------------- ----------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------- -------------------------------------------------- -------------------------------------------------- ----------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------- **2.5 km∙hr**^**-1**^ 0.19 ± 0.03 0.18 ± 0.03 0.19 ± 0.04 0.20 ± 0.03 0.18 ± 0.04 0.18 ± 0.03 0.18 ± 0.04 0.18 ± 0.03 **3.5 km∙hr**^**-1**^ 0.24 ± 0.04[\*](#t003fn002){ref-type="table-fn"} 0.25 ± 0.05[\*](#t003fn002){ref-type="table-fn"} 0.25 ± 0.04[\*](#t003fn002){ref-type="table-fn"} 0.28 ± 0.05[\*](#t003fn002){ref-type="table-fn"} 0.23 ± 0.04[\*](#t003fn002){ref-type="table-fn"} 0.25 ± 0.06[\*](#t003fn002){ref-type="table-fn"} 0.24 ± 0.05[\*](#t003fn002){ref-type="table-fn"} 0.24 ± 0.03[\*](#t003fn002){ref-type="table-fn"} **4.5 km∙hr**^**-1**^ 0.32 ± 0.04[\*](#t003fn002){ref-type="table-fn"} 0.32 ± 0.04[\*](#t003fn002){ref-type="table-fn"} 0.33 ± 0.04[\*](#t003fn002){ref-type="table-fn"}[^\#^](#t003fn003){ref-type="table-fn"} 0.39 ± 0.04[\*](#t003fn002){ref-type="table-fn"}[^\#^](#t003fn003){ref-type="table-fn"} 0.30 ± 0.05[\*](#t003fn002){ref-type="table-fn"} 0.33 ± 0.04[\*](#t003fn002){ref-type="table-fn"} 0.32 ± 0.04[\*](#t003fn002){ref-type="table-fn"}[^\#^](#t003fn003){ref-type="table-fn"} 0.34 ± 0.05[\*](#t003fn002){ref-type="table-fn"}[^\#^](#t003fn003){ref-type="table-fn"} **5.5 km∙hr**^**-1**^ 0.44 ± 0.05[\*](#t003fn002){ref-type="table-fn"} 0.44 ± 0.05[\*](#t003fn002){ref-type="table-fn"} 0.46 ± 0.06[\*](#t003fn002){ref-type="table-fn"}[^\#^](#t003fn003){ref-type="table-fn"} 0.55 ± 0.09\*[^\#^](#t003fn003){ref-type="table-fn"} 0.42 ± 0.06[\*](#t003fn002){ref-type="table-fn"} 0.44 ± 0.05[\*](#t003fn002){ref-type="table-fn"} 0.45 ± 0.06[\*](#t003fn002){ref-type="table-fn"}[^\#^](#t003fn003){ref-type="table-fn"} 0.49 ± 0.07[\*](#t003fn002){ref-type="table-fn"}[^\#^](#t003fn003){ref-type="table-fn"} Mean ± SD. VeDBA, vectorial dynamic body acceleration. \* indicates significant difference to 2.5 km∙hr^-1^ within angle (*P*\<0.05) ^\#^ indicates significant difference to straight walking within speed (*P*\<0.05) Overall $\overset{˙}{V}\text{O}_{2}$ during each condition was significantly correlated with straight (r^2^ = 0.61; *P* \< 0.001), corner (r^2^ = 0.58; *P* \< 0.001) and total mean *VeDBA* (r^2^ = 0.61; *P* \< 0.001), demonstrating a weaker but statistically significant relationship with peak $\overset{˙}{V}\text{O}_{2}$ (r^2^ = 0.36; *P* \< 0.001) and stature (r^2^ = 0.32; *P* \< 0.001). Discussion {#sec011} ========== This is the first study to investigate the interaction between speed and turn angle in determining the energy expenditure associated with walking. In agreement with one hypothesis, as speed increased for any given turning angle, the associated energy expenditure similarly increased. However, whether angle comprised a significant additional energy expenditure was dependent on the degree of turn angle. Specifically, irrespective of speed, 45° turns did not significantly increase energy expenditure, whilst 180° turns were always associated with a greater energy expenditure than straight line walking. Speed and angle demonstrated a significant interaction; 90° turns were only associated with significantly increased energy expenditure relative to straight line walking at 4.5 and 5.5 km∙hr^-1^. This synergistic interaction was further supported by the exponential relationship found to best represent the relationship between speed and angle \[[@pone.0182333.ref015]\]. These findings therefore highlight the importance of accounting for the quantity and magnitude of turn completed when estimating energy expenditure, particularly at higher speed and angles. In recent years there has been increasing recognition of the physiological demands engendered by turning 180° when running. Dellal et al. \[[@pone.0182333.ref031]\] reported a greater heart rate, blood lactate and ratings of perceived exertion (RPE) during intermittent shuttle runs involving 180° turns compared to straight line running at the same average running velocities, subsequently confirmed by Buchheit et al. \[[@pone.0182333.ref032]\]. Furthermore, Bekraoui et al. \[[@pone.0182333.ref033]\] found that covering the same distance at the same average speed resulted in a significantly greater physiological response when the course was 3.5m compared to 7.0m. These earlier findings were recently extended by Hatamoto et al. \[[@pone.0182333.ref017]\] who found that, even at running speeds as low as 3 km∙hr^-1^, thirty 180° turns per minute elicited a similar metabolic demand as straight line running at 6 km∙hr^-1^. In the present study, a significant increase in total energy expenditure relative to straight line walking was not observed at 2.5 km∙hr^-1^, but was observed at 3.5 km∙hr^-1^. Whilst these findings are largely in accord with those of Hatamoto et al. \[[@pone.0182333.ref017]\], it is pertinent to note certain methodological discrepancies, such as the training status of the sample population and turning frequencies utilised, which limit inter-study comparisons. Specifically, there were considerable differences in the number of turns completed, with Hatamoto et al. \[[@pone.0182333.ref017]\] utilising up to 30 turns per minute compared to the 35 turns in 3 minutes at 3.5 km∙hr^-1^ used in the present study. The greater energy expenditure associated with turning whilst walking is likely to be primarily attributable to the deceleration and subsequent acceleration required to make a turn, both of which necessitate eccentric and concentric muscle contractions, respectively \[[@pone.0182333.ref034]\]. Acceleration has been shown to engender a greater energy expenditure than travelling at a constant speed, with the energy expenditure dictated by the rate of acceleration \[[@pone.0182333.ref035]\]. A high acceleration rate requires a high degree of horizontal propulsion \[[@pone.0182333.ref036]\], therefore the change in acceleration is greater when performing a 180° turn at higher running velocities, thereby resulting in greater energy expenditure. The extent of the angle turned has also been shown to alter the biomechanical properties in running; a 90° turn exerts a significantly higher vertical, braking and propelling force than a 45° turn \[[@pone.0182333.ref037]\]. It would therefore be postulated that greater angles would also be associated with further increases in directional forces and thus energy expenditure during walking. In accord with this hypothesis, a linear relationship has been suggested between angle and energy expenditure when walking at 6 km∙hr^-1^ \[[@pone.0182333.ref018]\]. However, the present study suggests a synergistic interaction between speed and angle, with the influence of increasing angle within a speed only evident at 4.5 km∙hr^-1^ and above. This discrepancy may be attributable to differences in the walking velocities, the specific techniques used to turn, stature or training status \[[@pone.0182333.ref032], [@pone.0182333.ref038]\]. Indeed, both stature and peak $\overset{˙}{V}\text{O}_{2}$, an indicator of aerobic fitness and training status, were significant predictors of energy expenditure in the present study. Hatamoto et al. \[[@pone.0182333.ref016]\] previously suggested that ball game players, who are likely to be mainly running rather than walking and who turn more frequently anyway, were likely to have a more efficient turning technique. However, the mean $\overset{˙}{V}\text{O}_{2}$ of an individual turn was reported to be 0.34 ± 0.13 ml∙kg^-1^ and 0.55 ± 0.09 ml∙kg^-1^ at 4.3 km∙hr^-1^ and 5.4 km∙hr^-1^, respectively \[[@pone.0182333.ref016]\]. These values are substantially more than the values observed in the present study ([Fig 2](#pone.0182333.g002){ref-type="fig"}: 4.5 km∙hr^-1^ = 0.07 ± 0.03 ml∙kg^-1^; 5.5 km∙hr^-1^ = 0.13 ± 0.07 ml∙kg^-1^), despite the less trained status of the present participants. The reason for this discrepancy and its contradiction to the postulated role of aerobic fitness and technique are presently unclear, although it is perhaps pertinent to note the different methods of calculating the energy expenditure of an individual turn and the recent findings of Zagatto et al. \[[@pone.0182333.ref039]\] who found a lower metabolic power to be associated with more frequent changes of direction. It is interesting to note the apparent dissociation between *VeDBA* and turning angle in the present study, whereby increasing the angle of the turn was not associated with any significant increase in *VeDBA*. This could be attributable to the short duration of the turns, although the high measurement resolution makes this unlikely, or measurement error associated with the use of magnetometry to isolate the turn. However, whilst the magnitude of change in the signal was decreased at lower turn angles, this is unlikely to entirely explain the present findings. Rather, this finding may largely be attributable to the complex and individual-specific interaction between the surge, heave and sway components of *DBA* as well as muscular effort that involves generation of high forces without the dynamism typical of straight-line travel. Indeed, recent studies using force plates to investigate turn kinetics suggest that during a turn, the surge (inline) component of *DBA* is accompanied by a sway (perpendicular) component (Griffiths et al., In press). Furthermore, the surge component tends to 'average' zero over the straight sections (equal deceleration and acceleration phases) but during a turn section, the surge component becomes negative on average to provide the deceleration required to enter and execute the turn. In addition, the heave component (vertical) component of *DBA* may increase above and beyond normative walking values but this may depend on the turning technique being employed, e.g. some participants may elect to turn using a 'stop and reverse direction' method while others may prefer a 'gradual cornering' approach. The authors are of the opinion that this is by far the most likely explanation for the lack of sensitivity of *VeDBA* to turns. The present findings have significant implications within both sporting and health contexts given that few sporting, fitness or functional activities occur in a strictly linear fashion \[[@pone.0182333.ref037]\]. Indeed, whilst the present study only considered walking and caution should be exercised when extrapolating the findings to speeds associated with running and team sports, it is perhaps pertinent to note the similarity between the current findings and those reported elsewhere. Specifically, Dellal et al. \[[@pone.0182333.ref031]\] reported a greater heart rate, blood lactate and ratings of perceived exertion (RPE) during intermittent shuttle runs involving 180° turns compared to straight line running at the same average running velocities, subsequently confirmed by Buchheit et al. \[[@pone.0182333.ref032]\]. Furthermore, Bekraoui et al. \[[@pone.0182333.ref033]\] found that covering the same distance at the same average speed resulted in a significantly greater physiological response when the course was 3.5m compared to 7.0m. These earlier findings were recently extended by Hatamoto et al. \[[@pone.0182333.ref017]\] who found that, even at running speeds as low as 3 km∙hr^-1^, thirty 180° turns per minute elicited a similar metabolic demand as straight line running at 6 km∙hr^-1^. In the present study, a significant increase in total energy expenditure relative to straight line walking was not observed at 2.5 km∙hr^-1^, but was observed at 3.5 km∙hr^-1^. Whilst these findings are largely in accord with those of Hatamoto et al. \[[@pone.0182333.ref017]\], certain methodological differences should be considered, such as the training status of the sample population and turning frequencies utilised, which limit inter-study comparisons. Specifically, there were considerable differences in the number of turns completed, with Hatamoto et al. \[[@pone.0182333.ref017]\] utilising up to 30 turns per minute compared to the 35 turns in 3 minutes at 3.5 km∙hr^-1^ used in the present study. However, not all studies have found a significant influence of turning on energy expenditure, with Zamparo et al. \[[@pone.0182333.ref040]\] reporting no change in $\overset{˙}{V}\text{O}_{2}$ with increasing turn angle from 0 to 180°. This discrepancy may be related to the use of maximal running velocity during this study, thereby minimising the potential for further increases in $\overset{˙}{V}\text{O}_{2}$ to be elicited with increasing turn angle. Nonetheless, we would concur with Hatamoto et al. \[[@pone.0182333.ref016]\] that the energy expenditure associated with turning should be considered when estimating total energy expenditure during a football game in which more than 700 turns are typically completed per match \[[@pone.0182333.ref019]\]. From a health perspective, one important application of the present findings is in the design and interpretation of physical activity interventions. For example, the majority of energy expenditure prediction algorithms based on accelerometry data are derived from treadmill exercise. Such linear modes of locomotion are not cognisant of the additional metabolic costs associated with turning and this may, to some extent, contribute to the poor accuracy associated with the derived models during free-living conditions \[[@pone.0182333.ref041], [@pone.0182333.ref042]\]. Such inaccuracies are likely to be emphasised in certain populations, such as children, who are characterised by highly sporadic movements \[[@pone.0182333.ref043], [@pone.0182333.ref044]\]. Furthermore, accounting for the energy expenditure of turning could also be important in the evaluation of clinical trial effectiveness. Whilst the six-minute walking test is designed to be conducted over a 30m, straight line course with a 180° turn \[[@pone.0182333.ref020]\], reported distances covered range from 20 to 50 m \[[@pone.0182333.ref021], [@pone.0182333.ref022]\] due to space and resource limitations. Such discrepancies, using reference values reported by Chetta et al. \[[@pone.0182333.ref045]\] could result in the number of turns ranging from 12 to 32, which, according to the present data, may be associated with an additional $\overset{˙}{V}\text{O}_{2}$ expenditure ranging from 118 ml∙min^-1^ to 296 ml∙min^-1^. Swank et al. \[[@pone.0182333.ref046]\] demonstrated that a 6% improvement in peak $\overset{˙}{V}\text{O}_{2}$ was associated with a 5% decrease in risk of all-cause mortality in Congestive Heart Failure patients. Given the significantly lower peak aerobic capacity in patients, discrepancies arisen from failing to account for the energy expenditure of turning, which could be as much as 20% of a patients peak $\overset{˙}{V}\text{O}_{2}$, would considerably alter the interpretation of intervention efficacy. Future studies should seek to generate algorithms that account for distance and turns completed during a six-minute walk test, facilitating standardisation between centres. There are certain limitations associated with the current study that should be acknowledged, such as the walking velocities utilised. Previous studies have employed higher running speeds, whereas we employed speeds more typical of habitual physical activity. Whilst this increased the generalisability of our findings to health contexts, caution should be taken when extrapolating these findings to a sporting context. Furthermore, although a strength of the study to optimise interpretation of our results, the controlled nature of the protocol limits ecological validity. Finally, although the walking speeds were associated with a moderate intensity of exercise for most of the participants, some may not have achieved a steady state $\overset{˙}{V}\text{O}_{2}$ within the 3-minute bout, thereby influencing the mean $\overset{˙}{V}\text{O}_{2}$ observed. In conclusion, the present study demonstrated a synergistic interaction between speed and angle in determining the energy expenditure associated with walking. Specifically, 90° and 180° turns are associated with significant additional metabolic costs at 4.5 km∙hr^-1^ and above. These findings therefore highlight the importance of accounting for the quantity and magnitude of turns completed when estimating energy expenditure and have significant implications within both sport and health contexts. Supporting information {#sec012} ====================== ###### Minimum required dataset. (XLSX) ###### Click here for additional data file. [^1]: **Competing Interests:**The authors have declared that no competing interests exist.

{ "pile_set_name": "PubMed Central" }

Introduction {#sec1} ============ Selective serotonin reuptake inhibitors (SSRIs), including fluoxetine (Flx), are the most commonly prescribed antidepressants given their improved side effect profile compared with other major classes of antidepressants \[[@B1]\]. However, it is important to note that treatment with SSRIs such as Flx is known to exert potent anxiogenic behavioral effects, in particular during the acute phase of treatment, an observation that is also noted in animal models \[[@B2],[@B3]\]. The acute effects of Flx treatment differ from those noted following chronic administration, namely increased anxiety in the acute phase of treatment and a shift to anxiolytic effects observed on sustained treatment \[[@B3]\]. Furthermore, acute Flx is known to evoke enhanced activity of the stress-responsive, hypothalamus--pituitary--adrenocortical (HPA) axis leading to elevated circulating corticosterone levels, whereas sustained Flx administration is reported to normalize HPA axis dysfunction in animal models of depression and in clinical studies \[[@B4],[@B5]\]. The acute phase of Flx treatment through induction of a 'hyperserotonergic' state \[[@B6]\] may overlap to a certain degree with the nature of state induced by acute stress, which is known to activate the HPA axis, enhance serotonin release, and induce anxiety-like behavior. The contribution of specific serotonergic receptors to the acute anxiogenic effects of Flx treatment remains at present unclear, though prior studies have implicated the 5-HT~2~ receptor subtypes in contributing to the anxiogenic effects of acute SSRI treatment \[[@B7]\]. The G~q/11~-coupled serotonin~2A~ (5-HT~2A~) and serotonin~2C~ (5-HT~2C~)receptor are highly expressed in limbic neurocircuitry, and have been strongly implicated in the regulation of anxiety-like behavior \[[@B9]\]. Studies have raised the possibility that the acute anxiogenic effects of SSRIs, may involve a contribution of the 5-HT~2C~ receptor in the bed nucleus of the stria terminalis (BNST) \[[@B6]\] and basolateral nucleus of the amygdala (BLA) \[[@B8]\], with a down-regulation of the 5-HT~2C~ receptor suggested to mediate a shift toward anxiolysis following sustained SSRI administration \[[@B12]\]. Adjunct administration of the 5-HT~2A~ receptor selective antagonist (*R*)-(+)-α-(2,3-Dimethoxyphenyl)-1-\[2-(4-fluorophenyl)ethyl\]-4-piperinemethanol (MDL100907) along with SSRIs is reported to improve SSRI efficacy on specific behavioral tasks \[[@B13]\]. However, the contribution of the 5-HT~2A~ receptor to the acute anxiogenic effects of SSRIs still remains unclear. Here, we sought to examine the contribution of the 5-HT~2A~ receptor to the acute effects of the SSRI, Flx, on anxiety-like behavior and circulating corticosterone levels, using 5-HT~2A~ receptor knockout mice (5-HT~2A~^−/−^) \[[@B10]\]. Further, we addressed the degree of neural activation evoked by acute Flx in the paraventricular nucleus (PVN) of the hypothalamus and the prefrontal cortex (PFC) in 5-HT~2A~^−/−^ and WT mice, and gene expression of several activity-regulated, immediate early genes (IEGs) in the PFC. 5-HT~2A~^−/−^ mice have been previously reported to exhibit a baseline anxiolytic phenotype \[[@B10],[@B14]\], higher firing rates of dorsal raphe 5-hydroxytryptamine or serotonin (5-HT) neurones \[[@B15]\] and also to display treatment resistance to chronic Flx administration \[[@B16]\]. Our results reveal that the anxiogenic effects of acute Flx treatment, as well as the enhanced serum corticosterone levels are unaltered in 5-HT~2A~^−/−^ mice of both sexes. However, 5-HT~2A~ receptor deficiency does alter the pattern of neural activation within the PVN following acute Flx, and the baseline expression of several IEGs in the PFC in a sexually dimorphic manner. Experimental procedures {#sec2} ======================= Animals and drug treatment {#sec2-1} -------------------------- Serotonin~2A~ receptor (5-HT~2A~) knockout mice (5-HT~2A~^−/−^) \[[@B10]\] and wild-type (WT) littermate controls of both sexes (4--7 months) were maintained on a 129S6/SvEv background, and group housed on a 12-h normal light--dark cycle, with access to food and water *ad libitum* in the Tata Institute of Fundamental Research (TIFR) animal house facility. Genotypes were confirmed using PCR analysis as described previously \[[@B14]\]. All experimental procedures followed the guidelines of the Committee for Supervision and Care of Experimental Animals (CPCSEA), Government of India, and were approved by the TIFR Institutional Animal Ethics committee in accordance with the National Institutes of Health guide for the care and use of Laboratory animals (NIH Publication No. 8023, revised 1978). The experimental groups were as follows for both sexes: WT + saline (WT Sal), WT + Flx (WT Flx), 5-HT~2A~^−/−^ + saline (5-HT~2A~^−/−^ Sal) and 5-HT~2A~^−/−^ + Flx (5-HT~2A~^−/−^ Flx). Different cohorts were used for each behavioral (open-field test (OFT): males: *n*=10/group; females: *n*=9--10/group; elevated plus-maze test (EPM): males: *n*=11--15/group; females: *n*=12--14/group), cellular (FBJ murine osteosarcoma viral oncogene homolog peptide (c-Fos) profiling) (males: *n*=4--6/group; females: *n*=5--6/group) and molecular (quantitative real-time PCR (qPCR)) (males: *n*=8--12/group; females: *n*=7--10/group) characterization. Serum corticosterone analysis was performed on the same cohort used for cellular analysis (males: *n*=3--6/group; females: *n*=5--6/group). Mice were administered a single intraperitoneal injection of Flx (15 mg/kg, Sigma, U.S.A.) or vehicle (saline), and were subjected to behavioral testing or killed for gene expression analysis 2 h later, or serum corticosterone quantitation and immunohistochemistry analysis after 30 min. OFT and EPM {#sec2-2} ----------- Animals were assessed for anxiety-like response on the OFT as described previously \[[@B14]\]. Briefly, mice were placed in one corner of the arena and allowed to explore the arena for 10 min under dim light conditions. Behavioral tracking was done using Panlab SMART video tracking software (SMART 3.0) and percent distance in center, percent time in center, entries to center and total distances traversed were assessed. A separate cohort of mice was assessed for anxiety-like behavioral responses on the EPM consisting of adjacent open and closed arms for a duration of 10 min. The mice were placed at the intersection of the four arms facing the open arm, and allowed to explore the arena under dim light conditions. The movement was tracked using an overhead camera and Ethovision 3.1 (Noldus, The Netherlands) video tracking software. The total distance traversed in the EPM arena, percent distance traversed in the open arms, percent time spent in the open arms, and number of entries to the open arms were analyzed. Serum corticosterone assay {#sec2-3} -------------------------- Blood was collected by cardiac puncture 30 min post a single Flx (15 mg/kg) or saline injection from 5-HT~2A~^−/−^ and WT littermate control mice of both sexes. Serum was collected by transferring the supernatant into a fresh tube and stored at −80°C till further analysis. ELISA was performed to measure serum corticosterone levels using a commercially available ELISA kit (Abcam, Catalog number ab108821) as per the manual (male: *n*=3--6/group; female: *n*=5--6/group). c-Fos immunohistochemistry {#sec2-4} -------------------------- Animals were perfused with saline, followed by 4% paraformaldehyde (PFA) solution and brains were dissected out and stored in 4% PFA. Brains were sectioned on a vibratome (Leica VT1000 S, U.S.A.) at 40-μm thickness. Free-floating sections spanning the PVN of the hypothalamus (−0.58 to −1.22 mm from Bregma, five to six sections per animal) and PFC (2.46--1.34 mm from Bregma, five to six sections per animal) were subjected to processing for c-Fos immunohistochemistry. Briefly, sections were blocked in 10% horse serum, followed by overnight incubation with rabbit-anti-c-Fos antibody (1:500, Catalog number CST-2250S, Cell Signaling Technology). Sections were washed and incubated with a biotinylated horse-anti-rabbit IgG secondary antibody (1:500, Catalog number BA-1100, Vector Laboratories). Signal amplification and visualization involved exposure to an avidin--biotin complex solution (ABC solution, Vector Laboratories) followed by 3,3′-Diaminobenzidine (DAB) substrate (Sigma, U.S.A.). Sections were mounted and visualized at 200× magnification (Zeiss Axioskop, Germany). c-Fos positive cells were counted per section by an experimenter blinded to the treatment conditions. qPCR analysis {#sec2-5} ------------- Animals were killed 2 h post a single Flx (15 mg/kg) or saline injection by decapitation; the PFC was dissected out and snap-frozen in liquid nitrogen. RNA was extracted using Tri reagent (Sigma) and reverse transcribed using a cDNA synthesis kit (PrimeScript First Strand cDNA Synthesis Kit, Takara Bio). qPCR was performed with primers for the genes of interest (Supplementary Table S1) using a Bio-Rad CFX96 real-time PCR machine. Data were analyzed using the ΔΔ*C*~t~ method, as described previously \[[@B17]\]. Normalization was done using hypoxanthine guanine phosphoribosyl transferase (*Hprt*), whose level was unaltered across experimental groups. Gene expression of IEGs analyzed included activity regulated cytoskeleton associated protein (*Arc*), brain-derived neurotrophic factor (*Bdnf*), early growth response gene 1-4 (*Egr 1-4*), FBJ osteosarcoma gene (*Fos*), FBJ murine osteosarcoma viral oncogene homolog B (*Fosb*), Fos-like antigen 2 (*Fosl2*), Homer scaffolding protein 1-3 (*Homer 1-3*), Jun proto-oncogene (*Jun*), Jun B proto-oncogene (*Junb*), Jun D proto-oncogene (*Jund*). qPCR analysis was also performed for the 5-HT~2A~ gene (*Htr2a*) to further confirm our genotyping results. Statistical analysis {#sec2-6} -------------------- Statistical analysis was carried out using Prism 6 (GraphPad Software Inc, U.S.A.). Data were assessed using two-way ANOVA analysis followed by Bonferroni *post-hoc* group comparisons. Normality of data was verified using the Kolmogorov and Smirnov method. Significance was determined at *P*\<0.05. Results {#sec3} ======= 5-HT~2A~ ^−/−^ mice exhibit acute Flx-induced enhanced anxiety-like behavior in the OFT {#sec3-1} --------------------------------------------------------------------------------------- 5-HT~2A~ receptor knockout (5-HT~2A~^−/−^) mice have been reported to exhibit reduced anxiety-like behavior across both sexes (Supplementary Figure S1) \[[@B10],[@B14]\]. Acute Flx treatment is known to enhance anxiety-like behavioral responses \[[@B3]\]. Here, we sought to address whether the acute Flx-mediated increase in anxiety-like behavior is altered in 5-HT~2A~^−/−^ mice. Male and female 5-HT~2A~^−/−^ and WT littermate control mice were subjected to a single administration of Flx or saline, and were assessed for anxiety-like behavior in the OFT arena for 10 min ([Figure 1](#F1){ref-type="fig"}A). Two-way ANOVA analysis of OFT behavior in male 5-HT~2A~^−/−^ and WT mice revealed a significant main effect of acute Flx administration on percent distance traveled in center (*F*~(1,36)~ = 9.313, *P*=0.0043) ([Figure 1](#F1){ref-type="fig"}C), percent time in center (*F*~(1,36)~ = 4.164, *P*=0.048) ([Figure 1](#F1){ref-type="fig"}D), number of entries to the center (*F*~(1,36)~ = 7.608, *P*=0.0091) ([Figure 1](#F1){ref-type="fig"}E), and the total distance traveled in the OFT arena (*F*~(1,36)~ = 5.17, *P*=0.029) ([Figure 1](#F1){ref-type="fig"}F). Our results indicate a significant effect of acute Flx treatment on anxiety-like behavior in both WT and 5-HT~2A~^−/−^ mice ([Figure 1](#F1){ref-type="fig"}B--F). Interestingly, the baseline decrease in anxiety-like behavior previously noted in 5-HT~2A~^−/−^ mice \[[@B10],[@B14]\], was not observed following administration of a single saline injection, suggesting that the anxiety-like behaviors in 5-HT~2A~^−/−^ mice may be sensitive to both handling and injection stress. There was no significant main effect of genotype and no interaction effect (genotype × Flx). ![5-HT2A receptor knockout mice exhibit acute Flx-induced enhanced anxiety-like behavior in OFT\ Shown is a schematic of the experimental paradigm (**A**). 5-HT~2A~^−/−^ and WT littermate control male and female mice were administered acute Flx or saline (Sal), and were subjected to behavioral testing on the OFT 2 h post treatment (A). Shown are representative traces in the OFT arena from saline and Flx treated, WT and 5-HT2A^−/−^ male (**B**) and female (**G**) mice. Both 5-HT2A^−/−^ and WT male (B,**C--F**) and female (G,**H--K**) mice exhibited significant increases in anxiety-like behavior on the OFT following acute Flx treatment, with a reduction noted in percent distance traveled in center (males (C); females (H)), percent time in center (males (D); females (I)) and number of entries to the center (males (E); females (J)) of the OFT arena, as well as the total distance traveled in the OFT arena (males (F); females (K)). Results are expressed as the mean ± S.E.M. (*n*=9--10/group). Two-way ANOVA analysis, \**P*\<0.05, significant main effect of acute Flx treatment.](ns-03-ns20180205-g1){#F1} Similar to our observations in male mice, female 5-HT~2A~^−/−^ and WT littermate controls both showed enhanced anxiety-like behavior on the OFT following acute Flx treatment ([Figure 1](#F1){ref-type="fig"}G--K). Two-way ANOVA analysis of OFT behavior indicated a significant main effect of acute Flx treatment on percent distance traveled in center (*F*~(1,35)~ = 13.54, *P*=0.0008) ([Figure 1](#F1){ref-type="fig"}H), percent time in center (*F*~(1,35)~ = 7.454, *P*=0.0098) ([Figure 1](#F1){ref-type="fig"}I), number of entries to the center (*F*~(1,35)~ = 15.09, *P*=0.0004) ([Figure 1](#F1){ref-type="fig"}J), and total distance traveled (*F*~(1,35)~ = 5.592, *P*=0.0237) in the OFT arena ([Figure 1](#F1){ref-type="fig"}K). We noted no main effect of genotype and no significant interaction (genotype × Flx) effects. This indicates that the baseline anxiolytic responses on OFT reported in both sexes of 5-HT~2A~^−/−^ mice are not observed following handling and injection stress. Taken together, our findings indicate that 5-HT~2A~ receptor deficiency does not influence the increased anxiety-like behavior noted following acute treatment with the SSRI, Flx. 5-HT~2A~ receptor knockout mice exhibit acute Flx-induced enhanced anxiety-like behavior in the EPM {#sec3-2} --------------------------------------------------------------------------------------------------- We next examined the influence of acute Flx on anxiety-like behavior in 5-HT~2A~^−/−^ mice and their WT controls, in the EPM ([Figure 2](#F2){ref-type="fig"}A--K). ![5-HT~2A~^−/−^ mice exhibit acute Flx-induced enhanced anxiety-like behavior on EPM\ Shown is a schematic of the experimental paradigm (**A**). 5-HT~2A~^−/−^ and WT control male and female mice were administered saline (Sal) or acute Flx, and were subjected to behavioral testing on the EPM 2 h post treatment (A). Shown are representative EPM arena tracks from saline and Flx treated, WT and 5-HT~2A~^−/−^ male (**B**) and female (**G**) mice. Both 5-HT~2A~^−/−^ and WT male (B,**C**--**F**) and female (G,**H**--**K**) mice exhibited significant increases in anxiety-like behavior on the EPM following acute Flx treatment, with a reduction noted in total distance traveled in the EPM arena while percent distance traveled in open arms (males (C); females (H)), percent time in open arms (males (D); females (I)) were unaltered by Flx administration. The number of entries to the open arms of the EPM arena was significantly reduced by Flx in females (J) but not in males (E). Results are expressed as the mean ± S.E.M. (*n*=11--15/group). Two-way ANOVA analysis, \**P*\<0.05, significant main effect of acute Flx treatment.](ns-03-ns20180205-g2){#F2} Two-way ANOVA analysis of EPM behavior in males revealed no significant main effect of acute Flx on percent distance traveled ([Figure 2](#F2){ref-type="fig"}C), percent time spent ([Figure 2](#F2){ref-type="fig"}D), and number of entries ([Figure 2](#F2){ref-type="fig"}E) in the open arms. For the percent time spent in the open arms for males, we did note a trend toward an interaction of genotype × Flx (*F*~(1,50)~ = 3.39, *P*=0.0714) ([Figure 2](#F2){ref-type="fig"}D), which did not reach statistical significance. Flx treatment significantly reduced the total distance traveled in the EPM arena (*F*~(1,50)~ = 6.42, *P*=0.0145) ([Figure 2](#F2){ref-type="fig"}F), which is considered to be indicative of an enhanced anxiety-like behavioral response. We noted no main effect of genotype and no significant genotype × Flx interaction on any of the EPM parameters in male mice. Comparable with the male EPM behavioral response, female WT and 5-HT~2A~^−/−^ mice also exhibited enhanced acute Flx-mediated anxiety in the EPM ([Figure 2](#F2){ref-type="fig"}G--K). Two-way ANOVA analysis of EPM in females revealed no significant main effect of acute Flx on percent distance traveled ([Figure 2](#F2){ref-type="fig"}H), and percent time spent ([Figure 2](#F2){ref-type="fig"}I) in the open arms. Acute Flx treatment significantly reduced the number of entries in the open arms (*F*~(1,46)~ = 7.86, *P*=0.0074) ([Figure 2](#F2){ref-type="fig"}J) and the total distance traveled in the EPM arena (*F*~(1,46)~ = 25.3, *P*\<0.0001) ([Figure 2](#F2){ref-type="fig"}K), suggestive of enhanced anxiety-like behavior. We noted no main effect of genotype and no significant genotype × Flx interaction for the behavioral measures assayed on the EPM in females. 5-HT~2A~ receptor knockout mice exhibit elevated serum corticosterone levels and sexually dimorphic PVN neural activation following acute Flx administration {#sec3-3} ------------------------------------------------------------------------------------------------------------------------------------------------------------ Acute Flx treatment is known to increase serum corticosterone levels through activation of the HPA axis \[[@B18],[@B19]\]. Given that the 5-HT~2A~ receptor has been implicated in regulation of HPA axis activity via effects on stimulation of PVN neurones in the hypothalamus \[[@B20]\], we sought to address whether the acute Flx-mediated effects on serum corticosterone and neural activation in the PVN are altered in 5-HT~2A~^−/−^ mice ([Figure 3](#F3){ref-type="fig"}A). ![5-HT~2A~ receptor knockout mice exhibit elevated serum corticosterone levels and sexually dimorphic PVN c-Fos positive cell number following acute Flx administration\ Shown is a schematic of the experimental paradigm (**A**). 5-HT~2A~^−/−^ and WT control male and female mice were administered acute Flx or saline (Sal), and were killed 30 min later for serum corticosterone analysis or immunohistochemical analysis of c-Fos positive cells in the PVN of the hypothalamus (A). Male (**B**) and female (**E**) 5-HT~2A~^−/−^ and WT mice showed an increase in the serum corticosterone levels post Flx treatment. Shown are representative sections of c-Fos positive cells within the PVN of 5-HT~2A~^−/−^ and WT mice treated with saline or Flx (male (C); female (F)). 5-HT~2A~^−/−^ and WT male mice exhibited a significant increase in c-Fos positive cell numbers in the PVN (**D**). A significant genotype × Flx interaction effect was noted for c-Fos positive cell numbers in the PVN of 5-HT~2A~^−/−^ and WT female mice administered Flx or saline (**G**). Post-hoc Bonferroni multiple comparisons revealed a significant increase in PVN c-Fos positive cell numbers following acute Flx in WT, but not in 5-HT~2A~^−/−^, female mice (G). Results are expressed as the mean ± S.E.M. (*n*=3--6/group). Two-way ANOVA analysis, \**P*\<0.05, significant main effect of acute Flx treatment, ^§^*P*\<0.05, significant genotype × Flx interaction, Bonferroni post-hoc test, ^¢^*P*\<0.05, as compared with saline-treated WT female mice.](ns-03-ns20180205-g3){#F3} Two-way ANOVA analysis for serum corticosterone levels in males revealed a significant main effect of acute Flx treatment (*F*~(1,15)~ = 15.09, *P*=0.002) and genotype (*F*~(1,15)~ = 5.18, *P*=0.04), but no significance in interaction of genotype and Flx ([Figure 3](#F3){ref-type="fig"}B). We next assessed the number of c-Fos^+^ cells within the PVN of 5-HT~2A~^−/−^ and WT littermate control male mice as a measure of the degree of neural activation following acute Flx treatment. Two-way ANOVA analysis for number of c-Fos positive cells indicated a significant main effect of acute Flx treatment (*F*~(1,18)~ = 24.89, *P*\<0.0001) ([Figure 3](#F3){ref-type="fig"}C,D), with no main effect noted for genotype and no genotype × Flx interaction. This indicates that 5-HT~2A~ receptor deficiency in male mice does not alter either the acute Flx-mediated increase in serum corticosterone levels or the increased neural activation, as assessed using c-Fos positive cell numbers, within the PVN. Similar analysis for serum corticosterone levels carried out in female 5-HT~2A~^−/−^ and WT mice indicated a significant main effect of acute Flx treatment (*F*~(1,17)~ = 13.21, *P*=0.002) with elevated levels of circulating corticosterone noted following acute Flx administration ([Figure 3](#F3){ref-type="fig"}E). No main effect of genotype, or interaction between genotype and Flx treatment, was noted in the two-way ANOVA analysis. The degree of neural activation in the PVN of female 5-HT~2A~^−/−^ and WT mice following acute Flx treatment was assessed using c-Fos cell counting analysis ([Figure 3](#F3){ref-type="fig"}F,G). Two-way ANOVA analysis indicated a significant genotype × Flx interaction (*F*~(1,18)~ = 6.58, *P*=0.02), and significant main effects for acute Flx treatment (*F*~(1,18)~ = 20.57, *P*=0.0003) and genotype (*F*~(1,18)~ = 6.53, *P*=0.02) ([Figure 3](#F3){ref-type="fig"}G). *Post-hoc* group comparisons revealed a significant increase in c-Fos positive cell numbers in the PVN following acute Flx administration to WT, but not 5-HT~2A~^−/−^, female mice ([Figure 3](#F3){ref-type="fig"}G). This suggests that 5-HT~2A~ receptor deficiency in female mice results in a blunting of the degree of neural activation in the PVN following acute Flx administration. 5-HT~2A~ receptor deficiency does not alter neural activation patterns in the PFC following acute Flx administration {#sec3-4} -------------------------------------------------------------------------------------------------------------------- We next examined neural activation, using c-Fos cell counting analysis, within the subdivisions of the PFC, which are known to exert a strong top-down control on HPA axis activity \[[@B21]\], exhibit robust 5-HT~2A~ receptor expression and are known to regulate anxiety-like behavior \[[@B10],[@B22]\]. We assessed c-Fos positive cell numbers within the cingulate (Cg), prelimbic (PrL) and infralimbic (IL) subdivisions of the PFC in male and female 5-HT~2A~^−/−^ and WT littermate control mice treated with acute Flx or saline ([Figure 4](#F4){ref-type="fig"}A). ![Acute Flx treatment does not alter c-Fos positive cell number in the PFC of 5-HT~2A~^−/−^ and WT male and female mice\ Shown is a schematic of the experimental paradigm (**A**). 5-HT~2A~^−/−^ and WT control male and female mice were administered acute Flx or saline (Sal), and were killed 30 min later for immunohistochemical analysis of c-Fos positive cell number in the subdivisions of the PFC, namely the Cg, PrL, and IL cortex (A). Shown are representative sections of c-Fos positive cells within the PFC of 5-HT~2A~^−/−^ and WT mice treated with saline or Flx (male (B); female (G)). Cell counting analyses revealed no significant main effect of Flx treatment, genotype or genotype × Flx interaction on c-Fos positive cell numbers within the Cg (male (C); female (H)), PrL (male (D); female (I)), IL (male (E); female (J)) and total PFC (male (F); female (K)). Results are expressed as the mean ± S.E.M. (*n*=5--6/group). Two-way ANOVA analysis.](ns-03-ns20180205-g4){#F4} Two-way ANOVA analysis for c-Fos positive cell numbers in the Cg, PrL, IL and cumulatively for the entire PFC, in experiments done in 5-HT~2A~^−/−^ and WT male mice indicated no significant main effects of Flx, genotype or genotype × Flx interaction ([Figure 4](#F4){ref-type="fig"}B--F). Similarly, c-Fos positive cell numbers in the Cg, PrL, IL and for the total PFC were unaltered in 5-HT~2A~^−/−^ and WT female mice following acute Flx or saline treatment ([Figure 4](#F4){ref-type="fig"}G--K). Collectively, these findings indicate that neural activation patterns in the PFC subdivisions are not altered in 5-HT~2A~ receptor deficient mice of both sexes. 5-HT~2A~ receptor knockout mice exhibit sexually dimorphic alterations in PFC IEG expression {#sec3-5} -------------------------------------------------------------------------------------------- IEG expression within the neocortex, including the PFC, has been reported to be regulated by 5-HT~2A~ receptors \[[@B14],[@B23]\], hence we sought to address the effects of acute Flx on expression of several IEGs in the PFC of 5-HT~2A~^−/−^ and WT mice of both sexes ([Figure 5](#F5){ref-type="fig"}A). ![5-HT~2A~ receptor knockout mice exhibit sexually dimorphic changes in PFC IEG expression\ Shown is a schematic of the experimental paradigm (**A**). 5-HT~2A~^−/−^ and WT control male and female mice were administered acute Flx or saline (Sal), and were killed 2 h post treatment for qPCR analysis of IEG expression in the PFC (A). The heat map denotes the magnitude of regulation represented as percent of WT saline group (WT Sal), with up-regulated genes shown in red and down-regulated genes shown in green (key (A)). qPCR analysis for gene expression of the IEGs examined represented as percent of WT Sal ± S.E.M. (male (B); female (C)) (male: *n*=8--12/group; female: *n*=7--10/group). Two-way ANOVA analysis with significance determined at *P*\<0.05. IEGs analyzed included the Arc, Bdnf, Egr 1-4, Fos, Fosb, Fosl2, Homer 1-3, Jun, Junb, and Jund. We also performed qpCR analysis for the *Htr2a* to further confirm our genotyping results.](ns-03-ns20180205-g5){#F5} In the male PFC, two-way ANOVA analysis of qPCR results indicated significant main effects of genotype for several IEGs: *Bdnf, Egr2, Egr4, Fos, Fosb, Fosl2, Homer1-3*, and *Jun* ([Figure 5](#F5){ref-type="fig"}B). No significant main effects of genotype were observed for *Arc, Egr1, Egr3, Junb*, and *Jund* gene expression ([Figure 5](#F5){ref-type="fig"}B). 5-HT~2A~^−/−^ male mice exhibited a decline in PFC gene expression for several IEGs as compared with WT littermate control male mice ([Figure 5](#F5){ref-type="fig"}B). Acute Flx treatment did not alter the mRNA expression levels of the IEGs assessed, except for a strong trend for decline observed in gene expression levels of *Bdnf* and *Homer3* in the male PFC ([Figure 5](#F5){ref-type="fig"}B). We observed no significant interactions of genotype × Flx for gene expression of the IEGs in the PFC in 5-HT~2A~^−/−^ and WT male mice. In striking contrast with the males, PFC IEG expression was largely unperturbed in the 5-HT~2A~^−/−^ female mice indicating a sexual dimorphism in the effects of the 5-HT~2A~ receptor on PFC IEG expression. Two-way ANOVA analysis indicated no significant main effect of genotype on the gene expression of IEGs in the PFC in females, with the exception of a main effect of genotype observed for *Egr1* gene expression, and a strong trend toward a main effect of genotype in the expression of *Egr2* and *Junb* ([Figure 5](#F5){ref-type="fig"}C). Two-way ANOVA analysis revealed a significant main effect of Flx for *Bdnf, Egr1, Egr4*, and *Homer3* gene expression in the PFC ([Figure 5](#F5){ref-type="fig"}C), and a strong trend for a main effect of Flx in the prefrontal expression of *Fosl2* and *Homer1* ([Figure 5](#F5){ref-type="fig"}C). We observed no significant genotype × Flx interaction for gene expression of the IEGs in the PFC in 5-HT~2A~^−/−^ and WT female mice. Collectively, our results indicate that 5-HT~2A~ receptor deficiency exerts sexually dimorphic consequences on the expression of several IEGs in the PFC, with a decline in PFC expression of *Bdnf, Egr2, Egr4, Fos, Fosb, Fosl2, Homer1-3*, and *Jun* noted in 5-HT~2A~^−/−^ male, but not female mice. Discussion {#sec4} ========== Here, we report that 5-HT~2A~ receptor loss of function does not alter either the acute Flx-mediated increase in anxiety-like behavior, or the enhanced circulating corticosterone, in both male and female mice. However, 5-HT~2A~ receptor knockout mice do exhibit sexually dimorphic effects in the acute Flx-induced neural activation in the PVN with a blunted response to acute Flx noted in 5-HT~2A~^−/−^ female, but not male, mice. Furthermore, we noted sex-dependent changes in expression of several IEGs (*Bdnf, Egr2, Egr4, Fos, Fosb, Fosl2, Homer1-3* and *Jun*) in the PFC of 5-HT~2A~^−/−^ male, but not female, mice. 5-HT~2A~ receptor deficiency and acute Flx evoked anxiety-like behavior {#sec4-1} ----------------------------------------------------------------------- The SSRI Flx exerts paradoxical anxiogenic effects during the acute phase of treatment \[[@B3],[@B4]\]. These aversive effects of acute Flx on anxiety-like behavior are suggested to involve a role for the 5-HT~2~ receptor subtypes, with the 5-HT~2C~ receptor in the BNST \[[@B6]\] and BLA \[[@B8],[@B12]\], implicated in these anxiogenic effects. The 5-HT~2A~ receptor also plays an important role in the modulation of anxiety-like behavior \[[@B9]\], and co-administration of a 5-HT~2A~ receptor antagonist is reported to accelerate and augment antidepressant effects of SSRIs \[[@B13],[@B24]\]. Here, we examined whether the increased anxiety-like behavior evoked by acute Flx is altered in 5-HT~2A~^−/−^ male and female mice. Our results indicate that similar to WT mice, adult 5-HT~2A~^−/−^ male and female mice both exhibit acute Flx-induced increases in anxiety-like behavioral responses on the OFT and EPM, indicating that 5-HT~2A~ receptor deficiency does not influence the anxiogenic effects of acute Flx. It would be important in future experiments to address whether 5-HT~2A~ receptor deficient mice which continue to exhibit the acute anxiogenic effects of Flx, also respond in a similar manner to acute treatment with other SSRIs. Interestingly, while we did observe the previously reported baseline anxiolytic behavioral response in 5-HT~2A~^−/−^ male and female mice (Supplementary Figure S1) \[[@B10],[@B14]\], the vehicle-treated 5-HT~2A~^−/−^ cohorts used for the acute Flx experiment did not exhibit this baseline anxiolytic phenotype. The cohorts used for the present study were subjected to handling and injection of vehicle/Flx, and since the anxiolytic behavioral phenotype observed in naïve, unhandled 5-HT~2A~^−/−^ mice was lost, this suggests the possibility that 5-HT~2A~ receptor deficient mice may be highly sensitive to effects of handling and injection stress. Prior reports do indicate that handling and injection stress can alter anxiety-like behaviors \[[@B25],[@B26]\]. Further experiments are required to directly test whether 5-HT~2A~ receptor deficient mice exhibit such enhanced sensitivity to handling and injection stress. A previous report indicates that 5-HT~2A~^−/−^ mice exhibit treatment resistance to chronic Flx, failing to exhibit both antidepressant-like behavioral responses and enhanced hippocampal neurogenesis following chronic Flx treatment \[[@B16]\]. Taken together, these findings indicate that while the anxiogenic effects noted during the acute phase of Flx treatment are unaltered by 5-HT~2A~ receptor deficiency, 5-HT~2A~^−/−^ mice exhibit a treatment-resistant phenotype to the effects of chronic Flx \[[@B16]\]. Role of the 5-HT~2A~ receptor in the effects of acute Flx on the HPA axis {#sec4-2} ------------------------------------------------------------------------- Acute Flx treatment is known to influence the HPA axis, increasing circulating levels of corticosterone via enhanced activity of PVN neurones in the hypothalamus \[[@B18],[@B19]\]. 5-HT~2A~ receptors are densely expressed by PVN neurones, and their activation leads to release of adrenocorticotrophic hormone and enhanced circulating corticosterone \[[@B20],[@B27]\]. We find that the acute Flx-mediated elevation of serum corticosterone levels is observed in both WT and 5-HT~2A~^−/−^ male and female mice, indicating that the effects of acute Flx on enhanced corticosterone levels do not appear to involve a role for the 5-HT~2A~ receptor. Acute Flx administration also resulted in a clear increase in neural activation within the PVN, as revealed by enhanced numbers of c-Fos positive cells. This acute Flx-evoked increase in PVN c-Fos positive cells was noted in both WT and 5-HT~2A~^−/−^ male mice, and the magnitude of this effect was unaltered by 5-HT~2A~ receptor deficiency in males. Strikingly, 5-HT~2A~^−/−^ female mice showed a significant blunting of the acute Flx-mediated increase in c-Fos positive cells in the PVN. Previous studies indicate an interaction of sex hormones such as estrogen with SSRIs such as Flx \[[@B28],[@B29]\], and also with the 5-HT~2A~ receptor \[[@B30],[@B31]\]. This in turn provides support for the possibility that the effects of acute Flx on neural activation in the PVN may exhibit sex differences, possibly through interactions between the 5-HT~2A~ receptor, estrogen and acute Flx treatment. While we noted that the acute Flx-mediated increase in c-Fos positive cell number in the PVN was blunted in 5-HT~2A~^−/−^ female mice, we did not observe any significant genotype × Flx interaction for corticosterone levels in 5-HT~2A~^−/−^ female mice. This suggests that despite a substantial blunting of the neural activation in the PVN evoked by acute Flx in 5-HT~2A~^−/−^ female mice, this does not translate to a failure of acute Flx to increase circulating corticosterone in 5-HT~2A~^−/−^ female mice. Collectively, our findings indicate that the effects of acute Flx on enhancing corticosterone levels are unaltered by 5-HT~2A~ receptor deficiency, however 5-HT~2A~ receptor loss of function does alter the degree of neural activation induced by acute Flx in the PVN in a sex-dependent manner. 5-H~T2A~ receptor and the acute Flx-evoked regulation of neural activation, and IEG expression, in the PFC {#sec4-3} ---------------------------------------------------------------------------------------------------------- Acute Flx treatment is known to rapidly elevate serotonin levels within the PFC \[[@B32]\], a limbic brain region that exerts top-down control over the HPA axis and regulates anxiety behaviors \[[@B21]\]. The 5-HT~2A~ receptor is expressed at particularly high levels within the PFC, and the cortical 5-HT~2A~ receptor plays a key role in the modulation of anxiety behavior \[[@B10],[@B22],[@B33]\]. We sought to address whether 5-HT~2A~ receptor deficiency altered neural activation measured via assessing c-Fos positive cell number, or activity-dependent IEG expression, in the PFC, both under baseline conditions and in response to acute Flx administration. We observed no difference in c-Fos positive cell numbers in the PFC across groups 30 min post acute Flx administration. Our results do not preclude the possibility of differences in the nature and degree of neural activation of the PFC in WT and 5-HT~2A~^−/−^ mice at later time-points post Flx treatment. We also examined the expression of several IEGs in WT and 5-HT~2A~^−/−^ male and female mice, both baseline and post Flx treatment. While we did not observe any major effect of Flx treatment on IEG expression in the PFC, we noted robust baseline down-regulation of several IEGs (*Bdnf, Egr2, Egr4, Fos, Fosb, Fosl2, Homer1-3, Jun*) in the PFC of 5-HT~2A~^−/−^ male, but not female mice. Our results are largely in agreement with prior reports that acute Flx, unlike chronic treatment, does not appear to alter cortical IEG and plasticity-associated gene expression \[[@B34]\]. We have previously reported a sex-dependent, baseline regulation of several IEG and plasticity-associated genes in both the PFC and hippocampus of 5-HT~2A~^−/−^ mice \[[@B14]\]. This decline in basal expression of several IEGs in the PFC of 5-HT~2A~^−/−^ male mice may arise as a consequence of reduced excitability of PFC pyramidal neurones following 5-HT~2A~ receptor deficiency, however the fact that this change is not observed in 5-HT~2A~^−/−^ female mice is intriguing. Prior reports suggest an interaction between estrogen and the 5-HT~2A~ receptor with regard to the transcriptional regulation of *Bdnf* \[[@B31]\]. In particular, *Bdnf* expression is reported to be regulated by estrogen receptors via an estrogen response element \[[@B35]\], it is possible that such a regulation may contribute to the differential effects of 5-HT~2A~ receptor deficiency on cortical *Bdnf* expression in male compared with female mice \[[@B36]\]. Our findings suggest a sexual dimorphism in the 5-HT~2A~ receptor-mediated regulation of activity-dependent IEG expression, and motivate future experiments to address the nature of interaction between sex hormones and the 5-HT~2A~ receptor in this regard. Conclusion {#sec5} ========== The findings of our study indicate that the 5-HT~2A~ receptor does not appear to play a major role in the effects of acute Flx on anxiety-like behavior and corticosterone secretion. However, 5-HT~2A~ receptor deficiency evokes sex-dependent differences in both baseline cortical IEG expression, and in the acute Flx-mediated neural activation of the PVN of the hypothalamus. These findings highlight the sex differences in the effects of the 5-HT~2A~ receptor in the regulation of several genes, implicated in neuronal plasticity, within key brain regions such as the PFC. This suggests the possibility of an important interaction between the 5-HT~2A~ receptor, estrogen and activity-dependent IEG and plasticity-associated gene expression, motivating future experiments to uncover the mechanistic details of such an interaction. Supplementary Material ====================== ###### Supplementary Figures ###### Click here for additional data file. ###### Supplementary Table ###### Click here for additional data file. We thank Dr. Shital Suryavanshi for assistance with breeding and maintenance of the 5-HT~2A~ receptor knockout mouse colony. Competing interests {#sec6} =================== The authors declare that there are no competing interests associated with the manuscript. Funding {#sec7} ======= This work was supported by the intramural funds from the Tata Institute of Fundamental Research \[grant number 12P-0118\]. Author contribution {#sec8} =================== M.J. and V.A.V. conceptualized the experiments, provided critical inputs, and wrote the paper. M.J. performed most experiments, all analyses, and made the figures. T.B. performed elevated plus-maze experiment and helped in manuscript rebuttal. N.W. and J.A.G. made and provided the 5-HT~2A~ knockout mouse line. 5-HT~2A~ : serotonin~2A~ receptor 5-HT~2A~^−/−^ : 5-HT~2A~ developmental knockout 5-HT~2C~ : serotonin~2C~ receptor *Arc* : activity regulated cytoskeletal associated protein Avg : average *Bdnf*/BDNF : brain derived neurotrophic factor BLA : basolateral nucleus of the amygdala BNST : bed nucleus of the stria terminalis c-Fos : FBJ murine osteosarcoma viral oncogene homolog peptide Cg : cingulate cortex Flx : fluoxetine *Fos* : FBJ osteosarcoma oncogene *Fosb* : FBJ murine osteosarcoma viral oncogene homolog B *Fosl2* : Fos-like antigen 2 *Homer* : Homer scaffolding protein HPA axis : hypothalamus--pituitary--adrenocortical axis IEG : immediate early gene IL : infralimbic cortex *Jun* : Jun proto-oncogene OFT : open-field test PFA : paraformaldehyde PFC : prefrontal cortex PrL : prelimbic cortex PVN : paraventricular nucleus qPCR : quantitative real-time PCR SSRI : selective serotonin reuptake inhibitor WT : wild-type

{ "pile_set_name": "PubMed Central" }

1. Introduction =============== Multiple sclerosis (MS) is a chronic neurological disease affecting a significant number of young adults \[[@B1-ijms-16-13885],[@B2-ijms-16-13885]\]. Although the exact cause of MS is unknown, it is the most common multifocal inflammatory demyelinating disease of the central nervous system (CNS) \[[@B3-ijms-16-13885],[@B4-ijms-16-13885],[@B5-ijms-16-13885]\]. The neurodegenerative process with damage to axons and oligodendrocytes is thought to be the cause of permanent neurological impairment and disability \[[@B6-ijms-16-13885],[@B7-ijms-16-13885]\]. Due to the progressive neurodegenerative nature of MS, developing treatments that exhibit direct neuroprotective effects are needed. Currently, most available MS therapies are thought to exert their effects via immunomodulatory or immunosuppressive functions \[[@B5-ijms-16-13885],[@B8-ijms-16-13885],[@B9-ijms-16-13885]\]. Although these treatments are effective at inhibiting immune cell-driven inflammation and reducing the relapse rate, they are ineffective at controlling the predominantly neurodegenerative processes that occur later in the disease course \[[@B10-ijms-16-13885]\]. Methods of enhancing remyelination and decreasing axonal damage (*i.e.*, promoting neuroprotection) would be particularly useful for later stage and primary progressive MS forms. Fumaric acid esters (FAE) have been used since 1959 as a treatment for psoriasis \[[@B11-ijms-16-13885]\]. Tecfidera™ (BG-12) is an oral formulation of the FAE, containing the active metabolite dimethyl fumarate (DMF), which showed remarkable efficacy in lowering relapse rates in two phase III clinical trials of relapsing-remitting MS (RR-MS) treatment \[[@B12-ijms-16-13885],[@B13-ijms-16-13885],[@B14-ijms-16-13885],[@B15-ijms-16-13885],[@B16-ijms-16-13885]\]. However, its mechanism of action is not yet well understood. It has been shown that FAE and its primary metabolite DMF and monomethyl fumarate (MMF) were able to activate the transcription factor nuclear factor-erythroid 2-related factor 2 (Nrf2) pathway and subsequently induce the expression of antioxidant proteins \[[@B17-ijms-16-13885],[@B18-ijms-16-13885],[@B19-ijms-16-13885]\]. Oxidative stress plays a major role in multiple sclerosis (MS) and is readily apparent within experimental autoimmune encephalomyelitis (EAE), mouse model of MS, and MS lesions \[[@B15-ijms-16-13885],[@B20-ijms-16-13885]\]. Invading leukocytes contribute to cell damage and demyelination by producing excessive amounts of cytotoxic mediators, including reactive oxygen species (ROS). Under normal conditions, ARE/EpRE (antioxidant/electrophile response element) is not active, but under stress Nrf2 activates ARE/EpRE \[[@B21-ijms-16-13885]\]. Nrf2 is rapidly ubiquitinated and degraded by the help of the adaptor protein, kelch-like ECH-associated protein 1 (Keap1) \[[@B22-ijms-16-13885]\]. Glutathione, a major intracellular ROS scavenger, is decreased within CNS inflammatory foci, and a variety of antioxidant proteins are increased in MS lesions \[[@B20-ijms-16-13885]\]. Additionally, EAE in *Nrf2*^(−/−)^ mice is more severe and neuroprotective effects of DMF are absent in *Nrf2*^(−/−)^ mice \[[@B23-ijms-16-13885]\]. Furthermore, it has been shown that administration of MMF could protect motor neurons and astrocytes against hydrogen peroxide (H~2~O~2~)-induced oxidative stress \[[@B23-ijms-16-13885]\]. In a noninflammatory demylination model, DMF/MMF have little to no impact on demyelination or remyelination \[[@B24-ijms-16-13885]\]. Other studies also have shown Nrf2-dependent cytoprotection of neurons and astrocytes, but did not demonstrate decreased infiltrates \[[@B25-ijms-16-13885],[@B26-ijms-16-13885],[@B27-ijms-16-13885]\]. The therapeutic application of DMF in chronic EAE leading to a reduction in demyelination with relative preservation of myelin and axons suggests that the rescue of neurons and glial cells from oxidative stress-induced cell death is most likely through neuroprotection mediated by DMF-induced activation of Nrf2 pathway \[[@B23-ijms-16-13885]\]. The observation of proliferating neural cells in the adult rat brain made the possibility of CNS regeneration more feasible. Neural stem/progenitor cells (NPCs) reside in the adult CNS and support neurogenesis and gliogenesis throughout life \[[@B8-ijms-16-13885],[@B28-ijms-16-13885]\]. They are a heterogeneous population of perpetually self-renewing and multipotent cells that can spontaneously differentiate into neurons, astrocytes, or oligodendrocytes (post-mitotic daughter cells) \[[@B29-ijms-16-13885]\]. The ability of NPCs to reconstitute neurons makes them an important target for therapy in MS, as various drugs may interact with stem cells to promote neural protection and/or remyelination. It has been thought that the benefits seen in the clinical symptoms of EAE mice after stem cell transplantation are mediated by neural stem cells and immune cell interactions that promote neuroprotection. NPCs transplantation holds significant promise as a novel treatment strategy for MS \[[@B30-ijms-16-13885]\]. Neuroprotection has also been seen in excitotoxic neurodegeneration and Huntington's disease. DMF has also been shown to improve lifespan, reduce behavioral deficits, and preserve striatal and motor cortex neurons in two different genetic models of Huntington's disease in mice \[[@B31-ijms-16-13885]\], suggesting broad neuroprotective properties. Damage of cortical neurons and motor neurons are the main cause of dysfunction and disability in MS patients. MS disease occurs early and progresses later in life. It is currently unknown whether DMF exerts its neuroprotective effects directly on differentiated neuronal cells and/or through activation of NPCs. Furthermore, the effect of DMF at the level of NPCs has not yet been reported. However, the mechanism of DMF action in these neurons, in terms of signaling pathways in MS, is not clearly understood. In this study, we sought to investigate the neuroprotective effect of DMF in the context of two different neuronal cellular systems: NPCs and motor neurons. Our data showed that DMF protected NPCs in addition to differentiated motor neurons from oxidative damage (H~2~O~2~) through regulation of Nrf2, as well as several other novel genes involved in superoxide metabolism. Furthermore, the involvement of the signal transduction ERK1/2 MAPK pathway was identified in DMF-mediated neuroprotection. 2. Results ========== 2.1. DMF Increased Frequency of Mouse NPCs in Vitro and Protected Rat NPCs and Motor Neurons from Oxidative Damage ------------------------------------------------------------------------------------------------------------------ To examine the neuroprotective effects of DMF, we first used mouse NPCs to determine whether DMF has any effects on neurosphere formation and self-renewal *in vitro*. In the neurosphere formation assay, we found that DMF increased the frequency of the multipotent neurospheres in culture with the frequency of 30 ± 2 comparing to 22 ± 2.6 in control per 1000 cells ([Figure 1](#ijms-16-13885-f001){ref-type="fig"}A). Next we tested whether DMF plays a role in regulation of stem/progenitor cells self-renewal. We calculated the number of secondary neurospheres clonally derived from one primary neurosphere plated under low cell density condition and found that DMF treatment increased the number of secondary neurospheres in the self-renewal assay with 302 ± 11 comparing to 204 ± 27 in control ([Figure 1](#ijms-16-13885-f001){ref-type="fig"}B). To test whether DMF has neuroprotective effects with respect to oxidative damage, rat NPCs were under oxidative stress with hydrogen peroxide (H~2~O~2~) with or without DMF treatment, then an apoptosis assay was performed. DMF indeed decreased H~2~O~2~-induced apoptosis from 27.1% ± 3% to 12.6% ± 1.8%, effectively increasing the survival of rNPCs following oxidative stress treatment with H~2~O~2~. DMF treatment alone did not have any effect (11% ± 1%, [Figure 1](#ijms-16-13885-f001){ref-type="fig"}C). In addition, utilizing reactive oxygen species (ROS) assay, we showed that DMF reduced ROS production induced by H~2~O~2~ from 98 ± 0.002 to 0.73 ± 0.07 in rat NPCs ([Figure 1](#ijms-16-13885-f001){ref-type="fig"}D). To observe the efficacy of the neuroprotective features of DMF, dissociated rat motor neurons were cultured on electrospun poly-[l]{.smallcaps}-lactic acid (PLLA) nanofibers and the following features were measured: average number of neurites per image field, total neurite outgrowth, mean neurite length, and cell body area. The fibers serve as a substrate for growing the neurons since they direct neurite growth in a straight line, simplifying identification of neurites and neurite measurement. Neurons were exposed to oxidative stress with H~2~O~2~ after one day of culture and grown for either two or four days; half of these samples were treated with DMF. Both were compared to positive control samples that were unstressed and untreated. After two days of growth, total neurite outgrowth decreased on stress-treated neurons from 531 ± 16 to 438 ± 14 ([Figure 2](#ijms-16-13885-f002){ref-type="fig"}A), indicating poorer health of the neurons. However, addition of DMF abrogated this decrease and raised total neurite outgrowth from 438 ± 14 to 569 ± 16 ([Figure 2](#ijms-16-13885-f002){ref-type="fig"}A). No difference was observed amongst samples for cell body area or number of neurites per image field. This establishes that after one day of culture, neurons had begun to succumb to the stress of H~2~O~2~ treatment. After four days of growth (with three days of stress), no neurites were found on stressed motor neurons ([Figure 2](#ijms-16-13885-f002){ref-type="fig"}D). Addition of DMF protected neurons from stress as measured by average neurite number per image field (4 ± 0.3, *n* = 81, [Figure 2](#ijms-16-13885-f002){ref-type="fig"}C). This effect was also seen in all of our other measures: total neurite outgrowth, mean neurite length, and cell body area (data not shown). This substantiates the neuroprotective effects of DMF. ![Dimethyl fumarate (DMF) increased self-renewal and reduced oxidative stress-induced apoptosis and ROS levels in neural stem/progenitor cells (NPCs). (**A**) Frequency of mouse neuronal stem/progenitor cells was calculated by dividing the number of neurospheres in the well by the number of plated cells with (DMF) or without (C) DMF treatment (Mean ± SEM were determined from seven replicate samples, **\*** *p* = 0.027); (**B**) Self-renewal potential of mouse NPCs was calculated as number of secondary (2nd) neurospheres (NS) clonally derived from one primary (1st) neurosphere plated with (DMF) or without (C) DMF treatment (Mean ± SEM were determined from three replicate samples, **\*** *p* = 0.03); (**C**) Rat NPCs were grown in adherent cultures in self-renewal medium. After treatment under indicated conditions with control (C), DMF alone (DMF) for overnight, then H~2~O~2~ alone (H~2~O~2~) and DMF together with H~2~O~2~ (H~2~O~2~ + DMF) for another 6 h, the NPC cells were trypsinized, then apoptosis assay was performed as described under "Material and methods". Mean ± SEM were determined from three replicatesamples, **\*** *p* = 0.015, **\*\*** *p* = 0.007); (**D**) Cells treated same as in "C" and then incubated with 5 µM 2′,7′-dichlorofluorescein diacetate (DCFDA) for 15 min at 37 °C and oxidative stress was measured as ROS binding to DCFDA and quantified by flow cytometry. Mean ± SEM were determined from three replicate samples, **\*** *p* = 0.027.](ijms-16-13885-g001){#ijms-16-13885-f001} ![DMF promoted survival of rat motor neurons under oxidative stress. Isolated rat motor neurons were cultured on aligned, electrospun poly-[l]{.smallcaps}-lactic acid (PLLA) nanofiber scaffolds that direct neurite growth in straight lines along the fibers, allowing straightforward measurement of neurite length. Treatment with vehicle (control; labeled C), H~2~O~2~ alone (H~2~O~2~) and DMF together with H~2~O~2~ (H~2~O~2~ + DMF) was added after one day *in vitro*. Following fixation and staining, neurons were identified and their neurites were then counted and measured after two and four days *in vitro* (resulting in exposures of one and three days, respectively). Motor neuron survival was assessed by measuring the neurite length and counting the number of survived motor neurons on the slide view field. (**A**,**C**) mean ± SEM of three replicate measurements (*n* = 60--81); **\*\*** *p* \< 0.01, **\*\*\*** *p* \< 0.001 (**B**,**D**) representative images from Days 2 and 4 of culture. Scale bars = 200 µm.](ijms-16-13885-g002){#ijms-16-13885-f002} 2.2. Mechanism of Neuroprotective Effects of DMF May Be through Regulation of the Anti-Oxidative Stress Gene Nrf2 ----------------------------------------------------------------------------------------------------------------- The main mechanism of action that has been attributed to DMF involves direct effects on neuroprotection through activation of antioxidant response elements by the nuclear factor (erythroid-derived 2)-like 2 (Nrf2) pathway \[[@B23-ijms-16-13885]\]. To determine whether *Nrf2* gene and protein were regulated by DMF in rNPCs, we used RT-qPCR and Western blot analysis to evaluate Nrf2 expression levels of DMF-treated rNPCs *in vitro*. Using cDNA or whole cell lysates prepared from rNPCs treated under different conditions with or without oxidative stress, total RNA and protein lysates were prepared for RT-qPCR and Western blot analysis for Nrf2 expression. We found that Nrf2 at both mRNA and protein levels was increased by DMF treatment and this effect was significantly enhanced in the presence of oxidative stress induced by H~2~O~2~ treatment ([Figure 3](#ijms-16-13885-f003){ref-type="fig"}). This suggests that the neuroprotective effects of DMF may be through regulation of the anti-oxidative stress gene *Nrf2*. ![DMF increased Nrf2 RNA and protein expression in rat NPCs in the presence of oxidative stress. After treatment under indicated conditions with DMSO (C), DMF or H~2~O~2~ alone, or DMF and H~2~O~2~ together, Nrf2 RNA expression was determined using cDNAs prepared from rNPCs culture by RT-qPCR (**A**), **\*\*\*** *p* \< 0.001; Nrf2 protein expression was assessed using whole cell lysates prepared from rNPCs cells by Western blot analysis (**B**) as described under "Material and Methods"; (**C**) Mean ± SEM were determined from 3--4 individual experiments. **\*** *p* = 0.026.](ijms-16-13885-g003){#ijms-16-13885-f003} 2.3. Mechanism of Involvement of ERK1/2 MAPK in DMF-Mediated Neuroprotection ---------------------------------------------------------------------------- The ERK1/2 MAPK pathway plays an important role in cell growth and survival \[[@B32-ijms-16-13885]\]. To determine whether the ERK1/2 MAPK pathway is involved in DMF-mediated neuroprotection, we treated NPCs with H~2~O~2~ to induce oxidative stress. Due to the fact that Nrf2 is degraded through proteasomal cleavage, the proteasomal inhibitor, MG132, was used as a control \[[@B26-ijms-16-13885],[@B33-ijms-16-13885]\]. MG132 dramatically increased ERK1/2 MAPK phosphorylation as expected ([Figure 4](#ijms-16-13885-f004){ref-type="fig"}A). DMF treatment alone increased phosphorylation of ERK1/2 MAPK from 1.0 ± 0.09 to 1.7 ± 0.23 and further induction was seen when treated with a combination of DMF and H~2~O~2~ for 4 h (2.7 ± 0.33). We also compared serum-starved NPCs that were treated for 4 h without B-27 supplement with NPCs treated overnight with B-27 supplement. Data showed that overnight treatment did not have an effect on ERK1/2 MAPK phosphorylation as seen with 4 h treatment except with the MG132 positive control. For the subsequent assay ([Figure 4](#ijms-16-13885-f004){ref-type="fig"}B,C), we performed only 4 h treatment. Since different kinases have been reported to regulate Nrf2 expression \[[@B16-ijms-16-13885],[@B34-ijms-16-13885]\], to confirm the involvement of ERK1/2 MAPK pathway, we used MEK1 pathway (ERK1/2) MAPK inhibitor PD98059 \[[@B34-ijms-16-13885]\], which completely blocked DMF- and H~2~O~2~-induced ERK1/2 phosphorylation including DMSO control ([Figure 4](#ijms-16-13885-f004){ref-type="fig"}B,C). The total ERK was not changed indicating that cell viability was not reduced. To further confirm the ERK1/2 MAPK involved in the Nrf2 pathway, the effect of the ERK1/2 inhibitor PD98059 on blocking Nrf2 protein level was investigated ([Figure 5](#ijms-16-13885-f005){ref-type="fig"}). PD98059 completely blocked DMF-induced Nrf2 protein expression. These data indicated that the mechanism of DMF-mediated neuroprotection may be through the involvement of ERK1/2 MAPK upstream of Nrf2 expression. ![Involvement of ERK1/2 MAPK in DMF-mediated neuroprotection. (**A**) rNPCs cultures were serum starved in neurobasal medium with (24 h) or without (4 h) B-27 supplement and treated with medium alone (--), DMSO (C), MG132 (MG), DMF, H~2~O~2~, or both H~2~O~2~ and DMF (H + D). ERK1/2 phosphorylation was determined by Western blot analysis; (**B**) rNPCs cells were serum starved in neurobasal medium and treated with (PD98059) or without (V) the ERK1/2 inhibitor under the same treatment as "A" for 4 h; (**C**) ERK1/2 phosphorylation was presented as a ratio of normalized arbitrary units (a.u.) of phosphorylated ERK1/2 (P-ERK) over total ERK1/2 (T-ERK) in four individual experiments under same treatment as in "B". **\*** *p* = 0.028, **\*\*** *p* = 0.0042.](ijms-16-13885-g004){#ijms-16-13885-f004} ![Effect of ERK1/2 inhibitor on Nrf2 protein level. (**A**) Whole cell lysates were prepared from the rNPCs cultures treated under various conditions with (PD98059) or without (V) the ERK1/2 inhibitor for overnight. Western blot analysis for Nrf2 protein expression was performed as described under "Material and Methods"; (**B**) Mean ± SEM were determined from 3--6 experiments. **\*** *p* = 0.01, **\*\*** *p* = 0.0068.](ijms-16-13885-g005){#ijms-16-13885-f005} 2.4. Identification of Additional Anti-Oxidative Stress Genes by SuperArray Gene Screen Technology -------------------------------------------------------------------------------------------------- When activated, Nrf2 protein dissociates from Keap1 complex, translocates into nucleus, and then activates ARE-mediated phase II antioxidant gene expression \[[@B35-ijms-16-13885]\]. To further determine whether DMF has an effect on up- or down-regulation of antioxidant gene(s) expression, we utilized Oxidative Stress PCR SuperArray gene screen technology to identify additional anti-oxidative stress genes under our conditions with the treatment of DMF and oxidative stress. Our results showed that under oxidative stress with H~2~O~2~, DMF significantly up-regulated gene expression of genes including *Gstp1* (Glutathione *S*-transferase pi 1-, 5-fold increase), *Nqo1* \[NAD(P)H dehydrogenase, quinone 1-, 2-fold increase\], *Sod2* (Superoxide dismutase 2, mitochondrial, 2-fold increase), *Srxn1* (Sulfiredoxin1 homolog, 4-fold increase), and *Fth1* (Ferritin, 2-fold increase) ([Figure 6](#ijms-16-13885-f006){ref-type="fig"}A), which all play roles in reducing oxidative stress by breaking down ROS. Intriguingly, in NPC cultures, DMF also showed a trend of down-regulating genes including *Ccl5* \[Chemokine (C--C motif) ligand 5-, 3-fold decrease\], which is a ligand of CCR1, CCR3, and CCR5 with the potential to reduce the recruitment of leukocytes into inflammatory sites ([Figure 6](#ijms-16-13885-f006){ref-type="fig"}A). Using Real-Time qRT-PCR analysis, *Fth1* and *Gstp1* gene up-regulation were further confirmed ([Figure 6](#ijms-16-13885-f006){ref-type="fig"}B). ![Identification of additional anti-oxidative stress genes by SuperArray gene screening technology. (**A**) SuperArray gene screen technology for anti-oxidative stress genes was performed as described under "Material and methods". DMF significantly up-regulates gene expression of *Gstp1*, *Nqo1*, *Sod2*, *Srxn1*, and *Fth1*, with down-regulation of *Ccl5*; (**B**) Confirmation Real-Time qRT-PCR. Up-regulation of *Gstp1* and *Fth1* genes were further confirmed by Real-Time qRT-PCR. Mean ± SEM were determined from 3 replicate samples. \*\* *p* = 0.0015--0.0031.](ijms-16-13885-g006){#ijms-16-13885-f006} 3. Discussion ============= Our current study has shown that DMF protects different neuronal cell types including neural stem/progenitor cells (NPCs) and differentiated neurons (motor neurons) from oxidative damage through regulation of Nrf2 involving ERK1/2 MAPK pathway. Key *Nrf2* target genes and genes involved in superoxide metabolism were demonstrated to be induced by DMF in rNPCs. The possible role of this mechanism in reducing oxidative stress by breaking down ROS suggests that analysis of anti-oxidative stress mechanisms in NPCs may yield further insights into new targets for treatment of MS. Inflammation and oxidative stress are thought to promote CNS tissue damage in multiple sclerosis by affecting demyelination. Recent studies had described a potential neuroprotective mechanism of action for the fumaric acid ester, dimethyl fumarate (DMF), and its primary metabolite, monomethyl fumarate (MMF), on neurons, astrocytes and glial cells demonstrated by *in vitro* and *in vivo* data \[[@B23-ijms-16-13885],[@B31-ijms-16-13885]\]. Protection of oligodendrocyte and neurons from irreversible damage in CNS may exert neuroprotective effects both directly and indirectly in the periphery and CNS. It has been shown that DMF and MMF are able to activate Nrf2, the key transcription factor regulating the antioxidant response and induce expression of antioxidant proteins \[[@B19-ijms-16-13885]\]. Furthermore, administration of MMF could protect motor neurons and astrocytes against H~2~O~2~-induced oxidative stress \[[@B23-ijms-16-13885]\]. These results may point to an underlying functional cellular substrate for the neuroprotective response observed in mouse EAE, in which DMF treatment reduced oxidative damage and consequential nerve fiber demyelination, resulting in greater axonal preservation and improved motor function \[[@B23-ijms-16-13885],[@B36-ijms-16-13885]\]. Different CNS cell types were used to demonstrate the DMF- and MMF-dependent stabilization of Nrf2, indicating the relevance of these compounds in treating diseases that result in the degeneration of CNS cells. Reports have shown that DMF exerted anti-inflammatory and prometabolic effects in a variety of cell types in primary astrocytes and C6 glioma cells by modifying glutathione (GSH) levels that can induce expression of the anti-inflammatory protein HO-1 (heme oxygenase-1) \[[@B19-ijms-16-13885]\]. DMF treatment increased nuclear levels of Nrf2 in astrocytes but not C6 cells. DMF suppressed inflammatory activation in astrocytes and C6 glioma cells, but with distinct mechanisms, different dependence on GSH, and different effects on transcription factor activation. The effect of DMF on NPCs was not previously demonstrated. We add to this data that DMF directly enhanced NPCs neurosphere growth, and decreased oxidative stress-induced apoptosis in both NPCs and neurons treated with H~2~O~2~. It has been suggested that NPCs and perhaps neurons are capable of responding to DMF treatment, leading to induction of a prototypical Nrf2-dependent antioxidant response. One report also showed that DMF promoted cytoprotection of central nervous system cells against oxidative stress via Nrf2 \[[@B26-ijms-16-13885]\]. Some other studies have shown that antioxidant response proteins like Nrf2 are elevated in MS lesions \[[@B20-ijms-16-13885],[@B37-ijms-16-13885]\]. However, the degree of intrinsic antioxidant induction is apparently insufficient to ultimately prevent disease progression. Neural stem/progenitor cells (NPCs) proliferate and produce new neurons in neurogenic areas throughout the lifetime. While these cells represent potential therapeutic treatment of neurodegenerative diseases, regulation of neurogenesis is not completely understood \[[@B38-ijms-16-13885]\]. Our observations have confirmed that DMF can activate the Nrf2 pathway in rat NPCs and stabilize levels of Nrf2 protein, resulting in the accumulation of active transcription factor, thereby conferring neuroprotective responses against oxidative stress. All cells have an intrinsic mechanism for combating reactive oxygen species (ROS) that is dynamically controlled through the actions of Nrf2, a transcription factor that is the principal regulator of the phase II cellular antioxidant response \[[@B39-ijms-16-13885]\]. The studies described here demonstrated that DMF treatment can decrease apoptosis and increase cell survival in motor neurons under H~2~O~2~ oxidative stress which further supported the DMF neuroprotective effects on primary cultures of motor neurons and suggests a potential role for cell type-specific neuroprotective responses. In addition to characterization of Nrf2 pathway activation in NPC cells, the potential upstream and downstream functional mechanism by which DMF may promote resistance to oxidative stress has not been identified. Though the involvement of mitogen-activated protein kinases (MAPKs) in the Nrf2 pathway has been reported, different subgroups of MAPKs are involved, and their exact functions differ depending on cell types \[[@B35-ijms-16-13885],[@B40-ijms-16-13885]\]. MAPKs play crucial roles in cellular processes, including proliferation, stress responses, apoptosis, and immune defense \[[@B41-ijms-16-13885]\]. Some reports have shown that Nrf2 protein accumulation has been induced by activation of MAPKs. ERK1/2 is one of three major subgroups of MAPKs. An ERK pathway inhibitor prevented Nrf2 phosphorylation at Ser40 to retard Nrf2 nuclear translocation, thus decreasing antioxidant gene transcription \[[@B35-ijms-16-13885],[@B40-ijms-16-13885],[@B42-ijms-16-13885]\]. There is no previous report on DMF neuroprotection via ERK1/2 MAPK-induced Nrf2 accumulation in neurons and NPCs. Here we showed that DMF increased Nrf2 accumulation and oxidative stress enhanced this effect through ERK1/2 phosphorylation which was completely abolished upon treatment with the ERK1/2 inhibitor, PD98059 \[[@B43-ijms-16-13885],[@B44-ijms-16-13885]\]. The neuroprotective action of DMF on NPCs is therefore likely through increasing Nrf2 accumulation by activation of the ERK1/2 signaling pathway. This does not exclude the possibility that other signaling pathways may also be involved, such as the NFκB pathway, which also decreases apoptosis of neurons, astrocytes, microglia and oligodendrocytes \[[@B19-ijms-16-13885],[@B45-ijms-16-13885],[@B46-ijms-16-13885]\]. Currently, we cannot rule out the possibility that DMF may act as a Keap1 inhibitor that leads to the accumulation of Nrf2 protein by decreasing proteasomal degradation \[[@B47-ijms-16-13885],[@B48-ijms-16-13885],[@B49-ijms-16-13885],[@B50-ijms-16-13885]\]. This neuroprotective role of DMF may explain the decreased rate of brain atrophy and the slowed rate of disease progression in patients with MS treated with Tecfidera™ \[[@B51-ijms-16-13885],[@B52-ijms-16-13885]\]. Elucidation of this pathway suggests a potential role for cell type-specific neuroprotective responses. Multiple genes have been reported to be candidates for direct targets of Nrf2 activation in different cell types and tissues which are enhanced only in the presence of the *Nrf2* gene \[[@B53-ijms-16-13885],[@B54-ijms-16-13885]\]. The roles of these *Nrf2* target genes in protection from oxidative stress were unknown for NPCs. To our knowledge, this is the first report to screen *Nrf2* target genes and genes involved in superoxide metabolism in rat NPCs. Hence this is the first report in which five *Nrf2* targeted genes (*Fth1*, *Gstp1*, *Sod2*, *Nqo1* and *Srxn1*) have been identified as strong candidates for DMF-mediated Nrf2 activation in rat NPCs. Under oxidative stress, DMF significantly up-regulated the expression of these genes, which all play roles in reducing oxidative stress by breaking down ROS. Sod2 is a critical component of the mitochondrial pathway for detoxification of O~2~^−^ and targeted disruption of this locus leads to embryonic or neonatal lethality in mice \[[@B55-ijms-16-13885]\]. Gstp1 is a phase II detoxification enzyme that conjugates both endogenous and exogenous compounds to glutathione to reduce cellular oxidative stress. Decreased expression of Gstp1 has recently been implicated in dopamine neuron degeneration in Parkinson's disease progression \[[@B56-ijms-16-13885]\]. Nqo1 has been shown to be upregulated in Sulforaphane protection of cortical neurons against 5-*S*-cysteinyl-dopamine-induced toxicity through the activation of the ERK1/2-Nrf2 pathway \[[@B57-ijms-16-13885]\]. Srxn1 is an endogenous antioxidant protein that has neuroprotective effects. While the mechanisms of Srxn1 in oxidative stress have not been well studied, it has been shown that Srxn1 protects PC12 cells against oxidative stress induced by hydrogen peroxide \[[@B58-ijms-16-13885]\]. Srxn1 expression has also been shown to mediate protective effects of brain-derived neurotrophic factor (BDNF) against neurotoxicity in primary rat cortical cultures induced by 3-nitropropionic acid (3-NP), a mitochondrial complex II inhibitor \[[@B59-ijms-16-13885]\]. *Fth1* gene has been implicated in protection against neuronal cell death under various neurodegenerative stresses, such as hypoxia-ischemia and oxidative stress and its increased expression may have protective effects against neural damage that could otherwise result in cell death \[[@B60-ijms-16-13885]\]. DMF also showed a trend of down-regulation of Ccl5, a ligand of CCR1, CCR3, and CCR5, which could potentially reduce the recruitment of leukocytes into CNS inflammatory sites, thereby decreasing cell trafficking and inflammation \[[@B61-ijms-16-13885]\]. In particular, abnormal Ccl5 expression was detected in the cerebrospinal fluid (CSF) of patients suffering from MS and in the CNS of EAE mice \[[@B62-ijms-16-13885],[@B63-ijms-16-13885],[@B64-ijms-16-13885],[@B65-ijms-16-13885],[@B66-ijms-16-13885]\] implicating this chemokine in the onset of the demyelinating disease(s). 4. Materials and Methods ======================== All experiments were done in accordance with the NIH Guide for Care and Use of Laboratory Animals, as approved by the R&D Service of the VA Ann Arbor Healthcare Center on 21 November 2012 (ACORP 2012-090604) and the University Committee on Use and Care of Animals (UCUCA) of the University of Michigan on 6 June 2013 (PRO 00004815). 4.1. Preparation of Rat NPCs and Mouse Neurospheres --------------------------------------------------- NPCs were isolated based on neurosphere formation and confirmed through staining with the NPC marker nestin as previously described \[[@B67-ijms-16-13885],[@B68-ijms-16-13885],[@B69-ijms-16-13885]\]. Rat embryonic E15-18 pups were decapitated, and the heads were placed into a petri dish with ice-cold HBSS (Ca^2+^, Mg^2+^ free). The whole brain without cerebellum was dissected and washed with fresh HBSS solution. It was then digested with 25 µg/mL DNase I in ice-cold HBSS and filtered through a cell strainer (70 µm, Falcon, BD Biosciences, San Jose, CA, USA). Cells were counted and plated onto 100 mm dishes pre-equilibrated with self-renewal media (SRM) containing DMEM, Neurobasal-A media, 2-mercaptoethanol, Chick Embryo Extract (CEE), 1× N-2 supplement, 2× B-27 supplement, FGF (20 ng/mL), EGF (20 ng/mL), and Penicillin-Streptomycin. Mouse neurosphere preparation and growth were carried out as described previously \[[@B69-ijms-16-13885]\]. Briefly, neonatal mouse brain was dissected and washed with fresh cold HBSS solution. Two coronal slices (as thin as possible, \~1--2 mm) were cut starting at the olfactory bulb. The lateral wall of the lateral ventricle was removed and mechanically triturated in a 15 mL conical tube with 1:3 dilution of trypsin. These cells were incubated for 4 min at 37 °C, quenched in two volumes of medium containing 25 µg/mL DNase I and spun at 210× *g* at 4 °C for 5 min. The pellets were re-suspended in ice-cold HBSS, counted, and plated at a density of 1000 cells/well onto 6-well low binding tissue culture plates pre-equilibrated with SRM. After 8--10 days in culture the neurospheres were counted, measured, and selected for the self-renewal assay. The frequency of neurospheres was calculated with the numbers of neurospheres formation from originally plated 1000 cells. To assay self-renewal potential, individual primary neurospheres (\>50 μm in diameter) were dissociated and re-plated at clonal density in non-adherent secondary cultures. 4.2. Primary Rat Motor Neurons Preparation and Culture ------------------------------------------------------ Primary motor neurons were cultured as previously described \[[@B70-ijms-16-13885]\]. Briefly, perineural membranes were removed from spinal cords of E15 Sprague-Dawley pregnant rats and the tissue was chopped into 2-mm pieces. Cells were dissociated by incubating in 0.25% trypsin/EDTA for 15 min at 37 °C followed by gentle trituration for one minute with a serum-coated, fire-polished glass pasteur pipette. Motor neurons were isolated over 9% Optiprep in L-15 media by centrifugation at 1000× *g* for 15 min and then collected by taking 2 mL of supernatant. Cells were washed in L-15 media, then resuspended and plated in culture medium. Defined culture media consisted of Neurobasal (Invitrogen, Carlsbad, CA, USA) supplemented with 2× B27 and the following additives: 2.5 mg/mL albumin, 2.5 µg/mL catalase, 2.5 µg/mL superoxide dismutase, 0.01 mg/mL transferrin, 15 µg/mL galactose, 6.3 ng/mL progesterone, 16 µg/mL putrescine, 4 ng/mL selenium, 3 ng/mL b-estradiol, 4 ng/mL hydrocortisone, and 1× penicillin/streptomycin/neomycin. [l]{.smallcaps}-Glutamine (2 µM) was added to the medium immediately before plating. Cells were counted and plated at a density of 25 cells/mm^2^, so that neurons would not contact one another. 4.3. Immunocytochemistry and Morphological Analysis --------------------------------------------------- Immunohistochemistry techniques were performed as previously described \[[@B70-ijms-16-13885]\]. Briefly, cells were fixed in 4% paraformaldehyde at room temperature for 30 min. To block nonspecific antibody binding, samples were incubated in 2% goat serum/2.5% BSA/0.05% Triton-X-100 in 1× PBS for 30 min. Primary antibody (anti-β-Tubulin III diluted 1:1000, Sigma, St. Louis, MO, USA) was prepared in 10% goat serum/2.5% BSA/0.05% Triton-X-100/0.1% sodium azide in 1× PBS and incubated with cells overnight. Cells were then washed in 1× PBS and incubated in the secondary antibody Goat anti-mouse IgG (FITC) diluted 1:200 (Sigma) for 4 h. Prolong Gold (Molecular Probes/Invitrogen, Carlsbad, CA, USA), an antifade agent with 4′,6-diamidino-2-phenylindole (DAPI), was used to stain nuclei. The neurons were imaged using an EVOS fluorescent microscope and analyzed with Metamorph imaging software (Molecular Devices, LLC. Sunnyvale, CA, USA). Neurite number, total neurite growth, mean neurite growth, and cell body area were all recorded. 4.4. Treatments --------------- Dimethyl fumarate (DMF) and Hydrogen peroxide solution (H~2~O~2~, 30% *w*/*w* in H~2~O, contains stabilizer) was purchased from Sigma--Aldrich (St. Louis, MO, USA). DMF was dissolved in DMSO and aliquots were kept in dark in −20 °C freezer. H~2~O~2~ was kept at 4 °C refrigerator. Newly opened bottle of H~2~O~2~ was used and freshly prepared and diluted H~2~O~2~ stock solution was immediately added into cell culture medium to achieve the desired working concentration. The concentration responses for both H~2~O~2~ (ranging from 0, 10, 20, 40, 80, and 100 μM) and DMF (ranging from 0, 10, 20, and 40 μM) and time course for both H~2~O~2~ and DMF (ranging from 4, 6, 18, 24, 48, and 72 h) were studied in the culture medium and then the optimal conditions (40 µm H~2~O~2~ and 10 µm DMF) for studying DMF neuroprotective effect under H~2~O~2~-induced oxidative stress were determined for subsequent experiments. In most experiments unless otherwise noted, the cultures were divided into four different groups which were treated in the following way: (1) Control with same concentration of DMSO used to dissolve DMF; (2) DMF alone; (3) H~2~O~2~ alone with same concentration of DMSO; (4) DMF + H~2~O~2~. At least three separate cultures were set up for each experiment. 4.5. Survival Assay ------------------- Motor neurons were isolated from E15 pregnant Sprague Dawley rats and plated onto poly (lactic-*co*-glycolic acid) (PLGA) solvent-casted, poly-[l]{.smallcaps}-lactic acid (PLLA) electrospun nanofibers coated with poly-[l]{.smallcaps}-lysine (PLL) \[[@B70-ijms-16-13885]\]. Three separate groups of cells were cultured: a control group of motor neurons on fibers treated with 0.1% DMSO, a group of cells cultured with 40 µM H~2~O~2~ (the stress group), and a group of cells cultured with both 40 µM H~2~O~2~ and 20 µM DMF (the treatment for this experiment). Cells were cultured for one day before any addition of treatment (DMF) or stress H~2~O~2~. In a separate experiment, cells were subjected to 20 µM DMF alone and did not show any difference in growth when compared to controls. Concentrations of H~2~O~2~ and DMF were determined in pilot experiments on neurons for optimal destructive/neuroprotective results. 4.6. Electrospinning -------------------- The electrospinning methods were similar to those previously described \[[@B70-ijms-16-13885]\]. Briefly, 0.4 g of poly-[l]{.smallcaps}-lactic acid (PLLA) was dissolved in a total volume of 10 mL in parts 9:1 chloroform: dimethylformamide, to yield a 4.4% *w*/*v* solution. The mixture was loaded into a 3 mL syringe with a 22 gauge blunt-tip needle and placed on a syringe pump set to 0.22 mL/h. The tip of the needle was protruded through the center of a 10 × 10 cm folded aluminum sheet. The rotating disc collector was placed 30 cm away, such that the axis of rotation was perpendicular to the syringe. A high voltage power supply applied a 20 kV charge to the aluminum sheet via an alligator clip. A second power supply applied a counter charge of −2 kV to the rotating disc via a wire brush. Fibers were collected for 5 min or until a desired density was obtained. 4.7. Apoptosis Assay -------------------- Rat NPCs were grown in adherent cultures in self-renewal medium. Cultures were pre-treated with indicated concentration of DMF (or DMSO control) overnight. After H~2~O~2~ treatment, the NPC cells were trypsinized, washed with 1× Annexin binding buffer and stained with APC-Annexin V kit (BD Biosciences, San Jose, CA, USA), according to the manufacturer's instructions. 7-AAD viability staining solution (BioLegend, San Diego, CA, USA) was used to determine the necrotic cells. Staining was analyzed by flow cytometry and the apoptotic cell population was calculated by the percent of Annexin V positive cells. 4.8. ROS Assay -------------- Rat NPCs were grown in adherent cultures overnight following pretreatment with DMF or DMSO (control) in minimal medium containing Neurobasal-A medium plus 1:1:2 ratio of penicillin/streptomycin:[l]{.smallcaps}-Glutamine:glucose at 1:100 dilution. After induction of oxidative stress by addition of H~2~O~2~, cells were incubated with 5 µM 2′-7′-dichlorofluorescein diacetate (CM-H2DCFDA, Life Technology, Grand Island, NY, USA) for 15 min at 37 °C, washed, and the ROS levels were measured by flow cytometry \[[@B71-ijms-16-13885]\]. 4.9. Western Blot Analysis for Nrf2 Protein ------------------------------------------- After treatment, whole cell lysates were prepared in RIPA lysis buffer containing protease inhibitors, subjected to SDS-PAGE on an 8% mini-gel and transferred to Immobilon™-P transfer membrane as previously described \[[@B33-ijms-16-13885]\]. The membrane was blocked and cut at 75 kDa marker. The upper part of membrane was incubated with 1 µg/mL anti-Nrf2 antibody (R&D Systems Tools for Cell Biology Research) \[[@B72-ijms-16-13885]\] and the bottom part of membrane was incubated with anti-α tubulin (Sigma) as a loading control at a 1:10,000 dilution for overnight in a 4 °C cold room. After three consecutive washes with TBS plus 0.1% Tween 20 (10 min for each), the membranes were incubated with 1:10,000 secondary antibody (anti-mouse IgG-HRP, Sigma) for 1 h at room temperature. Prestained SDS-PAGE protein standards (Bio-Rad, Hercules, CA, USA) were used to determine the size of detected proteins. Proteins were visualized by chemiluminescence with SuperSignal West Pico (Pierce, Grand Island, NY, USA) and exposed to X-ray film. 4.10. ERK1/2 Phosphorylation Assay ---------------------------------- Rat NPCs were plated in a 24-well plate and changed to serum-free medium for 4 h before the assay. The ERK1/2 phosphorylation assay was performed as previously described \[[@B43-ijms-16-13885]\]. Briefly, cells were washed and removed from the wells, boiled, and then subjected to electrophoresis using a 12% SDS-PAGE mini-gel, followed by transfer to an Immobilon™-FL membrane (Fisher, Loughborough, UK) for Western blotting. After blocking, the blot was probed with mouse anti-phospho-p44/42MAPK (Thr202/Tyr204) antibody and rabbit anti-p44/42 MAPK antibody (Cell Signaling Technology, Danvers, MA, USA) overnight on a rocking shaker in a 4 °C cold room. Then, the blot was incubated with the secondary antibodies, anti-mouse IRdye 680 RD and anti-rabbit IRdye 800 CW, for 1 h at room temperature. Finally, images were acquired using an Odyssey FC imaging system (Li-COR Biosciences, Lincoln, NE, USA) and quantified using the manufacturer's analysis program. The ERK1/2 phosphorylation was calculated as the ratio of normalized arbitrary units (a.u.) of phosphorylated ERK1/2 over total ERK1/2. 4.11. RNA Purification and Gene Expression Analysis --------------------------------------------------- Rat NPCs were grown in adherent cultures and treated under various conditions described above. Then the total RNAs were prepared using RNeasy RNA purification kit (Qiagen, Valencia, CA, USA) according to the manufacturer's instruction. RNA integrity was assessed with the Agilent 2100 Bioanalyze at the University of Michigan microarray facility. RNA concentration was quantified using Nanodrop (Thermo Scientific, Grand Island, NY, USA) and the samples were tested for absence of genomic DNA contamination. Gene expression of oxidative stress related genes were assessed using Oxidative Stress PCR SuperArray \[[@B73-ijms-16-13885]\]. RNA samples (80 ng each) were reverse-transcribed to cDNA using the SuperArray RT^2^ First Strand Kit (Qiagen, Valencia, CA, USA). cDNA samples were next subjected to Oxidative Stress PCR SuperArray. PCR amplifications (Applied Biosystems 7900HT, Thermo Scientific, Grand Island, NY, USA) were completed with PCR plates (Superarray Cat No. PAHS-031A) assaying for 84 neurotrophic factors and receptors and five houkeeping genes. Results were quantified using (2^−ΔΔ*C*t^) method and expressed as a fold difference up/down with respect to the control condition. 4.12. Real-Time RT-qPCR ----------------------- The candidate genes that were either up- or down-regulated by DMF treatment were confirmed by real time-quantitative polymerase chain reaction (RT-qPCR) using Applied Biosystems 7500 (Thermo Scientific, Grand Island, NY, USA). Amplification consisted of 40 cycles (95 °C for 15 s and 60 °C for 1 min) with approximately 15 ng/μL cDNA, *TaqMan* Master Mix and specific primer pairs were purchased from Applied Biosystems (assay on demands primers). 4.13. Statistical Analysis -------------------------- Experimental data was analyzed using GraphPad Prism version 6.0 (GraphPad Software Inc., San Diego, CA, USA). The unpaired *t*-tests were performed for all experiments and values were expressed as mean ± SEM from at least three individual experiments. *p*-Values were considered significant at **\*** *p* \< 0.05 and highly significant at **\*\*** *p* \< 0.01 or **\*\*\*** *p* \< 0.001. ![Schematic model of DMF-induced neuroprotection in rat NPCs and differentiated neurons. Under oxidative stress (H~2~O~2~), DMF enhances Nrf2-mediated antioxidant response gene transcription through activation of ERK1/2 MAPK signaling pathway leading to neuroprotection in rNPCs and differentiated neurons.](ijms-16-13885-g007){#ijms-16-13885-f007} 5. Conclusions ============== DMF protected neural stem/progenitor cells and differentiated neurons from oxidative damage through regulation of Nrf2 involving the ERK1/2 MAPK pathway. We propose that under oxidative stress (H~2~O~2~), DMF enhances Nrf2-mediated antioxidant response gene transcription through activation of the ERK1/2 signaling pathway leading to neuroprotection in rat NPCs and differentiated neurons ([Figure 7](#ijms-16-13885-f007){ref-type="fig"}). The key *Nrf2* target genes and genes involved in superoxide metabolism are involved in reducing oxidative stress by breaking down ROS, decreasing apoptosis and increasing survival. Investigating the anti-oxidative stress mechanisms may yield further insights into new targets for treatment of multiple sclerosis (MS). The authors thank Catherine Dowling and Kelly VanDerberg for their critical reading the manuscript and also thank our collaborator Benjamin Segal for the support. The work of Y.M.D., Qin W., Qi W., S.C., and S.L. were supported by NIH (NINDS) 1R01NS080821-01A1 and NIH (NIAID) UM1 AI110557; A.R., S.J.T., and J.M.C. were supported by VA Rehab R&D Project B7025-R; S.T. was supported by NIH (NIAID) training grant T32 AI007413. Qin Wang, Sophina Taitano, Sergei Chuikov, Arjun Rastogi, Joseph M. Corey, and Yang Mao-Draayer. participated in research design; Qin Wang, Sophina Taitano, Sergei Chuikov, Arjun Rastogi, Qi Wu, and Samuel J. Tuck conducted experiments and data analysis; Qin Wang, Arjun Rastogi, Joseph M. Corey, Steven K. Lundy, and Yang Mao-Draayer wrote or contributed to the writing of the manuscript. The authors declare no conflict of interest. MS : multiple sclerosis MAPK : mitogen-activated protein kinase ERK1/2 : extracellular-signal-regulated kinases 1/2 FGF : recombinant human basic fibroblast growth factor 146 aa EGF : epidermal growth factor RIPA : radio-immunoprecipitation assay rNPCs : rat neural stem/progenitor cells ROS : reactive oxygen pecies CNS : central nervous system FAE : Fumaric acid esters MMF : monomethyl fumarate DMF : dimethyl fumarate H~2~O~2~ : hydrogen peroxide Nrf2 : transcription factor nuclear factor-erythroid 2-related factor 2 EAE : experimental autoimmune encephalomyelitis MOG : myelin oligodendrocyte glycoprotein ARE/EpRE : antioxidant/electrophile response element IL-10 : interlukin-10 Keap1 : kelch-like ECH-associated protein 1 CEE : chick embryo extract PLGA : poly lactic-*co*-glycolic acid PLLA : poly-[l]{.smallcaps}-lactic acid PLL : poly-[l]{.smallcaps}-lysine SRM : self-renewal medium EGFP : enhanced green fluorescent protein CM-H2DCFDA : chloromethyl derivative of 2′,7′-dichlorofluorescein diacetate RT-qPCR : real time quantitative reverse transcription polymerase chain reaction GSH : glutathione HO-1 : heme oxygenase-1 DMSO : dimethyl sulfoxide FITC : fluorescein isothiocyanate DAPI : 4′,6-diamidino-2-phenylindole *Gstp1* : glutathione *S*-transferase pi 1 *Nqo1* : NAD(P)H dehydrogenase, quinone 1 *Sod2* : superoxide dismutase 2 *Srxn1* : sulfiredoxin 1 homolog *Fth1* : ferritin 1 *Ccl5* : chemokine (C--C motif) ligand 5

{ "pile_set_name": "PubMed Central" }

1. Introduction {#sec1} =============== Odontogenic myxoma (OM) is a benign, locally invasive, aggressive, nonmetastasizing neoplasm of the jaw bones. It is the third most common odontogenic lesion with an annual incidence of \~0.07 new patients per million people \[[@B1], [@B2]\]. Although OM is a benign tumor, radical treatment is needed because the rate of local recurrence ranges from 10% - 33% \[[@B3]-[@B5]\]. Surgical resection with a minimum bone margin of 1 cm has been advocated \[[@B6]-[@B9]\], but emerging evidence suggests that a more conservative surgical approach will result in acceptable recurrence rates with less patient morbidity if follow-up can be over the long term \[[@B7], [@B10]-[@B12]\]. However, little has been reported about the actual long-term follow-up of patients with OM after conservative surgery. Here, we describe a mandibular odontogenic myxoma that was treated by conservative surgery and followed up for 100 months thereafter. 2. Case Report {#sec2} ============== 2.1. Patient Information {#sec2.1} ------------------------ A 56-year-old Japanese man was referred from a dental clinic for further examination of a radiolucent finding on the left side of his mandible in January 2010. The patient reported having no symptoms in his mouth including the left mandible area. His medical and dental history was noncontributory. An extraoral examination upon initial presentation revealed unremarkable findings and no complaints of paresthesia. An intraoral examination also confirmed the absence of redness and tender swelling of the left mandibular mucosa ([Figure 1](#fig1){ref-type="fig"}). Panoramic radiography revealed an extensive multilocular radiolucent area with imprecise borders and a "soap bubble appearance" ([Figure 2](#fig2){ref-type="fig"}). Computed tomography showed an approximately 39 × 19 × 11 mm tumor that extended to the roots of four teeth (\#33 - 36; Figures [3(a)](#fig3){ref-type="fig"} and [3(b)](#fig3){ref-type="fig"}). We considered that the odontogenic tumor was benign and an incisional biopsy was performed under local anesthesia. The histopathological findings revealed loosely arranged spindle-shaped cells in a myxoid fibrous stroma, indicating a clinical diagnosis of an odontogenic tumor. Segmental resection of the mandible was planned. The patient was given repeated and careful explanations about the high likelihood of recurrence, but he insisted upon a more conservative approach as he desired functional and cosmetic preservation. Conservative surgery then proceeded under general anesthesia after endodontic treatment of \#33 -- 36 was completed. The surgery consisted of extracting the second premolar from the left mandible, followed by total enucleation and vigorous curettage of the bone ([Figure 4(a)](#fig4){ref-type="fig"}). The surgical specimen ([Figure 4(b)](#fig4){ref-type="fig"}) revealed apparently benign, spindled-shaped cells in a loose and abundant mucoid stroma (Figures [5(a)](#fig5){ref-type="fig"} and [5(b)](#fig5){ref-type="fig"}). These findings confirmed the diagnosis of odontogenic myxoma. The immediate postoperative period and wound healing were uneventful. The patient underwent monthly clinical examinations for the first year thereafter, then every two months during the second year. Panoramic X-rays were obtained every three months for the first two years. Annual follow-up for eight years included panoramic X-rays and CT imaging (Figures [6(a)](#fig6){ref-type="fig"} and [6(b)](#fig6){ref-type="fig"}, respectively), which showed no clinical or radiological signs of recurrence. 3. Discussion {#sec3} ============= Odontogenic myxomas are very rare (\<10% all odontogenic tumors) benign tumors of the ectomesenchymal and/or mesenchymal origin \[[@B13], [@B14]\]. They are locally invasive and slow-growing, and their pathological characteristics in the tooth-bearing areas of the mandible and maxilla are well defined \[[@B13]-[@B17]\]. The radiographic features are described as those of a multilocular lesion with a "soap bubble" or "spider web" appearance \[[@B14]-[@B17]\]. Such lesions are often discovered incidentally during routine dental checkups, and about 60% of patients are in their second or third decade of life \[[@B4], [@B18]\]. Although the present patient was asymptomatic, panoramic radiography of the left mandible revealed an extensive radiolucent and multilocular area with imprecise borders that extended from the root of tooth \#33 to that of tooth \#36. The treatment strategy for OM remains controversial. Because OM is locally aggressive and it can potentially cause extensive bony destruction, the treatment of choice seems to be radical surgery such as segmental resection. Indeed, complete surgical removal with curettage and peripheral ostectomy alone seems insufficient since OM is not encapsulated and myxomatous tissue infiltrates adjacent bone \[[@B3], [@B6]-[@B9]\]. Thus, the only initial treatment option for extended OM in principle is surgical resection because of the high risk of recurrence which reportedly ranges from 10% to 33% \[[@B4], [@B5]\]. In addition, recurrence rates can reach 25% after simple enucleation and curettage alone \[[@B19], [@B20]\]. On the other hand, some reports described that the choice of current recommended therapy depends on the size of the lesion and its nature and behavior between curettage and radical excision \[[@B3], [@B7]\]. In the present case, CT had shown that the tumor included the \#33 - \#36 apex and extended to the inferior border of the mandible. Therefore, radical resection such as block resection for mandible was strongly recommended. However, the patient rejected this strategy because he desired functional and cosmetic preservation, and in fact, conservative treatment with enucleation, curettage, and marginal resection would confer several advantages over the radical approach. It is substantially less invasive, it can be surgically implemented via an intraoral approach, and it offers the possibility of preserved nerve function and aesthetics and a shorter stay in the hospital. Recent evidence suggests that a more conservative approach will result in acceptable recurrence rates with less morbidity to patients if long-term follow-up is provided \[[@B7], [@B10]-[@B12]\]. [Table 1](#tab1){ref-type="table"} describes 20 reports of 37 patients with that was treated by conservative surgery \[[@B4], [@B6], [@B7], [@B10], [@B11], [@B17], [@B18], [@B21]-[@B33]\]. The mean age of the patients was 27.0 ± 15.2, and 20 (54%) were female. The tumors in almost all of them were managed by enucleation and curettage and recurred in 7 (18.9%). One and six of the procedures with recurrence had undergone marginal resection and enucleation with curettage, respectively. The patients were followed up for a mean of 49.2 ± 42.9 (range, 2 -- 180) months. Including the present patient, only five have been followed up for over 100 months and the tumor recurred in one of them. The recurrence rate among all patients who were treated by conservative surgery was 19.0%. This rate is relatively lower than that previously described \[[@B19], [@B20]\]. Furthermore, our investigation of the literature indicated that the rate of recurrence decreases from 24.0% to 8.3% when the follow-up period exceeds 60 months. The main reason for recurrence is thought to be incomplete removal rather than the intrinsic biological behavior of the OM \[[@B34]\]. Although the tendency is towards a more conservative surgical approach for children and a more radical approach for adults, Kansy et al. does not support this management strategy because the recurrence rates between enucleation and segmental resection are similar \[[@B17]\]. Tumor size has recently been considered to be a factor in the choice of a radical or more conservative surgical approach \[[@B4], [@B10], [@B11]\]. Conservative surgical procedures such as enucleation and curettage are recommended when the diameter of OM is \<3 cm, whereas a radical approach such as segmental resection with immediate reconstruction is preferred when patients have larger tumors \[[@B11]\]. The conservative surgical recommendation is to enucleate a lesion with wide curettage of normal tissue or a generous amount of apparently intact tissue or even marginal resection of the mandible \[[@B12]\]. This approach has the advantage of preserving vital structures and maintaining oral function, and it could be applied to treat OM that recur after simple surgery \[[@B12]\]. However, one patient who developed recurrent OM 15 years after the initial procedure including tumor resection with the preservation of mandibular continuity has been described \[[@B17]\]. More radical surgery is inevitable for a large number of patients due to a tendency towards more extensive OM with significant destruction of key structures \[[@B17], [@B27]\]. Moreover, the rate of OM recurrence remains vague because few reports described long-term follow-up after conservative surgery. Thus, more evidence about long-term outcomes after conservative surgical treatment of OM is needed. 4. Conclusions {#sec4} ============== A conservative surgical approach comprising enucleation and curettage can be effective for OM management. Recurrence rates decreased from 24.0% to 8.3% among patients who were treated with conservative surgery and followed up for over 60 months. The risk of recurrence is likely to be considerable, and long-term follow-up is indispensable for the conservative management of OM. Conflicts of Interest ===================== The authors have no conflicts of interest to declare. ![Intraoral photography before surgery. The image shows no symptoms such as redness or swelling of mucosa in the mandible.](CRID2019-1634842.001){#fig1} ![Panoramic dental radiography before surgery. The image shows extensive multilocular radiolucent area with imprecise borders and "soap bubble appearance".](CRID2019-1634842.002){#fig2} ![Computed tomography (CT) image before surgery. The axial (a) and coronal (b) CT images show tumor infiltration of cortical bone extending to the inferior mandibular border.](CRID2019-1634842.003){#fig3} ![Surgical procedures and resected specimen. The total enucleation and wide curettage of the surrounding bone (a) and resected specimen (b).](CRID2019-1634842.004){#fig4} ![Pathophysiological findings of stained specimen. Hematoxylin and eosin stain ×100 (a) and ×400 (b) magnification.](CRID2019-1634842.005){#fig5} ![Images of the jaw at 96 months after surgery. The panoramic dental radiograph (a) and computed tomography image (b) show no signs of recurrence.](CRID2019-1634842.006){#fig6} ###### Clinical reports of mandibular odontogenic myxoma published after 1990. No. Author Year Age Sex Follow-up period (months) Treatment Recurrence Size (mm) ----- --------------------------------- ------ ----- ----- --------------------------- ----------- ------------ -------------- 1 Oliveira et al. \[[@B21]\] 2018 9 F 6 E/C None NA 2 Albanese et al. \[[@B22]\] 2012 25 F 6 E None 21.2 × 47.6 3 Mauro et al. \[[@B23]\] 2012 6 M 6 E/C None NA 4 Subramaniam et al. \[[@B7]\] 2016 16 --- 7 M None NA 5 Shivashankara et al. \[[@B24]\] 2017 13 M 12 E/N None 40 × 20 6 Subramaniam et al. \[[@B7]\] 2016 18 --- 12 E None NA 7 Miranda Rius et al. \[[@B25]\] 2013 55 M 12 E/C None 33 × 28 8 Hammad et al. \[[@B26]\] 2016 45 F 13 M None 50 × 30 9 Francisco et al. \[[@B27]\] 2017 9 F 14 E/C Recurrence NA 10 Francisco et al. \[[@B27]\] 2017 12 F 16 E/C Recurrence NA 11 Sumi et al. \[[@B28]\] 2000 48 M 22 E/C None 70 × 25 × 15 12 Mittal et al. \[[@B29]\] 2016 48 F 24 E/C Recurrence 25 × 20 13 Lin and Basile \[[@B30]\] 2010 25 F 24 E None NA 14 Lo Muzio et al. \[[@B4]\] 1996 21 M 24 E/C None NA 15 Lo Muzio et al. \[[@B4]\] 1996 28 M 24 E/C Recurrence NA 16 Bucci et al. \[[@B31]\] 1993 28 M 24 E/C None 43 × 40 17 Francisco et al. \[[@B27]\] 2017 7 F 26 E/C None NA 18 Francisco et al. \[[@B27]\] 2017 15 F 27 E/C Recurrence NA 19 Lo Muzio et al. \[[@B4]\] 1996 16 F 31 E/C None NA 20 Francisco et al. \[[@B27]\] 2017 30 F 34 E/C None NA 21 Boffano et al. \[[@B11]\] 2011 38 M 38 E/C None 25 22 Boffano et al. \[[@B11]\] 2011 42 F 40 E/C None 30 23 Boffano et al. \[[@B11]\] 2011 20 M 42 E/C None 20 24 Rajasekhar et al. \[[@B32]\] 2008 17 F 48 M None 70 × 30 25 Lo Muzio et al. \[[@B4]\] 1996 22 F 48 E/C Recurrence NA 26 Takahashi et al. \[[@B6]\] 2018 37 F 73 E/C None 40 × 19 × 12 27 Chaudhary et al. \[[@B33]\] 2015 25 F 84 E/C/M None NA 28 Li et al. \[[@B18]\] 2006 7 M 84 E/C None NA 29 Li et al. \[[@B18]\] 2006 32 M 84 E/C None NA 30 Lo Muzio et al. \[[@B4]\] 1996 65 F 84 E/C None NA 31 Francisco et al. \[[@B27]\] 2017 17 M 85 E/C None NA 32 Francisco et al. \[[@B27]\] 2017 11 F 98 E/C None NA 33 Present case 56 M 100 E/C None 39 × 19 × 11 34 Francisco et al. \[[@B27]\] 2017 27 F 117 E/C None NA 35 Kawase-Koga et al. \[[@B10]\] 2014 40 M 120 E/C None 40 × 30 × 15 36 Li et al. \[[@B18]\] 2006 37 M 132 C None NA 37 Kansy et al. \[[@B17]\] 2012 33 F 180 M Recurrence NA Abbreviations: M: male, F: female, E: enucleation, C: curettage, M: marginal resection, and NA: not available. [^1]: Academic Editor: Eugenio Maiorano

{ "pile_set_name": "PubMed Central" }

Introduction ============ Chronic insomnia is one of the most common sleep disorders, affecting one third of the population, and should receive greater recognition.[@B01],[@B02] Symptoms are considered clinically relevant if they occur at least three times/week and persist for 3 or more months despite adequate sleep opportunities, leading to sleep dissatisfaction.[@B03] Some insomnia patients underestimate total sleep time (TST). It is hypothesized that these individuals have higher cortical arousability and self-reported rates of insomnia, as well as other psychiatric symptoms. Other patients, however, might complain less but present objective sleep disturbance, being at higher risk of metabolic or cardiovascular outcomes.[@B04] From a mental health point of view, it is unlikely that insomnia, either chronic or recurrent, is the sole disorder, but is rather part of a complex multimorbidity. Untreated individuals are in a state of chronic sleep deprivation and have an increased risk of accidents,[@B05] work or social problems, and compromised health and quality of life.[@B01],[@B02] According to regulatory agencies, non-benzodiazepine GABA-A receptor agonists are a standard insomnia treatment, and prescription rates are on the rise.[@B01],[@B02],[@B06],[@B07] Zolpidem is the most popular non-benzodiazepine GABA-A receptor agonist. In therapeutic dosages, it reduces sleep latency and the frequency of nocturnal awakenings and increases TST and sleep efficiency. Pharmacokinetic parameters could differ for women and older adults, and have been associated with increased residual impairment and adverse events.[@B08] These findings have led the U.S. Food and Drug Administration to recommend lower initial dosages for these patients.[@B09] Nevertheless, mid- and long-term trials focusing on objective and subjective measures of effectiveness are scarce,[@B10] as are descriptions of complementary therapeutic schemes or dosing regimens.[@B11] The purpose of our study was to evaluate the mid-term safety and efficacy of a 5 mg sublingual dose of zolpidem for reducing sleep latency when administered at bedtime, and "as needed" following middle-of-the-night (MOTN) awakenings, compared to the traditional oral 10 mg dose. To achieve these goals, we designed a three-month randomized, controlled, and double-dummy trial, in which we tested the patients' clinical improvement, reported adverse events, and changes in subjective and objective sleep measures. Methods ======= Sampling and design ------------------- The prospective participants were adults (20-64 years) who reported nocturnal awakenings predominantly before 3:00 a.m. and who had not used psychoactive drugs in the 30 preceding days. Referrals and online subscriptions to the Sleep Institute's website were channels for recruitment. Exclusion criteria were: 1) a history of major neurological, psychiatric, or other medical disorders; 2) a history of substance abuse/dependence; 3) a history of daily alcohol use; 4) pregnancy, lactation, or intending to conceive; 5) an apnea and hypopnea index (AHI) \> 10 events/hour; 6) a periodic limb movement index \> 15 events/hour; and 7) blood alterations suggestive of an undiagnosed medical condition. The protocol included five onsite visits. The first was for screening, when participants gave written informed consent and were examined. The second was for randomization, scheduled the morning after polysomnography (PSG). The third, fourth, and fifth were follow-up visits. Performance was measured with the Psychomotor Vigilance Test (PVT) which was self-applied twice during the day: 30 minutes and 2 hours after awakening. After lab and tests results, the patients were reassessed for continued inclusion. Eligible participants were assigned (1:1) to receive active doses of zolpidem either as sublingual 5 mg or oral 10 mg tablets. Because sublingual and oral tablets cannot be made to look and taste alike, a double-dummy technique was used with matching placebos of both formulations, which assured participant and investigator blinding. The oral group received kits consisting of oral zolpidem 10 mg and sublingual placebo, which was used at bedtime and "as-needed." The sublingual group received sublingual zolpidem 5 mg and an oral placebo at bedtime, and sublingual zolpidem 5 mg "as-needed." Placebos were indistinguishable in size, shape, color, and taste. Visits 3, 4, and 5 were follow-ups scheduled at weeks 2, 6, and 13, when the kits were dispensed, and unused medication was collected and counted. The bottles provided to participants contained enough tablets for the period between visits. The sponsor provided boxed kits numbered according to a randomization list with sealed envelopes identifying treatment. Procedures and instruments -------------------------- All visits included a medical evaluation, application of questionnaires, and the distribution and collection of sleep diaries. PSG, the PVT, electrocardiogram, and blood testing were performed twice, once before randomization and again during the final visit. Patients were instructed to take a maximum of one tablet from each bottle per night (bedtime and as needed) and record it in their sleep diaries. Two sleep specialists performed the evaluations. At screening, the participants' history of insomnia, medication use, comorbidities, and related symptomatology were investigated, and adverse events and clinical improvement were assessed during follow-up. Heart rate, blood pressure, and body mass index were assessed in physical examinations. A 12-lead digital electrocardiogram system (Wincardio System, Micromed, Brasília, Brazil) was used. Blood tests included a complete cell count and assessed liver, renal, and thyroid functioning. ### Sleep diaries Sleep diaries were used to collect data on medication use and sleep schedules. Based on their data, the following were estimated: 1) medication time; 2) bedtime; 3) sleep duration; 4) time of rescue dose ("as-needed" tablet); 5) wake time; 6) rise time; 7) number of nights in which the regular dose was taken ; 8) number of nights in which a rescue dose was taken; 9) number of nights with sleep-onset latency \> 30 minutes; 10) number of nights that MOTN awakenings occurred; 11) perceived TST; 12) calculated TST (sleep time minus wake time); 13) sleep efficiency (TST/time-in-bed); 14) alcohol consumption; and 15) mood and sleep quality ratings.[@B12] ### Questionnaires The Pittsburgh Sleep Quality Index evaluates sleep schedule, frequency of insomnia/nighttime symptoms, and level of sleep-related problems in the last month. The global score varies from 0-21, and scores \> 5 indicate poor sleep quality.[@B12],[@B13] The Insomnia Severity Index includes seven Likert scales (scored 0-4) that evaluate symptoms and dissatisfaction/distress in the past two weeks. Scores \> 7 (up to 28) indicate clinically meaningful insomnia. A reduction of ≥ 8 points suggests at least moderate improvement after treatment.[@B12],[@B14] The Karolinska Sleepiness Scale measures alertness in the present moment on a scale from 1-9 ("extremely alert" to "very sleepy/great effort staying awake/fighting sleep"). Electrophysiological signs of sleepiness, such as slow eye movement and alpha-power density, correlate with ratings ≥ 7.[@B15] ### Objective sleep, alertness, and performance measures PSG was performed with a digital system (EMBLA^®^S7000, Embla-Systems, Broomfield, CO, USA). Sleep stages, respiratory events, and leg movements were scored according to American Academy of Sleep Medicine criteria.[@B16] The selected variables were TST, sleep-onset latency, wake after sleep onset, sleep efficiency, arousal index, percentage of each sleep stage, rapid eye movement (REM) sleep latency, AHI, minimum oxyhemoglobin saturation (SaO~2~), and periodic limb movement index. The PVT measures the response to visual stimuli (PVT-192, CWE, Inc., Ardmore, PA, USA). During the 10-minute test, patients are instructed to press a button when a light randomly appears on the screen every few seconds. Data are collected on sleepiness, number of errors, hits, lapses, and reaction time.[@B17] Outcome measures ---------------- Treatment compliance was defined as the use of bedtime medication at least 75% of the nights, which was verified using sleep diary data and tablet counts during follow-up visits. Clinicians rated adverse events by their level of association with the study medication (unrelated, possibly related, or probably/certainly related), and decided whether to continue or discontinue treatment. Means and frequencies from the diary data were computed weekly and for the entire treatment period. For delta values, means from the third and final phase of the treatment period (between weeks 6 and 13) were subtracted from baseline means, or from the first and second phase means (i.e., day 1 until week 6). Ordinal responses to the questionnaires were dichotomized to estimate changes in sleep satisfaction. Clinical improvement referred to positive changes in sleep and functioning, as well as symptom remission as rated by physicians during follow-up visits. Partial or no improvement included persistent symptomatology and/or sleep-related problems. Statistical analysis -------------------- Sample size estimation (30/group) aimed at finding differences in the proportion of patients with decreased MOTN awakenings and sleep onset latency (20% in oral 10 mg and 70% in sublingual 5 mg) with 80% power and a 5% error rate. Attrition was expected to be 10%. Descriptive statistics are presented as absolute frequency and percentage for categorical variables and mean ± standard deviation for continuous variables. Associations between the categorical variables were measured in cross-tabulations with chi-square or Fisher's exact tests. Analysis of variance/covariance was used to compare group means and were performed through general linear models. Confidence intervals were adjusted with Bonferroni correction, and effect size (Eta^2^) and observed power were used to evaluate type I and II errors. Mauchly's test of sphericity, with the Greenhouse-Geisse correction for degrees of freedom, was used for repeated measures without assuming equal variance. The frequencies of medication use and nights in which a rescue dose was taken were used as covariates in general linear models for PSG and PVT parameters, controlling for possible residual effects. Intention-to-treat analysis was performed and every randomized participant was included in the analyses. Missing data were replaced by using the last observation carried forward method, in which the last available measurement for each individual prior to discontinuation from the study was retained in the analysis.[@B18] Sleep diary data analysis included use of general estimating equations, which do not assume the normality of the dependent measure and allow the selection of different types of distribution probabilities. Since they also allow all recorded daily values to be examined, they were used to evaluate "daily" variations in perceived TST across all treatment phases and to control for possible interaction effects with bedtime, rescue medication, baseline insomnia severity, and age. Ethics statement ---------------- The study protocol was approved by the Universidade Federal de São Paulo research ethics committee (protocol 184.648), and the study was registered in ClinicalTrials.gov (NCT01896336). Results ======= The mean ± standard deviation number of days between screening and randomization was 6±4. Follow-ups at weeks 2, 6, and 13 occurred 15±2, 31±4, and 46±4 days after the previous visit, respectively, totaling 92±5 days of treatment and 98±5 days in the protocol (between March and November 2013). Study population ---------------- Of the 85 screened individuals, 18 (21%) were not randomized due to reasons presented in the study flow diagram ([Figure 1](#f01){ref-type="fig"}). The 67 participants were middle-aged (48±10 years), and most were non-obese women (n=53; 79%) whose average insomnia duration was 13±9 years and who took at least one dose of study medication. All participants attended the week 2 follow-up and were evaluated by clinicians, but not all had completed their sleep diaries and questionnaires or had PSG or PVT measures collected. Participants who completed the protocol (n=46; 69%) and those who did not had similar baseline characteristics. However, those who did not complete the protocol had more adverse events and fewer signs of improvement. The treatment groups also presented similar baseline characteristics ([Table 1](#t01){ref-type="table"}). Adverse events -------------- Seventeen of the 21 participants who did not complete the protocol reported adverse events (25% of the total group), seven did not wish to continue (10%), and 10 were discontinued on medical advice (15%). Of those who did not wish to continue, three were from the sublingual group and four were from the oral group, mostly women with mild-to-moderate headaches, dizziness and/or gastrointestinal symptoms who had lower compliance rates and concurrent life events. Of the 10 who discontinued based on medical advice, four presented moderate to severe events associated with emotional instability or gastrointestinal symptoms, but recovered within a few days of discontinuation: three of these were in the oral group and one was in the sublingual group. The other six presented persistent mild-to-moderate adverse events or were in need of medication that could interfere with the study's outcomes. The difference in discontinuation frequency between treatment groups was not significant ([Figure 1](#f01){ref-type="fig"}). The investigators registered 152 adverse events, 58 (38%) unrelated to the study medication, 69 (45%) possibly related, and 25 (16%) probably/certainly related. Headache, sleepiness, and dizziness were the most likely events to be treatment-related. The proportion of patients with adverse events did not differ between groups, but the number of treatment-related adverse events leading to discontinuation was higher in the oral group than the sublingual group (29%, n=24 vs. 13%, n=9; two-sided p = 0.02). See [Table 2](#t02){ref-type="table"} for further information. Efficacy analyses ----------------- ### Inter- and intragroup sleep diary variance before and after treatment Both treatments significantly decreased the number of nights per week with MOTN awakenings, on average by -3.1±2.3 days. Before treatment, the sublingual group reported MOTN awakenings an average of 75% (95%CI 62-88) of the nights each week, which was reduced to 36% (20-53) in the third phase of the treatment period. The oral group reported MOTN awakenings an average of 87% (71-100) of the nights before treatment, which was reduced to 43% (24-62) in the third phase of the treatment period. Accordingly, a mean increase of 1.5 hours perceived TST was reported by both groups, with a proportional increase in sleep efficiency ([Table 3](#t03){ref-type="table"}). Sleep-onset latency was the only variable from the sleep diary with an interaction effect and an intergroup difference. There was a greater decrease in sleep-onset latency in the sublingual group than the oral group, even though its baseline values were higher (78 vs. 51; p = 0.03; effect size = 0.10; observed power = 0.61) ([Table 3](#t03){ref-type="table"}). When comparing group means from the first and second phases of treatment vs. those from the third phase, no intergroup differences were detected in the percentage of nights in which a rescue dose was taken or the number of "as needed" tablets used per week of treatment. Complementary and detailed analyses of sleep diary data are presented in [Table 4](#t04){ref-type="table"}. ### Changes in insomnia severity and sleep quality indexes No significant differences were found in the general linear model between mean questionnaire scores at baseline and the third phase. The mean reduction in Pittsburgh Sleep Quality Index scores was modest but significant, both in the sublingual (mean difference: -3.4±4.1) and the oral group (-2.6±3.4). However, the changes in Insomnia Severity Index scores were more marked, -11.3±5.5 and -10.4±6.1, respectively ([Table 3](#t03){ref-type="table"}). In the per-protocol analysis, including only individuals who fully completed the trial, the frequency of sleep dissatisfaction according to the Insomnia Severity Index at week 13 was significantly lower in the sublingual group (16 vs. 43%; chi-square: p = 0.04). ### Objective sleep and signs of residual effects Findings for the PSG variables were similar to and corroborated those from the sleep diaries. For PSG TST, both groups had a mean increase of 30 minutes from baseline ([Table 3](#t03){ref-type="table"}). Differences in wake after sleep onset were not significant. As seen by the large standard deviations, individual changes varied consistently, particularly in the sublingual group (mean difference: -57±226) and in contrast with the oral group (-26±65). Sleep-onset latency was the only statistically significant factor: the sublingual group decreased by -14±42 minutes, while the oral group increased by 10±29 minutes (p = 0.03). The findings were similar after controlling for the percentage of rescue dose nights. Additionally, use of a rescue dose was found to have an interaction effect with changes in TST, sleep efficiency, REM latency, and the respiratory-related variables ([Table 3](#t03){ref-type="table"}). PVT scores showed minor signs of residual effects. Both groups had a slight increase in sleepiness 30 minutes after awakening, which reversed after 2 hours. A small group difference was observed in the number of lapses after 30 minutes, which also reversed after 2 hours. Sleepiness ratings in the Karolinska Sleepiness Scale corroborated the PVT results ([Table 3](#t03){ref-type="table"}). We further compared PVT variables between patients who had taken (n=8) as needed medication or not (n=38) on the night of PSG and found no differences in sleepiness (p = 0.47) or lapses (p = 0.37) 30 minutes after awakening. ### Intragroup sleep diary variability Patients handed in a total of 5,053 sleep diaries: 2,761 (55%) from the sublingual group, 2,292 (45%) from the oral group, and 578 (11%) from participants of both groups who discontinued the treatment protocol. All available data are described below. Of all the reported nights, 269 (5%) were from the baseline period (6±4 days), 1,027 (20%) from baseline until the week 2 visit (after 15±2 days), 1,657 (33%) from week 2 to week 6 (after 31±4 days), and 2,100 (42%) from week 6 until the final visit (after 46±4 days). Bedtime (11.1 p.m. ± 1.2h) and "as needed" (1.9 a.m. ± 2.8h) medication were reported for 96% and 12% of the nights, respectively. The total number of nights with reported "as needed" rescue doses were higher in the sublingual group than the oral group (n=366 nights \[14%\] vs. n=193 nights \[9%\]; p \< 0.05). [Table 4](#t04){ref-type="table"} presents the number of sleep diaries collected at each visit and the valid percentages of complete data reported for each variable, such as the percentage of nights with sleep-onset latency \> 30 minutes (39%), MOTN awakenings (45%), TST \> 6 hours (59%), and ratings of good mood and having had a good night's sleep after awakening (58%). [Figure 2](#f02){ref-type="fig"} compares histograms of sleep diary data throughout the protocol period for both treatment groups, including nightly variations in the number of patients reporting MOTN awakenings or not and whether they took a rescue dose or not. The pattern of rescue dose usage tended to decrease between the first and third phases of treatment in both groups, suggesting no dose escalation or abuse. According to repeated analysis of the sleep diary data with general estimating equations, controlling for bedtime, the use of rescue doses, age, and baseline insomnia severity, there was no difference between groups in baseline perceived TST. The TST increase in both groups differed somewhat over the treatment phases, there was an interaction effect (Wald χ^2^: 94.1; df: 7; p \< 0.001). In the oral group, there was a TST gain of 1.3 hours in the first phase of treatment, which was sustained until week 13. In the sublingual group, there was a TST gain of 1.2 hours in the first phase, followed by a mean, progressive, and cumulative increase of 0.2 hour (12 minutes) between the first and second phases, as well as between the second and third phases. MOTN awakenings that led to a rescue dose of medication were associated with a small decrease in perceived TST in the oral group (B: -0.11; Wald χ^2^: 11.3; degrees of freedom: 1; p = 0.001) but not in the sublingual group (B: -0.01; Wald χ^2^: 0.1; degrees of freedom: 1; p = 0.77). Bedtime (B: -0.03; Wald χ^2^: 23.6; degrees of freedom: 1; p \< 0.001), but not age or baseline insomnia severity, also exerted a small independent effect on perceived TST. There was clinical improvement in 37 (55%) participants after completing the protocol: 23 (68%) in the sublingual group and 14 (42%) in the oral group (p = 0.05). Most of these participants (n=23) improved after week 6. Additionally, 11 (16%) participants presented early improvement followed by symptom recurrence, and there was no clinical improvement in 19 (28%) participants. There was no symptom remission in 30 of the randomized participants by the end of the study. Discussion ========== Our main objective was to describe the safety and efficacy of different dosages and treatment schemes of zolpidem in a 3-month trial. Patients from both groups presented no abuse-related behavior and no adverse events severe enough to discourage either drug formulation, which agrees with previous findings.[@B08],[@B09],[@B19]-[@B21] Both treatments increased TST and sleep efficiency and decreased wake after sleep onset. Regarding sleep-onset latency, the sublingual version had faster absorption and distribution rates, reaching higher concentrations and inducing sleep more rapidly.[@B22]-[@B26] However, an important factor that we failed to control for was the impact of food/ingestion on pharmacokinetics.[@B27] Greenblatt et al. showed that taking zolpidem on a full stomach decreases the maximum concentration and increases the chance of residual sedation. The influence of gender is also significant.[@B08],[@B28]-[@B30] Our findings suggested a higher dropout rate among men, although we did not have a representative number of men, and thus could not detect whether this was an interaction effect. In women, especially older women, perhaps 10 mg was a high dose, as has been stated by the U.S. Food and Drug Administration. Moreover, a 5 mg dose may not have been effective in men. Taking rescue doses might have led to small changes in sleep architecture, such as changing REM latency or slow-wave sleep, or to changes in sleep quantity and quality, increasing TST and sleep efficiency. These results are supported by the findings of Roth et al.,[@B31] who found a dose-dependent effect on the same sleep parameters with even lower doses. These authors reported that improved sleep architecture led to better subjective outcomes, and our findings point in the same direction. It should be mentioned that our use of the last observation carried forward method to deal with missing values or intention-to-treat analyses was a conservative measure. It could have underestimated associations, not only with PSG results, but with PVT scores as well. Nevertheless, the participants' performance and sleepiness results did not suggest unmanageable residual effects. Studies using different methods, mostly in healthy individuals, have also found no signs of relevant residual sedation, though patients should always be advised about the risks of activities that require alertness early in the morning.[@B08],[@B21],[@B24],[@B26],[@B31],[@B32] Regarding clinical improvement, even though we performed no inter-rater agreement tests between clinicians, they were trained and experienced, as well as blinded. About half of the patients ended treatment with clinically significant symptom remission. Of those whose symptoms recurred, most were in the oral 10 mg group, which was also reflected in sleep dissatisfaction ratings in the questionnaires. The majority of those who had no meaningful improvement had discontinued treatment due to adverse events. A randomized trial by Morin et al.[@B20] evaluated cognitive behavioral therapy (CBT) for insomnia, applied either alone or associated with 10 mg oral zolpidem. After 6 weeks, the CBT-only group remained in the same treatment protocol. Patients in the combined therapy group either continued in the same treatment protocol or switched to CBT alone. After 12 weeks, symptom remission was higher among patients from the combined therapy group that switched to CBT alone than in those who remained in combined therapy (68 vs. 42%). We observed the same success rates in our treatment groups: 42% in the oral group (n=14) and 68% in the sublingual group (n=23) after the same treatment time, only without CBT. It is possible that the availability of rescue medication gave some patients a sense of control that contributed to lower sleep anxiety and MOTN awakenings. Perlis et al.[@B11] suggested something similar in a trial testing intermittent zolpidem doses with and without placebo. The present study has certain limitations. The heterogeneity and the night-to-night variability in sleep patterns of insomnia patients is hard to address and has been heavily investigated.[@B02],[@B04],[@B33],[@B34] Our sample included patients with different durations of insomnia and different histories of psychoactive drug use, and we did not control for possible exposure to previous behavioral interventions, although none of the participants reported previous CBT. Moreover, there was a higher proportion of women, which could increase the probability of slower pharmacokinetics, and we did not control for menopausal status. It is difficult to recruit patients with untreated insomnia, which contributed to our small sample size. We sought to address this limitation by applying general estimating equation models. Our study is one of the few to have used more robust statistical procedures that can adequately handle a large number of repeated measures, such as the data from the sleep diaries. With general estimating equations, we were able to identify "daily" predictors of changes in perceived TST throughout treatment and the "size" of those changes. In summary, and in agreement with recommended guidelines, our findings reaffirm that initial lower doses of zolpidem can achieve better overall treatment outcomes while lowering the occurrence of adverse events. For some patients, higher dosing regimens may be needed, for others, having rescue medication on hand might be helpful. In a meta-analysis on the effectiveness of non-benzodiazepine GABA-A receptor agonists, Huedo-Medina et al.[@B07] found that placebo response accounts for, on average, about half of the drug response. Similarly, according to findings by Perlis et al.,[@B11] alternating nights of drug-therapy and placebo, or stipulating the number of nights per week that a rescue dose can be taken, or other innovative strategies, are potential avenues for clinical and behavioral research and interventions among insomnia sufferers. In conclusion, both sublingual zolpidem 5 mg (with "as needed" medication) and oral zolpidem 10 mg showed comparable safety and efficacy. Both treatments reduced the number of nights with MOTN awakenings and increased perceived TST. Clinical improvement occurred in 55% of the participants, of whom 68% received the sublingual formulation. Sublingual zolpidem was more rapidly absorbed, induced sleep faster, and resulted in a higher frequency of sleep satisfaction. Disclosure ========== This study was conducted at Centro de Desenvolvimento de Estudos Clínicos Brasil and the Sleep Research Center of the Associação Fundo de Incentivo a Pesquisa (AFIP) and the Universidade Federal de São Paulo, and it was sponsored by EMS Pharmaceutical. EMS had no role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; preparation or approval of the manuscript; or the decision to submit the manuscript for publication. All authors have filled out a disclosure form: LSC has received honorariums and consults for EMS; CFS was a clinician involved in evaluating participants; ST presides over AFIP, which has received a grant from EMS for the present study; DP received a grant from EMS for this study and has received honorariums as a speaker from Sanofi-Aventis and Libbs Pharmaceuticals, for developing and delivering educational presentations and other expenses beyond this study. The authors report no other conflicts of interest. The authors would like to acknowledge the work of Magda Bignotto, study administrator and leader of the study's monitors and coordinators: Graciela Santos (research analyst), Alexandre Ito (pharmacist), and Merielen CPP Silva (research assistant). The authors would also like to thank Dr. Luciano Rotella and Dr. Erika Tremptow, clinicians involved in patient evaluation and screening, as well as Dr. Altay Alves Lino de Sousa, professor of statistics (psychobiology), for his advice. **How to cite this article:** Castro LS, Otuyama LJ, Fumo-dos-Santos C, Tufik S, Poyares D. Sublingual and oral zolpidem for insomnia disorder: a 3-month randomized trial. Braz J Psychiatry. 2020;42:175-184. <http://dx.doi.org/10.1590/1516-4446-2018-0389> ![Study flow diagram outlining subject selection and study design. The safety population included all randomized participants who took at least one dose of the study medication. The efficacy population included those who provided any information on treatment effect. Fisher's exact test: no group differences in discontinuation frequency (24 vs. 39%; chi-square = 2.08; two-sided p = 0.18). ECG = electrocardiogram; PSG = polysomnography.](bjp-42-02-175-g001){#f01} ![Histogram of sleep diary data: number of daily reports during the treatment period stacked according to middle-of-the-night awakenings with or without the use of "as needed" tablets. MOTN: middle-of-the-night awakenings.](bjp-42-02-175-g002){#f02} ###### Patient sociodemographic and clinical characteristics at baseline Sublingual zolpidem 5 mg (n=34) Oral zolpidem 10 mg (n=33) Total (n=67) Discontinued (n=21) ---------------------------------------------------------- --------------------------------- -------------------------------------------- -------------- --------------------- Age, mean (SD), years 46 (11) 49 (10) 48 (10) 48 (9) BMI, mean (SD), kg/m^2^ 26 (3) 26 (4) 26 (4) 26 (4) Women 27 (79) 26 (79) 53 (79) 16 (76) Marital status Married 15 (44) 19 (58) 34 (51) 10 (48) Single 11 (32) 6 (18) 17 (25) 5 (24) Divorced/widowed 8 (24) 8 (24) 16 (24) 6 (28) Education level Primary and secondary 22 (65) 21 (64) 43 (64) 12 (57) Higher 12 (35) 12 (36) 24 (36) 9 (43) Smoking status Never smoked 31 (91) 30 (91) 61 (91) 17 (81) Current or former smoker 3 (9) 3 (9) 6 (9) 4 (19) Clinical history Treated for insomnia 13 (38) 13 (41) 26 (39) 8 (38) Psychoactive drug use 3 (9) 9 (27)[\*](#TFN04t01){ref-type="table-fn"} 12 (18) 4 (19) Headache 15 (44) 16 (49) 31 (46) 7 (33) Muscle pain 10 (29) 5 (15) 15 (22) 5 (24) Gastrointestinal symptoms 10 (29) 9 (27) 19 (28) 4 (19) Type of symptoms DIS, DMS or EMA 14 (41) 13 (39) 27 (40) 10 (48) Two+ symptoms 20 (59) 20 (61) 40 (60) 11 (52) AHI, mean (SD) events/hour 2 (2) 4 (6) 3 (4) 3 (3) SaO~2~min, mean (SD), % 91 (3) 90 (4) 90 (3) 90 (3) Duration of insomnia, mean (SD), years 12 (9) 15 (10) 13 (9) 14 (8) ISI score, mean (SD)[†](#TFN05t01){ref-type="table-fn"} 17 (4) 18 (5) 17 (5) 18 (5) PSQI score, mean (SD)[‡](#TFN06t01){ref-type="table-fn"} 11 (3) 12 (3) 12 (3) 12 (3) Data presented as n (%), unless otherwise specified. AHI = apnea and hypopnea index; BMI = body mass index; DIS = difficulty initiating sleep; DMS = difficulty maintaining sleep; EMA = early-morning awakening; ISI = Insomnia Severity Index; PSQI = Pittsburgh Sleep Quality Index; SaO~2~min = minimum oxyhemoglobin saturation; SD = standard deviation. Statistical analyses were performed with two independent binary variables (sublingual zolpidem 5 mg vs. oral zolpidem 10 mg, and discontinued protocol vs. completed protocol). Statistically significant at p \< 0.05. ISI scores range from 0-28 (higher scores indicate more severe insomnia; clinical relevance \> 7). PSQI scores range from 0-21 (higher scores indicate poorer sleep quality; clinical relevance \> 5). ###### Number of reported adverse events in association with treatment discontinuation Adverse event types Sublingual zolpidem 5 mg 70 (31) Oral zolpidem 10 mg 82 (30) Total 152 (61) ---------------------------- ---------------------------------- ----------------------------- ---------------- -------- ---------- --------- Central nervous system 23 (16) 7 (3) 19 (16) 16 (7) 42 (32) 23 (10) Respiratory 15 (11) 0 (0) 15 (11) 0 (0) 30 (22) 0 (0) Gastrointestinal 7 (7) 2 (1) 14 (9) 3 (3) 21 (16) 5 (4) Neuromuscular and skeletal 10 (10) 0 (0) 4 (4) 2 (2) 14 (14) 2 (2) Cardiovascular 2 (2) 0 (0) 1 (1) 1 (1) 3 (3) 1 (1) Other 4 (4) 0 (0) 5 (3) 2 (2) 9 (7) 2 (2) Total 61 (27) 9 (4) 58 (22) 24 (8) 119 (49) 33 (12) Data presented as number of events (number of patients). Continued = continued in the protocol; Discontinued = discontinued participation in the study under medical advice. In the Total column, the numbers in parentheses may be higher than the total number of participants, since participants could report more than one adverse event during the study. ###### Inter- and intragroup variance in repeated subjective and objective measures Sublingual zolpidem 5 mg (n=34) Oral zolpidem 10 mg (n=33) Within subjects (p-values) --------------------------------------------------------------- --------------------------------- ---------------------------- ---------------------------- --------------- -------------- ----------- Sleep diaries Sleep onset latency, minutes 78 (57-99) 25 (17-34) 51 (27-75) 31 (20-41) **\< 0.001** **0.031** Nights with MOTN awakenings, % 75 (62-88) 36 (20-53) 87 (71-100) 43 (24-62) **\< 0.001** 0.707 Rescue dose nights[\*](#TFN05t03){ref-type="table-fn"}, % 22 (11-33) 15 (5-26) 20 (7-33) 10 (-3-22) 0.223 0.799 Total sleep time, hours 4.8 (4.3-5.3) 6.4 (5.9-6.9) 4.5 (3.9-5.0) 5.7 (5.2-6.3) **\< 0.001** 0.390 Sleep efficiency, % 62 (55-70) 81 (75-86) 60 (52-69) 76 (70-83) **\< 0.001** 0.096 Questionnaire scores ISI 17 (15-18) 6 (4-8) 18 (16-19) 8 (6-10) **\< 0.001** 0.470 PSQI 11 (10-12) 8 (7-10) 12 (11-13) 10 (9-11) **\< 0.001** 0.608 KSS, 30 minutes[†](#TFN06t03){ref-type="table-fn"} 5 (4-5) 5 (4-6) 5 (4-6) 5 (5-6) 0.088 0.605 KSS, 2 hours[†](#TFN06t03){ref-type="table-fn"} 4 (3-4) 3 (3-4) 4 (3-4) 4 (3-4) 0.178 0.233 PSG parameters Sleep-onset latency 27 (17-38) 17 (9-24) 18 (8-28) 24 (16-31) 0.502 **0.031** Wake after sleep onset 89 (39-139) 46 (34-58) 67 (17-118) 52 (39-64) 0.102 0.440 Total sleep time 326 (306-347) 352 (334-369) 333 (312-353) 350 (332-368) **0.001** 0.518 Sleep efficiency 80 (75-84) 83 (79-87) 80 (75-84) 82 (78-87) **0.048** 0.824 Arousal index 12 (10-15) 13 (11-16) 16 (13-18) 14 (12-17) 0.828 0.165 REM sleep 20 (18-22) 19 (17-20) 19 (17-21) 19 (17-20) 0.440 0.504 Slow wave sleep 22 (19-25) 24 (21-27) 22 (19-25) 23 (21-26) **0.057** 0.708 AHI 2.1 (0.6-3.6) 3.1 (1.7-4.5) 3.6 (2.1-5.1) 3.9 (2.5-5.3) 0.183 0.448 SaO~2~min 91 (89-92) 89 (88-90) 89 (88-91) 88 (87-90) **0.002** 0.591 PVT parameters[†](#TFN06t03){ref-type="table-fn"} Sleepiness, 30 minutes 5 (4-6) 6 (5-7) 5 (4-6) 6 (5-7) **0.043** 0.869 Sleepiness, 2 hours 3 (2-4) 3 (2-4) 4 (4-5) 4 (3-5) 0.171 0.328 Total hits, 30 minutes 92 (91-94) 90 (85-95) 93 (92-95) 91 (85-96) 0.169 0.885 Total hits, 2 hours 90 (86-94) 94 (90-98) 95 (91-99) 92 (88-96) 0.850 0.097 Lapses, 30 minutes 16 (10-22) 9 (5-12) 5 (0-11) 7 (3-11) 0.151 **0.054** Lapses, 2 hours 7 (3-12) 4 (2-7) 4 (0-9) 4 (2-7) 0.279 0.363 Reaction time, 30 minutes 412 (359-464) 368 (320-416) 340 (287-394) 346 (297-395) 0.285 0.162 Reaction time, 2 hours 336 (301-372) 309 (273-346) 300 (263-336) 313 (276-350) 0.625 0.151 Data presented as mean (95% confidence interval), unless otherwise specified. 95%CI = 95% confidence interval; AHI = apnea-and-hypopnea index; ISI = Insomnia Severity Index; KSS = Karolinska Sleepiness Scale; MOTN awakenings = middle-of-the-night awakenings; PSG = polysomnography; PSQI = Pittsburgh Sleep Quality Index; PVT = Psychomotor Vigilance Test; REM = rapid-eye-movement; SaO~2~min = minimum oxyhemoglobin saturation. Last observation carried forward (LOCF) = Missing values for discontinued participants were replaced by baseline or the last observed value. Within-subject statistics are presented for treatment effect alone (Factor) and interaction effect with the treatment group (x Group). Bold font indicates statistical significance. At baseline, rescue dose nights refers to the percentage of nights a rescue dose was taken until the visit at week 6, i.e. half of the treatment period. It was used as a covariate in the PSG and PVT models to control for residual effects. Subjective sleepiness and psychomotor performance were measured twice, 30 minutes and 2 hours after awakening from PSG. ###### Means and frequencies of data reported in all sleep diaries Sublingual zolpidem 5 mg Oral zolpidem 10 mg ------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- ---------------------------------------- Total diaries, n 139 507 912 1,203 130 520 745 897 Weekends, n (%) 48 (34) 220 (43) 395 (43) 510 (42) 42 (32) 231 (44) 324 (43) 377 (42) Nights medication taken \- 97 93 96 \- 84 95 99 Rescue dose nights \- 20 13 13 \- 14 7 9 Perceived TST ≥ 6 hours 33 65[\*](#TFN04t04){ref-type="table-fn"} 69[\*](#TFN04t04){ref-type="table-fn"} 68[\*](#TFN04t04){ref-type="table-fn"} 23 53[\*](#TFN04t04){ref-type="table-fn"} 49[\*](#TFN04t04){ref-type="table-fn"} 54[\*](#TFN04t04){ref-type="table-fn"} SOL ≥ 30 minutes 72 39 33[\*](#TFN04t04){ref-type="table-fn"} 35[\*](#TFN04t04){ref-type="table-fn"} 65 41 42[\*](#TFN04t04){ref-type="table-fn"} 42[\*](#TFN04t04){ref-type="table-fn"} MOTN awakenings 67[\*](#TFN04t04){ref-type="table-fn"} 52[†](#TFN05t04){ref-type="table-fn"} 41[\*](#TFN04t04){ref-type="table-fn"} 33[\*](#TFN04t04){ref-type="table-fn"} 89[\*](#TFN04t04){ref-type="table-fn"} 58[†](#TFN05t04){ref-type="table-fn"} 48[\*](#TFN04t04){ref-type="table-fn"} 41[\*](#TFN04t04){ref-type="table-fn"} Sleep efficiency ≥ 85% 37 64[†](#TFN05t04){ref-type="table-fn"} 71[\*](#TFN04t04){ref-type="table-fn"} 69[\*](#TFN04t04){ref-type="table-fn"} 44 57[†](#TFN05t04){ref-type="table-fn"} 59[\*](#TFN04t04){ref-type="table-fn"} 57%[†](#TFN05t04){ref-type="table-fn"} Data presented as %, unless otherwise specified. The chi-square and z-test of column proportions were adjusted for all pairwise comparisons within a row using the Bonferroni correction. Proportions were compared between treatment groups at each phase of the protocol period. MOTN awakenings = middle of the night awakening; SOL = sleep-onset latency; TST = total sleep time. Statistically significant at p \< 0.01. Statistically significant at p \< 0.05.

{ "pile_set_name": "PubMed Central" }

Introduction {#sec1-1} ============ The algorithm used for dose calculation plays a very important role in delivery of dose to patients undergoing radiation treatment.\[[@ref1]\] The beam configuration of analytical anisotropic algorithm (AAA) involves precalculated Monte Carlo data to determine all parameters to match the measured beam data.\[[@ref2]\] The AAA calculation uses accurate Monte Carlo-based three-dimensional pencil beam convolution superimposition for inhomogeneity correction. Algorithms can be divided into three classes based on their accuracy. Type "A" was proposed by Knöös *et al*.\[[@ref3]\] which were based on measurements and accounts for correction for patient contours and heterogeneities. Type "B" was proposed by Ojala *et al*.\[[@ref4]\] which were based on superposition and convolution techniques. The algorithm, Acuros XB (AXB), based on linear Boltzmann transport equation which was first implemented in the Eclipse treatment planning system (TPS) by Varian Medical Systems, Inc., (Palo Alto, CA, USA).\[[@ref5]\] comes under more accurate type "C"\[[@ref6][@ref7]\] algorithms. The origin of AXB algorithm designed for radiotherapy dose calculations is Transpire, Inc., (Gig Harbor, WA, USA).\[[@ref8][@ref9][@ref10][@ref11][@ref12]\] Lloyd and Ansbacher\[[@ref13]\] showed that the performance of AXB was better than existing clinical algorithm, AAA in high-density materials. Kan *et al*.\[[@ref14]\] showed the importance of options for dose to medium and dose to water in AXB in bone in the treatment of nasopharyngeal carcinoma by intensity modulated radiotherapy (IMRT) and RapidArc. Tomiyama *et al*.\[[@ref15]\] showed that Voxel Monte Carlo algorithm and AXB were better than AAA in stereotactic body radiotherapy (SBRT) of the lung. Similarly, Liu *et al*.\[[@ref16]\] showed that advantage of Acuros over AAA for treatment of SBRT lung. There were contradictory results on studies of lung cancers for planning target volume (PTV) coverage as well as sparing of the normal lung. Jiang *et al*.\[[@ref17]\] reported that volumetric modulated arc therapy (VMAT) was better than IMRT with better PTV coverage and sparing of normal lung by evaluating V20, V30, and mean dose (D~mean~) for the lung. Rao *et al*.\[[@ref18]\] showed comparable PTV coverage using VMAT and IMRT. Ong *et al*.\[[@ref19]\] reported that IMRT was better than VMAT by evaluating V5 for the lung. The current study evaluated the differences in PTV coverage and in sparing of OARs when using the two different dose calculation algorithms AAA and AXB in IMRT. Materials and Methods {#sec1-2} ===================== A 6 MV linear accelerator, Clinac 600C (Varian, Palo Alto, CA, USA) having forty-pairs multi-leaf collimator (MLC) with each leaf projecting 1 cm width at isocenter was used for the delivery of radiation treatments. A cohort of 15 patients diagnosed with non-small cell carcinoma of the left lung was taken for the study. All patients received a prescription dose of 50 Gy in 25 fractions. Thermoplastic sheet (Orfit) was used for immobilizing the patients. A Philips (big bore) computed tomography (CT) scanner was utilized for imaging patients, and CT images of 3-mm slice thickness were acquired with the patients lying in supine position. The CT images were transferred to the Eclipse TPS, version 13.6 (Varian Medical Systems, Palo Alto, CA, USA). The gross target volume (GTV), clinical target volume, the PTV, and the organs at risks (OARs) were contoured on the CT images by a qualified radiation oncologist following the guidelines of the International Commission on Radiation Units and Measurements report 83.\[[@ref20]\] Since the tumor was at the left side, the beams were selected from the left side only. Initially, the plans were created in AAA using IMRT technique with a grid size of 2.5 mm, and 7 mm margin was given around the GTV to account for lung motion during the treatment. For this, a template of dose constraints was created, and plan optimization was done using the created template. In addition, these constraints were changed to obtain minimum possible dose to critical organs without compromising the PTV coverage of at least 95% dose to 95% of PTV volume. One more set of plans was created using AXB algorithm, only by performing recalculation without doing any reoptimization. Thereby, in both the plans (AAA and AXB), all the plan parameters, namely, beam shape, beam angle, beam weight, number of beams, prescribed dose, normalization point, region of interest constraints, number of monitor units (MUs), and plan optimization were kept constant, and only the calculation algorithm was changed. The dose distributions in the PTV and the OARs in both the sets of plans were compared. [Table 1](#T1){ref-type="table"} shows dose constraints\[[@ref21]\] for PTV and OARs used in treatment planning. ###### Planning target volume and organs at risks-dose constraints for treatment planning ![](JMP-41-219-g001) Plan analysis for dose-volume histogram parameters {#sec2-1} -------------------------------------------------- The plans were evaluated and compared on the basis of following dosimetric parameters, namely, target coverage, dose to OARs, homogeneity index (HI), and conformity index (CI). In the current study, HI was evaluated using the formula:\[[@ref22][@ref23][@ref24][@ref25]\] HI = D5/D95 where D5 is dose to 5% of volume of PTV and D95 is dose to 95% of volume of PTV. A value of HI closer to 1 points to a more homogeneous dose inside the PTV. CI was evaluated using the formula\[[@ref26][@ref27][@ref28][@ref29]\] CI = (TV95/V95) × (TV95/TV). where TV95 is the volume of PTV receiving 95% of prescribed dose, V95 is the volume of tissue receiving 95% of prescribed dose, and TV is the volume of PTV. A value of CI closer to 1 represents more a conformal dose around PTV with less spillage of dose. Statistical analysis of the data sets for the two algorithms was done using paired, two-tailed *t*-test and computing the *P* value. When *P* \< 0.05, the difference between the two algorithms was considered as significant. Results {#sec1-3} ======= [Table 2](#T2){ref-type="table"} shows the location and dimensions of PTV for 15 cases. Mean of PTV was found to be 203.69 cc with standard deviation (SD: 84 cc). Mean values of OARs were left lung-GTV, 1163 cc (SD: 363 cc), contralateral lung, 1612 cc (SD: 564 cc), heart, 482 cc (SD: 202 cc), and spinal cord, 35.5 cc (SD: 18 cc), respectively. ###### Location and volume of planning target volume for 15 cases ![](JMP-41-219-g002) Tables [3](#T3){ref-type="table"} and [4](#T4){ref-type="table"} show the differences between the doses calculated by AAA and by AXB. [Figure 1](#F1){ref-type="fig"} shows the difference between dose coverage to PTV for a representative patient. [Figure 2](#F2){ref-type="fig"} shows the deviation in mean dose (D~mean~) for both the algorithms. ###### Comparison of planning target volume coverage for analytical anisotropic algorithm versus Acuros XB algorithm using intensity modulated radiotherapy technique ![](JMP-41-219-g003) ###### Comparison of organs at risk doses for analytical anisotropic algorithm versus Acuros XB algorithm using intensity modulated radiotherapy technique ![](JMP-41-219-g004) ![The difference between dose coverage to planning target volume for a representative patient (left side: Analytical anisotropic algorithm and right side: Acuros)](JMP-41-219-g005){#F1} ![Percent deviation in mean dose for Analytical Anisotropic Algorithm versus Acuros for 15 cases. The X-axis represents case number](JMP-41-219-g006){#F2} Of the various assessed dose-volume parameters for PTV, D95, maximum dose D~max~, TV95%, and HI showed statistical significance between AAA and AXB algorithms (*P* = 0.01, 0.00, 0.01 and 0.04). Similarly, among the several assessed dose-volume parameters for various OARs, significant differences between the two algorithms were observed only for V5, V20 and V30 in left lung-GTV, D~mean~ in right lung, D33 in heart, and D~max~ in spine (*P* = 0.00, 0.00, 0.00, 0.00 and 0.00 respectively) as shown in [Table 4](#T4){ref-type="table"}. Discussions {#sec1-4} =========== Dose at 95% of PTV, volume of PTV receiving 95% of dose, and HI were overestimated in AAA compared to AXB as reported previously.\[[@ref15][@ref16]\] Our results are comparable to that of Rana *et al*.\[[@ref30]\] who showed that more number of MUs are required for AXB to achieve similar target coverage in comparison to that of AAA. This typical dose coverage to PTV showing the difference in both algorithms is shown in [Figure 1](#F1){ref-type="fig"} for a representative patient. As the plans were recalculated with the same number of MUs in both the algorithms (AXB and AAA), PTV coverage at 95% level with AAA was good whereas it was not so with AXB. This can be also seen by observing the variation in percent deviation in D~mean~ for AAA versus AXB from [Figure 2](#F2){ref-type="fig"}. Even though there is a statistical significance between AAA versus AXB, the mean difference is very small for PTV coverage as shown in [Table 3](#T3){ref-type="table"}. Similarly, from [Table 4](#T4){ref-type="table"}, it can be observed that the OAR doses were slightly underestimated in case of normal lung (represents air) and slightly overestimated in case of the heart (represents muscle) and spine (represents bone) in AAA compared to AXB. Even though statistical significance does exist for V5, V10, V30 of (Lt.Lung-GTV), Rt.Lung (D~mean~), Heart (D33), Heart (D~mean~) and Spine (D~max~), the difference between the two algorithms is less than 3%. Accuracy of AXB over AAA is well documented in literature.\[[@ref13][@ref14][@ref15][@ref16]\] A limited study done by us, varying the size of PTV and OAR did not show any significant difference between the two algorithms. Conclusions {#sec1-5} =========== This study concludes that overall minor overestimation of PTV coverage in AAA compared to AXB algorithm. In case of OAR doses, mixed results were observed. Doses to the normal lung were slightly underestimated, and doses to the heart and spine were slightly overestimated in AAA compared to AXB. Even though statistically significant differences were observed between the two algorithms, the magnitude of the dose difference is too small to cause any clinical significance. Financial support and sponsorship {#sec2-2} --------------------------------- Nil. Conflicts of interest {#sec2-3} --------------------- There are no conflicts of interest. The authors wish to thank Prof. Komanduri Ayyangar, Emeritus Professor, University of Nebraska, Omaha, NE. USA, for helpful suggestions.

{ "pile_set_name": "PubMed Central" }

Introduction {#Sec1} ============ Hand, foot and mouth disease (HFMD) is a common childhood disease, that is characterized by rapidly ulcerating vesicles in the mouth and vesicular lesions on the hands, feet, and buttocks. In China, HFMD has exhibited the highest morbidity and mortality among the group C notifiable infectious diseases in the last decade (<http://www.nhfpc.gov.cn/jkj/s3578/new_list.shtml>). Enterovirus A71 (EV-A71) is a member of family Picornavieidae that is one of major etiological agent of HFMD and causes a variety of clinical manifestations including severe neurological complications including aseptic meningitis, brainstem encephalitis, neurogenic pulmonary edema, acute flaccid paralysis, and death^[@CR1]--[@CR3]^. EV-A71 was first isolated from a child with encephalitis in California, USA, in 1969^[@CR4]^. In the Asian-Pacific region, EV-A71-related outbreaks have been reported frequently and are associated with increased neurovirulence and fatalities^[@CR5],[@CR6]^. The dominant EV-A71 strains circulating in different countries and regions vary genetically. In mainland China, more than 600,000 HFMD cases and 126 deaths were reported in Fuyang City of Anhui province from March 2008 to June 2009. Subsequently this outbreak spread quickly to other regions and caused nationwide HFMD epidemics. An emerging recombinant EV-A71 C4 genotype was reported to be responsible for this outbreak^[@CR7]^. We have reported on the prevalence of EV-A71 during the HFMD epidemic in Beijing from 2007 to 2009, including the isolation of the CMU4232 strain, one of the circulating strains identified in 2008^[@CR8]^. Xing et al. reported that more than 90% of deaths were associated with EV-A71 in China from 2008 to 2012^[@CR9]^. However, the pathogenesis of EV-A71 infections is not fully understood, especially the cause of the severe neurological manifestation. Numerous animal models have been developed to study the pathogenesis of EV-A71 infection, and appropriate animal models are needed to better understand EV-A71-associated neuropathogenesis and to facilitate the development of effective vaccines and drugs. EV-A71-infected cynomolgus and rhesus monkeys develop similar clinical symptoms as those observed in humans, including neurological complications^[@CR10]--[@CR12]^. However, the application of EV-A71 nonhuman primate models is limited because of ethical and economic reasons. Mice models with the strategies of virus adaption and immunodeficient hosts are also developed, although they do not closely mimic human disease^[@CR13]--[@CR15]^. Moreover, mice that are more than 2-week-old are generally not susceptible to EV-A71. Human scavenger receptor class B member 2 (hSCARB2) is widely expressed in many human tissues and cell types, including neurons in the central nervous system (CNS)^[@CR16]^. Studies have showed that hSCARB2 serves as a functional receptor for EV-A71 in vitro and in vivo^[@CR17]--[@CR21]^. Although the mouse SCARB2 shares 85.8% homology to the human protein, it does not function as a receptor for EV-A71^[@CR18]^. To date, four reported transgenic (Tg) mice carrying hSCARB2^[@CR19],[@CR20],[@CR22]--[@CR24]^ have been established. Tg mice developed by Lin et al., appeared to be susceptible to EV-A71 up to 2 weeks old and exhibited pathological features similar to a suckling mouse model^[@CR20]^. In their model, B genotypes of clinically isolated EV-A71 led to HFMD-like diseases while C genotypes led to neuropathogenesis, such as limb paralysis (LP). Another hybrid (hSCARB2+/+/stat-1−/−) mouse strain established from crossbreeding SCARB2 transgenic and stat-1 knockout (KO) mice was also susceptible to EV-A71 infection up to 2 weeks of age^[@CR23]^. In the murine model developed by Fujii et al., young hSCARB2 Tg mice (3-week-old) were susceptible to infection by the EV-A71 Isehara strain (C genotype) and displayed features of neuropathology^[@CR19]^. A recent hSCARB2 knock-in (KI) mouse model established by Zhou et al. showed that the KI mice are not only susceptible to luciferase (Luc) recombinant EV-A71 pseudovirus infection (up to 12 weeks of age) but also exhibited partial HFMD clinical symptoms present in humans, combining with bioluminescent imaging (BLI) technique^[@CR24]^. In this study, we characterized of this hSCARB2 KI mice challenged by two clinical strains, including EV-A71-CMU4232, an endemic strain isolated in mainland China in 2008 and a clone-derived virus (CDV)-Isehara, whose original strain (Isehara/Japan/99) was among the circulating strains identified in Japan in 1999^[@CR25]^ and was used in another hSCARB2 Tg mouse model to study EV-A71 neuropathogenesis^[@CR19]^. We assessed the utility of these mice as appropriate small animal models for evaluating the pathogenicity and virulence of different EV-A71 strains. Results {#Sec2} ======= Characterization of CMU4232 and CDV-Isehara {#Sec3} ------------------------------------------- At 24-h post infection, the rhabdomyosarcoma (RD) cells infected with CMU4232 and CDV-Isehara displayed typical CPE, such as cell rounding, aggregation, detachment, and floatation (Fig. [1a](#Fig1){ref-type="fig"}). The production of specific VP0 and VP2 proteins (Fig. [1b](#Fig1){ref-type="fig"}) was detected in the lysates of the cells infected by CMU4232 or CDV-Isehara.Fig. 1Characterization of CMU4232 and CDV-Isehara.**a** Cytopathic effects at 1dpi displayed on RD cells infected with CMU4232 or CDV-Isehara. **b** Detection of viral structural proteins VP0 and VP2 of CMU4232 and CDV-Isehara by western blotting with antibody specific for EV-A71. β-actin was used as a internal control. **c** Growth curves of CMU4232 and CDV-Isehara in RD cells. RD cells were inoculated with CMU4232 or CDV-Isehara as indicated at either MOI of 1 or 0.1. Samples were collected at the times indicated and titrated by PFU assay. All assays were performed in triplicate. At each time point, titer values are means of three samples; error bars represent SEM. dpi days post infection, PFU plaque forming unit, SEM standard error of mean We next determined the growth kinetics of the two viruses in RD cells. The CMU4232 strain showed a similar proliferation rate to that of CDV-Isehara during the 24 h post-inoculation period when an multiplicity of infection (MOI) of 1 was used to inoculate the RD cells (Fig. [1c](#Fig1){ref-type="fig"}). The CMU4232 strain reached a peak viral titer (3.8 × 10^7^ plaque forming unit (pfu)) after 24 h and then decreased, whereas CDV-Isehara reached a peak viral titer (4.3 × 10^7^ pfu) at 48 h post-inoculation. We also determined the growth curve for both viruses at an MOI of 0.1. At a low MOI, CMU4232 and CDV-Isehara reached a peak viral titer at 48 h postinoculation. The maximum viral titer observed for CMU4232 and CDV-Isehara was 2.6 × 10^7^ pfu and 3.55 × 10^7^ pfu, respectively (Fig. [1c](#Fig1){ref-type="fig"}). These data indicated that the CDV-Isehara strain replicated relatively faster than that of the stain CMU4232 in RD cells. Sequences and phylogenetic analyses of the strains CMU4232 and CDV-Isehara {#Sec4} -------------------------------------------------------------------------- The complete genomes of CMU4232 and CDV-Isehara exhibited a nucleotide similarity of 83%. The nucleotide and amino acid similarities of the four structural genes (VP1 to VP4) between CMU4232 and CDV-Isehara were observed to be 89--90.8% and 98.3--100%, respectively ([S1 Table](#MOESM1){ref-type="media"}). Of the 297 amino acids of VP1 protein, five substitutions were identified, including H22Q, A145E, N237T, V249I, and A289T (Fig. [2](#Fig2){ref-type="fig"}), whereas only one substitution (I31M) was observed in the 254 amino acids of the VP2 protein. No substitutions were observed in the VP3 and VP4 proteins between the two strains ([S1 Table](#MOESM1){ref-type="media"}).Fig. 2Variation of the VP1 amino acid sequences between CMU4232 and CDV-Isehara strains.In total, there were 297 amino acids in the VP1 protein of EV-A71, and five residue variations (22 H/Q, 145 A/E, 237 N/T, 249 V/I, and 289 A/T) were found in the VP1 amino acid sequence between CMU4232 and CDV-Isehara strains In the phylogenetic analysis, the sequences of 30 prevalent EV-A71 strains reported from different regions of the world were compared, including the strains CMU4232 and CDV-Isehara used in this study. Consistent with the results of previous studies, all the representative strains isolated in China during 1998--2017 were clustered into the C4 genotype. The C4 cluster showed a stepwise evolutionary trajectory over time. There are two branches included in the genotype C4, the early C4b (1998--2004) and the later C4a (2005--2017). The C4a branch could be further divided into two separate groups, C4a-1 and C4a-2, which include strains isolated from 2005--2007 and 2008--2017, respectively. The results clearly show that CMU4232 belongs to the C4a-2 of genotype C4 and CDV-Isehara belongs to genotype C2 (Fig. [3](#Fig3){ref-type="fig"}).Fig. 3Phylogenetic relationship among 30 worldwide EV-A71 strains based on the complete VP1 gene.CMU4232, CDV-Isehara, 18 representative CHN strains, and 10 international representative strains of different subgenotypes are included in this dendrogram. The 18 representative CHN strains were selected from different isolation places in mainland China from 1998 to 2017 according to available data on internet. Details of all the EV-A71 strains included in the dendrogram are provided in Table [S1](#MOESM1){ref-type="media"}. CHN Chinese hSCARB2 KI mice are susceptible to CDV-Isehara intracerebrally (I.C.) infection {#Sec5} ------------------------------------------------------------------------------- When hSCARB2 KI mice of different ages were challenged, 76/105 showed ataxia, paralysis, or death at days 2--14 after infection (Figs. [4a, b](#Fig4){ref-type="fig"} and Supporting videos). A significant difference in the disease incidence and death rate was observed between KI mice younger and older than 3 weeks (*P* \< 0.0001) (Figs. [4c, d](#Fig4){ref-type="fig"}). In addition, the death rate of 1-week-old-infected KI mice was 100% (20/20) from 3 to 5 days post inoculation (dpi), while that of 2-week-old-infected Tg mice was 77.3% (17/22) from 4 to 11 dpi (*P* \< 0.0001) (Fig. [4d](#Fig4){ref-type="fig"}). Paralysis occurred in both the forelimb and hindlimb post infection, particularly in the hindlimb. Some of the KI mice died within 1--3 days after LP symptoms appeared, while some were still alive when the symptoms gradually decreased. The susceptibility of the hSCARB2 KI mice to CDV-Isehara infection decreased as the age of mice increased, although no significant difference was observed among groups of KI mice older than 3 weeks (*P* \> 0.05) (Figs. [4b--d](#Fig4){ref-type="fig"}). The development of disease symptoms in 4-week-old-infected KI mice was very similar to that of 3-week-old (Fig. [4b](#Fig4){ref-type="fig"}). In 4-week-old-infected KI mice, 33.3% (8/24) and 20.8% (5/24) experienced mild and severe symptoms, respectively, while 12.5% (3/24) died, whereas 17.2% (5/29) and 27.6% (8/29) of 3-week-old-infected KI mice exhibited mild and severe symptoms, respectively, while 13.8% (4/29) died.Fig. 4Susceptibility of hSCARB2 KI mice to CDV-Isehara infection.The 1, 2, 3, 4, and 6-week-old hSCARB2 Tg mice were infected intracerebral (I.C.) with CDV-Isehara strain at a dose of 4.8 × 10^6^ pfu and monitored daily after infection. **a** The CNS-like hindlimb paralysis, ruffled fur, and shrinking. **b** The clinical symptoms (asymptomatic, mild, severe, and dead), followed by the criteria described in the materials and methods section (**c**) asymptomatic (%) and (**d**) survival (%). Significant difference of asymptomatic (%) and survival (%) with different weeks old Tg mice infected with CDV-Isehara was shown as \*\*\*\**p*,0.0001. pfu plaque forming unit, CNS central nervous system, I.C. intracerebral Difference in clinical features of hSCARB2 KI mice challenged by CMU4232 and CDV-Isehara {#Sec6} ---------------------------------------------------------------------------------------- Differences in the clinical features of hSCARB2 KI mice caused by the CMU4232 and CDV-Isehara strains when using the same route of infection (I.C. or I.V.) were assayed for using 3-week-old hSCARB2 KI mice. Some of the KI mice infected with both EV-A71 strains led to diverse clinical signs and various degrees of manifestations, especially in the development of neurological symptoms. The peak period of disease onset was 3--7 dpi, and death was typically delayed by 1--2 days after this period. No clinical signs were observed in the mock-infected KI mice (Fig [5](#Fig5){ref-type="fig"}a[1](#Fig1){ref-type="fig"}, a[2](#Fig2){ref-type="fig"}, b[1,](#Fig1){ref-type="fig"} and b[2](#Fig2){ref-type="fig"}).Fig. 5The clinical symptoms, asymptomatic (%), survival (%), and body weight change (%) of different EV-A71 Tg mouse models.Three-week-old hSCARB2 KI mice were inoculated I.C. with DMEM/ EV-A71 (**a**) or inoculated intravenous (I.V.) with DMEM/EV-A71 (**b**) respectively. Three-week-old hSCARB2 KI mice were inoculated with CMU4232 (**c**) /CDV-Isehara (**d**) strain via I.C. or I.V. route respectively. (1) The clinical symptoms, (2) asymptomatic (%), (3) survival (%), and (4) body weight change (%) of the Tg mice were observed and recorded daily for 2 weeks after viral infection. \*\*\*p,0.000. I.C. intracerebral, I.V. intravenous The death rate of the CDV-Isehara/I.C. group was 17.6% (6/34), and 41.2% (14/34) of mice displayed mild or severe clinical signs, including ruffled fur, hunchbacked appearance, easily frightened, trembling, shrinking, limb weakness, mental bluntness, dispirited, slow in action, emaciation, and LP, whereas only mild symptoms (19%) occurred in the CMU4232/I.C. group (*N* = 21), with trembling, shrinking, and limb weakness observed, and these mild symptoms completely disappeared soon thereafter (Fig [5](#Fig5){ref-type="fig"}a[1](#Fig1){ref-type="fig"}). The clinical signs of the CDV-Isehara/I.C. group were observed in 20 of 34 mice from 3 to 11 dpi, while clinical signs were observed in 4 of 21 mice from 3 to 7 dpi for the CMU4232/I.C. group (Fig [5](#Fig5){ref-type="fig"}a[2](#Fig2){ref-type="fig"}). A significant difference in disease incidence was observed between these two groups (*P* \< 0.05) (Fig [5](#Fig5){ref-type="fig"}a[2](#Fig2){ref-type="fig"}). However, no significant differences were found in death rate between them (*P* \> 0.05) (Fig [5](#Fig5){ref-type="fig"}a[3](#Fig3){ref-type="fig"}). Among the groups infected via the I.C. route, a significant difference in body weight change was also observed (*P* = 0.023) (Fig [5](#Fig5){ref-type="fig"}a[4](#Fig4){ref-type="fig"}). Furthermore, significant differences were observed between the CDV-Isehara/I.C. and CMU4232/I.C. or DMEM/I.C groups (*P* = 0.042 and *P* = 0.047, respectively) (Fig [5](#Fig5){ref-type="fig"}a[4](#Fig4){ref-type="fig"}). The body weights of mice in the CDV-Isehara/I.C. group increased much slower than those of in the CMU4232/I.C. and DMEM/I.C. groups, demonstrating that CDV-Isehara infection affected the growth of infected KI mice more than CMU4232 when administered via the same I.C. route. Taken together, the disease severity induced by the strain CMU4232 in 3-week-old hSCARB2 KI mice was less than that caused by the strain CDV-Isehara. These results suggested that these KI mice were susceptible to EV-A71 and that there was a difference in sensitivity of the hSCARB2 KI mice to the two strains. When the onset of typical symptoms was compared, hSCARB2 KI mice infected with the strain CMU4232 via the I.V. route exhibited typical symptoms from 2 to 11 dpi, including ruffled fur, hunchbacked appearance, easily frightened, trembling, shrinking, limb weakness, mental bluntness, dispirited, slow in action, emaciation, and death, but no severe symptoms or paralysis occurred when the mice were followed up to 14 dpi. However, the symptoms and consequences of CDV-Isehara/I.V. infection were more severe from 2 to 14 dpi, with paralysis in the limbs and death observed in this group (Fig [5](#Fig5){ref-type="fig"}b[1](#Fig1){ref-type="fig"}, b[2](#Fig2){ref-type="fig"}). In the hSCARB2 KI mice I.V. infected with CMU4232, 86.4% (19/22) developed mild symptoms and 4.5% (1/22) died whereas 50% (9/18) and 27.8% (5/18) of mice in CDV-Isehara/I.V. group exhibited mild and severe symptoms, respectively, while 11.1% (2/18) died (Fig [5](#Fig5){ref-type="fig"}b[1](#Fig1){ref-type="fig"}). A significant difference in the disease incidence was observed between the CMU4232/I.V. or CDV-Isehara/I.V. groups and the DMEM/I.V. group (*P* \< 0.0005 and *P* \< 0.005, respectively) (Fig [5](#Fig5){ref-type="fig"}b[2](#Fig2){ref-type="fig"}). However, no significant differences in disease incidence and death rate were observed between CMU4232/I.V. and CDV-Isehara/I.V. groups. In the CDV-Isehara/I.V. group, two of 18 infected Tg mice died from 3 to 10 dpi, while one death was observed in the CMU4232/I.C. group at 4 dpi (Fig [5](#Fig5){ref-type="fig"}b[3](#Fig3){ref-type="fig"}). Furthermore, no difference in body weight change was observed between these two groups (*P* = 0.468), although a significant difference was observed between the CDV-Isehara/I.V. and DMEM/I.V. groups (*P* = 0.012) (Fig [5](#Fig5){ref-type="fig"}b[4](#Fig4){ref-type="fig"}). We next compared the disease development of hSCARB2 KI mice infected with the same strain via different routes (I.C. or I.V.). For KI mice infected with CMU4232, distinct symptoms were observed and were strongly dependent on the infection route. The severity of the disease caused by the I.V. route was more severe than that of I.C. route based on the observed clinical symptoms, disease incidence (*P* \< 0.005) and the body weight change (*P* \< 0.005) (Fig [5](#Fig5){ref-type="fig"}c[1](#Fig1){ref-type="fig"}--c[4](#Fig4){ref-type="fig"}). However, the severity of the disease induced by CDV-Isehara was similar for both infection routes, irrespective to the observed clinical manifestations, disease incidence, death rate, and the body weight change (*P* \> 0.05) (Fig [5](#Fig5){ref-type="fig"}d[1](#Fig1){ref-type="fig"}--d[4](#Fig4){ref-type="fig"}). Viral proliferation and pathological features in hSCARB2 KI mice {#Sec7} ---------------------------------------------------------------- To assess viral proliferation in vivo, the viral RNA in tissues of infected 3-week-old KI mice was quantified by real-time PCR (Fig. [6](#Fig6){ref-type="fig"}). Specific viral RNAs were detected in the brain of CMU4232/ I.C. KI mice (Fig. [6a](#Fig6){ref-type="fig"}), while in the CDV-Isehara/I.C. group, viral RNAs were detected in the brain, intestine, muscle, and heart tissues of mice, but no viral RNAs were identified in the lungs (Fig. [6b](#Fig6){ref-type="fig"}). The distribution of both EV-A71 strains in tissues of KI mice infected via the I.V. route was highly consistent. However, higher viral loads were detected in CMU4232-infected KI mice than was observed in CDV-Isehara-infected mice. Furthermore, the viral load in the muscle was higher than that observed in the brain, heart, lung, and intestine, demonstrating the susceptibility of different tissues of hSCARB2 KI mice to EV-A71 when infected via the I.V. route (Figs. [6c, d](#Fig6){ref-type="fig"}). No viruses were detected in any tissue of the control group, which was inoculated with RD cell lysate supernatant.Fig. 6Virus replication in some tissues of hSCARB2 KI mice by EV-A71 infection.Three-week-old hSCARB2 KI mice were inoculated with CMU4232/ CDV-Isehara via I.C. or I.V. route. EV-A71 genome copy numbers in the brain, heart, lung, intestine, and muscle of EV-A71-infected Tg mice were determined at several time points by quantitative RT-PCR. Mouse β-actin gene expression in each tissue was used as the internal control. Results represent the mean of three to five samples; error bars represent STDEV. **a** CMU4232/I.C. (*N* = 3); **b** CDV-Isehara/I.C. (*N* = 4); **c** CMU4232 /I.V. (*N* = 5); **d** CDV-Isehara/I.V. (*N* = 4) Next, we observed the pathological changes in the two major replication sites, the brain and skeletal muscle (Fig. [6](#Fig6){ref-type="fig"}). Different degrees of pathological changes were evident in the KI mice with clinical signs. These changes primarily included cellular damage (such as necrosis and neuronophagia) and inflammatory changes (such as gliosis and perivascular cuffing) (Fig [7.1--7.3](#Fig7){ref-type="fig"}) in the brain. In addition, degeneration and necrosis in muscle bundles and severe necrotizing myositis with a mass of inflammatory cell infiltration was observed (Fig [7.5--7.7](#Fig7){ref-type="fig"}), especially in KI mice infected with CDV-Isehara via the I.V. route. Viral antigens were also detected in the above affected tissues. Immunohistochemistry (IHC) staining revealed the presence of viral antigen in the tissues with pathological changes, indicating that the pathological findings were caused by EV-A71 infection. All mock mice tested negative for viral antigens, regardless of the infection routes (Fig. [8](#Fig8){ref-type="fig"}).Fig. 7Pathological analysis of brain and the skeletal muscle of EV-A71-infected mouse models.Sporadic neuronal necrosis and neuronophagia (arrowhead) in the brain (1) and Inflammatory cell infiltration (arrowhead) in the skeletal muscle (5) of mouse model with CDV-Isehara/I.C.; glial nodule (arrow) and neuronophagia (arrowhead) in the brain (2) and severe necrotizing myositis with a mass of inflammatory cell infiltration (arrow) in the skeletal muscle (6) of mouse model with CDV-Isehara/I.V.; perivascular cuffing (arrowhead), partial neuronal necrosis and neuronophagia in the brain (3) and degeneration and necrosis in muscle bundles with inflammatory cell infiltration (arrowhead) (7) in mouse model with CMU4232/I.V.; the brain (4) and the skeletal muscle (8) of MOCK mouse with DMEM/I.V. (magnification: 200 × ). Representative samples are shown for each groupFig. 8VP2 expression in the brain and the skeletal muscle of EV-A71-infected mouse models.The brain (1) and the skeletal muscle (5) of mouse model with Isehara/I.C.; the brain (2) and the skeletal muscle (6) of mouse model with CDV-Isehara/I.V.; the brain (3) and the skeletal muscle (7) of mouse model with CMU0804/I.V.; the brain (4) and the skeletal muscle (8) of MOCK mouse with DMEM/I.V. Arrowheads indicate EV-A71-positive cells (IHC staining against EV-A71 VP2, magnification: 200 × ). Representative samples are shown for each group Discussion {#Sec8} ========== For robust results, an adequate number of hSCARB2 KI mice (*N* = 181) were used to establish a mouse model that is sensitive to EV-A71 infection in this study. In this mouse model, we observed that 3 weeks is a crucial age, as infected mice younger than 3-week-old became extremely ill. The susceptibility of the KI mice to EV-A71 infection could be observed until 8-week in mice with viral challenged by EV-A71 at 6-week-old age. Fujii et al. reported that their Tg mice older than 6 weeks were also susceptible to EV-A71 infection, and the development of disease was similar to that of 3-week-old mice^[@CR19]^. The model based on 3-week-old weaning hSCARB2 KI mice in the present study displayed diverse clinical symptoms and was characterized as pathological tropism in the CNS and peripheral sites. Based on the reliable EV-A71 disease model, we evaluated the pathogenicity of an endemic strain (CMU4232) in mainland China, and compared it with an internationally published strain (Isehara)^[@CR19]^. We observed that disease manifestations induced by CDV-Isehara were more severe than those induced by CMU4232. Both the biological characteristics of specific viral strains and infection routes were observed to be important for the pathogenicity of EV-A71 in this hSCARB2 KI mouse model. Previous studies reported that different EV-A71 strains could display varying pathogenic effects^[@CR26]--[@CR28]^. When infected via the I.C. route, the disease induced by CDV-Isehara was more severe than that induced by CMU4232 at the same dose (Fig [5](#Fig5){ref-type="fig"}a[1](#Fig1){ref-type="fig"}) in our KI mice model. A significant difference in disease incidence between these two groups was observed (*P* \< 0.05) (Fig [5](#Fig5){ref-type="fig"}a[2](#Fig2){ref-type="fig"}). For mice in the CMU4232/I.C. group, only mild clinical symptoms (4/21) developed from 3 to 7 dpi, and these symptoms subsequently completely disappeared. In contrast, mice in the CDV-Isehara/I.C. group exhibited a longer onset (from 3 to 11 dpi) of various mild manifestations, paralysis, and death (6/34). In addition to the difference in disease incidence caused by the two strains, we also observed a difference in the presence of the viruses infected mouse tissues. Using qRT-PCR, viral RNAs were detected in the intestine, muscle, heart, and brain tissues of mice in the CDV-Isehara/I.C. group. We speculated that CDV-Isehara replicated well in the brain after I.C. inoculation and then entered into multiple organs via viremia, leading to viral infections that caused clinical manifestations (Fig. [6b](#Fig6){ref-type="fig"}). In contrast, viral replication of the strain CMU4232 was only detected in the brains of mice after infected via the I.C. route, and viral titers in other non-CNS organs were below detectable levels via qRT-PCR. This difference suggested that the strain CMU4232 may be less sensitive to or replicated less efficiently in the brain of this mouse model, resulting in the eventual elimination of the virus. Thus, low viremia failed to induce an effective infection in other non-CNS tissues. With I.C. administration, the brain is the most direct injury site. Both CMU4232 and CDV-Isehara strains can replicate at this site, indicating that both strains possess the characteristics of neurotropism. However, the neurovirulence of the strain CMU4232 appeared to be weaker than that of CDV-Isehara in this model. Considering that the CDV-Isehara strain originated from an infectious clone, it is unclear whether there could be some change in virulence compared with the original strain. However, the neurovirulence of the original Isehara/Japan/99 isolate in another hSCARB2 Tg mouse model was significantly superior to those of other EV-A71 strains from encephalitis and death cases^[@CR19]^. Based on a phylogenetic analysis, the two strains used in this study were determined to have the genotypes C4 and C2, respectively (Fig. [3](#Fig3){ref-type="fig"}). The different genetic backgrounds of the two strains may explain the difference in virulence observed between the two EV-A71 strains in our mouse model. We observed that the nucleotide similarity of the CMU4232 and CDV-Isehara complete genomes was only 83%. VP1 of EV-A71 is known to be an important immunodominant protein that mediates the binding of this virus to host receptor and determines its tissue tropism. Elizabeth et al. reported that a single mutation (K244E) in the VP1 is responsible for increased virulence and neurotropism in adult interferon-deficient mice^[@CR29]^. When the deduced amino acid sequences were compared in this study, difference in five residues (H22Q, A145E, N237T, V249I, and A289T) were observed between the two strains CMU4232 and CDV-Isehara (Fig. [3](#Fig3){ref-type="fig"}). The substitution of VP1-145G or -145Q for -145E has been shown to be responsible for murine adaptation and/or virulence, either alone or in combination with changes in other amino acids in different mouse models^[@CR30]--[@CR33]^. In our hSCARB2 KI mice, the strain CDV-Isehara harbors VP1-145E and induce more severe neurological manifestations and myositis than those induced by the strain CMU4232 which harbors VP1-145A (Figs. [7](#Fig7){ref-type="fig"} and [8](#Fig8){ref-type="fig"}). These results suggest that EV-A71 with VP1-145E can confer better in vivo fitness of the virus, and viruse carrying this mutation are more virulent than strains with other amino acids at this position (VP1-145A/G/Q) in mouse models. Furthermore, Kataoka et al. identified VP1-145 as a critical molecular determinant for the binding of EV-A71 to another specific cellular receptor, human P-selectin glycoprotein ligand-1 (PSGL-1)^[@CR34]^. The strong in vivo selection of VP1-145E variants and CNS spread have been suggested to occur in a PSGL-1-independent manner. The VP1-145E variant was primarily responsible for the development of viremia and neuropathogenesis in a cynomolgus monkey model^[@CR35]^. Previous studies showed that the amino acid VP1-145 is a variable residue among the clinical EV-A71 isolates, and it serves as a major site of positive selection in the evolution of EV-A71^[@CR3],[@CR36],[@CR37]^. Molecular epidemiological studies have shown that strains with VP1-145G/Q/R are more frequently observed in severe neurological cases in humans than VP1-145E isolates^[@CR8],[@CR38]--[@CR40]^. Our finding also suggested that VP1-145E may be a crucial residue that is responsible for increased virulence in mice through unclear mechanisms, which may involve the other receptors and proteins required for viral entry and host immune responses to EV-A71. Two latest studies conducted in Japan reported that the amino acid VP1-145E contributed to virulence determination by controlling attachment receptor and antibody sensitivity^[@CR41],[@CR42]^. The virus Isehara with VP1-145E replicated efficiently in the CNS and acted as a virulent phenotype in their hSCARB2 Tg mice. However, VP1-145G viruses were adsorbed by attachment receptors such as heparan sulfate during circulation in vivo, leading to abortive infection and less virulent phenotype in the same Tg mice model. The maintenance of viremia titer at a high level is necessary for virus entrance into CNS via the intravenous route^[@CR41]^. When cynomolgus monkey model was tested, VP1-145E, but not VP1-145G viruses induced neurological symptoms. VP1-145E viruses were frequently detected in the tissues of infected monkeys, but VP1-145G viruses were detected less frequently and disappeared quickly. Clearly, VP1-145E viruses were more resistant to neutralizing antibodies than VP1-145G viruses^[@CR42]^. Our results are in line with the two studies. However, the exact mechanism of the CMU4232 with VP1-145A in the infected hSCARB2 KI mice remains to be elucidated. The difference between the pathogenicity of EV-A71 in animals and humans caused by the VP1-145 variation may partly explain why the Isehara/Japan/99 strain from a HFMD patient was observed to be more pathogenic than other clinical strains from encephalitis and death cases in another hSCARB2 Tg mouse model^[@CR19]^. In addition to the VP1-145, two other variable sites observed in this study (VP1-22 and VP1-237) have been reported to be under positive selection pressure^[@CR3],[@CR36],[@CR39]^. We also observed a difference at VP2-I31M between these two strains. Huang et al. reported that amino acid changes in both A145E of VP1 and K149M of VP2 contributed to viral infectivity in vitro and mouse lethality in vivo^[@CR32]^. However, we did not observe an amino acid change at 149 K of VP2. It is known that both alanine (A) and glycine (G) are nonpolar neutral amino acids. However, it is not known whether CMU4232 VP1-145A could be adsorbed by heparan sulfate in tissues with abundant blood supply under the super higher viremia titer (10^8^ pfu), as VP1-145G viruses were tested by Kobayashi et al.^[@CR41]^ When assessing differences arising from the routes of viral infection (I.C. vs. I.V.), the CDV-Isehara models displayed no significant difference in disease incidence, survival rate or body weight change (Fig [5](#Fig5){ref-type="fig"}d[2](#Fig2){ref-type="fig"}--[5](#Fig5){ref-type="fig"}d[4](#Fig4){ref-type="fig"}), although differences in the distribution of the virus in the tissues of infected KI mice were observed (Figs. [6b](#Fig6){ref-type="fig"}, [6d](#Fig6){ref-type="fig"}). The replication of CDV-Isehara was detected in all tested tissues of mice via the I.V. route, which differed a report by Fujii et al. in which the Isehara/Japan/99 strain was only detected in CNS in another Tg mouse model when using the same I.V. infection route. However, the dose used by Fujii et al. was 1 × 10^6^ TCID~50~^[@CR19]^, whereas the dose used in our study was about 100 times higher. Therefore, the results obtained by Fujii et al. cannot be directly compared with those obtained in this study. It should be kept in mind that pathological tropisms observed in the CNS and peripheral sites cannot only be distinguished by clinical symptoms. The identification of virus distribution and viral load in different tissues still needs a combination of different assays used in this study. In contrast to the CDV-Isehara models, significant differences between the CMU4232/I.C. and CMU4232/I.V. group were observed with respect to disease incidence, body weight change (Fig [5](#Fig5){ref-type="fig"}c[2,](#Fig2){ref-type="fig"} c[4](#Fig4){ref-type="fig"}) and the distribution of the virus in tissues (Figs. [6a](#Fig6){ref-type="fig"}, [c](#Fig6){ref-type="fig"}). CMU4232 could only replicate in the brains of mice when 3.2--6.7 × 10^6^ pfu was administered via the I.C. route. However, via an I.V. administration of 3.2--7.5 × 10^8^ pfu, CMU4232 could infect multiple organs. Following viremia, CMU4232 disseminated from the peripheral to central, leading to the observed CNS manifestations finally and demonstrating that the viral dose/hyperviremia is crucial to the in pathogenicity of EV-A71. Our results are in agreement with a recent study in which epidemic Chinese EV-A71 strains isolated from 2008--2010 were compared in neonatal ICR mice and neonatal rhesus monkeys^[@CR27]^. It appears that the CMU4232/IV mouse model mimics not only the neurological manifestation and infection of multiple organs injury following hyperviremia but also the morbidity of EV-A71 infection in human beings to some extent. A previous study reported that the risk of neurological complications was 1.1% and the severe-case fatality risk was 3.0%, with \>90% of deaths associated with EV-A71 in the research of HFMD in China from 2008--2012^[@CR9]^. In this study, we also compared the replication capacity of the strains CMU4232 and CDV-Isehara in RD cells through a plaque-forming assay. The growth curves showed that CDV-Isehara replicated faster and generated more virions than CMU4232 (Fig. [1c](#Fig1){ref-type="fig"}). As reported by Sun et al., a difference in the replication capacity of EV-A71 strains in RD cells could indicate their pathogenicity in humans^[@CR43]^. There appeared to be some differences in viral loads of the two strains detected in vitro and in vivo. Under the I.C. inoculation, the virus CDV-Isehara was detected in different tissues including the brain, intestine, muscle, and heart (Fig. [6b](#Fig6){ref-type="fig"}) whereas the virus CMU4232 was only detected in the brain (Fig. [6a](#Fig6){ref-type="fig"}). Furthermore, the viral load in tissues infected by CDV-Isehara was higher than that infected by CMU4232. The finding was consistent to the growth curves in RD cells (Fig. [1c](#Fig1){ref-type="fig"}). Under the I.V. inoculation, both viruses were detected in the brain, heart, lung, intestine, and muscle (Figs. [6c,d](#Fig6){ref-type="fig"}). The highest viral loads of both viruses were observed in muscles, dissimilar to those observed under the I.C. route infection. In addition, the viral load in tissues infected by CMU4232 was higher than that by CDV-Isehara. However, we did not observe differences in rates of morbidity and mortality and in change of the body weight of mice between these two viruses after I.V. inoculation. In contrast, CDV-Isehara-infected mice appeared to have a high proportion of severe symptoms compared with those of CMU4232-infected mice. The result indicated that the virus CDV-Isehara seems to be more virulent than the CMU4232 in the hSCARB2 KI mouse model. On the other hand, although having had high viral load in different tissues, the virus CMU4232 seemed to cause mild symptoms, suggesting that the virus CMU4232 is less virulent than that of CDV-Isehara. In short, one possible explanation for the difference observed may be related to the infectious routes used. Another explanation may be due to characteristics of the same strain in vitro and in vivo. In addition, we could not exclude the effect of timing of disease onset when tissue samples were taken from the infected mice. EV-A71 strains have been shown to be continuously evolving. The strain CMU4232 used in this study is a member of the recent C4a-2 cluster, suggesting that CMU4232 could be a representative endemic strain after the HFMD outbreak in mainland China in 2008 (Fig. [3](#Fig3){ref-type="fig"}). EV-A71 strains with the genotype C4 have been endemic and epidemic in mainland China for 20 years. The prevalence of neutralizing antibodies against the EV-A71 strains has been high in China and the selection pressure for the virus from population immunity is strong^[@CR44],[@CR45]^. Indeed, as reported by Zhang et al. the evolution rate of C4a EV-A71 appeared to be faster than the average for all other EV-A71 genotypes^[@CR46]^. Therefore, persistent attention should be paid to the changes in virulence resulting from evolution of EV-A71 viruses and their related pathogenic characteristics. In summary, we showed that hSCARB2 KI mice are a sensitive and useful model for investigating the clinical outcomes caused by different EV-A71 strains. Diverse clinical symptoms characterized as bi-pathological tropism in the CNS and peripheral sites can be observed even when the "representative" Chinese strain CMU4232 was tested. The intracranial infection model appears to be suitable for evaluating of neurovirulence caused by EV-A71, whereas the venous infection model is useful for studying the pathogenicity of EV-A71. Moreover, impact of the change in key amino acids site (e.g., VP1-145) on pathogenesis in different hosts should be taken into account in future studies. Materials and methods {#Sec9} ===================== hSCARB2 KI mice and ethics statements {#Sec10} ------------------------------------- The human SCARB2 KI mice^[@CR24]^ used in this study were provided by the Institute for Laboratory Animal Resources, National Institute for Food and Drug Control, Beijing, China. The animal experiments were performed according to the recommendations in the national guidelines for the care and use of animals in scientific research "Regulations for the Administration of Affairs Concerning Experimental Animals". The study protocol was approved by the Capital Medical University Animal Experiments and Experimental Animals Management Committee (AEEI-2016-121). Termination of mice was performed under anesthesia, and all efforts were made to minimize suffering. All samples used in this study were anonymized. EV-A71 viruses and cells {#Sec11} ------------------------ CMU4232 (EV71/CMU4232-1/BJ/CHN/2008) was originally isolated from a HFMD patient by our group during the 2008 outbreak in Beijing^[@CR8]^, and it was passaged in RD cells (National Infrastructure of Cell Line Resource, Beijing, China) and stocked in our laboratory. An infectious cDNA clone of EV-A71 (pSVA14-Isehara ver4) was kindly provided by Professor Satoshi Koike, Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan. The RD cell line was cultured in DMEM containing 4.5 g/L glucose, L-glutamine, and sodium pyruvate (Corning, 10-013-CVR, USA) supplemented with 10% fetal bovine serum (FBS) (Corning, 35-076-CV, USA), 100 IU of penicillin, and 100 μg of streptomycin per ml. The cells were incubated at 37 ℃ and with 5% CO~2~. CMU4232 and the rescued virus, CDV-Isehara were harvested from RD cultures by freezing and thawing three times and were stored at −80 ℃. The titers of the virus stocks were tested using a modified plaque-forming assay and determining the CCID50. Growth curves of CMU4232 and CDV-Isehara in RD cells {#Sec12} ---------------------------------------------------- To assess the proliferation dynamics of both EV-A71 strains, the growth curves were determined as follows: RD cells cultured in 24-well plates were inoculated at an MOI of 1 or 0.1 in triplicate. After being allowed to attach for 2 h, the unbound viral particles were washed off with PBS. Next, the infected cells were cultured at 37 ℃ for 0, 8, 24, 48, 72, and 96 h and were titrated to determine pfu values after three consecutive freeze--thaw cycles. All assays were performed in triplicate. Western blot analysis {#Sec13} --------------------- Infected RD cells were collected and lysed using RIPA buffer. Viral and cellular proteins were separated by SDS-PAGE on a 12% gel and were electrotransferred onto a PVDF membrane. After blocking, the membrane was probed with a mouse anti-EV-A71 antibody (Millipore, MAB979, 1:1000 dilution) and an HRP-conjugated secondary antibody (Transgen Biotech, HS201-1, 1:5000 dilution). Reverse transcription (RT)-PCR and nucleotide sequencing of CMU4232 {#Sec14} ------------------------------------------------------------------- Viral RNA of the CMU4232 strain was extracted from infectious RD cell supernatant using a QIAmp^®^ Viral RNA Kit (Qiagen, Hilden, Germany). RNA was reverse transcribed into cDNA using a Superscript II Kit (Invitrogen, USA) according to the manufacturer's instructions. All sequencing was performed by Guangzhou Darui Biotechnology Co., Ltd. The accession number of the complete genome sequence of CMU4232 is MH373639. Nucleotide and deduced amino acid sequence identity analysis between CMU4232 and CDV-Isehara {#Sec15} -------------------------------------------------------------------------------------------- The complete genome sequence alignment of CMU4232 and CDV-Isehara was performed using the Bioedit and Clustal W programs in MEGA 4.0, and the deduced amino acid sequences of structural and nonstructural proteins were aligned and shown. Phylogenetic analysis {#Sec16} --------------------- Representative strains of each EV-A71 subgenotypes were chosen based on a previously published study, especially representative strains available at the Genbank reported in mainland China that were first screened from those isolated from different regions and time periods from 1998 to 2017. Phylogenetic analysis was based on viral VP1 protein-coding nucleotide sequences, and the phylogenetic tree was constructed using the neighbor-joining method using MEGA 4.0 software with 1000 bootstrap replications. Details of all the EV-A71 strains included in the phylogenetic tree are provided in Supplementary Table [S2](#MOESM1){ref-type="media"}. Infection of EV-A71 in hSCARB2 KI mice {#Sec17} -------------------------------------- To evaluate the susceptibility of the hSCARB2 KI mice to EV-A71 infection, we challenged 1-, 2-, 3-, 4-, and 6-week-old KI mice I.C. with the CDV-Isehara strain at a dose of 4.8 × 10^6^ pfu and observed the development of clinical symptoms daily for 2 weeks. In addition, we chose 3-week-old hSCARB2 KI mice to be inoculated with EV-A71 (CMU4232/ CDV-Isehara) at the indicated doses and routes (I.C. infection at a dose of 3.2--6.7 × 10^6^ pfu or I.V. infection at a dose of 3.2--7.5 × 10^8^ pfu per animal). Control mice were mock injected with an equivalent volume of RD cell lysate supernatant. All mice were monitored daily for clinical signs, survival, and body weight for 2 weeks. The activity level, mental status, degree of limb paralysis, and other symptoms were observed and recorded daily for each mouse. Furthermore, the clinical signs were scored as follows: asymptomatic (0), ruffled fur and/or hunchbacked appearance (1--2), easily frightened/ trembling/shrinking/mental bluntness/dispirited/slow in action (3--5), limb-shake weakness (6), wasting \> 3 consecutive days (7), limb paralysis (8), and moribund and death (9). Based on the scores, we classified the clinical manifestations at four levels, asymptomatic (0--2); mild symptoms (3--5); severe symptoms (6--8); and death (9). Experiments for group of 3-week-old mice were performed in triplicate. Quantitative detection of EV-A71 RNA by real-time RT-PCR in different tissues {#Sec18} ----------------------------------------------------------------------------- The infected mice were killed at the peak onset of disease development and were immediately processed for sample collection. Tissues were collected, including brains, hearts, lungs, skeletal muscle, and intestines. Total RNA was extracted from different tissues using TRIzol reagent (Invitrogen, 15596026, USA) following the manufacturer's instructions, and then was converted into cDNA with a RT-PCR kit (Transgen Biotech, AE301-02) followed by qPCR using a PCR system (Transgen Biotech, AQ101-02). PCR amplifications were performed at 94 ◦C for 30 s followed by 40 cycles of 94 ◦C for 5 s, 60 ◦C for 15 s, and 72 ◦C for 10 s. SYBR Green PCR Master Mix (Transgen Biotech, AQ131-02) was used to detect of viral load in different tissues using the primers of EV71-S /A (S3 Table). Mouse β-actin cDNA was used as an internal control. Each assay was carried out in triplicate. Histopathology examination (HE) and immunohistochemistry (IHC) analysis {#Sec19} ----------------------------------------------------------------------- Tissue samples from control and EV-A71-infected hSCARB2 KI mice were fixed with formalin buffered saline by perfusion and then dehydrated and embedded in paraffin and sectioned according to conventional procedures. Some sections were either stained with hematoxylin--eosin or were used for IHC analysis. Tissue sections were dewaxed and rehydrated in graded ethanol, and IHC was performed using a standard avidin--biotin immunoperoxidase technique. An anti-EV-A71 monoclonal antibody (MAB979; Millipore, USA), was used as a primary antibody. Electronic supplementary material ================================= {#Sec20} Supplementary Tables **Publisher's note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Electronic supplementary material ================================= **Supplementary Information** accompanies this paper at (10.1038/s41426-018-0201-3). The cDNA clone of EV-A71, pSVA14-Isehara ver4 was kindly provided by Prof. Satoshi Koike, Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan. This work was partly supported by National Key Technologies R&D Program for the 13th Five-year Plan (2017ZX10202101-004), Beijing Natural Science Foundation (7172016) and the National Science and Technology Major Projects of Infectious Disease (grants 2017ZX103304402). J.Z., C.F., and Q.H. conceived and designed the study. J.Z., N.C., K.Z., L.S., Y.Z., L.C., X.Z., Q.X., and Z.C. performed the experiments. S.Z. and C.W. participated in experiments. C.F. participated in the writing of the manuscript. J.Z. and Q.H. analyzed data and wrote the paper. All authors reviewed and approved the final paper. The authors declare that they have no conflict of interest.

{ "pile_set_name": "PubMed Central" }

Background ========== Alzheimer's disease (AD) afflicts millions of elderly people worldwide and is the most common cause of dementia among neurodegenerative diseases. AD symptoms include impaired memory, aphasia, and visuospatial deficits while the histopathological changes of AD are characterized by senile plaques, neurofibrillary tangles, and lipid granule accumulation. The major molecular component of senile plaques is aggregated amyloid-β (Aβ) peptide, which is generated from the amyloid precursor protein (APP) via sequential cleavage by β-secretase and the γ-secretase complex \[[@B1]\]. AD onset is assumed to be related to Aβ peptide polymerization. On the other hand, neurofibrillary tangles are composed of aberrantly phosphorylated tau protein \[[@B2]\]. In addition to the above pathological features, numerous studies reported a link between lipids and AD. In general, membrane lipids provide a milieu for transmembrane proteins and can modulate their function. Indeed, γ-secretase activity is affected by the lipid composition of the membrane with sphingolipids and cholesterol increasing and phosphatidylinositols decreasing its activity \[[@B3],[@B4]\]. Lipid carbon chain length and double bond position also affect γ-secretase activity. Hypercholesterolemia reportedly accelerates intraneuronal accumulation of Aβ oligomers and subsequent synapse loss, resulting in memory impairment in AD model mice \[[@B5]\]. Moreover, ω-3 polyunsaturated fatty acids were shown to have possible roles in AD prevention \[[@B6]\]. The docosahexaenoic acid (DHA) content of phospholipids (PLs) was lower in brain tissue and plasma of AD patients compared to those without cognitive impairment \[[@B7]\]. In addition, neuroprotectin D1, which is derived from DHA, was related to suppression of Aβ~42~-induced neurotoxicity \[[@B8]\]. Several lipidomic analyses for AD were recently reported. For instance, an analysis of postmortem brain tissues obtained from patients with late-onset AD showed enrichment in lysobisphosphatidic acids, sphingomyelins (SMs), ganglioside GM3, and cholesterol esters (ChEs) as well as novel region-specific lipid anomalies that were potentially linked to AD pathogenesis \[[@B9]\]. However, our current knowledge of the overall changes in lipids related to AD, especially for the time-course of their changes, remains low. Animal models are extensively used to study AD pathogenesis since brain samples can be obtained from pre-symptomatic to late stages of AD, which allows analysis of pathophysiological changes at different disease stages. A recent study showed that some lipid dysregulation occurred in postmortem AD brain tissues, including increased levels of ganglioside GM3 and ChEs, which was consistent with changes seen in AD mouse models \[[@B9]\], and supports the validity of using animal models to study AD disease mechanisms. In this study, we performed a global lipid metabolomic analysis using brain tissues and plasma obtained from transgenic mice expressing both mutated human APP and tau proteins \[[@B10]\] at 4 months (pre-symptomatic), 10 months (early disease stage), and 15 months (late stage). We focused not only on lipid changes observed in the brain but also on those that were consistently changed in both brain and plasma because these molecules could be candidate biomarkers for AD diagnosis. Materials and methods ===================== Animals ------- We used female transgenic mice hemizygous for transgenes encoding both human mutated APP and tau (APP/tau mice) as AD model mice. The mice were generated by crossing Tg2576 male transgenic mice hemizygously expressing human APP containing Swedish type mutations (Lys670Asn, Met671Leu) with JNPL3 female transgenic mice homozygously expressing mutant P301L four-repeat tau proteins. Wild-type tau females crossed with wild-type APP male mice were used as a control (no transgenes). Two-month-old mice were purchased from Taconics Farms, Inc. (Hudson, NY, USA) and raised to 4 (pre-symptomatic), 10 (early stage of symptoms), and 15 months (late stage) of age. At the given time point, mice were anesthetized with 2.0-2.5% isoflurane in the atmosphere and blood from the inferior vena cava was collected in tubes containing EDTA-2Na for plasma preparation. After blood collection, mouse brains were excised, and the cerebral hemispheres without the cerebellum were weighed and frozen in liquid nitrogen before storage at −80°C. Each group consisted of 5 mice except for the 15-month-old APP/tau mice (N=3). All animal experiments were performed in accordance with the Guidelines for Animal Experimentation and with approval from the Ethics Committee of Animal Care and Experimentation of the National Center for Geriatrics and Gerontology, Japan. Assay of human Aβ~40~ and tau proteins -------------------------------------- The right cerebral hemisphere from each animal (*ca*. 150 mg) was homogenized as described below in the section describing lipid extraction. After removal of methanol with a centrifugal vacuum concentrator (TOMY), Aβ~40~ in brain tissues (2 mg) was extracted by homogenization in 0.2 ml 70% formic acid with a small Dounce homogenizer and centrifugation at 100,000 × g for 1 h at 4°C in an ultracentrifuge (Optima MAX-E, Beckman Coulter, Tokyo, Japan). The supernatants were recovered and stored at −80°C until analysis. Frozen formic acid extracts were thawed on ice and neutralized with 20 volumes of 1 M Tris base. Within 1 h of neutralization, Aβ~40~ levels in the extracts were determined by an Enzyme-Linked Immunosorbent Assay (ELISA) kit for human Aβ~40~ (Wako Pure Chemical Industries, Ltd., Osaka, Japan) according to the manufacturer's instructions. Total protein in the extracts was assayed by a Pierce® BCA™ Protein Assay Kit (Thermo Fisher Scientific Ltd., Yokohama, Japan). Human tau protein in brain tissue from each mouse was analyzed according to the method described by Lewis et al. \[[@B10]\]. Two mg of EtOH-free homogenized brain tissue from each mouse was homogenized with a small Dounce homogenizer in 0.2 ml of tris-buffered saline (TBS) and centrifuged at 100,000 × g for 1 h at 4°C. The supernatant was collected as the TBS-soluble fraction. The pellet was homogenized similarly in 0.2 ml of 0.8 M NaCl and 10% sucrose in TBS, and after centrifugation at 150,000 × g for 15 min at 4°C, the supernatant was brought to 1% sarkosyl and incubated at 37°C for 1 h. The mixture was then centrifuged at 150,000 × g for 30 min at 4°C whereupon the supernatant and precipitate were collected as the sarkosyl-soluble and -insoluble fraction, respectively. TBS-soluble and sarkosyl-soluble fractions were combined and defined as the soluble fraction. Human tau levels in the soluble and sarkosyl-insoluble fractions were determined using an ELISA kit for human tau (Lifetechnology Japan Ltd., Tokyo, Japan). The sarkosyl-insoluble fraction was dissolved in 0.1 ml of 70% formic acid and neutralized with 20 volumes of 1M Tris base immediately before ELISA determination. Lipid extraction ---------------- Total lipids from right cerebral hemisphere tissues or plasma were extracted using Bligh & Dyer's (BD) method with minor modifications \[[@B11]\]. The right cerebral hemisphere from each animal (*ca*. 150 mg) was homogenized with zirconia beads in 1 mL of methanol at 4°C using a cell disruptor (TOMY), and the equivalent of 10 mg tissue was transferred into glass tubes. Plasma (100 μl) diluted with methanol was also transferred into glass tubes. A mixture of internal standards (ISs) was added to tissue homogenates or plasma: 1, 2-dipalmitoyl D6-3-*sn* glycerophosphatidylcholine (16:0/16:0PC-d6, 40 nmol/10 mg tissue or 20 nmol/100 μl plasma, Larodan, Malmo, Sweden), 1,2-caprylin-3-linolein (250 pmol and 20 nmol, Larodan), ^13^C-labelled tripalmitin (tripalmitin-1,1,1-^13^C~3~, 250 pmol and 2 nmol, Larodan), deuterated prostaglandin E~2~ (PGE~2~-d4, 5 ng in both tissue and plasma, Cayman Chemical, Ann Arbor, MI, USA) and deuterated leukotriene B~4~ (leukotriene B~4~-d4, 5 ng in both tissue and plasma, Cayman Chemical). Then, chloroform, methanol and 20 mM potassium phosphate (Kpi) buffer were added to achieve a volume ratio of buffer/methanol/chloroform = 0.8/2/1, and mixed vigorously for 5 min. Phase separation was achieved by adding 1 ml each of chloroform and 20 mM Kpi buffer and vortexing. The mixture was centrifuged at 1,000 × g for 10 min. For brain tissue, both the upper aqueous layer and bottom organic layer were collected. For plasma, 100 mM KCl/methanol/chloroform (48/47/3) was further added to the remaining organic samples after the upper aqueous layer was collected. After mixing vigorously for 5 min and centrifugation at 1,000 x g for 10 min, only the bottom organic layer was collected. Samples of the organic layer were dried under a gentle stream of nitrogen, and then dissolved in chloroform-methanol (1:1) at a concentration of 10 mg (tissue weight)/ml or 100 μL (plasma volume)/ml, and were stored at −90°C until use (BD sample). To distinguish alkenylacyl and alkyl PL species with the same exact mass as 34:1 ethanolamine plasmalogen (pPE) and 34:2 alkylacyl PE, a small aliquot of each BD sample was hydrolyzed using 0.5 N HCl as described previously (BD acid hydrolysis sample) \[[@B12]\]. Samples of the aqueous layer (3.2 ml) were subjected to solid extraction to obtain oxidative fatty acids. First, samples were diluted 10-fold using water adjusted to pH 3.0 with 1N HCl, and then applied to Oasis SPE cartridges (60 mg, Waters, Millford, MA, USA) preconditioned with 3 ml each of methanol and Milli Q water. After columns were washed with 3 ml of Milli-Q water followed by 3 ml hexane, the oxidative fatty acids were eluted with 3 ml methyl formate (MF). The MF fraction was dried under nitrogen, dissolved in chloroform-methanol (1:1) at a concentration of 10 mg (tissue weight)/ml or 100 μL (plasma volume)/ml, and stored at −90°C until use. Lipid metabolite analysis by reverse-phase liquid chromatography (RPLC)-time of flight mass spectrometry (TOFMS) ---------------------------------------------------------------------------------------------------------------- Total lipid samples (BD samples and BD acid hydrolysis samples) were analyzed by electrospray ionization (ESI)-TOFMS (LCT Premier XE; Waters Micro-mass, Waters) interfaced with an Acquity UPLC System (Waters). The MS was operated in a W-optics mode with 10,000 and 8,000 resolution in the positive and negative modes, respectively, using dynamic range extension. The scan range of the instrument was set at a mass-to-charge ratio (*m*/*z*) of 150--1200, and three functions (positive and negative ion modes, and another negative ion mode for in-source fragmentation) were recorded simultaneously for lipid species identification and quantification. The capillary voltage in the positive ion mode was set at 3.0 kV and the cone voltage 40 V, whereas in the negative ion mode the capillary and cone voltage was 2.5 kV and -40V, respectively. In in-source fragmentation of the negative ion mode, the aperture 1 voltage was set at 60 V. The desolvation gas was set to 600 L h^−1^ at a temperature of 350°C; the cone gas was set to 50 L/h and the source temperature was 120°C. The data acquisition rate was 0.5 s with a 0.01 s interscan delay. All analyses were acquired using the lock spray to ensure accuracy and reproducibility. A lock-mass of leucine enkephalin at 0.6 ng/mL in 50:50 acetonitrile:water containing 0.1% formic acid was used for the positive and negative ion modes (\[M+H\]^+^ = 556.2771 and \[M+H\]^-^ = 554.2615) with a flow rate of 5.0 μL/min via a lock spray ionization source. The data were collected in the centroid mode using MassLynx (Waters). Before the RPLC-ESI-TOFMS analysis, the dried BD and BD acid hydrolysis samples were redissolved in chloroform/methanol/isopropanol (1:2:4) supplemented with 0.2% formic acid and 0.028% ammonia at a final concentration of 10 mg tissue/ml solution or 100 μl plasma/ml solution, and filtered through a 0.2 μm pore size polyvinylidene difluoride membrane filter (Millipore, Bedford, MA, USA). The injection volume was fixed at 2 μL, and an ACQUITY UPLC BEH C18 column (1.7 mm i.d. × 150 mm, 1.7 μm) was used for separation. The column temperature was maintained at 45°C. The flow rate of the mobile phase was 70 μL/min. Mobile phase A consisted of acetonitrile/methanol/water = 18/18/4 (0.1% formic acid and 0.028% ammonia), while mobile phase B consisted of isopropanol (0.1% formic acid and 0.028% ammonia). The linear gradient increased from 0 to 15% B over 5 min, from 15% to 30% B for the next 20 min, from 30% to 45% B during the next 15 min, from 45% to 60% B for the next 10 min, from 60% to 80% B for the next 5 min, and immediately ramped to 85% B for the final 12 min. Blank runs were carried out randomly between samples to confirm that there was no chromatographic carryover. Structural analysis of lipid metabolites by RPLC-ESI-linear ion trap-MS ----------------------------------------------------------------------- For structural analysis of PL, SM and triacylglycerol (TAGs) species, BD samples were also analyzed using a Finnigan LTQ linear ion trap mass spectrometer (Thermo Fisher Scientific Inc. San Jose, CA, USA) interfaced with a Shimadzu Prominence HPLC system (Shimadzu, Kyoto, Japan). To analyze the fatty acid composition of PLs and SMs, data-dependent MS/MS (MS2) or MS3 analysis was performed in negative ion mode with an ion spray voltage of 4 kV and a scan range of *m*/*z* = 400--1000 as described previously \[[@B12]\]. For structural analysis of TAGs, data-dependent MS2 analysis was performed in the positive ion mode with an ion spray voltage of 4.5 kV and a scan range of *m*/*z* = 300--2000. The trap fill-time was 50 ms (full mass analysis) and 100 ms (MS^n^ analysis) in both ion modes. Nitrogen was used as a sheath gas (25 arbitrary units), and helium as a collision gas at a collision energy setting of 35%. BD samples were treated in the same way as RPLC-ESI-TOFMS analysis and applied onto an ACQUITY UPLC BEH C18 column (1.7 mm i.d. × 100 mm, 1.7 μm, Waters) kept at 45°C. LC conditions were the same as those in RPLC-ESI-TOFMS analysis. The classes of individual lipid species and their fatty acid compositions were determined with these data, together with RPLC-ESI-TOFMS data with exact mass and in-source fragmentation information. Targeted analysis of oxidative fatty acids by RPLC-ESI-triple quadrupole MS/MS ------------------------------------------------------------------------------ Oxidative fatty acids were measured as reported previously \[[@B13]\]. Briefly, MF fractions in solid extraction were analyzed by a 5500QTRAP quadrupole-linear ion trap hybrid mass spectrometer (AB Sciex, Framingham, MA, USA) interfaced with an ACQUITY UPLC System (Waters) equipped with an ACQUITY BEH C18 column (1.7 mm i.d. × 150 mm, 1.7 μm, waters), kept at 40°C. Samples were eluted with a mobile phase of water/acetate (100:0.1, v/v, solvent A') and acetonitrile/methanol (4:1, v/v, solvent B'). The gradient increased from 27 to 50% B' in five min, from 50 to 80% B' in the next 30 min, from 80 to 100% B' in the next five min, and kept at 100% B' for ten min with flow rates of 50 μL/min (0 -- 35 min), 50--100 μL/min (35 -- 40 min), and 100 μL/min (40 -- 50min). Analysis was performed with multiple reaction monitoring in the negative ion mode, with ion spray voltage, -4500 V; curtain (nitrogen), 10 arbitrary units; collision gas (nitrogen), 6 arbitrary units and a gas temperature of 350°C. We simultaneously measured 31 arachidonic acid metabolites, 14 eicosapentaenoic acid (EPA) metabolites, 13 DHA metabolites, 2 docosapentaenoic acid metabolites, and 2 linolenic acid metabolites. Declustering potential, collision energy and collision cell exit potential were optimized for each individual metabolite using authentic standard compounds (Cayman Chemical). In plasma samples from 15 month old control mice, one sample was omitted from the measurement due to instrumental problems (N=4). Data processing --------------- RPLC-ESI-TOFMS data were subsequently processed using the 2DICAL software package (2 Dimensional Image Converted Analysis of LCMS, Mitsui Knowledge Industry Co. Ltd., Tokyo, Japan) \[[@B14]\]. This software allowed detection and alignment of the ion-peak intensity derived from biomolecules identified with *m*/*z* and column retention times (RTs) in an amount of collected data. The main parameters of 2DICAL were set as follows: RT range 0.1--60.0 min and 0--37.5 min in the positive and negative ion modes, respectively, a mass range 400 to 1200 *m*/*z*, mass tolerance 0.3 *m*/*z*, mass window 0.3 *m*/*z*, an RT window of 0.60 min, and noise elimination level 2. The resulting three-dimensional matrix contains an arbitrarily assigned peak index (RT-*m*/*z* pairs) and ion intensity information (variables). The extracted ion peaks of each metabolite were normalized to that of the ISs. 16:0/16:0PC-d6 (Larodan Fine Chemicals) was used as an IS to normalize metabolites of PLs, lysoPLs, SMs, ceramides (Cers) and diacylglycerols (DAGs), which eluted at 0.1 to 37.5 min in RPLC. To normalize the TAG and ChE metabolites that eluted at 37.5 to 60 min, Tripalmitin-1,1,1-^13^C~3~ (Larodan Fine Chemicals) was used. Data for oxidative fatty acids were subsequently processed with Analyst™ (ver. 1.5.1, AB Sciex) software or MultiQuant™ Software (Version 2.1, AB Sciex). The integrated peak area of each metabolite was normalized to that of the IS (leukotriene B~4~-d4). Statistical and multiple classification analysis ------------------------------------------------ The IS-normalized peak list from RPLC-TOFMS analysis was first exported to the SIMCA-P+ for multivariate data analysis (Infocom Corp., Tokyo, Japan). Orthogonal partial least-squares-discriminant analysis (OPLS-DA) was used to select variables that strongly contributed to the difference between the two mice groups of the same ages. The Welch's *t*-test was then applied to assess statistical measures between the APP/tau and wild-type mice. For oxidative fatty acid analysis, Welch's *t*-tests were directly applied to the results with normalized peak areas. Results ======= Expression levels of Aβ~40~ and tau proteins in APP/tau mice ------------------------------------------------------------ We measured the protein expression levels of Aβ~40~ and tau in APP/tau mice by ELISA. In APP/tau mice, Aβ~40~ accumulated in the brain from 10 months of age (Additional file [1](#S1){ref-type="supplementary-material"}: Figure S1). On the other hand, soluble human tau protein was expressed at similar levels from 4 through 15 months of age. The levels of sarkosyl-insoluble tau appeared to be slightly higher at 10 and 15 months than at 4 months, but these differences did not reach statistical significance among all ages tested. In wild-type mice, both Aβ~40~ and human tau were under the detection limit at all tested ages. Analysis of brain tissues ------------------------- We first performed lipidomic analysis of BD samples extracted from brain tissues of APP/tau and wild-type mice at 4, 10, and 15 months of age using RPLC-ESI-TOFMS. When creating a two-dimensional map of RT versus *m/z* values of the individual precursor ion peaks in the positive ion mode, the intensities of a few spots that were apparently identified as ChEs were higher in APP/tau mice compared to wild-type mice at 4 months (Additional file [2](#S2){ref-type="supplementary-material"}: Figure S2). Likewise, at 10 and 15 months, these ChE spots retained the higher intensities in APP/tau mice compared to age-matched wild-type mice (data not shown). In the positive ion mode, a total of 803 peaks were detected and divided into two groups based on their retention time (RT). For RTs of 0.1--37.5 min, the 695 peaks contained mainly glycerophospholipids such as PLs, lysoPLs and sphingolipids such as SMs and Cers. For RTs of 37.5--60 min, the 108 peaks represented neutral lipids such as TAGs and ChEs. Peaks in each RT category were then separately processed by OPLS-DA (Additional file [3](#S3){ref-type="supplementary-material"}: Figure S3) because glycerophospholipids and neutral lipids are the two major classes of lipid metabolites with different physical properties. The OPLS-DA score plots for brain tissues indicated that APP/tau and wild-type mice could be reliably discriminated using established models at 4 (Additional file [3](#S3){ref-type="supplementary-material"}: Figure S3), 10 and 15 months (data not shown). To detect with high confidence the metabolites contributing to the discrimination between the two groups (APP/tau vs. wild-type) that were located at extreme ends of "S" of the OPLS-DA loading S-plot, we screened variables from this plot with threshold values of w\[[@B1]\] \> \|0.05\|, p(corr)\[[@B1]\] \> \|0.6\|, which represent the magnitude of contribution (weight) and reliability (correlation), respectively. For brain tissue from 4-month-old animals (pre-symptomatic phase), 9 and 12 metabolites in total (RT, 0.1-60 min) that exceeded this threshold value showed higher and lower levels in APP/tau compared to wild-type mice, respectively (Additional file [3](#S3){ref-type="supplementary-material"}: Figure S3). In addition, 8 and 8 metabolites at 10 months (early disease stage) and 5 and 58 metabolites at 15 months (late stage) had higher and lower levels in APP/tau compared to wild-type mice, respectively, and met the criteria (data not shown). The relative quantification data of the identified metabolites that meet the OPLS-DA loading S-plot criteria (unidentified peaks were omitted) were summarized in Additional file [4](#S4){ref-type="supplementary-material"}: Table S1. Of these metabolites, the levels of 2, 3 and 2 species were significantly increased and 4, 3 and 24 were significantly decreased in APP/tau mice compared to wild-type mice at 4, 10 and 15 months, respectively, by Welch's *t*-test (Table [1](#T1){ref-type="table"}). Particularly, docosahexaenoyl-ChE (22:6 ChE) and 38:2 pPE (p18:1/20:1) were consistently increased and decreased, respectively, across the tested ages in APP/tau mice compared to wild-type mice (Figure [1](#F1){ref-type="fig"}). Other ChE species such as 18:1ChE were under the detection limit (data not shown). Notably, many pPE species (e.g., 36:2pPE (p18:1/18:1), 36:4pPE (p16:0/20:4), 38:4pPE (p16:0/22:4), 38:4pPE (p18:0/20:4), 40:4pPE (p18:0/22:4), 40:6pPE (p18:0/22:6)) and SM species (e.g., 36:2SM (d18:2/N18:0), 38:1SM (d18:1/N20:0), 34:1SM (d18:1/N16:0)) were decreased at 15 months (Table [1](#T1){ref-type="table"}). Decreases in several phosphatidylcholines (PCs) and PEs mainly with polyunsatulated fatty acids (PUFAs) were also detected in the late symptomatic phase (Table [1](#T1){ref-type="table"}). As for triacylglycerol (TAG), the level of 60:12TAG (16:0/22:6/22:6) was increased (1.33-fold) in 10 month old animals, whereas most TAG levels were decreased in APP/tau mice compared to wild type mice at 15 months, and showed statistical significance for 52:3TAG (16:0/18:1/18:2, 16:1/18:1/18:1), 53:6TAG and 54:4TAG (16:0/18:0/20:4) (0.40, 0.54 and 0.31-fold, respectively) (Table [1](#T1){ref-type="table"} and Additional file [4](#S4){ref-type="supplementary-material"}: Table S1). ###### Fold changes in significantly changed lipids in brain tissues **Molecular species** **4 months** **10 months** **15 months** ------------------------------------------- -------------- --------------- --------------- ----------- ------ --------------- ***Phosphatidylcholine (PC)*** 30:0PC (14:0/16:0) 1.35 0.026\* 1.14 0.200 1.03 0.198 32:0PC (16:0/16:0) 1.02 0.815 0.87 0.093 0.83 0.011\* 34:2PC (16:0/18:2) 1.32 0.190 1.63 0.014\* 1.03 0.282 36:3PC (16:0/20:3) 1.06 0.550 1.07 0.441 1.28 0.029\* 38:1PC (18:0/20:1) 0.84 0.092 0.87 0.155 0.77 0.010\*\* 38:4PC (16:0/22:4) 1.07 0.462 0.92 0.279 0.86 0.011\* 38:5PC (18:1/20:4) 1.10 0.104 0.93 0.426 0.88 0.021\* 40:4PC (18:0/22:4) 1.05 0.596 0.91 0.188 0.84 0.003\*\* 40:7PC (18:1/22:6) 1.00 0.984 0.90 0.615 0.87 0.021\* 34:1ePC 1.04 0.542 0.92 0.338 0.86 0.029\* ***Phosphatidylethanolamine (PE)*** 38:1PE (20:0/18:1) 0.74 0.021\* 0.89 0.413 0.66 0.024\* 38:5PE (18:1/20:4) 0.99 0.942 0.94 0.352 0.85 0.023\* 40:4PE (18:0/22:4) 1.10 0.265 0.91 0.269 0.89 0.035\* ***Ethanolamine plasmalogen (pPE)*** 36:2pPE (p18:1/18:1) 0.76 0.299 0.88 0.183 0.86 0.042\* 36:4pPE (p16:0/20:4) 1.02 0.810 0.91 0.359 0.89 0.027\* 38:2pPE (p18:1/20:1) 0.78 0.037\* 0.83 0.037\* 0.76 0.034\* 38:4pPE (p16:0/22:4) 1.02 0.682 0.92 0.310 0.82 0.009\*\* 38:4pPE (p18:0/20:4) 0.94 0.819 0.93 0.796 0.87 0.029\* 40:4pPE (p18:0/22:4) 0.96 0.611 0.94 0.399 0.88 0.013\* 40:6pPE (p18:0/22:6) 1.02 0.822 0.89 0.182 0.87 0.043\* ***Sphingomyelin (SM)*** 34:1SM (d18:1/N16:0) 0.98 0.863 0.97 0.675 0.76 0.007\*\* 36:2SM (d18:2/N18:0) 0.94 0.335 0.81 0.010\* 0.86 0.027\* 38:1SM (d18:1/N20:0) 0.90 0.278 0.85 0.067 0.79 0.006\*\* 42:2SM (d18:1/N24:1) 0.78 0.048\* 0.86 0.388 0.60 0.066 ***Cerebroside*** 42:2cerebroside 0.81 0.045\* 0.92 0.334 0.84 0.137 ***Diacylglycerol (DAG)*** 38:4DAG 0.94 0.420 0.74 0.016\* 0.78 0.043\* ***Triacylglycerol (TAG)*** 52:3TAG (16:0/18:1/18:2, 16:1/18:1/18:1) 0.87 0.353 0.68 0.143 0.40 0.002\*\* 53:6TAG 1.05 0.765 0.91 0.648 0.54 0.003\*\* 54:4TAG (16:0/18:0/20:4) 1.09 0.463 1.04 0.801 0.31 \<0.001\*\*\* 60:12TAG (16:0/22:6/22:6) 0.99 0.935 1.33 0.011\* 0.87 0.544 ***Cholesterol ester (ChE)*** 22:6ChE 5.22 0.002\*\* 6.96 0.004\*\* 5.43 0.032\* ^1^ Fold change (ratio of APP/tau to wild-type) is the mean of N=3-5. ^2^ Welch's *t*-test (\**p*\<0.05, \*\**p*\<0.01,\*\*\**p*\<0.001). ![**Changes in 22:6ChE and 38:2pPE (p18:1/20:1) levels in brain tissue from APP/tau and wild-type mice.** The relative levels of metabolites are represented by the IS normalized peak heights using 16:0/16:0PC-d6 for 38:2pPE and tripalmitin-1,1,1-13C3 for 22:6ChE as ISs. These two metabolites were significantly changed in the brain across the tested ages. \**p*\< 0.05, \*\**p*\< 0.01 by Welch's *t*-test. Values are mean ± SD of N = 3--5.](1476-511X-12-68-1){#F1} Free PUFAs and their oxidized metabolites extracted from brain tissues were then measured using a multiple reaction monitoring system, and 16 species were detected (Additional file [5](#S5){ref-type="supplementary-material"}: Table S2). Of these, 9 species were significantly changed between APP/tau and wild-type mice as shown in Table [2](#T2){ref-type="table"}. The numbers of variable metabolites were 2, 7 and 0 species at 4, 10 and 15 months, respectively. No free PUFAs or their oxidized metabolites were consistently changed between APP/tau and wild-type mice across the ages tested (Additional file [5](#S5){ref-type="supplementary-material"}: Table S2). Pathway analysis of arachidonic acid metabolites is shown in Additional file [6](#S6){ref-type="supplementary-material"}: Figure S4. A marked decrease in prostaglandin D~2~ and 15-hydroxyeicosatetraenoic acid (HETE) was observed at 10 months (Figure [2](#F2){ref-type="fig"}). ###### Fold changes of the significantly changed oxidized fatty acids in brain tissues **Molecular species** **4 months** **10 months** **15 months** ------------------------------------------------------ -------------- ---------------- --------------- ---------------- ------ ------- ***Metabolites derived from arachidonic acid*** Prostaglandin D~2~ 0.99 0.941 0.46 \< 0.001\*\*\* 0.63 0.059 Thromboxane B~2~ 1.24 0.113 0.66 0.013\* 1.07 0.845 12-HHT 1.19 0.149 0.67 0.005\*\* 0.98 0.952 12-HETE 0.90 0.777 1.77 0.049\* 1.03 0.931 15-HETE 1.03 0.747 0.71 \<0.001\*\*\* 0.87 0.604 11,12-EpETrE 0.84 0.225 0.83 0.007\*\* 1.04 0.886 14,15-EpETrE 0.85 0.053 0.79 0.029\* 0.97 0.810 ***Metabolites derived from eicosapentaenoic acid*** 17,18-diHETE 2.12 \< 0.001\*\*\* 1.10 0.602 1.42 0.300 ***Metabolites derived from docosahexaenoic acid*** 19,20-diHDoPE 1.72 0.002\*\* 1.20 0.251 1.38 0.261 ^1^ Fold changes (ratio of APP/tau to wild-type) are means of N=3-5. ^2^ Welch's *t*-test (\**p*\<0.05, \*\**p*\<0.01,\*\*\**p*\<0.001). HHT, hydroxy-heptadecatrienoic acid; HETE, hydroxyeicosatetraenoic acid; EpETrE, epoxyeicosatrienoic acid; diHETE, dihydroxyeicosatetraenoic acid; diHDoPE, dihydroxydocosapentaenoic acid. ![**Changes of prostaglandin D**~**2**~**and 15-HETE levels in brain tissue from APP/tau and wild-type mice.** The relative levels of metabolites are represented by the IS normalized peak areas using leukotriene B~4~-d4 as an IS. \*\*\**p*\< 0.001 by Welch's *t*-test. Values are mean ± SD of N = 3--5.](1476-511X-12-68-2){#F2} Analysis of plasma ------------------ In the positive ion mode of BD samples from plasma, 385 and 308 peaks were detected on the basis of their RT, 0.1-37.5 min and 37.5-60 min, respectively. Next, we analyzed changes in lipid species levels in plasma by OPLS-DA loading S-plot (data not shown) in the same manner as for the brain tissues. At 4 months of age, 35 and 3 metabolites (RT, 0.1-60 min) with higher and lower levels in APP/tau mice than in wild-type mice, respectively, were detected. Forty-five and 9 metabolites displayed higher and lower levels in APP/tau mice compared to wild-type mice, respectively, at 10 months, and 32 and 22 were detected at 15 months. Relative quantification results of these metabolites are shown in Additional file [7](#S7){ref-type="supplementary-material"}: Table S3 (unidentified peaks were omitted) and the fold changes of metabolites with significantly different levels between APP/tau and wild-type mice are listed in Table [3](#T3){ref-type="table"}. The levels of 23 metabolites (22 were increased and 1 was decreased in APP/tau mice) at 4 months, 32 metabolites (26 and 6) at 10 months and 12 metabolites (7 and 5) at 15 months were significantly different between APP/tau and wild-type mice. As for PUFAs and their oxidized metabolites, the 34 species were detected (Additional file [8](#S8){ref-type="supplementary-material"}: Table S4) and, 7, 3 and 9 species levels were also significantly changed between the two mice groups at 4, 10, and 15 months, respectively (Table [4](#T4){ref-type="table"}). ###### Fold changes of the significantly changed lipids in plasma **Molecular species** **4 months** **10 months** **15 months** -------------------------------------------- -------------- --------------- --------------- --------------- ------ ----------- ***Phosphatidylcholine (PC)*** 16:1LPC 0.99 0.960 0.59 0.026\* 1.07 0.724 18:0LPC 1.50 0.019\* 1.27 0.281 1.01 0.921 20:5LPC ^3^ 1.45 0.011\* 0.82 0.440 1.26 0.190 22:6LPC 1.45 0.011\* 1.03 0.889 1.10 0.619 34:1PC (16:0/18:1) 1.24 0.037\* 1.00 0.983 1.27 0.071 34:2PC (16:0/18:2) 1.27 0.053 1.17 0.007\*\* 1.21 0.015\* 34:3PC (16:0/18:3) 1.06 0.698 0.65 0.034\* 0.98 0.936 35:2PC 1.68 0.030\* 1.55 0.013\* 1.27 0.035\* 36:1PC (18:0/18:1) 1.55 0.010\*\* 1.16 0.256 1.19 0.109 36:2PC (18:0/18:2) 1.48 0.024\* 1.35 0.017\* 1.11 0.139 36:5PC (16:0/20:5) 1.51 0.011\* 0.81 0.509 1.40 0.374 37:2PC (19:0/18:2) 1.33 0.144 1.47 0.008\*\* 1.26 0.071 38:3PC (18:0/20:3) 1.52 0.045\* 1.38 0.209 1.17 0.497 38:4PC (18:1/20:3) 0.99 0.957 0.63 0.031\* 0.82 0.137 38:5PC (18:0/20:5) 1.74 0.008\*\* 1.08 0.783 1.24 0.266 38:5PC (18:1/20:4) 1.13 0.331 0.79 0.018\* 0.84 0.149 38:7PC 1.08 0.644 0.48 0.019\* 0.94 0.807 40:4PC (20:0/20:4) 1.50 0.018\* 1.47 0.077 1.12 0.313 40:5PC (18:0/22:5) 1.71 0.007\*\* 1.25 0.182 1.30 0.349 40:6PC (18:0/22:6) 1.64 0.033\* 1.39 0.099 1.06 0.629 40:7PC (18:1/22:6) 1.06 0.642 0.72 0.022\* 0.73 0.072 42:6PC (20:0/22:6) 1.41 0.002\*\* 1.52 0.012\* 1.38 0.110 34:2ePC (16:0e/18:2) 1.52 \<0.001\*\*\* 1.81 \<0.001\*\*\* 1.47 0.011\* ***phosphatidylinositol (PI)*** 38:5PI (18:0/20:5) 1.53 0.024\* 0.92 0.748 2.00 0.002\*\* ***Sphingomyelin (SM)*** 34:1SM (d18:1/N16:0) 1.25 0.057 1.35 0.027\* 1.30 0.032\* 42:2SM 1.37 0.051 1.39 0.024\* 1.38 0.069 42:3SM (d18:2/N24:1) 1.48 0.030\* 1.22 0.061 1.30 0.141 ***Triacylglycerol (TAG)*** 52:3TAG (16:0/18:1/18:2) 1.21 0.557 1.21 0.558 0.58 0.045\* 52:4TAG (16:0/18:2/18:2) 1.57 0.250 1.41 0.421 0.40 0.032\* 52:5TAG (16:0/18:2/18:3) 1.04 0.903 0.74 0.314 0.49 0.011\* 54:3TAG (18:0/18:1/18:2) 0.64 0.553 0.62 0.411 0.48 0.043\* 54:4TAG (18:1/18:1/18:2) 0.93 0.919 0.97 0.950 0.37 0.038\* 56:6TAG (18:0/18:2/20:4) 2.00 0.209 2.75 0.016\* 0.82 0.500 58:10TAG (16:0/20:4/22:6) 1.40 0.311 2.07 0.020\* 0.88 0.509 58:11TAG 2.44 0.038\* 1.92 0.027\* 1.70 0.205 58:11TAG 1.87 0.030\* 1.69 0.176 1.64 0.383 58:12TAG (18:3/20:4/20:5) 2.06 0.032\* 1.91 0.168 2.65 0.005\*\* 60:10TAG 1.68 0.131 2.24 0.021\* 0.90 0.762 60:11TAG (16:2/22:4/22:5) 1.43 0.178 2.04 \<0.001\*\*\* 1.51 0.168 60:11TAG (16:0/22:5/22:6) 1.66 0.335 1.98 0.007\*\* 1.39 0.401 60:11TAG (18:2/20:3/22:6) 1.80 0.018\* 2.13 0.005\*\* 1.42 0.304 60:12TAG (18:2/20:5/22:5) 1.67 0.106 1.75 0.177 2.80 0.011\* 60:12TAG (18:3/20:4/22:5) 1.38 0.375 2.85 0.007\*\* 1.37 0.356 60:12TAG (18:3/20:4/22:5) 2.25 0.039\* 2.08 0.068 2.81 0.087 60:12TAG (18:3/20:4/22:5) 2.29 0.154 2.04 0.021\* 1.64 0.308 60:13TAG (18:2/20:5/22:6) 1.66 0.086 2.00 0.015\* 2.58 0.067 62:11TAG (18:2/22:4/22:5, 16:1/22:5/24:5) 1.64 0.063 1.77 0.031\* 1.58 0.232 62:12TAG (18:3/22:4/22:5) 1.56 0.147 2.27 0.003\*\* 1.35 0.370 62:13TAG (16:1/22:6/24:6, 20:2/20:5/22:6) 2.28 0.121 1.77 0.003\*\* 1.82 0.363 62:13TAG (18:1/22:6/22:6) 2.09 0.106 2.36 0.029\* 2.53 0.272 62:14TAG (18:2/22:6/22:6) 2.25 0.119 2.45 \<0.001\*\*\* 2.20 0.274 64:14TAG 1.57 0.215 2.06 0.008\*\* 1.95 0.251 65:14TAG 1.83 0.052 2.03 \<0.001\*\*\* 1.52 0.299 ***Cholesterol ester (ChE)*** 18:1ChE 0.74 0.016\* 0.78 0.089 1.12 0.423 18:2ChE 1.16 0.170 1.17 0.029\* 1.16 0.307 20:5ChE 1.38 0.041\* 0.93 0.762 1.33 0.138 ^1^ Fold changes (ratio of APP/tau to wild-type) are means of N=3-5. ^2^ Welch's *t*-test (\**p*\<0.05, \*\**p*\<0.01,\*\*\**p*\<0.001). ^3^ This specie was picked up not from OPLS-DA loading S-plot but manually. ###### Fold changes of the significantly changed oxidized fatty acids in plasma **Molecular species** **4 months** **10 months** **15 months** ------------------------------------------------------ -------------- --------------- --------------- --------- ------ ----------- ***Metabolites derived from arachidonic acid*** 17-HETE 1.59 0.018\* 1.41 0.112 1.36 0.111 18-HETE 1.46 0.021\* 1.48 0.105 1.67 0.011\* 5,15-diHETE 2.02 0.002\*\* 1.42 0.159 1.97 0.002\*\* ***Metabolites derived from eicosapentaenoic acid*** 8-HEPE 1.55 0.125 1.65 0.095 2.00 0.006\*\* 15-HEPE 1.20 0.314 1.47 0.197 1.51 0.020\* 17,18-EpETE 1.66 0.165 1.46 0.157 1.85 0.004\*\* 14,15-diHETE 2.16 0.032\* 1.80 0.051 2.21 0.122 17,18-diHETE 2.11 0.003\*\* 1.51 0.116 2.22 0.004\*\* ***Metabolites derived from docosahexaenoic acid*** 7-HDoHE 1.28 0.336 1.46 0.102 1.46 0.044\* 10-HDoHE 1.62 0.047\* 1.11 0.550 1.52 0.252 19,20-diHDoPE 1.67 0.010\* 1.42 0.027\* 1.80 0.033\* ***Metabolites derived from linolenic acid*** 9-HOTrE(alpha) 1.64 0.202 2.53 0.023\* 2.08 0.108 13-HOTrE(alpha) 1.50 0.147 1.87 0.041\* 1.73 0.008\*\* ^1^ Fold changes (ratio of APP/tau to wild-type) are means of N=3-5. ^2^ Welch's *t*-test (\**p*\<0.05, \*\**p*\<0.01,\*\*\**p*\<0.001). HETE, hydroxyeicosatetraenoic acid; diHETE, dihydroxyeicosatetraenoic acid; HEPE, hydroxyeicosapentaenoic acid; EpETE, epoxyeicosatetraenoic acid; HDoHE, hydroxydocosahexaenoic acid; diHDoPE, dihydroxydocosapentaenoic acid; HOTrE, hydroxyoctadecatrienoic acid . Concomitant changes in brain tissues and plasma ----------------------------------------------- Finally, we assessed whether the significant changes in lipid species were consistent for brain tissues and plasma. We found that 34:2PC (16:0/18:2) was concomitantly increased in brains (1.63-fold, *p* = 0.014) and plasma (1.17-fold, *p* = 0.007) of APP/tau mice compared to wild-type mice at 10 months (Tables [1](#T1){ref-type="table"} and [3](#T3){ref-type="table"}). Furthermore, 52:3TAG was decreased both in brain (0.4-fold, *p*=0.002) and plasma (0.58-fold, *p*=0.045) at 15 months (Tables [1](#T1){ref-type="table"} and [3](#T3){ref-type="table"}). The levels of 19,20-dihydroxy-docosapentaenoic acid (19,20-diHDoPE) and 17,18-dihydroxy-eicosatetraenoic acid (17,18-diHETE) in APP/tau mice at 4 months were significantly higher both in brain (1.72-fold, *p* = 0.002 for 19,20-diHDoPE and 2.12-fold, *p* \< 0.001 for 17,18-diHETE) and plasma (1.67-fold, *p* = 0.010 for 19,20-diHDoPE and 2.11-fold, *p* = 0.003 for 17,18-diHETE) than those from wild-type mice (Figure [3](#F3){ref-type="fig"}). ![**Consistent changes in the levels of two ω-3 fatty acid metabolites in brain tissue and plasma.** The relative levels of 19,20-diHDoPE and 17,18-diHETE in brain tissue (A) and plasma (B) are represented by the IS normalized peak areas using leukotriene B~4~-d4 as an IS. \**p* \< 0.05, \*\**p*\< 0.01, \*\*\**p* \< 0.001 by Welch's *t*-test. Values are mean ± SD of N = 3--5.](1476-511X-12-68-3){#F3} Discussion ========== In the present study we performed a lipidomic analysis of brain tissue and plasma from APP/tau mice and corresponding control (wild-type) mice at both presymptomatic and post-symptomatic phases using RPLC-ESI-TOFMS, RPLC-ESI-linear ion trap-MS, and RPLC-ESI-triple quadrupole MS/MS. AD brains display several characteristic pathological features, including accumulation of amyloid plaques composed of Aβ and neurofibrillary tangles composed of hyperphosphorylated tau protein. APP/tau mice are suitable animal models to study lipid dysregulation over very early to late stages of AD because these animals are destined to develop the disease based on currently proposed pathogenesis mechanisms. In brain tissue from our model mice, 22:6ChE levels were significantly increased in APP/tau mice compared to controls at all stages examined. In agreement with our results, 22:6ChE was reported to be increased in forebrains of mutated APP/presenilin-1 mice (11-fold, p \< 0.01) and mutated APP mice (3-fold, borderline significance), but not of presenilin-1 mice (0.9-fold) at 9--11.5 months of ages, compared with wild-type mice \[[@B9]\]. In postmortem entorhinal cortex of AD patients, increased levels of ChE (1.7-fold, p = 0.027) were also observed, although the predominantly increased species was different between AD humans (16:1ChE, 16:0ChE and 18:1ChE) and mice with mutated APP (22:6ChE) \[[@B9]\]. The elevation in ChE levels in our mouse model had already occurred at 4 months of age (Figure [1](#F1){ref-type="fig"}) when Aβ accumulation was not yet observed (Additional file [1](#S1){ref-type="supplementary-material"}: Figure S1), which supports the idea that intracellular ChE levels regulate Aβ generation \[[@B15]\]. Other notable lipid dysregulation observed in brains of APP/tau mice was the reduced levels of several pPE species, particularly at the late stage of AD (15 months). In brains from AD patients, plasmalogen levels were reported to be decreased \[[@B16],[@B17]\], suggesting that the mechanisms for this reduction could be common between humans and APP/tau mice. Marked increases in Aβ at 15 months of age might promote the formation of reactive oxygen species, which oxidize pPE, resulting in reduced pPE levels. It was reported that pPE decreases γ-secretase activities \[[@B18]\], and that Aβ destabilized the rate-limiting enzyme of plasmalogen synthesis, alkyl-dihydroxyacetonephosphate synthase, through peroxisomal dysfunction \[[@B19]\]. The levels of 38:2pPE (p18:1/20:1) were consistently decreased across the tested ages in APP/tau mice, although the pathophysiological significance of this lipid species in AD progression awaits evaluation. The elevated levels of 12-HETE observed in APP/tau mice at early stages of AD (10 months, Table [2](#T2){ref-type="table"}) were also consistent with a previous report showing that brains of AD patients exhibited higher levels of 12-HETE than control subjects \[[@B20]\]. As suggested in human cases, 12-HETE may be linked to oxidative stress and neurodegeneration in AD as well as being involved in learning and memory processes \[[@B21]\]. The decreased levels of SM species with medium-chain fatty acids at late stages of AD were also in agreement with a previous study using human postmortem AD brains in which medium-length SMs such as d18:0/N20:0 and d18:1/N20:0 were shown to be slightly down-regulated (*p* \< 0.05) \[[@B9]\]. These decreases might reflect demyelination as in the case of human AD brains described by Svennerholm and Gottfries \[[@B22]\]. Aβ42 was reported to accelerate SM degradation by activating neutral sphingomyelinase \[[@B23]\]. Increased levels of 60:12TAG bearing two DHAs (16:0/22:6/22:6, 1.33-fold) in 10 month old animals were in agreement with the previous report in which the levels of 56:7TAG containing docosahexaenoic acid (22:6FA) were increased in entorhinal cortices from AD patients. Since DHA-conjugated ChE was also increased in brains from APP/Tau mice at all ages tested (Figure [1](#F1){ref-type="fig"}), the AD brain appears to accumulate DHA in both TAG and ChE forms during the early symptomatic phase. These lipids might be involved in the production of DHA-derived messenger molecules such as neuroprotectin D1, which are reported to play neuroprotective roles in AD brain \[[@B8]\]. On the other hand, most APP/tau mice TAG levels were decreased compared with wild type mice at 15 months, particularly 52:3TAG (16:0/18:1/18:2, 16:1/18:1/18:1), 53:6TAG and 54:4TAG (16:0/18:0/20:4). A decrease in AA containing TAG (58:7TAG) was also observed in the prefrontal cortices of AD patients and in forebrains from 9--11.5 month old presenilin 1-APP mice compared to control animals \[[@B9]\]. Hydrolysis of these TAGs might thus be enhanced for their conversion into various structural and signaling lipid molecules, including eicosanoids, DAGs, monoacylglycerols (MAG) and phospholipids. In addition to the changes described above, significant decreases in PCs and PEs mainly with PUFAs were also detected during the late symptomatic phase. These phenomena might reflect fatty acid chain oxidation caused by the increased oxidative stress derived from increased Aβ peptide levels \[[@B24]\]. Several arachidonic acid metabolites in the brain also showed different levels when comparing APP/tau mice with their controls. Levels of PGD~2~, which mediates neural damage by Aβ~42~\[[@B25]\], were lowered at 10 months (Figure [2](#F2){ref-type="fig"}), reflecting a possible protective reaction in the brain. Inflammation is characteristic of AD brain tissues as aggregated Aβ and phosphorylated tau proteins are associated with increased levels of inflammatory cytokines such as interleukin-6 and tumor necrosis factor-α \[[@B26]\]. While levels of some cyclooxygenase (COX)-mediated metabolites such as TXB~2~ and 12-HHT were also decreased, others (PGE~2~, 6-keto PGF~1α~ and PGF~2α~) were not (Additional file [6](#S6){ref-type="supplementary-material"}: Figure S4). Overall, these results suggest that not all COX pathways but instead several downstream pathways (such as PGD~2~ synthesis) might be impaired in AD brains during the early symptomatic phase. 15-HETE is known to be a ligand for peroxisome proliferator-activated receptor γ (PPARγ) \[[@B27]\] and might have anti-inflammatory activity in the AD brain. In this context, decreases in 15-HETE levels in the early phase of AD might promote lowered resistance to AD progression. We also observed decreases in both 11,12-epoxy-5,8,11-eicosatrienoic acid (11,12-EpETrE) and 14,15-EpETrE during the early symptomatic periods (Table [2](#T2){ref-type="table"}). The roles of these decreases in AD progression should be evaluated in the future. Because metabolites that are consistently changed in both brain tissues and plasma from APP/tau mice compared to controls could be candidate diagnostic biomarkers for AD, we examined changes in lipid species levels in plasma. In both brain and plasma, 34:2PC (16:0/18:2) and 52:3TAG were consistently increased and decreased, respectively, in APP/tau mice compared with wild-type mice only after disease onset (10 or 15 months). Regarding oxidative fatty acid metabolites, 19,20-diHDoPE and 17,18-diHETE were consistently increased in both brain tissues and plasma from APP/tau mice before symptoms appeared at 4 months (Tables [2](#T2){ref-type="table"} and [4](#T4){ref-type="table"}, Figure [3](#F3){ref-type="fig"}), and these increases were essentially maintained at 10 and 15 months of age in plasma but not the brain. 19,20-diHDoPE and 17,18-diHETE are stable metabolites of DHA and EPA, respectively, that are mediated by cytochrome P450 (CYP) and soluble epoxide hydrolase (EH). In mouse brain, CYP2C29, CYP2C37, CYP2C38, and CYP2C40 were identified as arachidonic acid epoxygenases \[[@B28]\] and appear to be responsible for EPA and DHA epoxidation. Furthermore, soluble EH was reported to be expressed in various regions of mouse brain, including the cerebral cortex and hippocampus \[[@B29]\]. Levels of the intermediate metabolites, 19,20-epoxy-docosapentaenoic acid (19,20-EpDPE) and 17,18-epoxy-eicosatetraenoic acid (17,18-EpETE), which are generated by P450 epoxygenase in neuronal and glial cells, were below the quantification limit in brain (data not shown) and thus appeared to be quickly metabolized into their stable metabolites, 19,20-diHDoPE and 17,18-diHETE, respectively. Although the physiological functions of these metabolites in the central nervous system are unknown, 19,20-EpDPE and 17,18-EpETE were shown to act as vasodilators in cerebral arteries \[[@B30]\]. Since increased blood pressure is a risk factor for AD \[[@B31]\], these metabolites might counteract the vasoconstrictive effects of Aβ. Furthermore, 17,18-EpETE was reported to have anti-inflammatory properties mediated through PPARγ in lung tissues \[[@B32]\]. Therefore, 19,20-EpDPE and 17,18-EpETE in the brain might play a protective role against the onset of AD, but they are rapidly metabolized into their less active metabolites, 19,20-DiHDoPE and 17,18-diHETE. In this context, the expression of microsomal EH was reported to be significantly elevated in the hippocampus and associated cortex in AD patients \[[@B33]\]. However, the exact mechanisms leading to the consistent increases in these two compounds in brain and plasma remain unknown. During the pre-symptomatic phase, docosahexaenoyl lysoPC (22:6LPC) and eicosapentaenoy-lysoPC (20:5LPC) in plasma were significantly higher in APP/tau mice than in control mice (Table [3](#T3){ref-type="table"}). LysoPC may represent a preferred physiological carrier of fatty acids to allow passage across the blood--brain barrier \[[@B34]\], and thus brain DHA and EPA might be at least partly supplemented by plasma 22:6LPC and 20:5LPC. The contribution of the lysoPC species to supplementation of DHA and EPA in the brain needs to be investigated. A previous report showed that serum pPE levels were higher in Alzheimer's disease patients than in age-matched controls \[[@B35]\] and increased Cer and decreased SM levels in plasma were observed in patients with early-stage AD \[[@B36]\]. However, these changes were not observed in our APP/tau mice (Table [3](#T3){ref-type="table"}). Conclusions =========== We performed a lipidomic analysis using brain tissue and plasma obtained from APP/tau mice at three AD stages. Two lipid species, 38:2pPE (p18:1/20:1) and 22:6ChE, were consistently decreased and increased, respectively, across the tested ages in APP/tau mice compared to wild-type mice. As reported for human AD brains, brain tissue from APP/tau mice showed increased 12-HETE levels at the early symptomatic phase and decreased levels of both pPEs and SMs with medium chain fatty acids at the late symptomatic phase compared with control mice. In addition to these previously reported changes, 19,20-diHDoPE and 17,18-diHETE, the stable metabolites of 19,20-EpDPE and 17,18-EpETE that have a potential counteracting role for AD, were increased in APP/tau brain tissue at the pre-symptomatic phase, and concomitant increases in plasma were observed. Furthermore, 15-HETE, which has a potential protective role in AD, was decreased during the early symptomatic phase in APP/tau mice. Our results provide fundamental information on the further precise analysis of lipids in AD pathogenesis. These changes detected in APP/tau mice models of AD should be examined in patients in the near future. Abbreviations ============= AD: Alzheimer's disease; Aβ: Amyloid β; APP: Amyloid precursor protein; BD: Bligh & Dyer; Cer: Ceramide; ChE: Cholesterol ester; CYP: Cytochrome P450; DAG: Diacylglycerol; DHA: Docosahexaenoic acid; 19,20-diHDoPE: 19,20-dihydroxy-4Z,7Z,10Z,13Z,16Z-docosapentaenoic acid; 17,18-diHETE: 17,18-dihydroxy-5Z,8Z,11Z,14Z-eicosatetraenoic acid; EH: Epoxide hydrolase; ELISA: Enzyme-linked immunosorbent assay; 19,20-EpDPE: 19,20-epoxy-4Z,7Z,10Z,13Z,16Z-docosapentaenoic acid; 17,18-EpETE: 17,18-epoxy-5Z,8Z,11Z,14Z-eicosatetraenoic acid; EPA: Eicosapentaenoic acid; ESI: Electrospray ionization; GM3: Ceramide-lactose-N-acetylneuraminic acid; HETE: Hydroxyeicosatetraenoic acid; MAG: Monoacylglycerols; PC: Phosphatidylcholine; PE: Phosphatidylethanolamine; pPE: Ethanolamine plasmalogen; PUFA: Polyunsatulated fatty acid; MF: Methyl formate; m/z: Mass-to-charge ratio; OPLS-DA: Orthogonal partial least-squares-discriminant analysis; PL: Phospholipid; PPAR: Proliferator-activated receptor; RPLC: Reversed-phase liquid chromatography; RT: Retention time; SM: Sphingomyelin; TAG: Triacylglycerol; TOFMS: Time of flight mass spectrometry. Competing interests =================== The authors declare that they have no competing interests. Authors' contributions ====================== Conception and design: MK, TR, NS, TO, SY. Execution of the experiments: TY, IM, MK, MM, SY, OA, MR. Data analysis and interpretation: MK, NMT, NH, IK, AM, TR, NS, TO, SY. Manuscript writing: TY, MK, NS, TO, SY. All authors have read and approved the manuscript and agree with submission to your journal. Supplementary Material ====================== ###### Additional file 1: Figure S1 Expression levels of Aβ~40~ and tau proteins in APP/tau mice. The levels of human Aβ~40~, and soluble (TBS-soluble plus sarkosyl-soluble) and sarkosyl-insoluble APP/tau mice at 4, 10, and 15 months of age were determined by ELISA as described in Materials and Methods. Values are mean ± SD of N = 3--5. ###### Click here for file ###### Additional file 2: Figure S2 RPLC-ESI-TOFMS analysis of brain lipids from APP/tau and wild-type mice at 4 months of age. Total ion counts chromatograph (TIC) and two-dimensional map with retention time (RT) versus mass-charge (*m*/*z*) values of brain lipids measured by RPLC-ESI-TOFMS in the positive ion mode from wild-type (A) and APP/tau mice (B) at 4 months. The intensity of peaks is represented by color density spots. Lipid metabolites were eluted in the following order: lysophospholipids (Lyso PLs, LPLs) \> phospholipids (PLs) = sphingomyelins (SMs) = ceramides (Cers) \> triacylglycerols (TAGs) = cholesterol esters (ChEs), as indicated in the areas surrounded by the ellipses. Spots for ChEs were visible in APP/tau (B) but not wild-type mice (A). ###### Click here for file ###### Additional file 3: Figure S3 Multivariate statistical analysis of brain lipids between APP/tau and wild-type mice at 4 months of age. OPLS-DA score plots and loading S-plots for brain tissues from APP/tau vs. wild-type mice at 4 months, derived from the RPLC-ESI-TOFMS data set (A: 0.1-37.5min in RT; B: 37.5-60min in RT). In score plots (wild-type is shown in open circle vs. APP/tau in closed circle), the goodness-of-fit parameter R2 and the predictive ability parameter Q2 were 1.000 and 0.840, respectively, for (A) and 0.897 and 0.619, respectively, for (B). Loading S-plot showed covariance w against correlation p (corr) of variables for discriminating components of OPLS-DA model. Cut-off values for w\[1\] \> \|0.05\| and p (corr) \> \|0.6\| were used to select metabolites that strongly contributed to the discrimination between APP/tau and wild-type mice, which are surrounded by the red dotted line. ###### Click here for file ###### Additional file 4: Table S1 Relative quantification of lipid species meeting the OPLS-DA loading S-plot criteria in brain tissues. ###### Click here for file ###### Additional file 5: Table S2 Relative quantification of the detected oxidized fatty acids in brain tissues). ###### Click here for file ###### Additional file 6: Figure S4 Pathway analysis of arachidonic acid metabolites. HHT, hydroxy-heptadecatrienoic acid; HETE, hydroxyeicosatetraenoic acid; EpETrE, epoxyeicosatrienoic acid; TX, Thromboxane; COX, cyclooxygenase; LOX, lipoxygenase; CYP, cytochrome P450; PGFS, prostamide/prostaglandin F synthase; PTGIS, Prostacyclin synthase; PTGES, prostaglandin E synthase; PTGDS, prostaglandin D synthase; TBXAS, thromboxane-A synthase. ###### Click here for file ###### Additional file 7: Table S3 Relative quantification of lipid species meeting the OPLS-DA loading S-plot criteria in plasma. ###### Click here for file ###### Additional file 8: Table S4 Relative quantification of detected oxidized fatty acids in plasma. ###### Click here for file Acknowledgment ============== This work was supported by the Advanced Research for Products Mining Programme \[grant number 10--45\] from the National Institute of Biomedical Innovation of Japan.

{ "pile_set_name": "PubMed Central" }

1. Introduction {#sec1-molecules-23-02894} =============== Osimertinib (AZD9291, Merelitinib, Tagriiso^©^), *N*-(2-{2dimethylaminoethyl-methylamino}-4-methoxy-5-{\[4-(1-methylindol-3-yl)pyrimidin-2yl\]amino}phenyl)prop-2-enamide ([Figure 1](#molecules-23-02894-f001){ref-type="fig"}), a third-generation, highly selective, irreversible covalent inhibitor has been created by AstraZeneca for the clinical therapy of advanced non-small cell lung cancer (NSCLC) \[[@B1-molecules-23-02894],[@B2-molecules-23-02894],[@B3-molecules-23-02894],[@B4-molecules-23-02894],[@B5-molecules-23-02894],[@B6-molecules-23-02894]\]. NSCLC patients, who have epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor (TKI) resistance, are mostly subject to a mutation of EGFR. Therefore, this has been the active target for the osimertinib \[[@B7-molecules-23-02894],[@B8-molecules-23-02894],[@B9-molecules-23-02894]\]. In addition, the tablet formulation of osimertinib has been approved by the FDA (Food and Drug Administration of the USA) for NSCLC patients, who have progressed to or completed EGFR TKI therapy in 2015. To the best of our knowledge, several papers have established the methods for the determination of osimertinib in biological samples, and the utilized apparatuses are all ultra-performance liquid chromatography coupled with tandem mass spectrometry (UPLC-MS/MS) \[[@B10-molecules-23-02894],[@B11-molecules-23-02894],[@B12-molecules-23-02894],[@B13-molecules-23-02894]\]. There are some advantages of UPLC-MS/MS, including its high sensitivity, high stability, and short analytic time. Unlike UPLC-MS/MS, UPLC-TOF-MS has specific advantages such as its high working efficiency, wide measurable mass range, and high ratio of resolution \[[@B14-molecules-23-02894],[@B15-molecules-23-02894],[@B16-molecules-23-02894],[@B17-molecules-23-02894]\]. In addition, the capability of simultaneous quantitative analysis and qualitative analysis greatly benefits the analysts, and it is very useful for the further study of agents, such as their metabolism, enzymology, transportation and so on \[[@B18-molecules-23-02894]\]. However, to date, methods using UPLC-TOF-MS for the determination of osimertinib have not been reported. In the present study, we are the first to quantify osimertinib in rat plasma using UPLC-TOF-MS. The objective of this study was to investigate a specific, sensitive, rapid and reliable UPLC-TOF-MS method for osimertinib quantification in rat plasma samples. Meanwhile, we have successfully investigated the pharmacokinetic study of osimertinib in rats using this UPLC-TOF-MS method. 2. Results and Discussion {#sec2-molecules-23-02894} ========================= 2.1. UPLC-TOF-MS Method Development {#sec2dot1-molecules-23-02894} ----------------------------------- By using the product scan mode, we could find the method of pyrolysis of osimertinib and ion of sorafinib (IS) in rat plasma under the UPLC-TOF-MS condition that we have optimized (see [Section 2.3](#sec2dot3-molecules-23-02894){ref-type="sec"}) ([Figure 2](#molecules-23-02894-f002){ref-type="fig"}). The parent ion m/z of osimertinib was 500.2768 and the characteristic product ion was 72.0810. In addition, the m/z of the parent ion and product ion of sorafenib (IS) were 465.0953 and 270.0882, respectively. All these results are consistent with previous studies \[[@B4-molecules-23-02894],[@B5-molecules-23-02894],[@B6-molecules-23-02894],[@B7-molecules-23-02894],[@B19-molecules-23-02894]\]. However, the optimization of the UPLC condition became the key point in the process of investigating our UPLC-TOF-MS method. We used various proportions and gradients of water (containing 0.1% formic acid)--acetonitrile and water (containing 0.1% formic-ammonia formate)--acetonitrile to optimize the chromatographic conditions. However, the chromatographic peaks all displayed the trailing phenomenon ([Figure S1](#app1-molecules-23-02894){ref-type="app"}). Since the pKa of osimertinib was 13.64, it needs a basic environment to keep its molecular state. Therefore, we decided to use 0.1% ammonia water--acetonitrile as the mobile phase, which enabled us to finally obtain the symmetrical chromatographic peak ([Figure 3](#molecules-23-02894-f003){ref-type="fig"}). 2.2. Method Validation {#sec2dot2-molecules-23-02894} ---------------------- ### 2.2.1. Specificity and Selectivity {#sec2dot2dot1-molecules-23-02894} The typical chromatograms of blank plasma, the plasma sample spiked with osimertinib and IS and rat plasma after treatment are shown in [Figure 3](#molecules-23-02894-f003){ref-type="fig"}. [Figure 3](#molecules-23-02894-f003){ref-type="fig"}A--C displays the characteristic peaks of standard osimertinib in rat plasma, the pharmacokinetic plasma sample and blank plasma, and the chromatograms showed a good specificity. In addition, [Figure 3](#molecules-23-02894-f003){ref-type="fig"}D--F exhibits the good specificity of standard sorafenib (IS). Moreover, these results also showed that there was no significant chromatographic interference with the osimertinib and IS in rat plasma. ### 2.2.2. Calibration and Lower Limit of Quantification (LLOQ) {#sec2dot2dot2-molecules-23-02894} To investigate the linearity of osimertinib and IS, nine calibration concentrations were analyzed in each validation batch, and the results showed the good linearity (*r*^2^ \> 0.99) of the calibration curve of osimertinib and IS over the range of 1 to 500 ng/mL. The lower limits of quantification (LLOQ) of osimertinib and IS were both 1 ng/mL, and the ratios of signal-to-noise were considerably higher than 5. In addition, the LLOQ of this UPLC-TOF-MS method was sufficient for the determination of the osimertinib pharmacokinetic study. ### 2.2.3. Precision and Accuracy {#sec2dot2dot3-molecules-23-02894} The intra- and inter-day precision and accuracy are shown in [Table 1](#molecules-23-02894-t001){ref-type="table"}. In addition, the intra- and inter-day results of the HQC, MQC and LQC are investigated with the RSD ranging from 5.38--9.76% (intra-day) and 6.02--9.46% (inter-day), respectively. The results ranged within the standard acceptance limit of 15% and demonstrated the good accuracy and precision of osimertinib. ### 2.2.4. Extraction Recovery and Matrix Effect {#sec2dot2dot4-molecules-23-02894} [Table 2](#molecules-23-02894-t002){ref-type="table"} shows the extraction recovery of osimertinib and IS, and these results are sufficient for quantification. The matrix effects of osimertinib and IS range from 0.810 to 0.926 and from 0.798 to 0.934, respectively ([Table 3](#molecules-23-02894-t003){ref-type="table"}). The calibration curves of the final concentrations of IS were the same as those of osimertinib. The results showed the high extraction recovery and lack of significant matrix effect of this method for osimertinib and IS in the rat plasma. ### 2.2.5. Stability {#sec2dot2dot5-molecules-23-02894} [Table 4](#molecules-23-02894-t004){ref-type="table"} shows the stability of HQC, MQC and LQC of osimertinib under four different storage conditions. According to the results, osimertinib was found to have good stability at room temperature (25 °C) and the autosampler temperature (4 °C) for 24 h and remained stable following three freeze (−80 °C) and thaw (0 °C) cycles. Moreover, the plasma samples of osimertinib were also stable at the storage temperature (−80 °C) for at least 30 days. 2.3. Pharmacokinetic Application {#sec2dot3-molecules-23-02894} -------------------------------- After successfully establishing the analysis method of osimertinib by UPLC-TOF-MS, we applied this method into the pharmacokinetic study in SD rats. The dosage of the oral administration was 4.5 mg/kg and the mean plasma time-concentration of osimertinib in seven rats is shown in [Figure 4](#molecules-23-02894-f004){ref-type="fig"}. This showed that the osimertinib had the highest plasma concentration at 4.5 h after oral administration and the C~max~ was 28.49 ng/mL. In addition, [Table 5](#molecules-23-02894-t005){ref-type="table"} shows the following non-compartmental parameters of osimertinib in SD rats: a terminal half-life of (14.96 ± 3.44) h, a distribution volume of (233.82 ± 66.68) L/kg, and a clearance of about (10.84 ± 1.94) L/h/kg ([Table 5](#molecules-23-02894-t005){ref-type="table"}). These pharmacokinetic data of osimertinib can provide more information regarding its application in clinical treatment. 3. Materials and Methods {#sec3-molecules-23-02894} ======================== 3.1. Drugs and Materials {#sec3dot1-molecules-23-02894} ------------------------ Osimertinib (purity \>99%) and sorafenib (internal standard, IS) were purchased from Stanford Analytical Chemicals Inc. (Eugene, OR, USA). Dimethyl sulfoxide (DMSO) was purchased from Beijing Solarbio Science and Technology Co., Ltd. (Beijing, China). Ammonia and acetonitrile (HPLC grade) were provided by Tianjin Kemiou Chemical Reagent Co., Ltd. (Tianjin, China) and the Fisher Scientific Co., Ltd. (Pittsburgh, PA, USA), respectively. Ultrapure water was obtained from a milli-Q reagent water purification system (Millipore, Bedford, MA, USA). 3.2. Apparatus {#sec3dot2-molecules-23-02894} -------------- The UPLC-TOF-MS/MS method was performed on a system that includes a Qtrap 5600-TOF mass spectrometer (AB Sciex, MA, USA) and an UPLC chromatographic analysis system (Shimadzu, Kyoto, Japan). An Xterra MS C~18~ column (100 × 2.1 mm, 3.5 μm) (Waters Corp., Milford, MA, USA) was used for the analytical separation at the temperature of 40 °C. A TGL-16M high speed centrifuge was purchased from Cence Co., Ltd. (Changsha, China). 3.3. Solution Preparation {#sec3dot3-molecules-23-02894} ------------------------- DMSO was used to dissolve the accurately weighed standard of osimertinib and sorafenib (IS) to obtain the stock solutions of 1.0 mg/mL. Then, the working solutions of osimertinib were diluted serially with 50% acetonitrile in water to achieve 10, 20, 50, 100, 200, 500, 1000, 2000 and 5000 ng/mL. Next, 10 μL diluted solutions were diluted in 100 μL blank plasma to obtain the final calibration standard samples, and the range of the final concentrations of the calibration standards was from 1 to 500 ng/mL. The working solution concentration of IS was 500 ng/mL, which was dissolved using 50% acetonitrile water (*v*/*v*). The quality controls (QCs) were diluted to achieve the lower limit of the quantification (10 ng/mL, LLOQ), low (20 ng/mL, LQC), medium (200 ng/mL, MQC), and high (4000 ng/mL, HQC) concentrations of osimertinib. After that step, 10 μL of the QC solutions were dissolved by 100 μL of blank plasma to get the final concentrations of LLOQ (1 ng/mL), LQC (2 ng/mL), MQC (20 ng/mL), HQC (400 ng/mL) and 50 ng/mL of IS. All samples and working solutions were kept at −20 °C before use. 3.4. UPLC-TOF-MS Condition {#sec3dot4-molecules-23-02894} -------------------------- The chromatographic separation was performed using a C~18~ column, and its temperature was kept at 40 °C. The chromatographic separation consisted of a 0.1% ammonia (A) and acetonitrile (B) mixture and a 0.4 mL/min flow rate was maintained. The gradient ran linearly from 10% to 95% between 0 and 1.5 min, and then the mobile phase was kept at 95% for 5.0 min. 3.0 μL of samples was injected into the analysis system and the total analytical time of one sample was 5.1 min. The temperature of the autosampler was maintained at 4 °C. The TOF-MS spectrometer was set up in the positive ion full scan electrospray and high sensitivity mode with an m/z range from 100 to 1000 Da, and the accumulation time was set as 0.25s. The parameters of TOF-MS were as follows: nebulizer gas (gas 1), 55 psi; heater gas (gas 2), 55 psi; curtain gas, 35 psi; ion spray voltage, 5500 V; turbo spray temperature, 550 °C; declustering potential (DP), 100 V; and collision energy (CE), 35 eV. The conditions of the information-dependent data acquisition (IDA) criteria were as follows. The eight most intense fragment ions of each analyte in 100 cps were chosen as the product ions, and the m/z of the product ions ranged from 600 to 1300 over a 0.08 s accumulation time. In addition, the CE and collision energy scope (CES) of the product ions scan were set at 35 eV and 15 eV, respectively. 3.5. Pharmacokinetic Application {#sec3dot5-molecules-23-02894} -------------------------------- The pharmacokinetic study of osimertinib was applied to seven male SD rats using oral administration by gavage and the dosage was 4.5 mg/kg. Osimertinib was diluted and suspended by 0.5% sodium carboxymethylcellulose. No less than 0.3 mL rat plasma was sampled at 0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 24, 36, and 48 h after oral administration through the oculi chorioideae vein under the condition of light ether anesthesia. After 10 min centrifuging of all analytes at 5000× *g*, the supernatant was collected and frozen at −40 °C until analysis. The use of animals in the presented study was permitted by the Ethics Committee of the Hebei Medical University, and all animal studies were carried out according to the Guidance for the Care and Use of Laboratory Animals of the US National Institute of Health. 3.6. Sample Preparation {#sec3dot6-molecules-23-02894} ----------------------- Ten microliters of IS solution (500 ng/mL) was prepared in 100 μL rat plasma and vortex-mixed for 20 s. After that step, 500 μL of acetonitrile was added to precipitate the protein. Then, the mixture was vortexed for 1 min and centrifuged at 12,000 rpm for 10 min. The supernatant was transferred into a new Eppendorf tube and evaporated to dryness through nitrogen gas at 45 °C. After 100 μL of acetonitrile was added to reconstitute the residue and vortexed for 1 min, all samples were centrifuged at 12,000× *g* for 10 min. Finally, 80 μL of the supernatant was collected and injected into UPLC sample vials before use. 3.7. Method Validation {#sec3dot7-molecules-23-02894} ---------------------- According to the US Food and Drug Administration (FDA) guidelines regarding the bioanalytical method's validation \[[@B20-molecules-23-02894]\], the method validation was investigated and established, including the specificity and selectivity, the linearity and sensitivity, the recovery, the stability, and the precision, accuracy and matrix effect. ### 3.7.1. Selectivity and Specificity {#sec3dot7dot1-molecules-23-02894} The selectivity and specificity of the developed method was assessed by comparing the chromatography of the blank plasma and blank plasma spiked with the targets. ### 3.7.2. Linearity and Sensitivity {#sec3dot7dot2-molecules-23-02894} A series of calibration analytes from 1 to 500 ng/mL consisted of the calibration curve of osimertinib. In addition, the method to determine the linearity of calibration curve was the peak area ratios of osimertinib and IS, and these ratios were used to get a least-squares weighted regression (the weighting factor was 1/y, and y = peak area ratio of osimertinib/IS). The correlation coefficient (*r*^2^) of all calibration curves, which were desirable for this method, were better than or equal to 0.99. LLOQ was used to investigate the method sensitivity and follows these two criteria: (1) the comparison of the LLOQ and blank response should occur at least 5 times; and (2) the analyte peak of the LLOQ should be discrete, reproducible and identifiable, and its accuracy and precision should be at least 20%. ### 3.7.3. Precision, Accuracy and Matrix Effect {#sec3dot7dot3-molecules-23-02894} The intra-day accuracy and precision were investigated through the determination of six QC analytes of the high (HQC = 400 ng/mL), medium (MQC = 20 ng/mL), low (LQC = 2 ng/mL), and LLOQ (LLOQ = 1 ng/mL) concentrations. The inter-day accuracy and precision were conducted through the determination of the six replicates of the four levels of QCs using the same preparation on 3 separate days. The assay accuracy of the QC samples was compared to the corresponding standard calibration concentration. The precision of the replicates was evaluated by the RSD (relative standard deviation). It can be accepted that the mean values of accuracy should not exceed 15% at the HQC and MQC, and LQC concentrations and LLOQ should not exceed 20%. Similarly, the relative standard deviation of the precision for HQC, MQC, and LQC concentration levels should not exceed 15% and the limitation of LLOQ was 20%. The matrix effect was determined by dividing slopes of calibration curves of osimertinib in the rat blood matrix and mobile phase. ### 3.7.4. Recovery {#sec3dot7dot4-molecules-23-02894} The extraction recovery of osimertinib was evaluated by comparing the peak area ratios of standard solution samples and the same concentrations rat plasma samples through five replicates at HQC, MQC and LQC concentrations. ### 3.7.5. Stability {#sec3dot7dot5-molecules-23-02894} The stabilities of this method, which include the freeze--thaw stability, autosampler stability, short-term stability and long-term stability, were evaluated through three QC samples after the sample preparation method. The freeze (−80 °C)--thaw (room temperature) stability was conducted under the conditions of three free--thaw cycles. The autosampler stability of the plasma samples was investigated by the extracted QC samples that were kept in an autosampler (4 °C) for 12 h. In addition, the long-term stability was evaluated through the determination of three QC samples that were kept at −80 °C for 30 days. All samples were considered stable with RSDs \< ±15%. 3.8. Data Analysis {#sec3dot8-molecules-23-02894} ------------------ Dynamic background subtraction is a novel technique performed using the Analyst software (AB Sciex, Foster City, CA, USA). In addition, the data of the pharmacokinetic study of osimertinib were collected and calculated by the DAS 2.1.1 software in the non-compartmental mode (Mathematical Pharmacology Professional Committee of China, Shanghai, China). 4. Conclusions {#sec4-molecules-23-02894} ============== A sensitive UPLC-TOF-MS method for the determination of osimertinib has been established in this research. The method exhibited excellent precision, recovery and sensitivity. The results indicated that UPLC-TOF-MS could serve as a highly interesting analytical alternative for bioanalysis. This work was supported by the Liaoning Planning Program of Philosophy and Social Science (NO. L17BGL034), the Research Program on the Reform of Undergraduate Teaching in General Higher Schools in Liaoning (NO. 2016-346), the Key Program of the Natural Science Foundation of Liaoning Province of China (NO. 20170541027) and the Medical Education Scientific Research on the "12th Five-Year Plan" (Fifth Batch of Subjects) funded by China Medical University (NO. YDJK2015005). **Sample Availability:** Samples of the compounds are available from the authors. The following are available online. ###### Click here for additional data file. Conceived and designed the experiments: Y.Z., Z.-J.D. and S.-T.D. Performed the experiments: S.-T.D., Y.L., H.-T.Y., Y.W., Y.-J.L., C.-Y.D., L.M. and Z.-J.D. Analyzed the data: S.-T.D. Wrote the paper: Y.Z. and S.-T.D. All authors read and approved the final manuscript. The authors declared no conflict of interest. ![Structures of osimertinib (molecular weight = 499.619 Da) and sorafenib (molecular weight = 464.825 Da).](molecules-23-02894-g001){#molecules-23-02894-f001} ![Product ion scan of osimertinib 500.2768 → 72.0810 (**A**), and IS 465.0953 → 270.0882 (**B**).](molecules-23-02894-g002){#molecules-23-02894-f002} ###### Typical chromatograms of (**A**) standard osimertinib (20 ng/mL) in rat plasma, (**B**) pharmacokinetic plasma sample, (**C**) blank plasma, (**D**) standard ion of sorafenib (IS) (500 ng/mL) in rat plasma, (**E**) pharmacokinetic plasma sample, and (**F**) blank plasma. ![](molecules-23-02894-g003a) ![](molecules-23-02894-g003b) ![Plasma concentration-time profile after single oral administration of osimertinib (4.5 mg/kg) to rats. Data are expressed as the mean ± SD (*n* = 7).](molecules-23-02894-g004){#molecules-23-02894-f004} molecules-23-02894-t001_Table 1 ###### Intra- and inter-day precision and accuracy of osimertinib in rat plasma. RSD: relative standard deviation. Concentration (ng/mL) Intra-Day (n = 7) Inter-Day (n = 7) ----------------------- ------------------- ------------------- -------- ---------------- ------ -------- 400 395.12 ± 21.27 5.38 98.78 401.13 ± 24.15 6.02 100.28 20 21.08 ± 1.42 6.74 105.40 19.34 ± 1.83 9.46 96.70 2 2.05 ± 0.20 9.76 102.50 1.98 ± 0.17 8.59 99.00 molecules-23-02894-t002_Table 2 ###### Extraction recovery and matrix effect of osimertinib and IS in rat plasma. Analyte Concentration (ng/mL) Extraction Recovery (%) ------------- ----------------------- ------------------------- ------ Osimertinib 400 95.24 ± 3.01 3.16 20 96.14 ± 1.83 1.90 2 84.31 ± 3.18 3.77 Sorafinib 500 87.22 ± 4.23 4.85 molecules-23-02894-t003_Table 3 ###### The slope ratio of the solvent linear equation and the matrix linear equation. Analyte Calibration Curve R^2^ R~slope~ ------------- --------------------- -------- ---------- ------- ------- Osimertinib Y = 0.085X + 0.1102 0.9997 0.810 0.926 0.116 Sorafinib Y= 0.0781X + 0.2314 0.9996 0.798 0.934 0.136 R~slope~ = Slope of matrix standard calibration curve/slope of mobile phase standard calibration curve. molecules-23-02894-t004_Table 4 ###### Stability of osimertinib in rat plasma under various storage conditions. Storage Condition Concentration (ng/L) Mean ± SD RSD% ----------------------------------------------------- ---------------------- ------------- ------- Autosampler (4 °C) temperature for 24 h 2 2.12 ± 0.23 10.85 20 21.45 ± 1.81 8.44 400 406.81 ± 5.64 1.39 Room temperature (25 °C) for 24 h 2 2.21 ± 0.26 3.66 20 22.45 ± 2.18 9.71 400 407.28 ± 5.12 1.26 Storage temperature (−80 °C) for 30 days 2 2.29 ± 0.25 10.92 20 21.33 ± 1.74 8.16 400 406.34 ± 7.51 1.85 Three freeze--thaw cycles (each at −80 °C for 24 h) 2 2.27 ± 0.19 8.37 20 22.20 ± 1.92 8.65 400 406.17 ± 6.19 1.52 molecules-23-02894-t005_Table 5 ###### Pharmacokinetic parameters of osimertinib after oral administration of 4.5 mg/kg to rats. Pharmacokinetic Parameter Osimertinib --------------------------- ---------------- *AUC*~(0--t)~, ng/mL·h 382.00 ± 69.00 *AUC*~(0--∞)~, ng/mL·h 426.01 ± 81.73 MRT, h 14.51 ± 1.91 t~1/2z~, h 14.96 ± 3.44 t~max~, h 4.80 ± 1.10 C~max~, ng/mL 28.49 ± 3.97 V~z~/F, L/kg 233.82 ± 66.68 CL~z~/F, L/h/kg 10.84 ± 1.94 Data are expressed as the mean ± SD (*n* = 7).

{ "pile_set_name": "PubMed Central" }

1. Introduction {#sec1} =============== For any nation, understanding the epidemiology and clinical status of young people with diabetes is essential for training health professionals and also for conducting advocacy with the respective government to plan and improve the appropriate clinical services. Very little published material is available on diabetes in youth in Bolivia. The International Diabetes Federation (IDF) Atlas estimates of children \< 15 years with type 1 diabetes in Bolivia \[[@B1]\] are extrapolated from the data from Peru that is over two decades old \[[@B2], [@B3]\], a time during which the incidence of type 1 diabetes has been increasing at 2-3% or more per annum in many countries \[[@B1]\]. In 2005, the IDF Life for a Child (LFAC) Program \[[@B4]\], with the assistance of Rotary International, commenced support for the Centro Vivir con Diabetes (CVcD), a multidisciplinary clinic in Cochabamba. Satellite networks were established in five other cities. LFAC Program commenced professional mentoring and provision of insulin (initially with support of Insulin for Life Australia \[[@B5]\]). Later, blood glucose meters and test strips (three tests/day), HbA1c testing, and educational materials were provided. This study presents epidemiological data on all young people with diabetes known to CVcD from January 2005 to February 2017 and the clinical features of the cohort receiving care at the end of 2015 and documents the initial impact of the LFAC Program. 2. Materials and Methods {#sec2} ======================== The study was approved by the Board of Centro Vivir con Diabetes. 2.1. Epidemiology Data {#sec2.1} ---------------------- Data (age, gender, and diabetes type) were compiled on all people diagnosed with diabetes at age \< 25 years (y) and seen by CVcD and its five satellite clinics from January 2005 to February 2017. Diabetes was diagnosed according to the standard World Health Organization (WHO) criteria \[[@B6]\], and the diabetes type was determined by the local investigators based on clinical features. It is not possible to determine the exact national incidence as other diabetes care providers exist in Bolivia and no secondary ascertainment sources are available. However, in Cochabamba Province---comprising Cochabamba city and its immediate environs---where 63 (43.8%) of cases were living at diagnosis, CVcD is well-known and the ascertainment rate is estimated at 80%. A minimum incidence for Cercado Province was calculated according to census data \[[@B7]\], with population interpolated or extrapolated according to the annual growth rate of 1.6% \[[@B8]\]. The crude mortality rate was calculated as the total number of deaths in the cohort of 144 subjects divided by the total number of years from diagnosis until the last follow-up (February 2017) or, if they had died, date of death. Results are expressed as mortality per 1000 patient years. 2.2. Clinical Status {#sec2.2} -------------------- Clinical information was collated on all 141 subjects being followed from September to December 2015, who aged \<25 y at diabetes diagnosis and ≤35 y at the time of this assessment. These 141 subjects included some cases diagnosed before January 2005 and excluded some cases included in the incidence study who were now being seen by other centres. The data were collected via LFAC Annual Clinical Data Sheets. Data included sex, date of birth, diabetes diagnosis date, details of diabetes treatment regimen, and physical and biochemistry measures. Also, social parameters as to whether diabetes was limiting school attendance, if subjects were in the age-appropriate grade, and how well overall the young person was, in their doctor\'s opinion, psychologically coping with their diabetes (rated as good, with some problems, or poor) were also recorded. Body weight and height were measured by electronic scales and a stadiometer, respectively, with subjects wearing light-weight clothing and without shoes. Body mass index (BMI) was then calculated. BMI SD scores were calculated using the WHO standards for \<5 years \[[@B9]\], \>5 years, and \<19 years \[[@B10]\]. Blood pressure SD score was calculated from published data from the United States \[[@B11]\]. The presence of cataracts, retinopathy, and peripheral neuropathy was recorded. However, the methodology was not prescribed and therefore could have varied across centres and physicians. 2.3. Biochemistry {#sec2.3} ----------------- Glycosylated haemoglobin (HbA1c), serum creatinine, total cholesterol, triglycerides, and HDL cholesterol were measured in local biochemistry laboratories, with the exception of HbA1c at Cochabamba where it was measured with a Siemens DCA Vantage machine (Erlangen, Germany). Lipid tests were done on fasting samples. LDL cholesterol was calculated using the Friedewald equation and hence is only available on the subset of patients who had HDL cholesterol levels \[[@B12]\]. Estimated glomerular filtration rate (eGFR) was calculated for subjects ≥ 14 y by the CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) formula \[[@B13]\], adjusted for individual surface area. Initial impact of Life for a Child Program was measured by assessing all HbA1c levels of all subjects having data at baseline (program commencement in 2004) and then at either six- or 12-month follow-up. 2.4. Statistics {#sec2.4} --------------- Data and descriptive statistics were managed in Excel and analysed in R 3.3.1 (R Core Team, Vienna, Austria) with RStudio integrated development environment (RStudio Team, Boston, USA). 3. Results {#sec3} ========== 3.1. Epidemiological Data {#sec3.1} ------------------------- ### 3.1.1. Sex Distribution and Age of Onset {#sec3.1.1} Of the 144 subjects diagnosed at age \< 25 y between January 2005 and February 2017, 62 (43.1%) were males. Age of onset ranged from 0.3--22.2 y, with a peak at 11 years (see [Figure 1](#fig1){ref-type="fig"} histogram). Sixteen subjects (11.1%) were diagnosed at age 0--4 y, 42 (29.2%) were diagnosed at age 5--9 y, 62 (43.1%) were diagnosed at age 10--14 y, 19 (13.2%) were diagnosed at age 15--19 y, and 5 (3.5%) were diagnosed at age 20--24 y. Genetic studies are being arranged for the child diagnosed at 3 months to check for a monogenic form of diabetes. This child has no other medical problems. ### 3.1.2. Minimum Incidence {#sec3.1.2} Sixty-three cases of diabetes (diagnosed at \<25 years of age) occurred in the Cercado Province of Cochabamba Department in the 12.2 y study period from January 2005 to February 2017. Forty-nine were \<15 y at diagnosis and 14 were 15--24 y. This is a measured incidence of 2.2 per 100,000 children \< 15 y/y. Ascertainment is estimated as 80%, resulting in an estimated total incidence of 2.7 per 100,000 children \< 15 y/y. ### 3.1.3. Mortality Rate {#sec3.1.3} Of the 144 subjects followed up for 0.1--12.2 y (mean ± SD 5.9 ± 3.4 y), two had died, a male aged 22.8 y from gastric ulcer haemorrhage and a female aged 13.4 y from ketoacidosis (who had been abandoned by her family). The crude mortality rate was 2.3/1000 patient years. 3.2. Clinical Status {#sec3.2} -------------------- ### 3.2.1. Demographics {#sec3.2.1} One hundred and forty-one T1D subjects diagnosed at \<25 years of age and ≤35 y were being followed up by CvCD at the end of 2015. Sixty-four (45%) patients were male. Sixty (43%) were from Cochabamba Department; 31 (22%), Santa Cruz; 27 (19%), La Paz; 12 (8.5%), Sucre; 6 (4.3%), Tarija; and 5 (3.5%), Potosi. One child also had cerebral palsy, one had Down syndrome, and one child also had epilepsy. Age at diagnosis and diabetes duration are shown in [Table 1](#tab1){ref-type="table"}. ### 3.2.2. Diabetes Care {#sec3.2.2} Ninety-four subjects (67%) were adjusting their insulin dose (data available on 141 subjects), with the remainder giving a fixed insulin dose set in agreement with their treating clinicians. Regarding insulin types (*n* = 138), 99 (72%) were on short- and long-acting human insulin only. Twenty-five (18%) were using analogue insulin, of which nine (7%) were using analogue only. Thirteen (9%) were on premixed insulin, and one patient was receiving long-acting human insulin only. For the number of insulin injections per day (*n* = 122), 18 subjects (158%) were taking two injections per day; 70 (57%), three injections; 24 (20%), four injections; and the remaining 10 (8%), five injections. Insulin usage ranged 0.35--1.76 units/kg/day (mean ± SD 0.99 ± 0.29) (*n* = 136). The frequency of self-blood glucose tests per week (*n* = 139) was 14--28 (mean ± SD 21.3 ± 1.4). The frequency of clinic visits in the last year (*n* = 131) was 1--8 (mean ± SD 3.5 ± 1.3). ### 3.2.3. Social Parameters {#sec3.2.3} School/college attendance was regular in 93 (96%) of the 97 patients for whom this information was available, and the year of schooling was age-appropriate in 94%. Out of 105 patients for whom a response was recorded on the question of how well they thought their patients were coping with their diabetes, 62 (59%) of the responses were "good," 38 (37%) were with "some problems," and four (4%) were "poor." ### 3.2.4. Physical Measurements {#sec3.2.4} Height and weight information was available for 137 patients, and 81 of these were under 19 years, permitting BMI SD score calculation (shown in [Table 1](#tab1){ref-type="table"}). Blood pressure was measured on 21 subjects \< 20 years, with systolic and diastolic SD scores shown in [Table 1](#tab1){ref-type="table"}. Both systolic and diastolic mean SD scores were above zero (*p* \< 0.001 for both, by one-sided *t*-test). None of the patients for whom SD scores were available were noted to have hypertension (as defined by local clinical standards) or be on antihypertensive medications. ### 3.2.5. HbA1c {#sec3.2.5} HbA1c was measured in the 135 subjects with duration of diabetes \> 6 months. Mean and median HbA1c levels were 8.5% (69 mmol/mol) and 8.2% (66 mmol/mol), respectively (see also [Table 1](#tab1){ref-type="table"}). Forty-four subjects (33%) had HbA1c levels in the recommended target range of \<7.5% (58 mmol/mol); 35 (25%), 7.5%--8.49% (58--69 mmol/mol); 48 (36%), 8.5%--12% (69--108 mmol/mol); and the remaining eight subjects (5.8%), \>12% (108 mmol/mol). Twenty-six percent of those aged \<15 y and 30% of those aged 15--24 y had target HbA1c \< 7.5% (58 mmol/mol). A series of one-way ANOVAs on log-transformed HbA1c levels were performed, finding trends towards relationships between HbA1c levels, the number of insulin injections per day (*p* = 0.07), and the use of analogue insulin (*p* = 0.06), but not for gender (*p* = 0.38). A pattern of increased HbA1c levels and variance during adolescence was observed from a plot of HbA1c by age ([Figure 2](#fig2){ref-type="fig"}). For those aged 0--13 y, the HbA1c mean ± SD was 7.9% (63 mmol/mol) ± 1.3% (14 mmol/mol); for those aged 13--20 y, the HbA1c mean ± SD was 9.4% (78 mmol/mol) ± 2.0% and for those aged \>20 y, the HbA1c mean ± SD was 7.9% (63 mmol/mol) ± 1.7% (18 mmol/mol). This indicates a significant difference in HbA1c levels between age groups (*p* = 0.0002 by one-way ANOVA on log-transformed values). However, Levene\'s test for heterogeneity of variance was not statistically significant (*p* = 0.24). Mean/median HbA1c levels by a diabetes clinic site were 9.0%/8.4% (75/68 mmol/mol) in Cochabamba, 7.9%/7.6% (63/60 mmol/mol) in La Paz, 8.0%/7.4% (64/57 mmol/mol) in Potosi, 8.5%/8.4% (69/68 mmol/mol) in Santa Cruz, 8.5%/8.3% (69/67 mmol/mol) in Sucre, and 7.1%/6.8% (54/51 mmol/mol) in Tarija, with no statistical difference between clinics on ANOVA. ### 3.2.6. Lipids {#sec3.2.6} Lipid levels are shown in [Table 1](#tab1){ref-type="table"}. No subjects were on lipid-lowering agents. For total cholesterol, 97 patients (80%) had \<200 mg/dL (5.2 mmol/L), 16 patients (13%) had 200--239 mg/dL (5.2--6.2 mmol/L), and nine (7.3%) had ≥240 mg/dL (≥6.2 mmol/L). Total cholesterol levels were not correlated with concurrent HbA1c levels (*p* = 0.14 by Pearson\'s correlation). Triglyceride levels (*n* = 121) were \<1.7 mmol/L (150 mg/dL) in 99 patients (82%), 1.7--2.2 mmol/L (150--199 mg/dL) in 11 (9%) patients, and 2.3--5.6 mmol/L (200--499 mg/dL) in 11 (9%) patients (all subjects had \<4.5 mmol/L (450 mg/dL). Log-transformed triglyceride levels were correlated with concurrent HbA1c levels (*p* = 0.03 by Pearson\'s correlation). HDL cholesterol was measured in 27 patients from the Department of La Paz (see [Table 1](#tab1){ref-type="table"}). None of these patients were on renal replacement therapy. Fifteen patients (56%) had HDL \< 1.0 mmol/L (40 mg/dL) for men or 1.3 mmol/L (50 mg/dL) for women. There were no significant correlations between HDL cholesterol and either HbA1c or log-transformed triglyceride levels. Seven patients (26%) had calculated LDL cholesterol levels \< 100 mg/dL (\<2.6 mmol/L), 10 patients (37%) had 100--129 mg/dL (2.6--3.3 mmol/L), and 10 patients (37%) had ≥130 mg/dL (3.3 mmol/L). [Table 2](#tab2){ref-type="table"} shows the percentage of subjects meeting target levels for cardiovascular risk factors (BMI, triglycerides, total cholesterol, and HbA1c), for the 115 subjects that had values for all four measurements. ### 3.2.7. Renal Function {#sec3.2.7} Serum creatinine values are shown in [Table 1](#tab1){ref-type="table"}. Five patients (4%) had serum creatinine levels above the recommended level of 1.3 mg/dL (115 *μ*mol/L) for males or 1.1 mg/dL (97 *μ*mol/L) for females. eGFR for the 115 subjects ≥ 14 y ranged from 10.4--178.9 mL/min/1.73m^2^ (mean ± SD 104.9 ± 31.0 mL/min/1.73m^2^). Three patients on renal replacement therapy (see below) were excluded from this analysis. Two subjects were receiving haemodialysis (age 28.5 and 28.8 y) (duration of diabetes to start of haemodialysis 16.5 and 1.25 y), one had received a renal transplant after 0.3 y on haemodialysis (age 30.4 y, duration of diabetes to time of haemodialysis 17.3 y) and the other was noted as having chronic renal failure (24.4 y, duration 20.5 y, eGFR 12.8 mL/min/1.73m^2^). Three other subjects had eGFR \< 50 mL/min/1.73m^2^, all with diabetes duration of at least 16 years. Two of these three were receiving treatment for hypertension, and one was also treated with erythropoietin injections for associated anaemia. Seventeen subjects ≥ 14 y had an eGFR of ≥140 mL/min/1.73m^2^, suggestive of hyperfiltration. ### 3.2.8. Other Complications {#sec3.2.8} One subject had hypothyroidism which was being treated. One subject had experienced a severe hypoglycemia episode in the past resulting in severe cognitive and motor impairment. Three patients were recorded as having diabetic retinopathy, and five subjects had cataracts (two of whom also had retinopathy). Neuropathy (loss of vibration and light touch) was recorded in one subject. 3.3. Initial Impact of Life for a Child Program {#sec3.3} ----------------------------------------------- HbA1c results at the start of the LFAC Program in 2004-2005 were available for 57 subjects. Values ranged from 4.9--15% (30--140 mmol/mol), with mean ± SD 9.6 ± 2.2% (81 ± 24 mmol/mol). For six-month paired data, mean ± SD HbA1c at baseline was 9.4 ± 2.3% (79 ± 25 mmol/mol), and mean ± SD HbA1c at six months was 8.1 ± 1.8% (65 ± 20 mmol/mol) (*n* = 29, *p* = 0.006 by one-sided paired *t*-test). For 12-month paired data, mean ± SD HbA1c at baseline was 9.7 ± 2.1% (83 ± 23 mmol/mol), and mean ± SD HbA1c at twelve months was 8.7 ± 2.5% (72 ± 27 mmol/mol) (*n* = 36, *p* = 0.007 by one-sided paired *t*-test) (see [Figure 3](#fig3){ref-type="fig"}). 4. Discussion {#sec4} ============= Globally, the incidence rates of type 1 diabetes in children vary widely due to differences in genetic susceptibility, combined with inadequately understood but powerful and evolving environmental factors \[[@B1]--[@B3]\]. There have been no previous studies on type 1 diabetes incidence in Bolivia. Rates vary substantially in Central/South American and other Hispanic populations. Estimates for Bolivia in the IDF Atlas \[[@B1]\] are extrapolated from a 1990--1991 study in Lima, Peru, which found an incidence of 0.5 per 100,000 children \< 15 y/y \[[@B2]\]. Other studies in South America in the early 1990s (all \<15 y) showed rates per 100,000 ranging from 0.1 in Caracas, Venezuela, to 8.3 in Montevideo \[[@B2], [@B3]\]. Countries with both recent and older data are seeing sharp rises in type 1 diabetes incidence---in Mexico, from 3.4 (per 100,000) in 2000 to 6.2 in 2012 for young people \< 20 y in the Instituto Mexicano del Seguro Social (IMSS) health insurance service \[[@B14]\]; in Chile, from 5.7 in 2006 to 12.1 in 2012 for those aged \<20 y \[[@B15]\], with the sharpest increase in the \<5 y age group; and in Brazil, a trend for increased incidence in those aged \<15 y from 1986 to 2006 \[[@B16]\]. Further afield for Hispanic populations, rates (\<15 y) are 20.6 in Spain (\<15 y, 1995--2001) \[[@B1]\], 13.2 in Portugal (\<15 y, 1994--1998) \[[@B1]\], and 14.1 in Hispanic Whites in the USA (\<15 y, 2002--2003) \[[@B17]\], with a recent study showing an increase of 4.2% per year in 2003--2012 on numbers \< 20 y \[[@B18]\]. Some of the variance in South American populations is thought to be due to lower rates of type 1 diabetes in Central/South American indigenous peoples as compared to those of Hispanic White/mixed descent \[[@B2], [@B19]\]. Larenas et al. \[[@B20]\] found 3.8 times higher rates in Caucasian Chileans than in Mapuche (native Chileans). Bolivia has, by a substantial margin, the highest proportion of indigenous peoples in Central/South America \[[@B21]\]. This lower rate in indigenous populations is likely largely genetic in origin due to differences in HLA haplotypes \[[@B19], [@B22], [@B23]\]. However, differing exposure to environmental factors may also be potent influencers: rates of type 1 diabetes have increased in many countries as standards of living have increased. Incidence rates were found to be higher in high socioeconomic communes in Santiago de Chile \[[@B24]\]. The observed type 1 diabetes rate in Cercado in this study---2.2 per 100,000 children \< 15 y/y with an estimated ascertainment of 80%, giving an estimated incidence of 2.7 per 100,000---is five times higher than the figure previously used for Bolivia (based on the Peruvian data from 1990--1991 \[[@B2]\]. It is likely that type 1 incidence is increasing in Bolivia; however, further studies will be needed to confirm this. This is still a low incidence rate compared to global figures. The female preponderance seen in Bolivia is common in low incidence studies \[[@B25]\] and is believed to be due to fathers with type 1 susceptibility genes being more likely to survive and pass on diabetogenic genes than mothers, with fathers\' genes more likely to cause type 1 diabetes in female offspring and mothers\' genes to cause diabetes in male offspring. In addition, it is thought that fathers\' genes may be more diabetogenic than similar genes in mothers \[[@B26], [@B27]\]. Monogenic diabetes has been reported from various countries, and it is likely that the child diagnosed at 3 months of age has this form of diabetes---genetic testing is being arranged. If there is a gene defect, alternate noninsulin therapy may be possible depending on the specific gene \[[@B28]\]. Like various other less-resourced countries \[[@B29]\], the Bolivian government health service does not cover the cost of care for people with diabetes at any age. Most families cannot afford the cost of care, which can be prohibitive in such countries for families with lower incomes \[[@B30]\], and this can translate into the premature death of the child or young adult with diabetes, as occurred once in this Bolivian cohort. Furthermore, in Bolivia, there are very few paediatric endocrinologists and, due to its relative rarity, limited knowledge of type 1 among general paediatricians. Therefore, the support from a centre such as CvCD ± LFAC is crucial for families with limited resources. CvCD provides a multidisciplinary service of not just health professionals in diabetology but also health professionals in ophthalmology, nutrition, psychology, social work, foot care and podiatry, physiotherapy and rehabilitation, laboratory services, and pharmacy. There is also a young group or people with type 1 diabetes which encourage and support each other through meetings and the use of the social media platform "WhatsApp." The IDF Life for a Child Program provides insulin and, since 2009, also provides blood glucose test strips, educational materials, and point-of-care HbA1c testing in Cochabamba. The initial impact seen in the LFAC follow-up data from 2004--2005 has been sustained. Mortality is low at 2.3 per 1000 patient years, and the mean/median HbA1c of 8.5/8.2% (69/62 mmol/mol) is surprisingly good, considering the substantial challenges faced in caring for these young people. Twenty-seven percent of those aged \<15 y and 28% of those aged 15--24 y were achieving the recommended HbA1c target of \<7.5% (58 mmol/mol). This is not substantially different from the mean of 29/30% (male/female) for subjects aged \<15 y and from the mean of 24/20% (male/female) for those aged 15--24 y in a large international study in which 15 of 16 countries were high-income nations \[[@B31]\]. The higher HbA1c values seen in adolescents/emerging adults are similar to those of other reports \[[@B32]\] and likely reflect lifestyle and behavioural factors and growth spurt-related insulin resistance. As in other countries, thoughtful medical attention and support are warranted during this period, along with provision of insulin supply security. It is of great concern that three young people with diabetes required renal replacement therapy and four others have substantial renal impairment (eGFR \<50 mL/1.73m^2^). Furthermore, 17 had hyperfiltration, which may be a risk factor for diabetic nephropathy, although recent evidence suggests this may not be the case \[[@B33], [@B34]\]. The presence of ESRD at a young age suggests that blood glucose control must have been poor for some years, and we note that for most of the Bolivian subjects evaluated herein, regular self-blood glucose monitoring has only been possible since 2009. Anderzén et al. \[[@B35]\] has shown the importance of a good early start in terms of blood glucose control, if later complications are to be reduced. Further assessment should be done for early signs of nephropathy such as microalbuminuria and hypertension, with management (with ACE inhibitors or LDL-lowering "statin" drugs) being informed by the forthcoming results of the Adolescent Type 1 Diabetes Cardio-Renal Intervention Trial (ADDIT) \[[@B36]\]. Given the number with renal complications, it is likely that a thorough systematic assessment would lead to more subjects being found with other microvascular complications such as neuropathy and retinopathy, for whom early intervention with risk factor control and increased screening rates is appropriate. The common cause of death of people with type 1 diabetes is cardiovascular disease, for which dyslipidaemia is a major risk factor and is treatable, with significant LDL-C, cardiovascular event, and mortality reduction with statins \[[@B37]\]. The substantial numbers with one or more lipid risk factors such as high triglycerides and total cholesterol and low HDL cholesterol levels, combined with the suggestion of blood pressure elevation, raise the possibility that some patients have features of the metabolic syndrome, which is recognised in type 1 diabetes and associated with higher rates of diabetes vascular complications \[[@B38], [@B39]\]. This "double diabetes" may benefit from the addition of metformin to insulin and risk factor control \[[@B40]\], though clinical trials with vascular end-points are still awaited. The study has three other limitations. It is possible that with further study (autoantibodies and C-peptide, unfortunately not available in this clinical context), some of the subjects diagnosed with type 1 diabetes may be recategorised as having type 2 diabetes. Secondly, HbA1c measurements were not standardised across clinics, and so it is possible that results could be affected by the different methods used. Finally, the 80% ascertainment is an estimated figure, and Cercado Province may not be fully representative of Bolivia---incidence may vary in other parts of the country due to ethnic or socioeconomic factors. 5. Conclusion {#sec5} ============= This study has estimated the incidence of type 1 diabetes in Bolivia, confirming that incidence is relatively low compared to that in many other countries. The results demonstrate that reasonable blood glucose control and low mortality can be achieved in many subjects in a low-resourced setting with international support. However, there is a significant risk of early long-term complications in some subjects, as evidenced by renal impairment and substantial levels of cardiovascular risk factors. The authors acknowledge the dedicated efforts of all at Centro Vivir con Diabetes and the determination of all these young people with diabetes and their families. They also thank the supporters of the IDF Life for a Child and Insulin for Life Programs and Rotary International (Rotary Club of Cochabamba, Quillacollo, District 6940, and the Rotary Foundation). Finally, they thank Sarah Garnett for helping with the SD scores. This work was supported by the Leona M. and Harry B. Helmsley Charitable Trust. Gabriel Andrew Gregory was partly funded by the University of Sydney Summer Research Scholarship. Alicia Josephine Jenkins is funded by the NHMRC (Australia) Practitioner Fellowship and the Sydney Medical School Foundation Fellowship. Ethical Approval ================ The study was reviewed and approved by the Board of Centro Vivir con Diabetes. Conflicts of Interest ===================== None of the authors have any conflicts of interest in regard to this study. Authors\' Contributions ======================= Elizabeth Duarte Gómez collected the data and contributed to the manuscript, Gabriel Andrew Gregory analysed the data and wrote the manuscript, Miriam Castrati Nostas collected the data and contributed to the manuscript, Angela Christine Middlehurst assisted in the data analysis and contributed to the manuscript, Alicia Josephine Jenkins helped with interpretation of the data and contributed to the manuscript, and Graham David Ogle designed the study and cowrote the manuscript. ![Age at diagnosis of type 1 diabetes in Bolivia, 2005--2017.](JDR2017-8454757.001){#fig1} ![Relationship of age to HbA1c.](JDR2017-8454757.002){#fig2} ![Initial impact at commencement of Life for a Child support.](JDR2017-8454757.003){#fig3} ###### Clinical and biochemical characteristics. Parameter *n* Mean (SD) Range ----------------------------------------- ----- ------------------ ------------------ Age at diagnosis (years) 141 9.9 (4.4) 0.3--22.2 Diabetes duration (years) 141 8.1 (5.7) 0.1--26.9 BMI SD score 137 0.9 (1.1) −2.2--3.7 Systolic SD score 21 0.7 (0.8) −1.4--2.0 Diastolic SD score 21 1.1 (0.7) −0.1--2.6 HbA1c (%)^∗^ 135 8.5 (1.9) % 5.4--14.0% HbA1c (mmol/mol)^∗^ 135 69 (21) mmol/mol 36--130 mmol/mol Serum creatinine (mg/dL) 130 0.92 (0.69) 0.5--6.2 Total cholesterol (mg/dL) 122 167 (43.2) 81--334 Triglycerides (mg/dL) 121 112 (63) 42--497 HDL cholesterol (mg/dL) 27 43 (14.2) 7.9--59 LDL cholesterol (mg/dL, calculated) 27 124 (38) 43--191 Non-HDL cholesterol (mg/dL, calculated) 27 149 (44) 58--216 ^∗^HbA1c not included on six subjects assessed within six months of diagnosis. ###### Risk factors for cardiovascular disease. Risk factor Definition *n* \% ------------------------------------ -------------------------------------------------------- ----- ------ High BMI For \<19 years, SD score ≥ 2.0; if ≥19 years, BMI ≥ 25 22 19.1 Elevated fasting triglycerides ≥150 mg/dL (≥1.7 mmol/L) 21 18.3 Elevated fasting total cholesterol ≥175 mg/dL (≥4.5 mmol/L) 45 39.1 HbA1c above target range ≥7.5% (58 mmol/mol) 74 64.3 One risk factor --- 44 38.3 Two risk factors --- 34 29.6 Three risk factors --- 14 12.2 Four risk factors --- 2 1.7 (*n* = 115). [^1]: Academic Editor: Konstantinos Papatheodorou

{ "pile_set_name": "PubMed Central" }

Introduction {#s1} ============ Cells secrete different types of small membrane vesicles. Exosomes are one of the vesicles with 30--100 nm diameter and physicochemically distinct from other secreted vesicles [@pone.0013247-Thry1]. Inside exosomes, cellular gene products including proteins, mRNAs, and microRNAs (miRNAs) are packaged and these molecules can be transferred to recipient cells to deliver their molecular signaling such as oncogenesis [@pone.0013247-Valadi1]--[@pone.0013247-AlNedawi1] and immune response [@pone.0013247-Thry1], [@pone.0013247-Davis1]. Exosomes are released by many types of cells [@pone.0013247-Thry2] which include tumor cells, epithelial cells, and hematopoietic cells such as reticulocytes, cytotoxic T lymphocytes, Epstein-Barr virus-transformed B cells, mastocytes, dendritic cells, and platelets. Secreted exosomes have been isolated and characterized *in vitro* from cultured cell lines [@pone.0013247-Simpson1] along with *in vivo* in body fluids [@pone.0013247-Simpson1] including blood [@pone.0013247-Rabinowits1], urine [@pone.0013247-Pisitkun1], saliva [@pone.0013247-Michael1], amniotic fluid [@pone.0013247-Keller1], and malignant pleural effusions [@pone.0013247-Andre1]. Since observed in many proliferating cell types, it is conceivable to exacerbate tumor cells, as evidenced by their increased presence in plasma and pleural effusions of patients with cancer [@pone.0013247-Rabinowits1], [@pone.0013247-Andre1]. This increased presence in non-invasive body fluids of cancer patients has accelerated to profile molecular components in the exosomes for discovering clinically useful tumor markers and biomarkers [@pone.0013247-Skog1], [@pone.0013247-Simpson1], [@pone.0013247-Taylor1]. miRNAs are a class of noncoding small RNAs that are involved in post-translational regulation of gene expression by inhibiting both stability and translation of mRNAs [@pone.0013247-Ghildiyal1]. Recent evidence has shown that miRNA mutations or misexpression correlate with various human cancers and indicate that some miRNAs can function as oncogenes or tumor suppressors [@pone.0013247-Ahmed1], [@pone.0013247-Croce1]. To analyze RNAs, it is always to consider their stability from degradation by RNase. Recent findings indicate that endogenous plasma miRNAs in blood samples are stably detectable in a form that is resistant to RNase activity [@pone.0013247-Mitchell1], evidenced by identification of miRNAs in body fluids such as blood [@pone.0013247-Mitchell1]--[@pone.0013247-Tanaka1], urine [@pone.0013247-Hanke1], and saliva [@pone.0013247-Michael1], [@pone.0013247-Park1]. Cultured cancer cells have been used to search for tumor markers. In particular, identifying proteins and peptides secreted into the culture media has developed by proteomics-based approach [@pone.0013247-Sasaki1], [@pone.0013247-Ogura1]. As for molecular signature in the exosomes, proteomics as well as transcriptomics analyses have been performed to reveal tumorigenesis and identify tumor marker candidates [@pone.0013247-Valadi1]--[@pone.0013247-AlNedawi1], [@pone.0013247-Simpson1], [@pone.0013247-Ji1]. Here, to identify miRNA related to tumorigenesis and metastasis, we performed extensive miRNA analysis in three cellular fractions including cells, exosomes, and medium from cultured cells. Ranking data of these intracellular and extracellular miRNAs obtained by microarray analysis, we found that let-7 miRNA family is rich in all the fractions from AZ-P7a cells, a metastatic gastric cancer cell line, which produces homogeneous and condensed morphology, and high recovery rate of exosomal miRNAs. These findings were distinct from other cell lines including lung cancer cell lines (SBC-3, NCI-H69, and DMS53), colorectal cancer cell lines (SW480 and SW620), and AZ-521, the parental cell line of AZ-P7a. Considering that let-7 miRNA family functions mainly as tumor suppressor genes [@pone.0013247-Roush1] to target oncogenes such as *RAS* and high mobility group A2 (*HMGA2*) [@pone.0013247-Boyerinas1], we propose that AZ-P7a cells selectively secrete let-7 miRNAs into the extracellular environment via exosomes to maintain their tumorigenic and metastatic propensities. Results {#s2} ======= Isolation of exosomes from various cancer cell lines {#s2a} ---------------------------------------------------- Exosomes are produced from inner cells to the extracellular environment via an exocytosis-like pathway [@pone.0013247-Thry1]. In addition to body fluids such as serum and plasma from peripheral blood [@pone.0013247-Simpson1], [@pone.0013247-Rabinowits1], exosomes are found in the medium of cultured cells [@pone.0013247-Simpson1], facilitating identification of exosomal molecules including proteins, mRNAs, and miRNAs for the aim for their clinical use. Profiling such exosomal molecules produced in cultured cancer cells have led to discover a novel candidate marker and antigen for cancer diagnostics and immunotherapy [@pone.0013247-Simpson1]. In this study, for the purpose of profiling exosomal miRNAs, we first isolated exosomes from culture media of 46 cancer cell lines with various tissue types, which include 8 for stomach, 16 for lung, 5 for colon, 9 for pancreas, 3 for prostate, and 5 for breast. After cells were grown for 48 h, exosomes were collected with a combination of successive centrifugation and molecular weight cut-off membranes as described in [Materials and Methods](#s4){ref-type="sec"}. Their purity of exosomal fractions was assessed by analyses of transmission electron microscopy and western blotting. Transmission electron microscopy showed round-shaped vesicular membrane structures approximately within the size of 100 nm in diameter. Among three cell lines, SBC-3, AZ-521, and AZ-P7a, it is of interest that the morphology from AZ-P7a cells was more homogeneous and condensed than other cells ([Figure 1A](#pone-0013247-g001){ref-type="fig"}). Immunoelectron microscopy showed that the extracellular particles isolated from the culture medium of AZ-P7a cells had immunoreactivity with an antibody against CD63, one of the exosomal membrane proteins, on the capsular membranes ([Figure 1B](#pone-0013247-g001){ref-type="fig"}). The presence of known exosomal proteins including CD29/β1-integrin, Aip1/Alix and tumor susceptibility gene 101 (Tsg101) was confirmed by western blot analysis, while a protein localized to endoplasmic reticulum, Bip/Grp78, was undetectable ([Figure 1C](#pone-0013247-g001){ref-type="fig"}). This result indicates that the contamination of material derived from other cellular compartments in the exosomal fractions was minimal. The yield of exosomal proteins from AZ-P7a cells was 10 times higher than that from AZ-521 cells; ∼2.5 mg proteins were obtained from 1×10^8^ AZ-P7a cells. ![Characterization of exosomes.\ **A.** Morphological characterization of exosomes derived from AZ-P7a, AZ-521, and SBC-3 cells by transmission electron microscopy. **B.** Immunoelectron micrographs of AZ-P7a exosomes labeled with immunogold anti-CD63. Exosomes with black colloidal gold particles on the capsular membranes were observed as positive CD 63 staining under the transmission electron microscope (arrowheads). **C.** Molecular characterization of exosomes derived from AZ-P7a, AZ-521, and SBC-3 cells by western blot analysis. Protein extracts (10 µg) prepared from cells (C) or exosomes (Ex) were assessed using antibodies against exosomal protein markers (CD29/β1-integrin, Aip1/Alix, and Tsg101) and an endoplasmic reticulum marker (Bip/Grp78).](pone.0013247.g001){#pone-0013247-g001} Enrichment of exosomal RNAs derived from AZ-P7a cells {#s2b} ----------------------------------------------------- After isolation, exosomal total RNAs were isolated from 46 cultured cell lines. Interestingly, the RNA yields from AZ-P7a cells were much greater than those from other cells ([Figure 2A](#pone-0013247-g002){ref-type="fig"}). The distribution in length showed the presence of large amounts of small RNAs in exosomes (center panel, [Figure 2B](#pone-0013247-g002){ref-type="fig"}). It should be noted that there is a significant difference in total exosomes and the exosomal miRNA levels between patients with lung adenocarcinoma and controls [@pone.0013247-Rabinowits1]. AZ-P7a cell line was established by repeated inoculation of the parental AZ-521 cells in mice, resulted in high peritoneal-metastatic potential [@pone.0013247-Nishimori1], [@pone.0013247-Fukui1]. Thus, the high yields of both RNA and protein along with the morphological characteristics in AZ-P7a cells may be attributed to their high tumorigenic and metastatic propensities. Since miRNAs have been detected in cell-free body fluids [@pone.0013247-Mitchell1]--[@pone.0013247-Gilad1], we also performed direct isolation of total RNAs from culture medium without procedure for isolating exosomes. The amounts of total RNAs from culture media were generally greater than those from exosomes ([Figure 2A](#pone-0013247-g002){ref-type="fig"}). RNA distribution in length showed that miRNA fractions were detected in the range from 10 to 40 nucleotide length in RNAs prepared from culture medium (right panel) as well as exosomes (center panel) from AZ-P7a cells ([Figure 2B](#pone-0013247-g002){ref-type="fig"}). According to the manufacture\'s protocol for Bioanalyzer, peaks with longer than 40 nucleotide length contain other small RNAs including tRNAs at 70∼90, 5S ribosomal RNA at 100, and 5.8S ribosomal RNA at 150, as observed in the intracellular fraction (left panel). These results suggest that miRNAs may exist in the culture medium outside the exosomal fraction although RNAs are believed to be protected from RNases as being packaged in exosomes. Particularly, compared to adherent cells, the increment of RNA yields was remarkably greater in floating cells (NCI-H69 and Lu-135) or cells with mix populations of floating and adherent cells (Colo205), which may be reflected by contamination of RNAs shed from lysed cells that have been continuously left in culture media. ![Total RNAs in exosomes and culture media from various cancer cell lines.\ **A.** Amounts of total RNAs recovered from exosomes (lower panel) and culture media (upper panel). The amount of total RNAs per cell was shown. **B.** Distribution in length of RNA. Isolated total RNAs from the exosomes and culture medium of AZ-P7a cells were detected using Bioanalyzer. The result of cellular total RNAs was shown for comparison.](pone.0013247.g002){#pone-0013247-g002} miRNA profiling by microarray analysis {#s2c} -------------------------------------- miRNA profiles in the intracellular and extracellular fractions were determined by microarray analysis for seven cancer cell lines including AZ-521, AZ-P7a, SBC-3, NCI-H69, DMS53, SW480, and SW620. Samples from culture media as well as exosomes provided hybridization signals ([Figure 3](#pone-0013247-g003){ref-type="fig"}), indicating the existence of miRNAs in these samples. Up to the present, analyzing microarray data for miRNA among samples, no clear methods have been made for normalization of signal intensity since it is difficult to find an appropriate internal standard material [@pone.0013247-Pradervand1]. Thus, miRNA microarray data have been in general analyzed without normalization, assuming that the total amounts of input RNAs are constant among samples [@pone.0013247-Pradervand1], [@pone.0013247-Sato1]. In this study, after averaging miRNA microarray data between two replicates, we ranked miRNAs according to their signal intensities. We found that let-7 miRNA family such as let-7a, let-7b, let-7c, let-7d, let-7e, let-7f, let-7g, and let-7i were detected with relatively high signals throughout most of the intracellular fractions from the seven cell lines ([Table 1](#pone-0013247-t001){ref-type="table"}, [Figure 3](#pone-0013247-g003){ref-type="fig"}). However, no or little signal intensities for the let-7 miRNAs were detected in the extracellular fractions except both extracellular fractions from AZ-P7a cells and culture medium fraction from NCI-H69 cells ([Table 1](#pone-0013247-t001){ref-type="table"}, [Figure 3](#pone-0013247-g003){ref-type="fig"}). Particularly, in both the extracellular fractions from AZ-P7a cells, all the eight let-7 miRNAs were detected although the rank of let-7g was lower than other seven let-7 miRNAs. Based on the distinct patterns of let-7 miRNA levels among the three cellular fractions, we divided the seven cell lines into three groups ([Figure 3](#pone-0013247-g003){ref-type="fig"}) as follows; (1) AZ-P7a, cells that let-7 miRNAs were found in all the three fractions, (2) AZ-521 along with SBC-3, DMS53, SW480, and SW620, cells that let-7 miRNAs were generally found in only intracellular fractions, (3) NCI-H69, cells that let-7 miRNAs were found in the intracellular fractions as well as in the culture medium to some extent. ![miRNA profiling in the intra- and extra-cellular fractions from AZ-P7a, AZ-521, NCI-H69, SBC-3, DMS53, SW480 and SW620 cells.\ miRNA profiles in cells (C), exosomes (Ex), and culture media (CM) were obtained by miRNA microarray analysis. Y axis represents log~2~ of hybridization signals, shown by bars and filled boxes. The arrows on the left side from the bunch of signals represent signals corresponding to let-7 miRNA family including let-7a (red), let-7b (yellow), let-7c (light green), let-7d (dark green), let-7e (sky blue), let-7f (blue), let-7g (navy blue), and let-7i (purple).](pone.0013247.g003){#pone-0013247-g003} 10.1371/journal.pone.0013247.t001 ###### Ranking of let-7 miRNA family in the intra- and extra-cellular fractions based on signal intensities from microarray data. ![](pone.0013247.t001){#pone-0013247-t001-1} Cell line Sample Rank by signal intensity ----------- -------- -------------------------- ---- ----- ---- ---- ---- ----- ----- ----- AZ-P7a Cells 2 9 8 14 20 6 29 22 202 Ex 10 4 9 34 28 15 90 51 212 CM 30 13 22 71 83 46 147 63 209 AZ-521 Cells 3 4 5 13 17 6 22 18 171 Ex -- 53 130 -- -- -- -- -- 153 CM -- 99 -- -- -- -- -- 158 225 NCI-H69 Cells 1 2 6 25 27 4 44 23 261 Ex -- 41 -- -- -- -- -- -- 102 CM 36 16 24 86 91 55 -- 69 131 SBC-3 Cells 1 31 56 83 17 7 115 -- 132 Ex -- 38 -- -- -- -- -- -- 56 CM -- -- -- -- -- -- -- -- 186 DMS53 Cells 2 8 20 16 -- 9 29 10 85 Ex -- 25 -- -- -- -- -- -- 42 CM -- -- -- -- -- -- -- -- 86 SW480 Cells 1 2 15 13 36 6 23 25 124 Ex -- 33 -- -- -- -- -- -- 47 CM -- 72 -- -- -- -- -- -- 106 SW620 Cells 2 1 13 14 23 7 21 16 107 Ex -- -- -- -- -- -- -- -- 32 CM -- -- -- -- -- -- -- -- 88 Ex = exosomes. CM = culture medium. miRNAs = the total number of miRNAs detected. A dash (-) indicates no detection. RT-PCR analyses of intracellular and extracellular let-7 miRNA family {#s2d} --------------------------------------------------------------------- We then performed quantitative RT-PCR for the eight let-7 miRNA family to validate the microarray data. let-7 miRNAs were readily detected in the intracellular and extracellular fractions from AZ-P7a cells ([Figure 4A](#pone-0013247-g004){ref-type="fig"}) and AZ-521 cells ([Figure 4B](#pone-0013247-g004){ref-type="fig"}). The levels of let-7 miRNAs per input RNAs were generally higher in the extracellular fractions from AZ-P7a cells than from AZ-521 cells. For normalizing the levels of target miRNAs obtained by RT-PCR, U6 snRNA has been generally used as an internal control. We observed the presence of U6 snRNA in all the three fractions from AZ-P7a cells and AZ-521 cells ([Figure 4C](#pone-0013247-g004){ref-type="fig"}) and then compared the levels of let-7 miRNAs between the corresponding fractions from these two cell lines. After normalization, the levels of let-7 miRNAs including let-7a, let-7b, let-7c, let-7d, let-7e, and let-7i in both the fractions of exosomes and culture media from AZ-521 cells were lowered as compared with those from AZ-P7a cells. On the contrary, there was no large difference in the levels of the intracellular let-7 miRNAs. We next compared the level of exosomal let-7a from AZ-P7a cells with those from other cancer cell lines including SBC-3, NCI-H69, DMS53, SW480, and SW620, and found that the let-7a level only from SW620 cells was relatively close (0.7 times) to the level from AZ-P7a cells ([Figure 4E](#pone-0013247-g004){ref-type="fig"}). It should be noticed that the intracellular level of let-7a from SW620 cells was approximately 3.5 times more abundant than that from AZ-P7a cells. Based on these results, we assumed localization of let-7a in the fractions of cells and exosomes ([Figure 5](#pone-0013247-g005){ref-type="fig"}), and conducted further analysis. Normalized by the U6 levels, we calculated relative levels of let-7a packaged in exosomes per the levels of cellular let-7a ([Figure 4F](#pone-0013247-g004){ref-type="fig"}). As a result, the amount of exosomal let-7a was much greater in AZ-P7a cells than other six cells, suggesting that AZ-P7a cells were selectively and actively secreted let-7 miRNA family into the extracellular environment via exosomes. ![RT-PCR analyses of intra- and extra-cellular let-7 miRNA family.\ **A**, **B.** Levels of mature let-7 miRNA family including let-7a, let-7b, let-7c, let-7d, let-7e, let-7f, let-7g, and let-7i in cells (shaded bars), exosomes (filled bars), and culture medium (open bars) from AZ-P7a cells (A) and AZ-521 cells (B). Each miRNA level was analyzed using quantitative RT-PCR. The cycle threshold (Ct) value is presented as the mean ± SD (n = 4). **C.** Levels of U6 snRNA in cells (shaded bars), exosomes (filled bars), and culture medium (open bars) from AZ-P7a and AZ-521 cells detected by quantitative RT-PCR analysis. The cycle threshold (Ct) value is presented as the mean ± SD (n = 4). **D.** Relative amounts of let-7 miRNA family in cells (shaded bars), exosomes (filled bars), and culture medium (open bars) from AZ-521 cells versus AZ-P7a cells. The dotted line and arrow represents the levels of let-7 miRNA family in AZ-P7a cells, shown as 100%. The levels of U6 snRNA in each sample were used as an internal standard for normalization amounts of let-7 miRNAs. Except let-7f and let-7g, the levels of let-7 miRNAs were reduced in the extracellular fractions. **E.** Relative amounts of let-7a in cells (shaded bars) and exosomes (filled bars) from 7 cancer cell lines including AZ-P7a, AZ-521, NCI-H69, SBC-3, DMS53, SW480 and SW620 cells. The dotted line represents the levels in AZ-P7a cells, shown as 1. The levels of U6 snRNA in each sample were used for normalized amounts of let-7a. The value of fold change is presented as the mean ± SD (n = 3) from samples independently prepared from cell culture. **F.** Relative amounts of exosomal let-7a in 7 cancer cell lines. The dotted line represents the level in AZ-P7a cells, shown as 100%.](pone.0013247.g004){#pone-0013247-g004} ![Models on difference in localization of let-7 miRNA family.\ Based on the results obtained by microarray and RT-PCR analyses, these models were drawn. In comparison between AZ-P7a cells (left) and AZ-521cells (center), normalized by the amount of U6 snRNA (open triangles), the amount of exosomal let-7a miRNA (filled triangles) in AZ-521 cells was approximately 3% of that in AZ-P7a cells while the intracellular amount in AZ-521 cells was rather 1.4 times greater than that in AZ-P7a cells ([Figure 4E](#pone-0013247-g004){ref-type="fig"}). These models are applied to six of eight let-7 miRNA family tested, including let-7a, let-7b, let-7c, let-7d, let-7e, and let-7i. In SW620 cells (right), the amount of let-7a was 3.5 times greater than that in AZ-P7a cells while the amount of exosomal let-7a was 0.7 times less than in AZ-P7a cells ([Figure 4E](#pone-0013247-g004){ref-type="fig"}).](pone.0013247.g005){#pone-0013247-g005} Discussion {#s3} ========== In the present study, we revealed that let-7 miRNA family was secreted into the extracellular environment via exosomal transport, which was specific to a metastatic gastric cancer cell line, AZ-P7a. let-7 miRNA family in human consists of 10 sequences from 13 precursors [@pone.0013247-Roush1]. In general, let-7 miRNAs act as a tumor suppressor by targeting oncogenes such as *RAS* and *HMGA2* and let-7 miRNAs are downregulated in many cancers from solid organs [@pone.0013247-Boyerinas1]. AZ-P7a cells possess a metastatic ability with peritoneal dissemination in nude mice [@pone.0013247-Nishimori1], [@pone.0013247-Fukui1]. Thus, we propose that the exosomal release of let-7 miRNAs into the extracellular environment results in decrease of anti-tumorigenic effect within the cells, which lead to maintain their oncogenesis and invasiveness. On the other hand, let-7 miRNAs less frequently play an oncogenic function, due to increase of expression by hypomethylation at the let-7 locus [@pone.0013247-Brueckner1] or targeting caspase-3 mRNA [@pone.0013247-Tsang1]. In this case, the exosomal release of let-7 miRNAs may cause transformation in target cells by transferring their oncogenic properties. In patients with lung cancer, the total levels of exosomes and their miRNAs increase compared to controls [@pone.0013247-Rabinowits1]. It has been demonstrated that tumor exosomes induce immune escape mechanism in cancers by spontaneous T cell apoptosis [@pone.0013247-Poutsiaka1]--[@pone.0013247-Ichim1]. In this study, we have shown homogeneous morphology and enrichment of RNAs in exosomes from metastatic AZ-P7a cells. This may be reflected by their tumorigenicity. miRNAs are very stable in blood plasma and serum since protected from RNases [@pone.0013247-Mitchell1]. Thus, it makes miRNA levels tested for clinical diagnosis as tumor markers and biomarkers. How does it happen? Chin et al. have suggested two hypotheses for the source of miRNAs on circulating blood samples; it may be due to a result of tumor cell death and lyses or release by tumor cells into the extracellular microenvironment of blood vessels [@pone.0013247-Chim1]. On the contrary to adherent cells, we observed that the amounts of total RNAs in culture media were relatively higher than in exosomes for cells with floating property such as NCI-H69, Lu-135, and Colo205. This increment probably resulted in the contamination of RNAs from dead cells in culture media. For the discovery of novel blood tumor markers and biomarkers, omics approaches including proteomics and transcriptomics are conducted either by direct analysis of blood samples or application of profiling in tissue samples. There are contradictory reports if whether the profiles of miRNA and mRNAs in bloodstream are parallel with the tumor\'s profile. The signature of exosomal miRNAs reflects that of the originating tumor cells in patients with lung adenocarcinoma [@pone.0013247-Rabinowits1] and breast cancer [@pone.0013247-Heneghan1]. On the other hand, Skog et al. have shown that microarray analysis for mRNA derived from glioblastoma cells and the corresponding exosomes revealed mRNAs exclusively detected in microvesicles, speculating that mRNAs are localized to a specific region of cytoplasm [@pone.0013247-Skog1]. In addition, Tanaka et al. have demonstrated that the level of miR-92a decreased in the blood plasma of acute leukemia patients while its strong expression was found in the tissue samples [@pone.0013247-Tanaka1]. Our results on selective enrichment of let-7 miRNA family in AZ-P7a cells-derived exosomes fit with latter case. Unless target molecules are miRNAs derived from circulating tumor cells, it suggests careful investigation are needed to use miRNA profiles in tissue samples for detection in serum or plasma samples. Since exosomes are produced in hematopoietic cells as well as epithelial cells, circulating blood contains exosomes derived from both type of cells. In general, to investigate exosomal profiles of proteins, mRNAs, and miRNAs in serum and plasma, exosomes derived from tumors with epithelial characteristics are separated from those from hematopoietic cells [@pone.0013247-Rabinowits1], [@pone.0013247-Taylor1]. This is performed using affinity purification with antibodies against molecules such as EpCAM that express on the surface of epithelial cells. Our findings that there were similar levels of let-7 miRNA family between exosomes and culture media from AZ-P7a cells suggest that these miRNA levels may be measured without isolating exosomes from clinical samples such as serum and plasma. In conclusion, this study demonstrated that let-7 miRNA family is rich in exosomes from a metastatic gastric cancer cell line, AZ-P7a cells. Since let-7 miRNAs are involved in proliferation process [@pone.0013247-Boyerinas1], their exosomal secretion may play an important role in tumorigenesis and metastasis. Materials and Methods {#s4} ===================== Cell culture {#s4a} ------------ Human stomach cancer cell lines (AZ-521, MKN45, KATOIII, and NUGC-3), human lung cancer cell lines (SBC-3, Lu-65, Lu-135, and KNS-62), and human pancreatic cancer cells (Suit-2) were purchased from Japanese Collection of Research Bioresources. Human stomach cancer cells (SNU-1), human lung cancer cell lines (DMS-53, DMS114, NCI-H69, A549, NCI-H1299, NCI-H1155, NCI-H520, NCI-H1560, SK-MES-1, and HCC827), human mesothelioma cells (MSTO-H211), human colorectal cancer cell lines (SW480, SW620, Colo205, LoVo, and HCT116, human pancreatic cancer cell lines (BxPC-3, AsPC-1, Panc-1, MIAPaca-2, HPAF-II, PL45, and Capan-2), human prostate cancer cell lines (PC-3, SU145, and LNCaP), and human breast cancer cell lines (T-47D, McF7, SK-BR-3, BT-474, and MDA-MB-231) were purchased from American Type Culture Collection. MKN28 (human stomach cancer cells) and PSN-1 (human pancreatic cancer cells) were purchased from Immuno-Biological Laboratories and European Collection of Cell Culture, respectively. Human stomach cancer cell lines, AZ-P7a [@pone.0013247-Nishimori1], [@pone.0013247-Fukui1] and MKN45P [@pone.0013247-Iinuma1] were gift from Drs. Yutaka Yonemura and Yoshio Endo. In AZ-521 and AZ-P7a cells, no RT-PCR products were observed for 14 retroviruses including HPV16, EBV, CMV, SV40, HBV, HCV, BKV, JCV, HHV7, HTLV1, HTLV2, HIV1, HIV2, and ADV type 1 (data not shown), excluding infection by these retroviruses in both the cell lines. Cells were maintained in RPMI-1640 medium (Sigma-Aldrich) supplemented with 10% heat-inactivated fetal bovine serum (Invitrogen), 100 U/ml penicillin, and 0.1 mg/ml of streptomycin. Production and isolation of exosomes {#s4b} ------------------------------------ Adherent cells were seeded to 150 mm-dishes at appropriate number of cells ranging from 3×10^6^ to 1×10^7^ cells per dish, while floating cells were inoculated into 75 cm^2^-flask at 2∼2.5×10^7^ cells per flask. Cells were cultured in complete RPMI-1640 medium of 25 ml and 50 ml for dishes and flasks, respectively, for 24 h at 37°C and 5% of CO~2~, washed twice with phosphate-buffered saline (PBS), and incubated 48 h in phenol red-free RPMI-1640 medium (Sigma-Aldrich) containing 10% heat-inactivated fetal bovine serum which was previously depleted contaminating microvesicles by overnight centrifugation at 100,000×g. Exosomes were isolated from the total cells ranging from 6×10^7^ to 3×10^8^ cells as previously described [@pone.0013247-Ji1]. Briefly, culture medium was collected and centrifuged at 800×g for 5 min and additional 2,000×g for 10 min to remove lifted cells. The supernatant was subjected to filtration on a 0.1 µm pore polyethersulfone membrane filter (Corning) to remove cell debris and large vesicles, followed by concentration by a 100,000 Mw cut-off membrane (CentriPlus-70, Millipore). The volume of supernatant was reduced from approximately 250--500 ml to approximately 30 ml. The supernatant was then ultracentrifuged at 100,000×g for 1 h at 4°C using 70Ti rotor (Beckman Coulter). The resulting pellets were resuspended in 6 ml PBS and ultracentrifuged at 100,000×g for 1 h at 4°C using 100Ti rotor (Beckman Coulter). Transmission electron microscopy {#s4c} -------------------------------- The pelleted exosomes were mixed with equal quantities of freshly prepared 2% glutaraldehyde in PBS, incubated overnight at 4°C, postfixed with 1% osmium tetroxide in PBS at 4°C for 2 h, and dehydrated in a graded series of ethanol. Following dehydration, the samples were transferred to propylene oxide and embedded in epoxy resin Quetol 812 (Nisshin EM). Ultrathin sections were cut with Ultracut UCT (LEICA), stained with uranyl acetate and lead citrate, and observed with the JEM-1230 Electron Microscope (JEOL). Immunoelectron microscopic analysis was performed according to the method described by Xiao et al. [@pone.0013247-Xiao1] with minor modifications. In brief, exosomes were mixed and incubated with mouse anti-human CD63 monoclonal antibody (catalog no. sc-51662, Santa Cruz) for 1 h at room temperature. The sample was dropped onto the membrane surface of a copper mesh (collodion/carbon coated 400 mesh, Nisshin EM) and incubated for 1 h at room temperature. After wash with PBS, immunogold conjugated goat anti-mouse IgG (catalog no. EM.GMHL15, BBInternational) diluted at 1∶20 was dropped onto the membrane. The copper mesh was floated in the droplets with its membrane surface faced down at room temperature for 30 min. After wash with PBS, uranyl acetate drops was put onto the copper mesh surface and stained at room temperature for 30 s. Exosomes with black colloidal gold particles on the capsular membranes were marked as positive under the transmission electron microscope. Western blot analysis {#s4d} --------------------- Cells and exosomal fractions were washed, resuspended in a lysis buffer \[7.5 M urea (Sigma-Aldrich), 2.5 M thiourea (Sigma-Aldrich), 12.5% glycerol (Wako), 50 mM Tris, 2.5% n-octyl-beta-D-glucoside (Sigma-Aldrich), 6.25 mM Tris(2-carboxyethyl)phosphine hydrochloride (Sigma-Aldrich), 1.25 mM protease inhibitor (catalog no. P2714, Sigma-Aldrich)\], and incubated for 1 h at 4°C using the Rotator RT-50 (TAITEC). After centrifugation at 14,000×g for 60 min at 4°C, the supernatant was collected and protein concentration was determined by the Bradford Protein Assay Kit (Bio-Rad). A portion of proteins (10 µg) were denatured by boiling in Laemmli sample buffer, run on 10% SDS-polyacrylamide gels, and transferred to Immobilon-P PVDF membranes (0.45 µm pore size, Millipore). The blots were blocked for 1 h with 5% non-fat dry milk in Tris-buffered saline containing Tween 20 (10 mM Tris--HCl, pH 7.5, 150 mM NaCl and 0.01% Tween 20). After blocking the membranes were incubated with each primary antibody or antiserum including rabbit anti-Tsg101 serum (catalog no. T5951, Sigma-Aldrich), goat anti-Aip1/Alix serum (catalog no. sc-49268, Santa Cruz), mouse monoclonal anti-CD29/β1-integrin (catalog no. sc-610468, BD Biosciences), and mouse monoclonal anti-Bip/Grp78 (catalog no. 610979, BD Biosciences) at a dilution of 1∶200 for 1 h at room temperature. The blots were then incubated with the corresponding anti-IgG secondary antibody conjugated with horseradish peroxidase (Jackson Laboratories) at a dilution of 1∶2,000 for 1 h at room temperature. Specific proteins were visualized using the ECL system (GE Healthcare) and the FUJIFILM Luminescent Image Analyzer LAS3000 (Fuji Film). The protein molecular weight was deduced using the Precision Plus Protein Standards (Bio-Rad Laboratories). RNA Isolation {#s4e} ------------- For RNA isolation from culture medium, cells were seeded as described above. 48 h after grown, culture medium was collected and centrifuged at 800×g for 5 min and additional 2,000×g for 10 min. The supernatant was subjected to filtration on a 0.22 µm pore polyethersulfone membrane filter (Corning), followed by concentration with a 5,000 Mw cut-off membrane (Centricon Plus-70, Millipore). The final volume of supernatant was approximately 10--20 ml from the original volume of 250--500 ml. The supernatant was mixed with equal volume of QIAzol Lysis Reagent (Qiagen) and followed the procedure for RNA isolation described below. Total RNA was isolated from cells, exosomes, or culture media using the miRNeasy Mini Kit (Qiagen) according to the manufacture\'s instructions. To remove genomic DNA, 40 µg of cellular total RNA were incubated with 40 units of RQ1 RNase-Free DNase (Promega) at 37°C for 30 min in the presence of 40 units of RNasin Plus RNase Inhibitor (Promega). For total RNA from exosomes and culture media, all amounts of crude products were treated in the same procedure except the use of 5 U DNase. RNA was purified using the RNeasy MinElute Cleanup Kit (Qiagen) according to the manufacturer\'s instructions. The RNA samples were quantified with a NanoDrop spectrophotometer (Thermo Fisher Scientific) and assessed using the Agilent Small RNA Kit and the Agilent 2100 Bioanalyzer (Agilent Technologies). Microarray analysis {#s4f} ------------------- For miRNA profiling, 100 ng and 20 ng of cellular and extracellular total RNAs, respectively, were fluorescence-labeled using the miRNA Complete Labeling Reagent and Hyb Kit (Agilent Technologies) according to manufacture\'s protocol (Agilent miRNA microarrays version 2.2). The comprehensive miRNA analysis was performed using the Human miRNA Microarray Kit (8×15 K) Ver.3.0 (Agilent). Hybridization signals were detected with the DNA Microarray Scanner (Agilent Technologies). The scanned images were analyzed using the Agilent Feature Extraction software and data analysis was performed using the GeneSpring GX software (Agilent Technologies). The data discussed in this manuscript have been deposited in NCBI\'s Gene Expression Omnibus (GEO) and are accessible through GEO Series accession number GSE21350: <http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE21350>). Reverse transcription (RT) and quantitative RT-PCR for miRNA analysis {#s4g} --------------------------------------------------------------------- Quantitative miRNA levels were determined using real-time RT-PCR with the Applied Biosystems 7900 HT Sequence Detection System (Applied Biosystems), TaqMan® Gene Expression Assay (Applied Biosystems) for human let-7a (assay ID 000377), let-7b (assay ID 002619), let-7c (assay ID 000379), let-7d (assay ID 002283), let-7e (assay ID 002406), let-7f (assay ID 000382), let-7g (assay ID 002282), let-7i (assay ID 002221), and U6 snRNA (assay ID 001973) as an endogenous control. Ten ng of total RNA were subjected to reverse transcription with TaqMan® Universal PCR Master Mix No AmpErase (Applied Biosystems) and the respective TaqMan® reagents for target genes. RT-PCR was carried out in a total volume of 20 µl reaction mixture according to the manufacture\'s protocol. Amplification was carried out as follows: 95°C for 10 min, 40 cycles of 95°C for 15 s and 60°C for 60 s. Samples were analyzed in triplicate as biological replicate or quadruplicate as technical replicate. The miRNA levels were defined from the cycle threshold (Ct), the comparative Ct method, and normalization by the level of U6 snRNA in each sample. Fold increases or decreases in each miRNA level of cell lines tested were determined by reference to the level of AZ-P7a cell line. We thank the members of the Shizuoka Cancer Center Research Institute for their many helpful suggestions. We also thank Drs. Yutaka Yonemura and Yoshio Endo for gift of AZ-P7a and MKN45P cell lines. **Competing Interests:**The authors have declared that no competing interests exist. **Funding:**The work was supported by a grant in cooperation with Innovative Technology and Advanced Research in Evolutional Area (CITY AREA) from the Ministry of Education, Culture, Sports, Science and Technology of Japan. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: KO. Performed the experiments: KI AF KK YW KM. Analyzed the data: KO KH YF SiO KY TM. Contributed reagents/materials/analysis tools: KY TM. Wrote the paper: KO KY TM. [^2]: Current address: Integrated Research Institute, Tokyo Institute of Technology, Yokohama, Japan

{ "pile_set_name": "PubMed Central" }

Achim Quaiser, Mart Krupovic, Alexis Dufresne, André-Jean Francez, and Simon Roux. Diversity and comparative genomics of chimeric viruses in Sphagnum-dominated peatlands. *Virus Evolution*, 2016, 2(2): vew025 The publisher wishes to apologize for mistakenly having abbreviated "Chimeric viruses" to "CHIV" throughout the article where this was not always appropriate. The article has been updated accordingly.

{ "pile_set_name": "PubMed Central" }

In an isolated model of myocardial stunning, cardiac myocytes demonstrate significant swelling and reduced contractility during exposure to 3 different stresses: hypothermic hyperkalemic cardioplegia (CPG), hypo-osmotic stress, or metabolic inhibition.[@b1]--[@b3] The ATP-sensitive potassium channel (K~ATP~) opener, diazoxide (DZX), ameliorates these stress-induced derangements in mouse, rabbit, and human tissue.[@b1]--[@b5] A role of K~ATP~ channels in ischemic preconditioning, by coupling membrane conductance of K^+^ ions to the metabolic state of the cell, has also been well defined in multiple animal models.[@b6]--[@b10] However, no precise mechanism of cardioprotection has been identified. K~ATP~ channels are formed as hetero-octamers of subunits that include 4 potassium inward rectifying (Kir) subunits and 4 regulatory subunits from the sulfonylurea receptor family (SUR1/2).[@b11]--[@b13] Different combinations of Kir and SUR isoforms exist in different tissues, each with unique properties and function. For example, the predominant combination of subunits found in murine ventricle is Kir6.2 and SUR2A, whereas murine atria mainly consist of Kir6.2 and SUR1.[@b14] Identification of these subunits has allowed for genetic manipulation to more specifically evaluate the mechanism of DZX. Using K~ATP~ subunit-knockout animal models, we have shown that the cardioprotection afforded by DZX requires the regulatory subunit, SUR1, but the pore-forming subunit involved remains elusive.[@b15],[@b16] We have also demonstrated that the sarcolemmal K~ATP~ channel (sK~ATP~) is implicated in myocyte swelling secondary to stress, and that knockout of the pore-forming subunit of this channel (Kir6.2) provides resistance to myocyte swelling.[@b2],[@b17] However, there is evidence that DZX cardioprotection does not require the sK~ATP~ channel, leading us to propose mechanisms that may include non-K~ATP~ channel or mitochondrial K~ATP~ channel (mK~ATP~) locations of action.[@b15],[@b16],[@b18]--[@b21] Regulation of mitochondrial volume through a purported mK~ATP~ channel is one of the proposed mechanisms of action of DZX.[@b22]--[@b24] Because no molecular composition of mK~ATP~ has been determined, identification of mK~ATP~ channel subunits has to date involved indirect methods, such as pharmacological inhibition or observed changes in mitochondrial volume (proposed assay of mK~ATP~ activity).[@b24],[@b25] Using such strategies, the inward rectifier potassium channel subunit, Kir1.1 (renal outer medullary kidney channel; ROMK) has been implicated as a pore-forming subunit of the mK~ATP~ channel utilizing the nonspecific inhibitor Tertiapin Q (TPN-Q).[@b24] The present study was conducted to assess potential roles of Kir1.1, Kir3.1, and Kir3.4 in the cardioprotection of DZX at the cellular and mitochondrial levels using the Kir subunit inhibitor, TPN-Q. Methods ======= All animal procedures were approved by the Animal Studies Committee at Washington University School of Medicine, and all animals received humane care in compliance with the National Institute of Health's (NIH) Guide to Care and Use of Laboratory Animals.[@b26] Mitochondrial Isolation ----------------------- Mice (both sexes, 6 weeks to 5 months and 15 to 30 g in weight) were anesthetized with 3% Avertin (0.3 g of 2,2,2-tribromoethanol, 0.186 mL of 2-methyl2-butanol, and 9.814 mL of sterile water) intraperitoneally, and rapid cardiectomy was performed. To isolate mitochondria, atria were excised and discarded whereas ventricular tissue was rapidly minced in cold buffer (in mmol/L: 10 HEPES (*N*-\[2-hydroxyethyl\][@b1],[@b3]--[@b5] piperazine-*N*-\[4-butanesulfonic acid\]), 1 EDTA-potassium, 250 sucrose, adjusted to a pH of 7.1 with 20% KOH and transferred to a 10-mL glass tube with a Teflon pestle (Glas-Col Homogenizer, Terre Haute, IN), and volume was adjusted to 7 mL with buffer. Tissue was mechanically homogenized with a Teflon pestle driven by a low-speed motor drive shaft set at 120 rpm. The homogenate was transferred to 6 microcentrifuge tubes and centrifuged at 900*g* for 10 minutes at 4°C. The supernatant was combined into a clean test tube and mixed to get a homogeneous solution that was divided equally between 6 clean microcentrifuge tubes and centrifuged at 5000*g* for 15 minutes. One pellet was resuspended in 100 μL of buffer, and 10 μL was taken in duplicate for the Bradford protein assay (Thermo Scientific; Rockford, IL) to determine total protein. Each pellet was maintained on ice and was resuspended in test media volume to equal 0.3 μg/μL in order to standardize protein content. Mitochondrial Volume Measurement -------------------------------- The volume of isolated mitochondria was measured after suspension in test solution: (1) isolation buffer (No ATP; n=13); (2) 200 μmol/L of ATP (n=12); (3) 200 μmol/L of ATP and 100 μmol/L of DZX (Sigma-Aldrich, St. Louis, MO) (n=12); (4) 200 μmol/L of ATP, 100 μmol/L of DZX, and 500 nmol/L of TPN-Q (n=7); or (5) 200 μmol/L of ATP, 100 μmol/L of DZX, and 100 nmol/L of TPN-Q (n=6). Isolation buffer (10 mmol/L of HEPES, 200 mmol/L of mannitol, 50 mmol/L of sucrose, 1 mmol/L of EGTA; pH 7.2) was used as a control solution. ATP has been shown to close mitochondrial K~ATP~ channels, so 200 μmol/L of ATP was used to slow the initial rate of mitochondrial swelling (0% Mito K~ATP~ activity).[@b22] Conversely, DZX activates "Mito K~ATP~" channels, and 100 μmol/L of DZX was added to achieve maximal activation of mitochondrial K~ATP~ channels (100% Mito K~ATP~ activity). We were unable to reproduce the TPN-Q dose-response relationships on mitochondrial volume demonstrated by other investigators using 4 TPN-Q concentrations (0.5, 10, 90, or 1000 pmol/L) despite attempts utilizing 3 different vehicles for TPN-Q: 20% acetonitrile, water, and HEPES.[@b24] In the present study, water was used as a vehicle for TPN-Q because of its stability in the medium and because 20% acetonitrile alone resulted in myocyte swelling owing to its cyanide moiety. Mitochondrial matrix volume measurements were obtained using a light-scattering technique,[@b27] where the absorbance, at 520 nm, of a solution of isolated mitochondria was obtained every 14 seconds for the period of 3 minutes using UV Probe 2.33 (Shimadzu Scientific Instruments, Columbia, MD) and a spectrophotometer (UV-1700 Spectrophotometer; Shimadzu Scientific Instruments, Columbia, MD). Myocyte Isolation ----------------- Ventricular myocytes were isolated from C57BL/6J mice of either sex (age 6 weeks to 5 months and 15 to 30 g in weight), as previously described.[@b15] Mice were anesthetized with 2.5% Avertin intraperitoneally. Heparin (0.1 mL) was administered intraperitoneally. Rapid cardiectomy was performed and solution A (described below) was perfused through the aorta for 5 minutes. The heart was then perfused at 37°C for 12 to 20 minutes with solution B (described below). Ventricles were removed, minced, and placed into solution C (described below) and gently dispersed by glass pipette. Cells were allowed to centrifuge by gravity, and serial washings were performed every 10 minutes for 15 to 20 minutes. Cells were used within 5 hours and randomized to treatment group. A typical yield of viable myocytes was 65% to 75%. Solution A consisted of (in mmol/L, except as noted) 116 NaCl; 5.36 KCl; 0.97 Na~2~HPO~4~; 1.47 KH~2~PO~4~; 21.10 HEPES (*N*-\[2-hydroxyethyl\] piperazine-*N*′-\[4-butanesulfonic acid\]); 11.65 glucose; 26.50 μmol/L of phenol red (Sigma-Aldrich); 3.72 MgCl~2~; 4.40 NaHCO~3~; essential vitamins (100×, 10 mL; GIBCO, Grand Island, NY); and amino acids (50×, 20 mL; GIBCO). Solution B consisted of solution A plus 10 μmol/L of CaCl~2~; 1.2 mg/mL of collagenase (Type 2; Worthington Biochemical Corporation, Freehold, NJ). Solution C consisted of solution A plus 5 mg/mL of BSA (Sigma-Aldrich); 1.25 mg/mL of taurine; and 150 μmol/L of CaCl~2~. Myocytes were exposed to 37°C control TYR for 5 minutes to obtain baseline volume. Any changes in cell volume secondary to the isolation would be evident during this period. Myocytes were then exposed to test solution (10 minutes) followed by re-exposure to TYR solution (5 minutes). Test solutions included the following groups (n=12 for each): (1) TYR; (2) CPG; (3) CPG+100 μmol/L of DZX; (4) CPG+100 μmol/L of DZX+TPN-Q 200 nm/L; (5) TYR+TPN-Q; and (6) CPG+TPN-Q. CPG consisted of (in mmol/L): NaCl 110, NaHCO~3~ 10, KCl 16, MgCl~2~ 16, and CaCl~2~ 1.2 and was equilibrated with 95% O~2~ to 5% CO~2~ and titrated to the pH of 7.3 with 10% NaHCO~3~ solution. Diazoxide (7-chloro-3-methyl-1,2,4-benzothiadiazine-1,1-dioxide[@b4]; Sigma-Aldrich) dose of 100 μmol/L was utilized because it was effective in ameliorating cell swelling secondary to stress in previous studies.[@b1]--[@b4] A stock solution of DZX was made by dissolving DZX in 0.1% dimethyl sulfoxide (DMSO), at which concentration DMSO has no effect on cell volume.[@b28] Myocyte Volume Measurement -------------------------- Myocytes were used on the day of isolation and were not cultured. Myocytes were visualized on an inverted microscope stage (IX-51; Olympus, Tokyo, Japan), as previously described.[@b15] After 5 minutes, the chamber was perfused at a rate of 3 mL/min with TYR (in mmol/L): NaCl 130, KCl 5, CaCl~2~ 2.5, MgSO~4~ 1.2, NaHCO~3~ 24, Na~2~HPO~4~ 1.75, and glucose 10 (buffered to a pH of 7.4 using 95% O~2~ to 5% CO~2~). After viability was confirmed, myocyte images were captured using video-based edge detection software (IonOptix, Milton, MA) and volume measured every 5 minutes, as previously described.[@b5] Myocyte Contractility --------------------- Myocyte contractility was measured using a video-based edge detection system (IonOptix). Cells were paced using a field stimulator (MyoPacer; IonOptix) at a voltage of 105 above threshold at a frequency of 1 Hz with a 5-ms duration to avoid occurrence of fusion beats. After 5 minutes of stimulation, data were obtained from 12 to 30 consecutive beats and averaged. Parameters of contractility included percentage of cell shortening, maximal velocity of shortening, and percentage of cell relengthening, as previously described.[@b5] Contractility was measured at baseline and after 5 minutes of re-exposure to TYR. Cells that showed less than 7% cell shortening at baseline were excluded. Statistical Analysis -------------------- Data were analyzed using SYSTAT 13 (Systat Software, Inc, Point Richmond, CA). All data are presented as mean±SEM relative to baseline. A repeated-measures ANOVA was used for sequential time-based measurements for each test solution against its own baseline and control values. Using Fisher's least significant different test, post-hoc multiple comparisons were done between different test groups at different time points during test solution and re-exposure periods. Probability values \<0.05 were considered significant. A Shapiro-Wilk test was used to test for normality. If the data failed the normality test, a nonparametric (Friedman's nonparametric repeated-measures comparison) was used. Statistical analysis was performed using SyStat 10.2 (Systat Software). Results ======= Mitochondrial Matrix Volume --------------------------- Isolated mitochondrial volume is represented in [Figure 1](#fig01){ref-type="fig"}. There were no statistically significant differences in mitochondrial volume between any of the groups. ![Mitochondrial matrix volume. Isolated mouse mitochondria were exposed to: (1) isolation buffer (no ATP); (2) 200 μmol/L of ATP; (3) 200 μmol/L of ATP and 100 μmol/L of DZX; (4) 200 μmol/L of ATP, 100 μmol/L of DZX, and 500 nmol/L of TPN-Q; or (5) 200 μmol/L of ATP, 100 μmol/L of DZX, and 100 nmol/L of TPN-Q and volume measured using light scattering (at 520 nm) for 2 minutes. Mitochondrial matrix volume is represented as 1/percent change in absorbance (mitochondrial matrix swelling is inversely related to absorbance measured at 520 nm) over time. Data are represented as mean±SEM. DZX indicates diazoxide; TPN-Q, tertiapin Q.](jah30004-e002016-f1){#fig01} Myocyte Volume -------------- Myocyte volume over time is represented in [Figure 2](#fig02){ref-type="fig"}. Myocytes demonstrated no significant volume change from baseline during exposure to TYR with or without TPN-Q. Myocytes did demonstrate significant swelling when exposed to CPG (*P*\<0.05 vs. TYR). This swelling was prevented by addition of DZX (*P*\<0.05 vs. CPG). The further addition of TPN-Q prevented this benefit (volume homeostasis) of DZX (CPG+DZX+TPN-Q *P*\<0.05 vs. CPG+DZX). As expected, the interaction between experimental groups and time was significant (*P*\<0.001). ![TPN-Q prevents beneficial maintenance of myocyte volume during stress by DZX. Isolated myocytes were exposed to control Tyrode's physiologic solution (TYR) for 5 minutes (time 0 to 5), stress (CPG, CPG+DZX, CPG+DZX+TPN-Q, CPG+TPN-Q, and TYR+TPN-Q) for 10 minutes (time 5 to 15), followed by TYR for 5 minutes (time 15 to 20). Myocyte volume was compared. Data are represented as mean±SEM. \**P*\<0.05 vs. TYR; ^Ψ^*P*\<0.05 vs. CPG+DZX. CPG indicates hypothermic hyperkalemic cardioplegia; DZX, diazoxide; TPN-Q, tertiapin Q.](jah30004-e002016-f2){#fig02} Myocyte Contractility --------------------- Myocyte contractility is presented in [Figure 3](#fig03){ref-type="fig"}. Contractility remained unaltered in the TYR group. Myocytes demonstrated a significant decline in all 3 parameters of contractility after exposure to CPG (*P*\<0.05 vs. TYR) that was worsened by the addition of TPN-Q. The addition of DZX to CPG prevented the significant reduction in contractility observed in the CPG group. TPN-Q inhibited the protection provided by DZX. ![TPN-Q prevents beneficial maintenance of myocyte contractility after stress by DZX. Isolated mouse myocytes were exposed to control Tyrode's physiologic solution (TYR) for 5 minutes (time 0 to 5), stress (CPG, CPG+DZX, CPG+DZX+TPN-Q, CPG+TPN-Q, and TYR+TPN-Q) for 10 minutes (time 5 to 15), followed by TYR for 5 minutes (time 15 to 20). Myocyte contractility was compared. Data are represented as mean±SEM. \**P*\<0.05 vs. TYR; ^Ψ^*P*\<0.05 vs. CPG+DZX. CPG indicates hypothermic hyperkalemic cardioplegia; DZX, diazoxide; TPN-Q, tertiapin Q](jah30004-e002016-f3){#fig03} Discussion ========== In previous studies, we have shown that myocyte swelling and reduced contractility in response to stress (hyperkalemic CPG, hypoosmotic stress, and metabolic inhibition) are ameliorated by addition of the K~ATP~ channel opener, DZX.[@b1]--[@b5] Our efforts to localize the precise mechanism of cardioprotection afforded by DZX have implicated an involvement of the K~ATP~ channel, SUR1,[@b16],[@b19] but not the Kir6.2 subunit.[@b2] Other investigators have provided evidence in support of the idea that Kir1.1 (the ROMK) is a pore-forming subunit of the mK~ATP~ channel.[@b24] In the current study, changes in mitochondrial and myocyte volume were observed in the presence of ROMK blocker TPN-Q in the presence of DZX. A dose effect on mitochondrial volume was not detectable using TPN-Q in the presence of DZX and multiple doses of TPN-Q. The stability of TPN-Q in the vehicle medium was thus evaluated in the media, in 20% acetonitrile, HEPES, and in water. However, no significant change in mitochondrial volume was observed after the addition of TPN-Q, even at high concentrations, in contrast to other investigators.[@b24] Thus, we conclude that TPN-Q had no effect on mitochondrial volume in the presence of DZX, suggesting a location of action of DZX that was distinct from the channels targeted by TPN-Q. Myocytes demonstrated significant swelling in response to CPG that was prevented by the K~ATP~ channel opener, DZX, which is consistent with previous results.[@b2],[@b3] Similarly, a correlation between myocyte volume changes and contractility changes was consistent with previous work, supporting this myocyte model of stunning.[@b4],[@b5] TPN-Q prevented this beneficial effect of DZX, thus implicating TPN-Q targets---Kir1.1, Kir3.1, or Kir3.4---in the cardioprotection of DZX at the cellular level. Other non--Kir TPN-Q targets or undefined channel subunits sensitive to TPN-Q may play a role (GIRK1/4, K~ACH~, and voltage-dependent Ca^2+^-activated K^+^ channels) in cardioprotection and will be the subject of future investigations.[@b29]--[@b31] The present study further characterizes the mechanism of cardioprotection provided by DZX. TPN-Q did not alter DZX-induced mitochondrial swelling, but it did inhibit myocyte cardioprotection provided by DZX during stress. Given that TPN-Q inhibits Kir1.1-, Kir3.1-, and Kir3.4-dependent K-channel activities, these data support that any of these subunits (as well as undefined subunits sensitive to TPN-Q) could be involved in the cardioprotection afforded by DZX. However, these data also suggest that mitochondrial swelling by DZX does not involve Kir1.1, Kir3.1, or Kir3.4 and likely results from a yet to be identified mechanism. Future work to identify the site and mechanism of action of K~ATP~ channel openers will involve indirect methods until genetic identification is accomplished. Sources of Funding ================== This study was supported by NIH RO1 HL098182-01A1 (Lawton), NIH 5T32HL007776 (Henn, Janjua), and the Barnes Jewish Hospital Foundation (Lawton). Disclosures =========== None. [^1]: Presented at the American Heart Association Scientific Sessions, November 15--19, 2014 in Chicago, IL.

{ "pile_set_name": "PubMed Central" }

Background {#Sec1} ========== Suicide is a particular concern in mental health settings because of its strong association with mental illness \[[@CR1]\]. Although suicides rarely occur during in-patient care, these events are clinically important and are among the most concerning patient safety incidents in the mental health sector \[[@CR2]--[@CR4]\]. Suicide prevention is one of the primary tasks of health care professionals practicing in psychiatric wards \[[@CR4]\]. In-patient suicide prevention is a high-priority in many countries \[[@CR5]--[@CR7]\]; however, its practice remains poorly understood. The ethical and pragmatic problems posed by including suicidal patients in research have contributed to the currently limited research regarding the treatment of high-risk and hospitalized suicidal patients \[[@CR8]\]. To understand safety in health care services, information must be obtained from multiple sources, including the patient's perspective. As such, patients can provide insight regarding care and can contribute important information when other sources of evidence are limited \[[@CR9]\]. Patients can also provide unique information on adverse events in hospitals \[[@CR10], [@CR11]\] as well as useful descriptive feedback regarding safety, in particular sensitive safety-related topics \[[@CR12]\]. Patient experiences are considered one of the three pillars of health care quality, along with clinical safety and effectiveness of outcomes \[[@CR13]\]. Qualitative studies of patient experiences with psychiatric in-patient care have been reviewed within certain areas, such as involuntary hospitalizations \[[@CR14]\], physical restraint \[[@CR15]\], acute wards \[[@CR16]\], seclusion practices \[[@CR17]\], locked doors \[[@CR18]\] and service user expectations \[[@CR19]\]. However, no reviews to date have examined studies regarding suicidal in-patients. Therefore, this review aimed to summarize empirical qualitative studies by exploring suicidal patients' psychiatric in-patient care experiences to better understand their perspectives toward safety. Review question {#Sec2} --------------- A literature review was conducted to answer the following review question: How can we describe suicidal patients' experiences regarding safety during psychiatric in-patient care? Methods {#Sec3} ======= The selected studies were systematically reviewed using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines \[[@CR20]\]; the articles were then synthesized using thematic analysis \[[@CR21]\] and assessed further via quality appraisal \[[@CR22]\]. The objectives, inclusion criteria, analysis methods and search strategy were specified and documented in a protocol reviewed by the three authors prior to the database search. The authors are researchers with backgrounds in psychology (SHB), mental health nursing (KR) and safety science (SHB and KAA). Inclusion and exclusion criteria {#Sec4} -------------------------------- The eligibility criteria for inclusion in the review pertained to the following three characteristics: *Type of study*: Qualitative peer-reviewed studies in English with empirical data on patients' experiences regarding safety were eligible. *Participants*: Studies examining a sample of suicidal in-patients who were interviewed during their hospitalizations or after discharge were eligible. "Suicidal in-patients" included patients hospitalized after a recent suicide attempt, described as suicidal during hospitalization or with serious suicidal thoughts or ideations; self-harming behaviour was excluded. The final criteria related to *Setting:* Experiences regarding care in psychiatric hospital wards, including psychiatric emergency wards and psychiatric long-term in-patient care, were eligible. Studies in multiple hospital settings were included if information regarding psychiatric in-patient care experiences could be extracted. Patient experiences pertaining to outpatient clinics, community mental health care, home care, forensic psychiatric services, emergency care and medical care were excluded. Studies describing patient experiences with adverse side effects from pharmacological treatment were excluded. Studies with mixed patient samples and studies involving health care professionals' experiences were included if information regarding patient experiences could be extracted. Search strategy and study selection {#Sec5} ----------------------------------- To increase sensitivity, limitations on publication date were not imposed during the database search. The selection of databases, search terms and search methodology were determined in collaboration with a university librarian. The databases included in the systematic search were MEDLINE and the Academic Search Premier, CINAHL, SocINDEX with Full-Text and PsycINFO Ovid databases. Systematic database searches were conducted between June and December 2014 and in July 2016. Search terms were identified in relevant studies during the planning of the systematic review. The terms were selected from qualitative studies of patient experiences in mental health care and from qualitative studies of suicidal patients' experiences. All identified search terms were included to increase search sensitivity. The full electronic search strategy for PsycINFO is outlined in Additional file [1](#MOESM1){ref-type="media"}. We also screened reference lists and conducted author searches in EMBASE and Google Scholar. We systematically searched all of the above databases using the following terms: patient\* satisfaction\*, patient\* preference\*, in-patient\* experience\*, patient\* experience\*, patient\* perception\*, patient\* view\*, patient\* perspective\*, patient\* opinion\*, user\* experience\*, consumer\* experience\*, consumer participation, suicide, suicidal, feeling safe and feeling unsafe. The study selection process was conducted according to the eligibility criteria displayed in the flow diagram in Fig. [1](#Fig1){ref-type="fig"}. First, all titles were screened, and the abstracts were read by one author (SHB). Ineligible studies were excluded. Full-text articles were obtained for the eligible studies. Two authors (SHB and KR) independently assessed the full-text articles for eligibility in a standardized manner. A third author (KAA) validated the assessments. The level of agreement was generally high; however, setting was often discussed, as the studies were conducted in mixed settings. Agreement was reached by re-reading the articles to determine whether information on patient experiences with psychiatric in-patient care could be extracted from the studies in question. All authors were in agreement regarding the final inclusion and exclusion of all articles. A data extraction sheet was developed to guide study selection. Information from all full-text articles was added to the sheet. All studies were assessed based on the abovementioned eligibility criteria and colour-coded as red (no), orange (maybe) or green (yes).Fig. 1PRISMA (2009) flow diagram Synthesis of results {#Sec6} -------------------- Thematic analysis, as proposed by Thomas and Harden \[[@CR21]\] and Braun and Clarke \[[@CR23]\], was used to facilitate the synthesis of the results of the included studies. The thematic synthesis consisted of two stages. The first stage entailed coding the text "line by line", condensing the meaning units and developing descriptive themes. An inductive approach was used in which the descriptive themes remained close to the original findings of the studies \[[@CR23]\]. The second stage developed relationships between the descriptive themes and patient safety to generate analytical themes \[[@CR21]\]. Thematic mapping was used to identify relationships between meaning units, descriptive themes and analytical themes \[[@CR21], [@CR23]\]. Connections between patients' needs, expectations, experiences, reported outcomes (such as experiencing increased or decreased suicidal behaviour) and use of the term 'safety' were studied in the analytical stage. Coding and preliminary theme development were conducted by one author (SHB) and reviewed by all three authors. The analysis yielded 83 meaning units, nine descriptive themes and three analytical themes ("Connection", "Protection" and "Control"). Forty-nine of the 83 meaning units were found in the "Connection" theme, which was thus considered the most comprehensive theme. An example of theme condensation is presented in Table [1](#Tab1){ref-type="table"}. A full overview of the meaning units and themes is provided in Additional file [2](#MOESM2){ref-type="media"}.Table 1Example of theme condensationExample of extracted dataMeaning unitsDescriptive themeAnalytical themeLack of acknowledgment from observers; these perceptions sometimes overlapped with perceptions of a lack of empathy. Such behaviors included observers' reading books, appearing distracted or uninterested in the participant, and acting like the participant was a burden \[[@CR40]\].61. Lack of observer support manifests as lack of empathy and acknowledgementReceiving support from the observersProtectionFeelings of objectifications in formal observation without interpersonal engagement...It's a scary thing going somewhere where you feel like you're isolated and locked away. (Claire)...Being watched like that; it's freaky...a bit invasive...that separation, that 'us and them'. It's a bit tricky. (Kate) \[[@CR26]\].62. Feeling objectified and detached without observer support"They don't care. You get that feeling quite often. It just kind of supports that hopeless kind of feeling that life isn't worth living and nobody cares about anything." Such encounters did little to alleviate hopelessness, and six participants noted that they increased their anxiety or aggravated their dysphoria \[[@CR40]\].63. Feeling objectified increases stress and hopelessness Two authors (SHB and KR) independently assessed the methodological quality of the included studies and rated the studies based on Malterud's \[[@CR22]\] checklist for qualitative research. Malterud's guidelines for assessing qualitative studies and an example of a scored article are provided in Additional file [3](#MOESM3){ref-type="media"}. An overview of the quality assessment of the included papers is presented in Additional file [4](#MOESM4){ref-type="media"}. Results {#Sec7} ======= Study selection {#Sec8} --------------- The study selection process utilized the PRISMA guidelines \[[@CR20]\] (Fig. [1](#Fig1){ref-type="fig"}) and identified a total of 1,097 records through database searches. Additional searches yielded 29 records. After removing duplicates, the remaining 984 records were screened. Title screening and abstract reading resulted in the exclusion of 904 records that did not meet the eligibility criteria. Eighty full-text articles were read, and relevant information was extracted and entered into the information sheet, assessed according to the inclusion criteria and coded (yes/maybe/no). Sixty records were excluded for not meeting the inclusion criteria, and we ultimately included the remaining 20 studies in the systematic review (Table [2](#Tab2){ref-type="table"}).Table 2The list of included studiesAuthor/year/originAimSampleSettingData collection methods and analysisKey points related to in-patient careVatne & Nåden, 2016 \[[@CR38]\]. NorwayTo develop a deeper understanding of suicidal patients in the aftermath of suicide attemptsTen patients considered with serious suicidality after a suicide attempt. Non-psychotic. Interviewed after suicide attempt.Two emergency psychiatric wards and one crisis resolution team.Semi-structured interviews. Analysed using thematic analysis inspired by Braun and Clarke. Gadamerian hermeneutic approach.• Connectedness, someone who cares\ • Hospital admission important for staying alive\ • Support from family and friendsLees, Procter and Fassett, 2014 \[[@CR26]\]. AustraliaTo explore the experiences and needs of mental health-care consumers who had a suicidal crisis (shortened).Nine patients recovered from a recent suicidal crisis where they received mental health in-patient care.Setting not specified. Experiences of psychiatric in-patient care are described.In-depth, semi-structured interviews collected as part of a larger multi-method study. Analysed with a constant comparative method and classical content analysis.• Therapeutic engagement central to quality of care\ • Isolation, loss of control, objectificationMontross Thomas et al., 2014 \[[@CR28]\]. USATo better understand suicide experiences from the perspective of patients diagnosed with serious mental illness.23 patients hospitalized after a suicide attempt. Diagnosed with serious mental illness. Interviewed after discharge.Veterans Affairs Hospital, mental health program.Qualitative interviews with audio/videotaping. Analysed using van Manen's phenomenological framework.• Need for clinicians' empathy, compassion and listening skills\ • Addressing problems underlying suicide attemptVatne & Nåden, 2014 \[[@CR32]\]. NorwayTo explore the experiences of being suicidal and encounters with health care personnel.Ten patients considered seriously suicidal. Psychosis excluded. Interviewed after suicide attempt.Psychiatric emergency ward, sub-emergency psychiatric wards and one crisis resolution team.Semi-structured interviews. Analysed using thematic analysis inspired by Braun and Clarke. Gadamerian hermeneutic approach.• Openness and trust\ • Someone who addresses the matter\ • Being met on equal terms, humiliatedCutcliffe et al., 2012a \[[@CR41]\]. Unknown origin.To better understand the observed increased risk for suicide following discharge from an in-patient psychiatric service. Key theme one.20 patients admitted to the hospital with suicidal ideation and/or a lifetime history of suicidal behaviour. Interviewed after discharge.In-patient psychiatric service.Hermeneutic interviews. Analysed using van Manen's phenomenology.• Anxiety to go back to life without having a sense of control\ • Need to be involved in discharge planningCutcliffe et al., 2012b \[[@CR42]\]. Unknown origin.To better understand the observed increased risk for suicide following discharge from an in-patient psychiatric service. Key theme two.20 patients admitted to the hospital with suicidal ideation and/or a lifetime history of suicidal behaviour. Interviewed after discharge.In-patient psychiatric service.Hermeneutic interviews. Analysed using van Manen's phenomenology.• Patients still suicidal at discharge\ • Disorientation concerning what to do with their life\ • Need for post-discharge supportPavulans et al. 2012 \[[@CR27]\]. SwedenTo explore the experience of being suicidal, including a suicide attempt, and identify possible implications for health care professionals.Ten patients interviewed after a suicide attempt while hospitalized in a psychiatric ward.Psychiatric in-patient care at one university hospital.Semi-structured interviews. Analysed using van Manen's phenomenology and qualitative content analysis.• Being in need of control\ • Re-establish control before the point of no return\ • Control related to problem-solving and insightVatne & Nåden, 2012 \[[@CR29]\]. NorwayTo explore experiences of persons after a suicide crisis or a recent suicide attempt.Ten patients considered seriously suicidal. Psychosis excluded. Interviewed after suicide attempt.Psychiatric emergency ward, sub-emergency psychiatric wards and one crisis resolution team.Qualitative interviews. Analysed using thematic analysis. Gadamerian hermeneutic approach.• Losing touch with the world\ • Someone to see, listen and understand\ • Desperation increases with involuntary hospitalizationHolm & Severinsson, 2011 \[[@CR31]\]. NorwayTo explore how recovery processes facilitate changes in suicidal behaviour in women with borderline personality disorder.13 patients with suicidal behaviour. Borderline personality disorder.Recruited from different settings within mental health. Experiences of psychiatric in-patient care were described.In-depth interviews. Data analysed with thematic analysis.• *Changing suicidal behavior by feeling confirmed, safe, and trusted.*Cutcliffe et al, 2006 \[[@CR36]\]. EnglandTo determine if psychiatric/mental health nurses provide meaningful caring responses to suicidal people, and if so, how was it achieved.20 patients with experiences from a serious suicide attempt.Crisis care in emergency psychiatric services.Semi-structured interview. Data analysed with constant comparative method. Glaserian grounded theory approach.• Reconnecting the person with humanity\ • Guiding the individual back to humanity, learning to liveSun, et al 2006b \[[@CR25]\]. TaiwanPresentation of a nursing care theory developed to guide the care given to people with suicidal ideas and those with a previous suicide attempt.15 patients with either suicidal ideas or attempted suicide. Interviewed while hospitalized.Psychiatric hospital ward.Semi-structured interviews and participant observation. A grounded theory approach.• Safe and compassionate care giving via the therapeutic relationshipSun et al, 2006a \[[@CR24]\]. TaiwanTo investigate nurses' and patients' perceptions of psychiatric wards (the context of care) and the professionals' response (the intervening conditions) that may impact the delivery of suicidal nursing care.15 patients with either suicidal ideas or attempted suicide. Interviewed while hospitalized.Psychiatric hospital ward.Semi-structured interviews and participant observation. A grounded theory approach.• Protective environment\ • Access to lethal items\ • Group support, spiritual supportTalseth, Gilje & Nordberg, 2003 \[[@CR30]\]. NorwayTo describe a process of consolation revealed by two suicidal patients' experiences.Two patients. Interviewed after a suicide attempt (from the Talseth et al., 1999 \[[@CR34]\] study).Psychiatric hospital ward.Qualitative interviews. Phenomenological hermeneutic study inspired by Ricoeur's philosophy.• Vulnerability and deep despair\ • Closeness\ • Connection\ • The dialogue with HCPsWiklander, Samuelsson, & Åsberg, 2003 \[[@CR33]\]. SwedenTo extract and analyse the interview data concerning experiences of shame.13 patients with experiences from attempted suicide. Interviewed after discharge.Specialized psychiatric in-patient care.Qualitative semi-structured interviews. Transcripts analysed using qualitative methods (not specified).• Sensitive to attitudes and behaviours of HCPs\ • Shame reactions related to aspects of careTalseth, Jacobsson & Nordberg, 2001 \[[@CR39]\]. NorwayTo illuminate the experience of being treated by physicians.21 patients expressing the wish to die or attempted to commit suicide. Interviewed while hospitalized.Psychiatric emergency wards, psychiatric sub-emergency wards and one psycho-geriatric ward.Qualitative interviews interpreted using a phenomenological hermeneutic approach inspired by Ricoeur's philosophy.• Need for confirmation in interactions with physiciansSamuelsson et al., 2000 \[[@CR35]\]. SwedenTo describe the attempted suicide patients' perceptions of receiving specialized in-patient psychiatric care.18 patients. Interviewed after a suicide attempt near the time of discharge.Specialized psychiatric in-patient care.Qualitative interviews. Analysed for qualitative content using methods inspired by Burnard.• Perception of care and caregivers, a sense of security\ • Confirmation and lack of confirmation\ • Commitment and respectCardell & Pitula, 1999 \[[@CR40]\]. USATo explore patients' experience of constant observation to determine whether they derived any therapeutic benefits beyond the intended protective benefit.20 patients placed under constant observation for suicidality.Psychiatric hospital ward and a general medical centre with a psychiatric in-patient unit.Extensive in-depth interviews. Analysis of themes consistent with Hutchinson's recommended management of grounded theory data.• Constant observation not merely a protective intervention, but with therapeutic potential.\ • Need for engaged and supportive observersFletcher, 1999 \[[@CR43]\]. UKTo explore the perceptions of staff regarding the constant observation of a suicidal patient in mental health settings.24 patients at risk for suicide, constantly observed for at least 48 h.Acute psychiatric hospital.Ethnographic study with participant observation and semi-structured interviews. Data transcribed onto cards and subjected to content analysis.• Patients' negative feelings of being under constant observation related to staff actionsMcLaughlin, 1999 \[[@CR37]\]. UKTo explore psychiatric nurses' and patients' opinions regarding the care offered to suicidal patients and how the care for suicidal patients could be improved.17 patients admitted for depression, suicidal ideation or overt suicidal behaviour.Three psychiatric hospital wards.Observation and semi-structured interview. Data analysed using content analysis by Field and Morse.• The need to address difficulties\ • Help with problem-solvingTalseth et al., 1999 \[[@CR34]\]. NorwayTo illuminate the meaning of suicidal psychiatric in-patients' experiences of being cared for by mental health nurses.21 patients admitted with suicidal ideations or after a suicide attempt.Psychiatric emergency wards, psychiatric sub-emergency wards and one psycho-geriatric ward.Qualitative narrative interviews. A phenomenological--hermeneutic method inspired by Ricoeur used in the data analysis.• Being confirmed\ • Lack of confirmationAbbreviations: *HCP* health care professional Study characteristics {#Sec9} --------------------- The review consisted of 20 articles published between 1999 and 2016. The patients' ages ranged from 16 to 63 years. The most frequently occurring diagnoses in the sample were affective disorders, of which major depression was the most prevalent, followed by schizophrenia spectrum diagnoses and personality disorders. Patients reported different experiences and needs depending on their symptoms and level of functioning; however, these parameters could not be analysed because of the presence of mixed samples. All patients had experienced suicidal crises, and the majority had attempted suicide prior to hospitalization. The studies originated primarily from Western mental health care settings, with the exception of studies by Sun et al. \[[@CR24], [@CR25]\], which were conducted in Taiwan. Themes representing patients' experiences regarding safety {#Sec10} ---------------------------------------------------------- The results of the 20 studies were synthesized and organized under analytical and descriptive themes (Table [3](#Tab3){ref-type="table"}). The results of this synthesis are described in greater detail in the following text.Table 3Analytical and descriptive themesAnalytical themeDescriptive themeConnectionMeeting someone who cares\ Receiving a confirmation of feelings\ Being acknowledged as a human beingProtectionBeing protected from death\ Receiving support from the observersControlGaining insight\ Coping with difficulties and symptoms\ Attaining discharge readiness ### Connection {#Sec11} The "Connection" theme illustrates how connections with health care professionals were vital for patient recovery and feelings of safety. A lack of connection was also experienced by the patients and had potentially fatal consequences. The sample of suicidal patients included in this review reported multiple and diverse causes of their suicidal crises \[[@CR26], [@CR27]\], but all patients experienced feelings of overwhelming suffering and increased vulnerability \[[@CR27]--[@CR31]\]. Patients experienced increased emotional sensitivity regarding how they were perceived and approached by health care professionals, and this sensitivity affected their perceptions of themselves, their recent suicide attempt, their therapeutic relationships \[[@CR26], [@CR32], [@CR33]\] and their feelings of safety in the hospital \[[@CR31], [@CR34], [@CR35]\]. Patients' connections with health care professionals enabled them to feel valued as human beings by *meeting someone who cares*; to feel understood by *receiving a confirmation of feelings;* and to feel respected and trusted by *being acknowledged as a human being.* Meeting someone who cares {#Sec12} ------------------------- Suicidal patients expressed feeling lonely, being alone with their despair, being separated from the external world and feeling a need to be connected with others \[[@CR28]--[@CR30], [@CR34]\]. A sense of being cared for could be achieved by meeting the patient's basic needs, such as bodily contact, fresh air, food, hygiene, sleep and rest \[[@CR34]\]. Patients also felt cared for when they engaged with health care professionals who were active and empathetic listeners, who spent time with them, and who showed interest in them as well as compassion for their situation \[[@CR26], [@CR28], [@CR34], [@CR36]--[@CR38]\]. These interpersonal interactions and the physical presence of the health care professionals helped patients feel that they were valuable \[[@CR30], [@CR34], [@CR39]\] and that they mattered and belonged in the world \[[@CR30], [@CR36]\]; these feelings reduced their suicidal ideations \[[@CR36]\] and made them feel safe in the psychiatric ward \[[@CR34], [@CR35]\]. Cutcliffe (\[[@CR36]\], s. 797) described this recovery process as a "re-connection with humanity" driven by connecting with and feeling cared for by nurses. Some patients felt that their health care providers had neither time nor compassion for them \[[@CR25], [@CR34], [@CR37]\], and these feelings had potentially fatal consequences. These patients experienced that their health care providers spent little time with them because the providers were busy performing other tasks or were interrupted during patient visits. Some patients experienced having no one to talk to, feeling ignored or feeling that they were being stored away as though they were an object \[[@CR34], [@CR39]\]. When met with a lack of interest and disengagement from health care professionals, patients lost confidence in their providers \[[@CR34]\], refrained from seeking help and felt unsafe in the ward \[[@CR35]\]. The experience of being isolated and alone on the ward raised feelings of hopelessness and worthlessness \[[@CR39]\]. Some patients felt redundant and started to plan ways to take their lives on the ward \[[@CR34]\]. Receiving a confirmation of feelings {#Sec13} ------------------------------------ Patients indicated that they needed someone who could listen to and understand their story and situation \[[@CR29], [@CR32], [@CR34]\] and provide confirmation of their feelings \[[@CR24], [@CR34], [@CR36]\]. They also expressed a need to be taken seriously in their suffering, to be allowed to express their feelings \[[@CR33]--[@CR35]\] and to be able to talk about their suicidality \[[@CR28], [@CR32]\]. The patients positively described their experiences being asked directly about their suicidal thoughts and plans, as they longed for opportunities to talk about difficult questions \[[@CR32]\]. Patients felt confirmed when they perceived that their mental health providers understood their situation and their need to step away from the demands of their lives \[[@CR33]\] and supported their need for hospitalization \[[@CR35]\]. The quality of the patient-physician relationship depended on patients' experience of this confirmation, as it enabled them to feel safe and understood \[[@CR34], [@CR36]\] and mitigated the despair and shame elicited by their suicide attempts \[[@CR30], [@CR33], [@CR34]\]. Patients experienced a lack of confirmation when health care professionals denied their feelings, neglected their illness, diverged from topics that the patients wanted to address, did not address difficult feelings \[[@CR33]--[@CR35]\], merely emphasized their positive resources \[[@CR32]\], or did not provide adequate or empathetic responses when they disclosed sensitive issues \[[@CR33]\]. Some patients reported that their health care professionals did not spend sufficient time with them to properly understand the reasons for their suicide attempts or that the professionals avoided talking about their suicide attempt \[[@CR34], [@CR39]\]. Other patients felt that their nurses were concerned only about their symptoms or the effects of their medications and thus did not allow them opportunities to share their thoughts and feelings \[[@CR34], [@CR39]\]. Patients perceiving these types of non-responsive attitudes with respect to sensitive or important topics experienced worsening feelings of shame and humiliation \[[@CR32], [@CR33]\] that exacerbated their suicidal ideations and, in some cases, resulted in subsequent suicide attempts \[[@CR32], [@CR35]\]. Being acknowledged as a human being {#Sec14} ----------------------------------- Patients stated that it was important for providers to meet them on equal ground in order for them to feel acknowledged as a human being \[[@CR26], [@CR33], [@CR34]\]. This meant being treated non-judgementally \[[@CR24], [@CR28], [@CR33], [@CR36]\] -- being empowered and understood as individuals rather than as objects, cases or diagnoses \[[@CR30], [@CR31], [@CR33]\]. When the patients felt that they were achnowledged as a human being, they were able to feel trusted, respected, and safe in the ward and were thus receptive to help \[[@CR26], [@CR30], [@CR31], [@CR35]\]. Through these feelings, patients regained their sense of human dignity and thereby felt that it was worthwhile to be alive \[[@CR26], [@CR33], [@CR36]\]. Not being seen as a human being was related to feelings of inequality \[[@CR32], [@CR34]\], e.g., patients whose providers overused medical jargon or limited their visits to discussions about medications and diagnoses \[[@CR31], [@CR34]\], as well as the feeling of being punished by health care professionals through the use of ward rules, verbal expressions or body language to exert their power \[[@CR33]\]. Not being seen as a human being was also related to feelings of disempowerment, e.g., being talked about when they were present \[[@CR32]\], not being informed about ward routines \[[@CR33]\] or who their primary nurse was \[[@CR25], [@CR37]\], not being informed about their own arrangements \[[@CR35]\], or experiencing that their opinions, information or histories were not considered important \[[@CR32], [@CR39]\]. Suicidal patients with borderline personality disorder experienced that they were able to recover by experiencing feelings of safety and trust during their encounters with health care professionals. However, when treated as inferior, the patients did not feel safe in the hospital \[[@CR31]\]. ### Protection {#Sec15} The "Protection" theme pertained to patients' experiences when under constant observation and their struggles to feel safe from themselves and their invasive suicidal impulses \[[@CR31], [@CR40]\]. Patients felt safe from themselves and their suicidal impulses and *protected from death* during constant observation. *Receiving support from the observers* was the most important aspect during constant observation, as patients lacking these relationships felt detached and objectified, and their anxiety and symptoms worsened \[[@CR26], [@CR40]\]. Being protected from death {#Sec16} -------------------------- During constant observation, some patients experienced a state of mind in which they continually searched for available means to attempt suicide. Some experienced feeling powerless against their suicidal thoughts, whereas others experienced command hallucinations related to suicide \[[@CR40]\]. Patients perceived constant observation as a means of altering their suicidal ideations and self-destructive behaviour. Patients considered this practice life-saving because of the presence of vigilant observers, the limited availability of objects to use for suicide attempts, the passage of time \[[@CR40]\] and the distraction and escape from the outside world \[[@CR24]\]. Patients struggled to feel safe from themselves and to assume responsibility for their own lives when they lacked protection during acute suicidal crises \[[@CR31], [@CR41]\]. Adequate protection was also related to their perceptions of the hospital as a safe place \[[@CR41], [@CR42]\]. Accordingly, patients who easily found ways to attempt suicide in the ward and those who did not receive safety searches or monitoring often felt unsafe in the hospital \[[@CR25]\]. However, one patient explained that not being able to end his life actually increased his suffering, as he believed that being able to end his suffering in the event that it became unbearable was a source of comfort that helped him cope with his situation \[[@CR29]\]. Patients experienced a lack of freedom and privacy under constant observation \[[@CR25], [@CR40], [@CR41]\], and most were happy when it was discontinued because of its invasiveness. Some patients even lied about their suicidality to discontinue their observation \[[@CR40]\]. Receiving support from observers {#Sec17} -------------------------------- Cardell and Pitula \[[@CR40]\] concluded that the relationship with care providers was at the heart of constant observation and highlighted the importance of patients having supportive observers as opposed to impersonal and detached observers. Patients experienced observer support as vital for decreasing their suicidality during constant observation \[[@CR36], [@CR40]\], as these relationships facilitated reduced suicidality. It was important for the observers to have an optimistic attitude, encourage problem-solving, enable patients to gain self-esteem, acknowledge patients as unique and meaningful human beings \[[@CR40]\], and try to understand patients by talking with them about their feelings \[[@CR43]\]. By interacting with supportive observers, the patients internalized what the observers projected and felt worthy as human beings and thus worthy of being alive \[[@CR40]\]. Some patients experienced a lack of acknowledgement and a lack of interpersonal engagement under constant observation, in which the observers appeared disinterested or distant or behaved as though their patients were a burden \[[@CR26], [@CR40]\]. When attempting to start a conversation, the observers either did not respond or displayed hostile facial expressions, which was perceived as a lack of empathy \[[@CR40]\]. Lees \[[@CR26]\] observed that having minimal interpersonal engagements limits the therapeutic potential of interventions, such as formal observation and medications. Patients deprived of interpersonal engagement felt objectified and separated from their health care professionals \[[@CR26]\] or that nobody was there for them or acknowledged their existence \[[@CR40]\]; these experiences exacerbated their feelings of anxiety and hopelessness and supported their perceptions that nobody cared about them and that their lives were not worth living \[[@CR40]\]. ### Control {#Sec18} The "Control" theme involved patients' need to re-establish a feeling of control over their lives \[[@CR27]\]. Suicidal patients experienced a sense of not being in control, a desire to regain control and a sense of losing control during suicidal crises \[[@CR26], [@CR27]\], which they often described as periods of overwhelming emotional suffering that left them unable to cope with life \[[@CR27]--[@CR31]\]. Patients whose health care professionals enabled them to *gain insight* and *cope with difficulties and symptoms* were able to regain control of themselves. This sense of control was important for attaining *discharge readiness* and feeling safe from themselves. Patients without this sense of control experienced increased suicidal thoughts. Gaining insight {#Sec19} --------------- Gaining insight into their illnesses enabled patients to regain control after their suicide attempt \[[@CR27], [@CR31], [@CR41]\]; patients who understood themselves were able to address the difficulties in their life without attempting suicide \[[@CR27]\] and also felt safer from themselves \[[@CR31]\], which helped them feel in control of their lives \[[@CR41]\]. Coping with difficulties and symptoms {#Sec20} ------------------------------------- Patients felt that a sense of control could be achieved by being able to manage difficulties and by learning new problem-solving and help-seeking skills, as well as by receiving adequate treatment for mental health problems and obtaining assistance for social and economic problems. Patients who were able to manage difficulties were able to visualize a way back to their lives \[[@CR27], [@CR36], [@CR37]\]. Variations in coping strategies related to different support and independence needs were described, as some patients expressed a need for others to "fix" their problems, some expressed a need for a break from any type of demand, and others emphasized a need to strengthen their self-efficacy to more effectively cope with their life situations \[[@CR27], [@CR33], [@CR41]\]. Some patients experienced that their problems were best addressed through one-on-one conversations with health care professionals \[[@CR35], [@CR37]\], whereas others preferred group support \[[@CR25], [@CR28], [@CR37]\], spiritual support \[[@CR25]\], or family or friend support \[[@CR25], [@CR38]\]. Patients needed health care professionals who could adapt to their needs and coping strategies \[[@CR38]\]. Attaining discharge readiness {#Sec21} ----------------------------- Patients expressed the expectation that their admission would result in a cure for or solution to their problems; this belief represented a major disconnect between patients' expectations and the treatment provided during short-term hospitalization \[[@CR41]\]. At discharge, some patients felt that their problems were unsolved \[[@CR37]\] and that they lacked the skills and tools for coping with their problems and their unchanged circumstances; this feeling resulted in increased distress and suicidal thoughts \[[@CR41], [@CR42]\]. At discharge, patients experienced unaddressed problems related to their suicidality \[[@CR32], [@CR37], [@CR41]\]. Thus, they did not feel prepared for discharge and feared that leaving the hospital would lead to subsequent suicide attempts \[[@CR41]\]. These patients experienced the feeling that the system was failing them and indicated that they did not know where to seek support in the event that formal mental health services could not help \[[@CR41]\]. Patients' sense of control was strengthened by having a post-discharge support plan and by being able to contact the ward after discharge if necessary \[[@CR27], [@CR35], [@CR41]\], as well as by being prepared for the upcoming change in their freedom by feeling empowered and supported prior to discharge \[[@CR31], [@CR41]\]. Thus, it was important for patients to be allowed to participate in decision making regarding their post-discharge support, as this reduced their fears and anxieties at discharge when being sent "back to the lion's den" (\[[@CR41]\], s. 24). Discussion {#Sec22} ========== This paper posed the following review question: "How can we describe suicidal patients' experiences regarding safety during psychiatric in-patient care?" Suicidal patients' experiences with safety during psychiatric in-patient care were described in 20 studies that addressed whether their needs were met during their hospitalization. This review argues that patients define safety in terms of "feeling safe" and that connection, protection and control play vital roles in their safety-related experiences. Fulfilment of these needs are experienced as essential for recovery from their suicidal crises, in addition to the ability to feel safe during their encounters with health care professionals and to feel safe from their suicidal impulses. When experiencing unmet needs, the patients not only felt unsafe but also exhibited increased suicidal thoughts and feelings. For some patients, these experiences were characterized as triggers for another suicide attempt. The patient experiences discussed in our review are related to the relational and emotional aspects of hospital care and are consistent with the findings of other studies regarding patient experiences \[[@CR10], [@CR13]\]. Our findings also resonate with those of psychiatric in-patient care studies, in which patients identified psychological safety as the most common safety issue \[[@CR44]\]. The *connection* and *protection* components discussed herein emphasize the importance of the therapeutic relationship in not only establishing feelings of safety but also optimizing patient outcomes, such as those related to increases or decreases in patient suicidality. The suicidal patients in this review addressed the vital importance of the therapeutic relationship in helping patients both feel safe and be safe. These findings are consistent with those of studies highlighting the therapeutic alliance in effective suicidal patient assessments and management \[[@CR45]--[@CR47]\] and studies identifying the staff--patient relationship as important to patients' feelings of safety \[[@CR44], [@CR48], [@CR49]\]. Poor staff-patient relationships were found to play key roles in preventable suicides and were attributed to poor communication and relationship quality \[[@CR50]\]. This review highlights the importance of addressing the control component to enable suicidal patients to feel and be safe after discharge from the hospital ward. The *control* component demonstrates the importance of supporting external and internal processes that help suicidal patients feel a sense of control and of understanding the individual from an ideographic point of view. Consistent the results of this review, Connell \[[@CR51]\] found that, for mental health patients, a sense of control was linked to feelings of safety. The level of desired dependence or independence varied according to each patient's current circumstances and differed over time. Undrill \[[@CR52]\] stated that psychiatric risks should be perceived as manifestations of suffering. Thus, maintaining high-quality core activities during care and acknowledging suicidal patients' suffering through trust and therapeutic closeness should be the primary methods of addressing patients' suicide risk and improving their safety. In accordance with Undrill's \[[@CR52]\] findings, our review indicates that ensuring patient safety entails addressing patients' therapeutic needs and psychological safety in addition to their physical safety. Although integrating relational and technical patient safety measures into psychiatric care is challenging \[[@CR53], [@CR54]\], safety is dependent on this integration. The link between feeling safe and being safe is vital for suicidal patients; suicidal patients' physical safety cannot be ensured if they do not feel safe. A system that is designed to physically prevent patients from committing suicide but that neglects their need for a connection with health care professionals may not be successful, as patients may exhibit increased suicidality despite the implementation of procedures to prevent this outcome. Furthermore, patients may not only feel unsafe, but they may also be unsafe because of an increased suicide risk imposed by the complex dynamics between emotionally vulnerable patients and their health care professionals. A broader perspective regarding patient safety that integrates therapeutic needs, psychological safety and physical safety is therefore needed. Limitations {#Sec23} ----------- There were a few limitations to this review. There is a risk of missed studies due to a lack of common nomenclature. To address this limitation, the search terms and strategy were designed to increase the sensitivity to relevant literature. Furthermore, the systematic search included only published peer-reviewed studies, resulting in the exclusion of possibly valuable grey literature and unpublished papers. Although there is a risk of reviewer bias, efforts were made to minimize this bias by applying systematic search methods and by following the PRISMA guidelines for systematic reviews. The review was limited to studies regarding psychiatric in-patient care. Studies examining the experiences of suicidal patients when receiving emergency care and outpatient treatment were excluded, as were studies regarding the experiences of patients without access to psychiatric care. These types of studies should be included in future reviews that aim to explore patient pathways and continuity of care, as poor continuity of psychiatric care has been associated with preventable suicides \[[@CR50]\]. Implications for research and practice {#Sec24} -------------------------------------- The literature included a diverse group of patients characterized by suicidal behaviour. These different patient groups may present distinct experiences, thus limiting the general understanding of suicidal patients as a group. To account for the diversity of patients in suicidal crises, more studies involving the elderly, youths, low-income countries and non-Western health care settings are necessary. There is also a need to explore the experiences of suicidal patients in different diagnostic groups, such as suicidal in-patients with/without psychotic symptoms and patients with/without chronic suicidality or borderline personality disorder. The similarities and differences between the experiences of suicidal patients and non-suicidal patients must be elucidated to identify the generic versus group-specific characteristics that determine patient safety in psychiatric care. Additionally, patients may have different needs during different stages of their suicidal crises. For example, Rise et al. \[[@CR55]\] observed that patients indicated different safety-related needs depending on their symptoms. However, this distinction was not addressed in the studies included in our review and represents a direction for further research. We recommend the following changes regarding in-patient care practices for suicidal patients based on the results of our systematic review:Patient experiences should be considered an integral part of suicidal patients' safety to guide clinical practice and the design of patient safety measures.Suicidal patients' need for connection with health care personnel indicate that the relational component of patient safety is considered the most vital aspect of care and should thus be integrated into measures such as constant observation, suicide risk assessments, clinical supervision, ward therapeutic environments and encounters with health care personnel groups.Suicidal patients' need for protection highlights the importance of constant observation in suicidal crises and the need for skilled professionals in close proximity to patients.Suicidal patients' need for control emphasizes the need for therapeutic interventions that increases the patient's insight and problem-solving skills as well as shared decision making regarding treatment plans, crisis plans, support systems and post-discharge follow-up activities. Conclusion {#Sec25} ========== Our review addresses the importance of having a broader view of safety for suicidal patients rather than merely understanding safety in technical terms. When considering suicidal patients' experiences, safety appears to be related to more than the absence of suicide risk and the need for physical protection. Safety for the suicidal patient is highly dependent on patients' perceptions of their connections with health care professionals, the fulfilment of their needs during care and their psychological safety. To be safe, patients must feel safe through their *connections* with health care professionals; they must be *protected* against their suicidal impulses and they must have a sense of *control* over their lives. These components should serve as the basis of future efforts designed to understand the ontology of safety for suicidal patients during in-patient psychiatric care. Additional files {#Sec26} ================ Additional file 1:Search strategy for PsychINFO. (DOCX 12 kb) Additional file 2:Table of themes and meaning units. (DOCX 24 kb) Additional file 3:Malterud's \[[@CR22]\] "Guidelines for authors and reviewers of qualitative studies -- an example of checks and scores". (DOCX 15 kb) Additional file 4:An overview of the included studies' scores: high, middle or low. (DOCX 16 kb) We thank all the psychologists, doctors, nurses and patients in the Division of Psychiatry at Stavanger University Hospital who provided their reflections during the manuscript preparation. Funding {#FPar1} ======= The Western Norway Regional Health Authority funded this project under grant agreement no. 911846. Availability of data and materials {#FPar2} ================================== All data generated or analysed during this study are included in this article. The additional files include supplementary information. The protocol and a matrix of the data analysis of the excluded studies are available from the corresponding author upon reasonable request. Authors' contributions {#FPar3} ====================== All three authors (SHB, KR and KAA) devised the search strategy and eligibility criteria. SHB conducted the database searches and the primary exclusion of studies. SHB and KR performed the eligibility assessments of the full-text articles. SHB completed the synthesis of the results, and KR and SHB conducted the quality appraisal. All three authors (SHB, KR and KAA) validated the results synthesis and quality appraisal. SHB drafted the manuscript, and KR and KAA provided critical revision of intellectual content. All authors approved the final manuscript. Authors' information {#FPar4} ==================== SHB is a Ph.D. scholar in risk management and societal safety at the University of Stavanger and a psychologist at Stavanger University Hospital. KR has a Ph.D. and is a researcher in Psychiatric Nursing at Stavanger University Hospital. KAA is a professor and head of the research group "Quality and safety in health care systems" in the Department of Health Studies at the University of Stavanger. Competing interests {#FPar5} =================== The authors declare that they have no competing interests. Consent for publication {#FPar6} ======================= Not applicable. Ethics approval and consent to participate {#FPar7} ========================================== Not applicable.

{ "pile_set_name": "PubMed Central" }

Introduction ============ Aneurysmal subarachnoid hemorrhage (SAH) is a devastating disease with substantial morbidity and mortality. Prognostic modeling is an important instrument to identify high-risk patients in both clinical practice and research settings. Recently, a prognostic model to predict 60-day case fatality after aneurysmal SAH was developed with data from the International Subarachnoid Aneurysm Trial (ISAT) \[[@B1]\]. Our aim was to externally validate this model in a retrospective cohort of consecutive SAH patients. Methods ======= We included consecutive aneurysmal SAH patients admitted to one university hospital between October 2007 and October 2011. Exclusion criteria were: age \<18 years, hospital admission \>28 days after SAH, nonaneurysmal SAH, explicit objection by the patient to view the medical data and missing data on 60-day case fatality. The model\'s predictors were age, maximum lumen size of the aneurysm, Fisher grade and World Federation of Neurological Surgeons (WFNS) grade. Two versions of the model were validated: one with WFNS grade scored on admission and the other with WFNS grade assessed at the time of treatment decision, as a proxy to WFNS grade at randomization used in the ISAT. The outcome was 60-day case fatality. Model performance was assessed by studying discrimination, expressed by the area under the receiver operating characteristic curve (AUC), and calibration. Results ======= A total of 307 patients were included in the validation cohort. The observed 60-day case fatality rate was 30.6%. Discrimination was good, and was considerably better for the model with WFNS grade at treatment decision (AUC = 0.89) compared with the model with WFNS grade on admission (AUC = 0.82). Calibration was poor, with mean predicted probabilities of 17.0% for the model with WFNS grade on admission and 17.7% for the model with WFNS grade at the time of treatment decision. Conclusion ========== Our results indicate that the ISAT prediction model is generalizable, since the model showed adequate performance in an independent, unselected cohort of aneurysmal SAH patients. The model discriminated well between patients who died and those who survived the first 60 days after SAH. Additionally, use of WFNS grade at the time of treatment decision of the ruptured aneurysm improved model performance. However, since predicted probabilities were lower than observed probabilities, the ISAT prediction model needs to be adapted before use in clinical practice.

{ "pile_set_name": "PubMed Central" }

Introduction ============ The atherogenicity of type 1 diabetes has been increasingly recognized \[[@B1]\]. Patients with diabetes show a 2- to 10-fold risk for developing atherosclerotic lesions compared with the normal population \[[@B2],[@B3]\]. Even if these complications become manifest in the adult diabetic patient, the process of vascular changes starts much earlier \[[@B4],[@B5]\]. The most significant changes in early subclinical period of atherosclerotic disease are endothelial dysfunction and increase in intima-media thickness observed in all arterial beds \[[@B6]\]. Common carotid artery intima-media thickness (CIMT), measured by high-resolution B-mode ultrasonography, is a noninvasive marker of subclinical atherosclerosis \[[@B7]-[@B9]\]. Normative values for CIMT are available for subjects aged 10 to 20 years \[[@B10],[@B11]\]. Mechanisms involved in the increased oxidative stress in diabetes include not only oxygen free radical generation due to nonenzymatic glycosylation (glycation), autooxidation of glycation products, but also changes in the tissue content and activity of antioxidant defense systems. Increased levels of the products of oxidative damage to lipids have been detected in serum of diabetic patients, and their presence correlates with the development of complications \[[@B12]-[@B18]\]. Oxidative stress plays a pivotal role in the development of diabetes microvascular and cardiovascular complications \[[@B19]\]. In the diabetic macrovascular and in the heart, this appears to be a consequence of increased oxidation of fatty acids, resulting in part from pathway-specific insulin resistance \[[@B19]\]. Nitric oxide has cellular antioxidant and pro-oxidant actions \[[@B20]\]. The endothelial nitric oxide (NO) system plays a pivotal role in vascular physiology and pathology. NO is a potent vasodilator agent with anti-hypertensive, anti-thrombotic, anti-atherogenic, and anti-smooth muscle proliferative properties \[[@B21]\]. However, high amounts of NO produced by inducible NO synthase (iNOS) and/or peroxynitrite (ONOO-), a reactive intermediate of NO with superoxide anion are involved in pro-inflammatory reactions and tissue damage as well \[[@B22]\]. Total Antioxidant Capacity (TAC) is capable of serving as a parameter to monitor diabetes in patients with type 1 DM \[[@B23]\]. A depletion of the total antioxidant capacity is associated with a higher incidence of diabetic complications \[[@B24]\]. The relation between oxidative stress markers and carotid intima media thickness in type 1 diabetic children and adolescents was not extensively studied. Hence, this study aimed at assessment of carotid intima media thickness in children and adolescents with type1 diabetes mellitus in relation to plasma nitric oxide and plasma total antioxidant capacity levels and with diabetes duration, glycemic control and microvascular complications. Patients and methods ==================== This case control study was conducted at the Pediatric Diabetes Clinic, Children's Hospital, Ain Shams University, Cairo, Egypt in the period from April 2011 to February 2012.It included 50 patients with type 1 diabetes mellitus(DM) regularly attending the clinic. They were 37 females (74%) and 13 males (36%). Their ages ranged from 6--16 years with a mean age of 9.7 ± 3.4 years. Their duration of illness ranged from 1--13 years with mean diabetes duration of 4.5 ± 3.5 years. Patients were further subdivided into two groups according to the duration of diabetes: (Group I) included 22 children and adolescents with diabetes duration of 5 years or more, (Group II) included 28 children and adolescents with diabetes duration of less than 5 years. Fifty age and sex matched healthy individuals were included as a control group. They were 14 males and 36 females. Their age ranged from 2--14 years with mean age of 9.8 ± 3.14 years. The study was approved by the Ethical Committee of Ain Shams University Faculty of Medicine. Inclusion criteria ------------------ Patients were included in the study only if they have type 1 diabetes mellitus on regular insulin therapy regularly visiting the Clinic. Exclusion criteria ------------------ Included Type 2 diabetes mellitus, hypertension, patients with malignancy, connective tissue diseases, liver dysfunction, renal dysfunction (serum creatinine \> 1.2 mg/dl), congenital or acquired cardiovascular disorders, administration of drugs other than insulin (such as oral hypoglycemics, antihypertensives, antiplatelets or lipid lowering medications, aspirin, or vitamin supplements) at the time of the study and none of them was cigarette smoker. All patients were subjected to detailed history taking, thorough clinical examination, measurement of Glycosylated Hb (HbA1c) by HPLC (high performance liquid Chromatography). Patients were considered under optimal glycemic control when their HbA1c was \< 7.5% \[[@B25]\]. Microalbuminuria was assayed using SERA-PAK immuno-microalbumin Kit (Bayer Corporation, Benedict Ave, Tarry town, NY, USA). Persistent microalbuminuria was defined when two of three samples showed urinary albumin excretion rate of 30--300 μg/mg creatinine \[[@B26]\] fasting serum triglycerides, serum cholesterol using Synchron CX7 (Brea, California, USA) \[[@B27]\]. Serum total antioxidant capacity by colorimetric method \[[@B28]\], serum NO by colorimetric method \[[@B29]\] and Carotid intimal --media thickness (CIMT) assessment using B mode ultrasonography \[[@B30]\]. Measurment of serum Total Antioxidant Capacity (TAC) by colorimetric method. The determination of antioxidative capacity was performed by the reaction of antioxidants in the sample with a defined amount of exogenously provide hydrogen peroxide (H2O2) the antioxidants in the sample eliminate a certain amount of the provided hydrogen peroxide. The residual H2O2 was determined clorometrically by an enzymatic reaction which Involves the conversion of 3,5,dichloro \_2\_ hydroxyl benzensulphonate to a colored product \[[@B28]\]. Measurement of serum Nitric oxide (NO) was done using colorimetric method. This assay determines nitric oxide concentrations based on the enzymatic conversion of nitrate to nitrite by nitrate reductase. The reaction is followed by colorimetric detection of nitrite as an azo dye product of the Griess Reaction. The Griess Reaction is based on the two-step diazotization reaction in which acidified NO2^-^ produces a nitrosating agent, which reacts with sulfanilic acid to produce the diazonium ion. This ion is then coupled to N-(1-naphthyl) ethylenediamine to form the chromophoric azo-dervative which absorbs light at 540--570 nm \[[@B29]\]. Statistical analysis -------------------- Demographic and clinical data are presented as means ± SD or proportions. Differences in continuous variables between males and females were tested with the Student *t* test for normally distributed data and the Mann--Whitney *U* test for non-normally distributed data. The χ2 test for contingency tables with different degrees of freedom was obtained to detect associations between categorical independent variables. Adjustment for multiple confounding was done using linear regression analysis with a manual backward procedure. Multivariate analyses were preceded by estimation of the correlation between potential confounders. A significance level of \< 0.05 was used. All statistical analysis was done using the SPSS software package for Windows, version 15.0 (SPSS, Chicago, IL). Results ======= -- Patients mean systolic blood pressure (SBP) was103.500 ± 18.077 mmHg and mean diastolic blood pressure (DBP) was 68 ± 10.302 mmHg. Three (6%) of our patients had SBP and DBP \> 95% percentile while 97% were within the normal ranges. -- Diabetic patients showed significant increase in nitric oxide and decrease in total antioxidant capacity level than controls (P value \< 0.001)\] (Table [1](#T1){ref-type="table"}), Figure [1](#F1){ref-type="fig"}. ![Comparison between patients and controls regarding values of nitric oxide and total antioxidant capacity.](2251-6581-12-50-1){#F1} -- NO level was higher in patients with suboptimal glycemic control than those with optimal glycemic control \[(23.38 ± 6.7 μmol/L) vs (16.33 ± 5.6 μmol/L)\] (P value = 0.017) and in patients with nephropathy than patients without nephropathy \[(22.00 ± 4.14 μmol/L) vs (16.67 ± 6.5 μmol/L)\] (P value = 0.033). -- No statistically significant difference was found in TAC level between diabetic patients with and without nephropathy or between patients with suboptimal and optimal glycemic control (p value \>0.05) -- Serum total cholesterol and triglycerides were significantly higher in studied patients than control group, although within normal range. -- Carotid intima media thickness was significantly increased in diabetic patients compared to normal controls \[(mean 0.46 ± 0.04 mm) vs (mean 0.39 ±0.02 mm)\] (P \< 0.001) (Table [1](#T1){ref-type="table"}), Figure [1](#F1){ref-type="fig"}. -- Carotid intima media thickness was significantly increased in diabetic patients with suboptimal than those with optimal glycemic control \[(0.46 ± 0.04 cm) vs (0.40 ± 0.02 cm)\] (P value 0.04), in patients with diabetes duration \> 5 years than those with diabetes duration \< 5 years \[(0.50 ± 0.03 mm) vs (0.43 ± 0.02 mm)\] (P value \<0.001)\] (Table [2](#T2){ref-type="table"}) and in diabetic patients with than patients without nephropathy \[ (0.49 ± 0.02) vs (0.40 ± 0.04)\] (P value = 0.009). -- A statistically significant positive correlation was found between carotid intima media thickness and age, duration of diabetes mellitus, systolic and diastolic blood pressure and nitric oxide (P value \< 0.001 and \<0.001, 0.002, 0.007 and 0.01 respectively) and a significant negative correlation was found between carotid intima media thickness and total antioxidant capacity (P value = 0.02) (Table [3](#T3){ref-type="table"}). No significant correlation was found between carotid intima media thickness and mean HbA1c, cholesterol or triglycerides (P value \>0.05). ###### Comparison between diabetic patients and control group regarding studied parameters **Groups** **t-test** ----------------------- ---------------- ---------------- ---------- ----------- Age (years) 9.78 ± 3.45 9.86 ± 3.14 -0.12 0.90 BMI(kg/m2) 20.120 ± 4.155 18.740 ± 3.87 t = 1.72 0.09 Cholesterol (mg/dl) 147.35 ± 38.92 120.68 ± 35.26 t = 3.5 0.0005\* Triglycerides (mg/dl) 80.86 ± 25.10 62.40 ± 20.50 t = 4.03 0.0001\* NO (μmol/L) 17.07 ± 6.36 12.57 ± 4.74 3.98 \<0.001\* TAC (mmol/L) 0.41 ± 0.29 0.87 ± 0.23 -8.87 \<0.001\* CIMT (mm) 0.47 ± 0.04 0.40 ± 0.02 9.66 \<0.001\* Gender 37/13 36/14 0.437\# 0.509 *NO* nitric oxide, *TAC* Total antioxidant capacity, *CIMT* carotid intima media thickness; \#= Chi-Square. \*Indicate statistical significance. ###### Comparison between patients with diabetes duration \<5 years and \>5 years regarding studied parameters **Patients with diabetes duration ≤ 5 years(n = 28)** **Patients with diabetes duration \> 5 years(n = 22)** **T-test** ----------------------- ------------------------------------------------------- -------------------------------------------------------- ------------ ----------- Age (years) 7.64 ± 2.80 12.50 ± 1.946 -6.92 \<0.001\* Mean HbA~1~c% 8.47 ± 1.98 9.73 ± 2.07 -2.19 0.03\* Cholesterol (mg/dl) 134.36 ± 38.34 162.50 ± 35.23 -1.94 0.07 Triglycerides (mg/dl) 79.13 ± 24.74 81.86 ± 25.54 0.38 0.7 NO (μmol/L) 16.27 ± 6.70 18.09 ± 5.89 -1.01 0.32 TAC (mmol/L) 0.45 ± 0.31 0.36 ± 0.25 1.18 0.25 CIMT (mm) 0.44 ± 0.03 0.50 ± 0.03 -7.09 \<0.001\* *NO* nitric oxide, *TAC* Total antioxidant capacity, *CIMT* carotid intima media thickness. \*Indicate statistical significance. ###### Correlation between CIMT, serum NO and serum TAC in diabetic patients **CIMT (mm)** **NO (μmol/L)** ------------------------------ ----------- --------------- ----------------- Age (years) r 0.741 P-value \<0.001\* Duration of diabetes (years) r 0.656 P-value \<0.001\* SBP (mmHg) r 0.126 P-value 0.002\* DBP (mmHg) r 0.379 P-value 0.007\* NO (μmol/L) r 0.40 P-value 0.01\* TAC(mmol/L) r -0.34 -0.53 P-value 0.02\* \<0.001\* *CIMT* carotid intima media thickness, *SBP* systolic blood pressure, *DBP* diastolic blood pressure, *NO* nitric oxide, *TAC* Total antioxidant capacity. \*Indicate statistical significance. Discussion ========== The present study showed a significant increase in serum nitric oxide (NO) level in diabetic patients than controls which came in agreement with many studies \[[@B31]-[@B33]\]. Pitocco et al. \[[@B34]\] suggested a reduced asymmetric dimethyl arginine (ADMA) inhibition of NOS as possible mechanism involved in the pathogenesis of oxidative stress in female subjects with a short duration and uncomplicated type 1 diabetes. Serum NO level was significantly increased in diabetic patients with than those without nephropathy which came in accordance to previous studies \[[@B35],[@B36]\]. Prabhakar, \[[@B37]\] reported that the enhanced NO production may contribute to hyperfiltration and microalbuminuia at early diabetic nephropathy. NO level was significantly higher in patients with suboptimal than optimal glycemic control which came in agreement with others \[[@B38]\]. A significant decrease in total antioxidant capacity level was found in patients than controls which came in agreement with several studies \[[@B39]-[@B42]\]. Lack of difference in TAC level between our studied patients with and without nephropathy is concordant with El-desoky et al. \[[@B43]\] who found that plasma TAC in patients of diabetes mellitus type I with nephropathy was not significantly decreased as compared with diabetes mellitus type 1 with no nephropathy and they concluded that TAC cannot be used to differentiate between diabetes mellitus type 1 with and without nephropathy. Serum total cholesterol and triglycerides were significantly higher in studied diabetic patients compared to control group although within the normal range. This agreed with Margeisdottir et al. \[[@B44]\] who attributed those findings to the young age of the patients. Diabetic patients showed a significant increase in CIMT mean values compared to controls. Those results came in accordance to several studies \[[@B45]-[@B47]\]. A significant increase in CIMT in patients with nephropathy compared to patients without nephropathy came in concordance with Gul et al. \[[@B47]\] suggesting that diabetic microangiopathy is related with macroangiopathy and in patients with suboptimal glycemic control compared to patients with optimal glycemic control which came in concordance with Abdelghaffar et al. \[[@B46]\]. CIMT was directly correlated with age in studied diabetic patients which agreed with many studies \[[@B46]-[@B48]\]. A strong direct correlation was found between mean CIMT and patients systolic and diastolic blood pressure which agreed with several studies \[[@B46],[@B48],[@B49]\]. The relationship between increased CIMT and blood pressure suggests that smooth muscle proliferation plays a role in the early diffuse thickening of the arterial wall \[[@B49]\]. No significant correlation was found between mean CIMT and metabolic control parameter HbA1c, this agreed with many studies \[[@B47],[@B50]\]. Data from the literature indicated that, in contrast to the functional impairment of the endothelium, structural changes are not correlated to single parameter such as the HbA1c at a young age \[[@B50]\]. A statistically significant positive correlation was found between carotid intima media thickness and nitric oxide level which agreed with Dursun et al. \[[@B51]\] who explored the relation between CIMT and oxidative stress markers in type 2 diabetic patients on maintenance hemodialysis. Zineh et al. \[[@B52]\] reported the association between NOS3 polymorphisms and arterial stiffness in children with type 1 diabetes. Insulin acts by modulating the release of vasodilator substances, such as nitric oxide and prostaglandins, from vascular endothelium, by both stimulating and inhibiting the sympathetic nervous system and by protecting smooth muscle cells in blood vessel from apoptosis induced by oxidative stress \[[@B53]\]. Thus the vasodilatory and antioxidant effects of insulin are depressed in case of insulin deficiency (i.e. type 1 diabetes) \[[@B54]\]. A negative correlation was found between CIMT and total antioxidant capacity level. Lower serum TAC levels were observed in atherosclerotic coronary artery disease patients with an inverse association with number of damaged coronary vessels \[[@B55]\]. Other research reported increased DNA damage in the nucleus of coronary cells and decreased plasma TAC level in coronary artery disease patients \[[@B56]\]. Conclusion ========== The significant elevation in nitric oxide and reduction in total antioxidant capacity in children and adolescents with type 1 diabetes mellitus together with their correlation with carotid intima media thickness may reflect the role of oxidative stress in the development of atherosclerosis in young type 1 diabetic subject. Further studies to measure other antioxidant markers like glutathione or antioxidants or markers of oxidative stress like malondialdehyde in relation to carotid intima media thickness are warranted. Abbreviations ============= CIMT: Carotid intima media thickness; DM: Diabetes mellitus; HbA1c: Glycosylated hemoglobin; NO: Nitric oxide; TAC: Total antioxidant capacity; SBP: Systolic blood pressure; DBP: Diastolic blood pressure; ELISA: Enzyme linked immune sorbent assay. Competing interests =================== The authors declare that they have no competing interests. Authors' contributions ====================== MHE supervised the work and reviewed the manuscript. RMM designed the study, analyzed the data and drafted the manuscript. OIY carried out the carotid intima media thickness assessment, contributed in data analysis and drafting the manuscript MAS carried out the laboratory studies. NAK collected the data. All authors approved the manuscript. Acknowledgements ================ We are grateful to the staff of the Diabetes Clinic and echocardiography unit, Children\'s Hospital, Ain Shams University, Cairo Egypt.

{ "pile_set_name": "PubMed Central" }

Taking into account the SARS-coronavirus 2 (SARS-CoV-2 also know as COVID-19) spreading and invasiveness in this critic period and the lack of therapeutic drugs and antiviral vaccines, it is necessary for all of us researchers to discuss about new therapeutic devices and quickly define if agents, other than that conventionally used, could offer true clinical benefits. The interest in statins as an inexpensive, readily available and easily tolerated drugs useful to reduce the morbidity and mortality caused by influenza and viral diseases, has been provided from numerous scientific publications over the past 15 years \[[@bib0005],[@bib0010]\] but for some reason it has never gone deep. Statins, inhibitors of the HMG-CoA reductase an enzyme that limits the pathway of mevalonate and cholesterol, have pleiotropic effects, as we described in several publications and especially in a Thematic Issue \"Statin: new life for an old drug\" published in 2014 on Pharmacol Res in collaboration with the discoverer of these compounds, prof. A. Endo \[[@bib0015]\]. These are substances with anti-inflammatory and immuno-modulatory effects, which could certainly benefit patients with influenza and viral pathologies. Experimental evidence suggested the effectiveness of statins in viral infections and their potential mechanisms of action ([Table 1](#tbl0005){ref-type="table"} ) \[[@bib0010]\]. In some viruses, statin-induced reduction of cholesterol in the plasma membrane results in lower viral titres and failure to internalize the virus. Such data suggest that in some viruses, cholesterol affects the early stages of infection. In particular, in the initial stage of infection, viruses bind specific receptors concentrated within lipid rafts, areas of plasma membrane rich in cholesterol. Statins, by reducing the percentage of cholesterol present in the membrane, alter the assembly of the receptors and dramatically reduce the possibility of adhesion of the viral agent to the host. Lipid rafts are also involved in the viral replication phases, as they constitute packets of vesicles capable of concentrating virus replication factors. These processes are also highly destabilized by statins \[[@bib0010]\].Table 1Potential mechanisms triggered by statins as anti-virus compounds (see reference \[[@bib0010]\] for detailed studies).Table 1StatinsExperimental ModelsMechanism of action and observed effectsAtorvastatin, RusovastatinMadin-Darby Canine Kidney (MDCK) cells infected with influenza A (strains- H3N2 and H1N1) viruses-down-regulation of Rho/Rho kinase pathway;\ -inhibition of the virus proliferationAtorvastatinC57BL/6 mice infected with influenza A virus (strains H3N2 and H1N1)-reduced lung virus titers;\ -reduced mortality rates in infected miceAtorvastatin, SimvastatinCrandell Feline Kidney (CrFK) cells infected with influenza A virus (strain H1N1)-reduced TNF-α and IL-6 in supernatants of infected cellsSimvastatinPrimary normal human bronchial epithelial (NHBE) cells stimulated with synthetic dsRNA viral analogue;\ Human lung epithelial cell line A549 stimulated with synthetic dsRNA viral analogue-suppressed dsRNA-induced STAT3 activation;\ -inhibition of RANTES expressionSimvastatinsynthetic dsRNA-induced pneumonia in BALB/c mice-reduced STAT3 activation;\ -reduced RANTES release;\ -reduced neutrophilia in the lungsAtorvastatinMadin-Darby Canine Kidney (MDCK) cells infected with influenza A virus (strain H1N1)-inhibition of the early stage of virus multiplication;\ -reduced virus infectivity by decreasing the virus titer from infected cells;\ -increased cell viability of virus infected cellsSimvastatinMadin Darby canine kidney (MDCK) cells infected with Influenza A virus (strain H1N1)-inhibition of the early stage of virus multiplication;\ -decreased virus-induced cytotoxicity in infected cells;\ 2-fold decrease of secreted pro-inflammatory cytokines (TNF-α, IL-6, INF-γ);\ decrease of virus replication through inhibition of Rab/RhoA GTPase activity and LC3 membrane localizationLovastatinHEp-2 cells infected with Respiratory Syncytial Virus (RSV, strain A2)\ C57BL/6 and BALB/c mice infected with RSV-reduced RSV replication in HEp-2 cells;\ -reduced cell-to-cell fusion in cell culture (through inhibition of RhoA);\ -decreased RSV virus replication in mice;\ -decreased virus-induced weight loss and illness in miceSimvastatinBALB/c mice infected with Influenza A virus (strain H5N1)-decrease of secreted pro-inflammatory cytokines and chemokines (IFNγ, IL-10, TNFα);SimvastatinLeukocytes from HIV-infected patients-depletion of cell membrane cholesterol and dissociation of lipid-rafts;\ -decrease of subpopulations and macrophages acting as Antigen Presenting Cells\ -decrease of highly efficient HIV-1 infection transferred by macrophages to CD4**^+^**T cellsMevastatin, SimvastatinHuman hepatocarcinoma (Huh7) cells containing subgenomic HCV replicons-dose-dependent inhibition of HCV replicon replication (measured as luciferase signal);\ -additive antiviral activity in short-term used in combination with IFNα or HCV nonstructural (NS)5B polymerase or NS3 protease inhibitors Many viral pathogens encode substrates for mammalian prenylation pathway (mevalonate pathway). It is hypothesized that the pre-dilation of viral proteins is beneficial for the propagation of the virus and that in some viral infections the statins, by inhibiting HMG-CoAR and reducing the expression of prenylated proteins, carry out an evident antiviral activity \[[@bib0010]\]. A recent study has shown that the influenza A virus induces the formation of numerous lipid droplets in infected cells. In the experimental model used, Madin-Darby dog kidney cells were treated with atorvastatin (ATV, 5microM) before inducing the infection. Pretreatment reduced the infectious capacity and, consequently, the production of new influenza A viruses, with a percentage higher than 95 % \[[@bib0020]\]. Studies performed in other viral strains have shown that different statins perform antiviral activity with variable efficacy, probably due to the specific pharmacological and biochemical properties of the statins used in the analysis and, therefore, to the different antiviral power observed \[[@bib0025]\]. Furthermore besides statins there are also a series of mevalonate pathway inhibitors that could then be used in the next future with a more specific target and effectiveness. It is certainly worth studying in detail the mechanism by which statins inhibit virus replication. Although unsuitable for prophylaxis in all patients \[[@bib0030]\], to study their efficacy as antivirals in vitro could provide a valuable tool to identify their mechanism of action in preventing virus reproduction and, therefore, pandemic spread. These hypotheses are strongly supported by substantial meta-analyzes and retrospective studies showing reduced hospitalization rates and mortality in subjects already users of statins affected by flu diseases \[[@bib0035]\]. Of course, studies are needed to evaluate, in cellular and animal models, the direct effect of statins on different viral strains, and more than anything on coronavirus, obviously followed by clinical studies, largely dictated by the current clinical emergency. The analysis of the interaction with statins of viruses at very high costs, health and social, such as SARS-CoV-2, represents a valuable possibility. There is also to consider that the effects of statins on the family members of Coronaviridiae has never been investigated. Positive results from these studies could provide the Health Services with an effective, safe and low-cost tool to improve the prognosis of this viral pathology, prevent pandemics and improve antiviral prophylaxis in "frail patients". In conclusion with this letter, we want draw the attention from all researchers and physicians to this therapeutic perspective not yet and fully investigated. Declaration of Competing Interest ================================= The authors declare no conflicts of interest.

{ "pile_set_name": "PubMed Central" }

Gastrointestinal (GI) disorders are among the most frequent complaints in patients with HIV disease. Before combination antiretroviral therapy, it was suggested that 50-93% of all patients with HIV disease had marked GI symptoms during the course of their illness \[[@ref1],[@ref2]\]. Inflammatory changes, direct mucosal invasion of the GI tract by HIV, opportunistic infections and neoplasms are well recognized causative factors \[[@ref3]\]. GI manifestations of HIV disease may be debilitating and include anorexia, weight loss, diarrhea, nausea, vomiting, dysphagia and odynophagia, abdominal pain, anorectal disease, GI bleeding, jaundice and hepatomegaly, and GI tumors \[[@ref4]\]. Progressive immunosuppression is associated with increasing prevalence of GI symptoms \[[@ref5],[@ref6]\] and thus the evaluation of specific GI complaints must be based on an assessment of the degree of immunosuppression. However, antiretroviral therapy may change the nature of HIV disease and can lead to improvement in gastrointestinal symptoms for patients with advanced immunosuppression \[[@ref5],[@ref6]\]. Appropriate symptom management requires improved symptom recognition and should be an essential component of patient care at all stages of any disease worldwide \[[@ref7]\]. However, limited data exist regarding the incidence of various GI manifestations in HIV-infected patients in certain areas of the world including the Caribbean. The paper by Thompson et al in this issue of *Annals of Gastroenterology* \[[@ref8]\] investigated the prevalence of common GI symptoms in the ambulatory Jamaican HIV infected population compared to those seen in a healthy population of blood donors not known to have HIV. This study revealed that in HIV-infected patients the presence of symptoms was directly related to CD4 count similarly to previous studies \[[@ref9]\] and that GI symptoms were common in ambulatory HIV-infected patients and controls. Heartburn, belching and nausea were more common in the control patients whereas early satiety was the only symptom significantly more common in the HIV-infected patients. However, pain on swallowing, and vomiting were significantly more common in patients with CD4 \<350 cells/μL than in controls. The authors conclude that the relatively low incidence of significant GI symptoms in HIV-infected patients may reflect the changing manifestations of HIV infection due to the success of antiretroviral therapy. However, although the incidence of certain GI symptoms such as dysphagia and weight loss in the Jamaican population was similar to the incidence of GI symptoms reported in previous studies \[[@ref10]\], it was surprising that 35% of control patients had weight loss. Indeed, this present study has several limitations. Firstly, the weight loss as reported in this study was subjective and thus the estimated incidence may be inaccurate. Similarly, the subjective reporting of other GI symptoms such as belching and nausea may also explain why these symptoms were more common in the control patients. Secondly, the used questionnaire did not seek to determine if weight loss was intentional. Thirdly, the number of patients and controls was relatively small. Finally, the sample of HIV-infected patients represented only the ambulatory patients and was not entirely representative of the general population of HIV-infected patients. It would also be of interest to look at the possible association of specific antiretrovirals with the incidence of GI symptoms, similarly to previous studies \[[@ref5],[@ref10]\]. In conclusion, improved symptom recognition of GI symptoms in HIV patients should be an essential component of patient care worldwide and an astute clinical acumen is often required to avoid subjective interpretation of certain GI symptoms. Conflict of Interest: None David Geffen School of Medicine, University of California, Los Angeles, U.S.A.

{ "pile_set_name": "PubMed Central" }

Background {#Sec1} ========== Introduction {#Sec2} ------------ With the Sustainable Development Goals (SDGs), the global community sees greater than ever consensus about the crucial role that the health workforce plays in realizing goals to achieve universal health coverage \[[@CR1], [@CR2]\]. Qualified health workers, trained to work in effective teams within and across professional cadres to address the biomedical and social determinants of health, are critical to achieving health goals. While great progress has been made in maternal and child health and HIV/AIDS, continued shortages of adequately trained health workers raise the question of whether health professional education (HPE) systems are producing the health workforce needed to meet outstanding and emerging global health challenges. In 2010, the independent Commission on Education of Health Professionals for the 21st Century (henceforth "the independent Commission") released a comprehensive report on HPE. The Commission's report spurred a growing movement for HPE reform in many regions, including Africa, Asia, and the Americas \[[@CR3]\]. The report calls for institutional and instructional HPE reforms to target a variety of systemic problems \[[@CR3]\], including:Outdated, fragmented, and content-oriented curricula that produce graduates with narrow contextual understanding and insufficient knowledge, skills, and competencies to understand social and other determinants of health and diseasePoor teamwork and inadequate collaboration within and across health professional cadresEpisodic encounters with patient illnesses rather than continuous and holistic health careA predominant hospital orientation at the expense of primary careAn imbalance between health workforces and health needsWeak leadership in improving health system performance The independent Commission's report also mentions persistent gender stratification of professional status as a systemic deficiency and puts forth as one of its proposed reforms the need to pay particular attention to ensuring equal opportunities through more flexible working arrangements and career paths that accommodate temporary breaks, actively addressing gender discrimination and subordination \[[@CR3]\]. Yet gender-related deficiencies in HPE are not limited to gender stratification and include other forms of often invisible gender discrimination and inequalities as described in Table [1](#Tab1){ref-type="table"}. This paper identifies how particular gender-related deficiencies in HPE can be addressed as part of both instructional and institutional governance reforms. The definitions of the terms used in the paper are in Table [2](#Tab2){ref-type="table"}.Table 1Gender discrimination in health workforce systems \[[@CR4]\]Form of discriminationDescriptionSexual harassmentUnwanted, unwelcome, or offensive conduct that changes the terms and conditions of school or work, where either a person's rejection of, or submission to, such conduct is used explicitly or implicitly as a basis for a decision that affects that person's education or career (quid pro quo), or conduct that creates an intimidating, hostile, or humiliating work environment for the recipient (hostile environment). A form of violence as well as discrimination.Pregnancy discriminationExclusions, restrictions, or distinctions made on the basis of pregnancy, childbirth, or related conditions, such as unwillingness to hire, promote, or retain female students or workers who may get pregnant and leave school or the workforce or who require maternity leave and benefits. This type of discrimination is related to:Family responsibility discriminationExclusions, restrictions, or distinctions against individuals (such as pregnant women, mothers and fathers of young children, parents of disabled children, and individuals who care for their aging parents or sick spouses/partners) based on their responsibilities to care for family members.\ Pregnancy and family responsibility discrimination are related forms that target a broad range of reproductive functions and may be viewed together as *caregiver* (*or reproductive role*) *discrimination*. These related forms of discrimination usually target women of childbearing age who are not able to equally access opportunities for education, hiring, or promotion.Occupational gender segregationConcentration of men and women in different jobs (horizontal) or at different levels (vertical) in a job hierarchy. What has been called gender stratification \[[@CR3]\] may refer to vertical or horizontal segregation or both.Gender stereotypingOvergeneralized characterizations of persons in a particular group, occurring when the personal characteristics deemed necessary for a job are inconsistent with characteristics generally associated with a particular sex.Table 2Key definitions*Bias* \[[@CR43]\]\ An inclination or prejudice for or against one person or group, especially in a way considered to be unfair, that often results in discrimination.*Discrimination in employment and occupation* \[[@CR33]\]\ Practices that place individuals in a subordinate or disadvantaged position in school, the workplace, or the labor market because of characteristics (e.g., race, sex, age, religion, or other attribute) that bear no relation to the person's competencies or the inherent requirements of the job. Discrimination occurs when bias is enacted.*Equal opportunity and non-discrimination* \[[@CR44]\]\ The offering of employment, pay, or promotion to all, without discrimination as to sex, race, color, disability, and so forth.*Gender-blind* \[[@CR23]\]\ Gender-blind policies and programs that are designed without prior analysis of culturally defined economic, social, and political roles, responsibilities, rights, entitlements, obligations, and power relations associated with being female and male and the dynamics between and among men and women, boys and girls. Gender-blind policies and programs ignore gender considerations altogether.*Gender discrimination* \[[@CR45]\]\ Any distinction, exclusion, or restriction made on the basis of socially constructed gender roles and norms that prevents a person from enjoying full human rights.*Gender equality* (*in the health workforce*) \[[@CR46]\]\ A condition where women and men can enter the health occupation of their choice, develop the requisite skills and knowledge, be fairly paid, enjoy fair and safe working conditions, and advance in a career, without reference to gender. It implies that health professional education schools and workplaces are structured to integrate family and work to reflect the value of caregiving for women and men.*Gender inequality*\ Denotes the gender-based differences that result from gender discrimination and serve to diminish or enhance individuals' opportunities, access, power, conditions, and/or income.*Gender transformative* \[[@CR23]\]\ Policies and programs that seek to transform gender relations to promote equality and achieve program objectives. This approach attempts to promote gender equality by (1) fostering critical examination of inequalities and gender roles, norms, and dynamics; (2) recognizing and strengthening positive norms that support equality and an enabling environment; and (3) promoting the relative position of women, girls, and marginalized groups and transforming the underlying social structures, policies, and broadly held social norms that perpetuate gender inequalities.*Special measures* \[[@CR19]\]\ Programs, policies, and laws that seek to neutralize and redress embedded structures of discrimination and preferences for privileged groups that are already built into social institutions. Such affirmative measures place women or other marginalized groups in a situation of comparative advantage for a limited period, with the aim of achieving substantive equality in the long term.*Substantive equality* \[[@CR19]\]\ Takes into account the effects of past discrimination and recognizes that rights, entitlements, opportunities, and access are not equally distributed throughout society and therefore the need to sometimes treat people differently (through special measures) to achieve equal results.*Systemic structural discrimination* \[[@CR19]\]\ Patterns of behavior, policies or practices, and social, economic, or cultural background conditions that are part of the structures of institutions, which create or perpetuate disadvantage for members of a marginalized group relative to other groups in society or organizations.*Intersectionality*\ A feminist theory and analytical tool for understanding and responding to the ways in which gender intersects with other identities. The experiences of marginalization and privilege are not only defined by gender but by other identity factors, such as race, class, and sexual orientation to name a few---all of which are determined, shaped by, and imbedded in social systems of power. Intersectional paradigms view race, class, gender sexuality, and ethnicity among others as mutually constructing systems of power (\[[@CR47], [@CR48]\]; references [@CR36], [@CR49]--[@CR52] present the theoretical and methodological issues and opportunities related to this theory). Gender discrimination---whether culturally, socially, or structurally driven---is sometimes evident in health worker education and employment systems and sometimes so normative as to be invisible. Where gender discrimination and inequality exist, they can hinder production of the robust and competent health workforces needed to achieve health and development goals, by subjecting students and faculty to gender-based exclusions or restrictions and diminishing access and opportunities. As systemic problems that stem from (among other things) institutional cultures, norms, and policies, gender discrimination and inequality should be addressed in HPE reforms. To date, however, the HPE reform movement has paid scant attention to gender discrimination toward and among students and faculty. The new SDGs include a standalone goal that aims for achieving gender equality and empowering all women and girls, including a target to end all forms of discrimination against all women and girls \[[@CR1]\]. This gives gender discrimination and inequality a new primacy in development policy. As the independent Commission's report emphasizes, a key outcome of educational reform should be transformative learning \[[@CR3]\], which can cultivate a "new professionalism" and establish "enlightened change agents" who have the status, authority, ability, and willingness to challenge the numerous structural and cultural factors that keep gender-based discrimination in place. Transformative learning to address gender discrimination must necessarily be gender transformative. Equal opportunity, non-discrimination, gender equality, and respect for human rights should be core health professional values and competencies, promoted and enacted by HPE leaders, embedded in curricula, and enacted through HPE instruction, professional socialization, and institutional governance as foundations for the professional practice of future health workforce managers and frontline service providers. This paper summarizes the findings of an expert review that sought to assess interventions to combat gender discrimination and inequality in HPE settings and rank them according to whether they counter two broad types of gender discrimination in transformative ways: discrimination based on pregnancy and on family responsibilities (hereinafter called "caregiver discrimination") and sexual harassment. We also discuss challenges to implementing the identified interventions, recommendations for addressing the challenges, and implications for HPE reform. Gender discrimination and inequality {#Sec3} ------------------------------------ The global literature on gender and human resources for health (HRH) has demonstrated that gender discrimination and inequality are key barriers to entry, reentry, and retention in employment systems, especially for female health workers \[[@CR4]--[@CR6]\]. The common forms of gender discrimination documented in health workforce employment systems (Table [1](#Tab1){ref-type="table"}) also appear to operate in HPE settings, affecting HPE students' opportunities, treatment, and ability to complete their studies (Table [3](#Tab3){ref-type="table"}), and limiting faculty members' career satisfaction, advancement, and economic opportunities (Table [4](#Tab4){ref-type="table"}). We focus on sexual harassment and caregiver discrimination because they were apparent from the interventions described in the literature and because they are relevant to female HPE students and faculty \[[@CR7]\].Table 3Students' experience of gender discrimination and inequalitiesPhase of academic life cycle (students)Examples and resultsCareer selection• Gender stereotypes and segregation in health professional cadres such as nursing and nutrition (Kenya) \[[@CR10]\]Admission and entry• Negative attitudes against girls and women pursuing training and scholarship opportunities \[[@CR6]\]\ • Lower admission rates of female students at tertiary education institutions (Rwanda) \[[@CR53]\]Course participation and completion• Threats of failing grades against female students made by male faculty if students refuse sexual advances, leading to difficulty concentrating on studies or failed courses (Ghana, Kenya, Nigeria, Uganda, Zimbabwe) \[[@CR10], [@CR20], [@CR54]--[@CR56]\]\ • Demotion fees levied against students for taking time off for pregnancy and falling behind in their programs (Kenya) \[[@CR10]\]Career progression• Sexual harassment during medical training, affecting selection of medical specialty and residency programs (Japan, Sweden, USA) \[[@CR5], [@CR57]--[@CR60]\]\ • Attitudes discouraging female medical residents from becoming pregnant (USA) \[[@CR8]\]Retention and graduation• Unsafe living conditions, limiting students' ability to safely access university facilities and contributing to dropouts of female students \[[@CR11]\]\ • Threats of failing grades against female students by male faculty if students refuse sexual advances, leading to graduation delays or dropouts (Ghana, Kenya, Nigeria, Uganda, Zimbabwe) \[[@CR10], [@CR20], [@CR53]--[@CR56]\]\ • Required suspension or termination of studies when female students become pregnant (Namibia, Kenya) \[[@CR9], [@CR10]\]\ • Insufficient time for students with family and domestic responsibilities to participate in educational opportunities, contributing to dropouts (Tanzania, Uganda, UK, USA) \[[@CR11]--[@CR13]\]\ • Higher dropout rates of female students at all educational levels (Rwanda) \[[@CR53]\]Table 4Faculty members' experience of gender discrimination and inequalitiesPhase of academic life cycle (faculty)Examples and resultsRecruitment• Challenges in balancing work and family obligations, contributing to faculty leaving their positions or turning down employment offers (USA) \[[@CR31]\]\ • Hiring preferences for male faculty due to belief that female faculty taking maternity leave will be disruptive (Kenya) \[[@CR10]\]Career advancement opportunities• Discriminatory promotion decisions against female staff due to resistance to sexual advances (Nigeria) \[[@CR20]\]\ • Lower number of publications by female medical faculty with children (USA) \[[@CR31]\]\ • Taking reduced workload or time off for family responsibilities seen as a reflection of lower commitment to work (Australia, USA) \[[@CR16], [@CR61]\]Leadership• Higher numbers of male faculty in senior leadership, even in cadres traditionally considered female occupations, such as nursing (Kenya) \[[@CR10]\]Satisfaction and retention• Lower rates of career satisfaction among female medical faculty with children than among male medical faculty with children (USA) \[[@CR15], [@CR17]\] ### Students {#Sec4} For HPE students, gender discrimination and inequalities are apparent at different points throughout educational careers, from admission to career track designation to graduation. At the outset of the educational continuum, cultural and gender norms and stereotypes related to childbearing and childrearing can discourage girls and women from pursuing HPE training and scholarship opportunities \[[@CR6]\]. Once students are enrolled, gender-blind institutional policies and practices (i.e., policies and practices that do not take gender considerations into account) may prevent female HPE students from participating in classes, practicums, and other curricular offerings by failing to consider potential conflicts between educational requirements and students' caregiving responsibilities. Such gender blindness typically results in a lack of instrumental support for students, creating barriers to students' ability to equally access education or remain enrolled. Some professional programs also discourage women from becoming pregnant while they are students \[[@CR8]\]. In several countries, pregnant students are required to take mandatory time off before returning to school or may even face expulsion \[[@CR9]\]. Caregiver discrimination also may play out in the form of demotion fees for pregnant students who take time off and fall behind in their courses and practicums \[[@CR10]\]. Caregiving responsibilities have been shown to play a major role in attrition rates in countries such as Kenya, Tanzania, Uganda, the UK, and the USA \[[@CR10]--[@CR13]\]. In higher-education systems as a whole, insufficient and sometimes insecure living conditions can limit female students' ability to safely access university facilities and further contribute to decisions to drop out \[[@CR11]\]. Sexual harassment and assault have been documented in primary and secondary schools and universities in high- and low-resource settings \[[@CR14]\]. For HPE students, sexual harassment, threats, or assault by other students or teachers (whether quid pro quo or in a hostile environment) can make it difficult for the targeted student to concentrate on or complete coursework \[[@CR10]\]. Sexual harassment can fundamentally change students' educational environment and opportunities and may contribute to a student's decision not to pursue a particular career track. ### Faculty {#Sec5} For faculty, gender discrimination and associated inequalities often relate to conditions that structurally disadvantage members of one sex (typically women) in the academic system, such as requiring training that involves travel in order to obtain promotions \[[@CR6]\]. An academic culture of long working hours and implicit biases against faculty with family responsibilities can affect promotion and tenure decisions in both HPE and general higher-education institutions \[[@CR15]--[@CR17]\]. Recent organizational research found that pregnant women were perceived as less competent, less committed to their jobs, and furthest from meeting male "ideal worker" norms \[[@CR18]\]. Such forms of invisible bias and structures of discrimination can be embedded in social institutions over time and become culturally normative \[[@CR19]\]. Sexual harassment of faculty members can also affect career advancement. For example, Nigerian female academic staff reported that their refusal of university officials' sexual advances led to exclusion from promotion and other benefits \[[@CR20]\]. This type of discrimination lowers faculty self-confidence, career satisfaction, and retention, which in turn can affect the quality of education being provided at HPE institutions and contribute to faculty attrition \[[@CR5]\]. HPE institutions may enable (and fail to regulate) sexual harassment due to cultural norms that tolerate harassment and gender-blind policies \[[@CR21]\]. The failure to counter sexual harassment in HPE systems can have a serious and detrimental impact on both student and faculty experiences, as seen by system dysfunctions such as sex in exchange for grades or academic career advancement \[[@CR20]\]. The following sections describe our review of interventions to counter caregiver discrimination and sexual harassment in high- and low-resource school-based settings. Methods {#Sec6} ======= Review and assessment of gender-transformative potential {#Sec7} -------------------------------------------------------- A panel of experts with expertise in gender and health system strengthening reviewed and rated 52 distinct school-related interventions to address sexual harassment and caregiver discrimination in high- and low-resource settings (Fig. [1](#Fig1){ref-type="fig"}).Fig. 1Stages in the review, rating, and ranking of interventions There were 87 distinct educational institutions/organizations/programs that were included in the literature review, most of which were universities. Of the 87 institutions, the geographical breakdown was:North America and the Caribbean, 33 (38 %)West, East, and Southern Africa, 32 (36.8 %)Asia, 10 (11.5 %)Middle East and North Africa, 5 (5.75 %)Europe, 4 (4.6 %)Australia, 2 (2.3 %)South America, 1 (1.2 %) Interventions that had been previously identified in a review of 379 articles from peer-review and non-peer-review literature on HPE and general tertiary education \[[@CR22]\] provided the initial pool of interventions. The experts then assessed interventions' gender-transformative potential using criteria described in Table [5](#Tab5){ref-type="table"} by a given type of discrimination, focusing on key aspects such as location(s) where implemented, intervention features, and results of any formal evaluations or informal assessments. One practice was subsequently removed from the analysis because there was not enough information about the intervention itself, resulting in a final count of 51 interventions: 18 to counter sexual harassment, 27 to counter caregiver discrimination, and 6 to address gender equality more generally. Detailed, contextual information related to the review and assessment of interventions is available \[[@CR22]\].Table 5Criteria used to rate gender-transformative interventions in health professional educationCriteriaSexual harassment Take measures to end impunity for perpetrators of sexual harassment and other forms of gender discrimination Top critical criterion to counter sexual harassment Introduce, make use of, or further legal protections against gender discrimination Second critical criterion to counter sexual harassment Provide information and education about discrimination or rights Third critical criterion to counter sexual harassmentCriteriaCaregiver Discrimination Transform family, school, and/or work arrangements so that women are not penalized or disadvantaged for caregiving Top critical criterion to counter caregiver discrimination Challenge and change common discriminatory gender beliefs or norms Second critical criterion to counter caregiver discrimination Attempt to change imbalance of power or otherwise level the playing field Third critical criterion to counter caregiver discrimination Most interventions lacked the type of evaluation data related to inputs, processes, and outcomes that would allow the reviewers to determine their effectiveness, feasibility, or sustainability. Thus, reviewers rated the interventions' *potential* to counter gender discrimination and inequality by applying the six gender-transformative criteria in Table [5](#Tab5){ref-type="table"}. These criteria were formulated by the review team based on the USAID-supported Interagency Gender Working Group definition of gender-transformative policies and programs (see Table [2](#Tab2){ref-type="table"}) \[[@CR23], [@CR24]\]. The criteria were considered to be minimum standards for classification of a gender-transformative intervention in HPE settings. The reviewers discussed the meaning of the criteria and marked "Yes" or "No" for each criterion for each intervention to be rated. Intervention rating and ranking {#Sec8} ------------------------------- For interventions targeting the two types of discrimination (i.e., sexual harassment and caregiver discrimination), each reviewer was assigned a weight based on the reviewer's area of expertise, with assigned weights totaling 100 %. For each of the 51 practices, reviewers rated interventions applying the six criteria, with 0 representing "No" and 1 representing "Yes." The researchers then derived a weighted average for each intervention. The expert panel decided that an intervention had at least some gender-transformative potential if its weighted average was 0.5 or above. To rank interventions, one that had a 0.5 rating would be ranked higher than one with a 0 rating. Interventions were ranked by taking inventory of the reviewers' ratings using these critical criteria \[[@CR22]\]. The reviewers met to discuss the ranking process and final rankings. During these meetings, they refined their application of the six gender-transformative criteria, developed recommendations about core sets of interventions, and developed cross-cutting recommendations related to countering the two types of discrimination. Results {#Sec9} ======= Core sets of intervention {#Sec10} ------------------------- Of the 51 interventions reviewed, 13 interventions were identified as having significant stand-alone transformative potential in terms of the transformative criteria (see Table [5](#Tab5){ref-type="table"}) and more so when implemented in combination, that is, in "core sets." The reviewers formulated these core sets to include those interventions that were necessary (though not sufficient) to counter a particular form of discrimination in HPE settings. The core sets of priority interventions to counter sexual harassment and caregiver discrimination were those that met the critical criteria such that an intervention that met, for example, the top two critical criteria was ranked higher than an intervention that met only the top critical criterion. The panel also identified implementation challenges for each set where such information was available and formulated recommendations for addressing those challenges. ### Core set to counter sexual harassment {#Sec11} The expert reviewers identified a core set of three gender-transformative interventions with the potential to counter sexual harassment in HPE, shown in Table [6](#Tab6){ref-type="table"}. Establishing a sexual harassment policy and a grievance procedure appears to be feasible across high- and low-resource settings, as evidenced by implementation of the two practices in a number of African and North American universities \[[@CR22]\].Table 6Interventions included in the core set to counter sexual harassment in health professional educationInterventionDescriptionSexual harassment policy• Includes a single code of conduct for students, faculty, and staffGrievance or reporting procedure• Is confidential, outlines consequences for perpetrators, and takes concrete action to end impunity and reduce victims' fear of or vulnerability to retributionEducation and awareness raising• For students, faculty, and staff ### Key implementation challenges for interventions to counter sexual harassment {#Sec12} Although many HPE and other higher-education institutions included in the review had implemented one or more of the interventions in the core set to counter sexual harassment, the review identified a number of challenges:Sexual harassment policies may outline strong principles and institutional responsibilities, but the practical implementation of such policies can differ widely from their intentions. For example, the University of Stellenbosch's (South Africa) policy mandates a sexual harassment advisory and disciplinary committee, yet an assessment found that not only were many managers unaware of the policy but committee members' workloads made trainings on the policy difficult to schedule and implement \[[@CR25]\].The lack of awareness of grievance procedures (and of sexual harassment policies), along with inadequate individual and institutional training, can contribute to anemic use of grievance procedures in settings where sexual harassment is normative.Most policies explicitly prohibit retaliation against victims who report sexual harassment, but flawed grievance procedures and prevailing environments of intimidation or impunity can render anti-retaliation policies ineffective.Fear of retribution and lack of accountability discourage many victims of sexual harassment from using grievance procedures \[[@CR25]--[@CR28]\]. Assessments of Chancellor College in Malawi \[[@CR26]\] and the University of Botswana \[[@CR27]\] noted that when cases were reported, significant errors occurred in handling investigations, maintaining confidentiality, assuring that alleged harassers showed up, coordinating with responsible agencies, and even following the prescribed procedures, which caused students to lose confidence in the process. Given the possibility of culturally normative and unregulated sexual harassment in HPE settings, grievance procedures are an important intervention. To address the identified challenges, grievance procedures should pay special attention to confidentiality, guidance for documenting and reporting, clearly outline consequences for the perpetration of sexual harassment and retaliation, avoid an inadvertent chilling effect on reporting that may result from an overemphasis on false reporting, and take concrete action to both decrease and eliminate fear of retribution. Equally important, strategies must be implemented and enforced through strong institutional leadership, vigilant oversight, and timely follow-up and resolution. ### Core set to counter discrimination based on caregiver responsibilities {#Sec13} The reviewers also identified two core sets of interventions for students and faculty to counter caregiver discrimination (Table [7](#Tab7){ref-type="table"}). Practices included in these core sets have been shown to be feasible in some settings, with institutions in South Africa, Tanzania, and other countries offering child care \[[@CR29], [@CR30]\]. However, of the institutions reviewed, only the University of California and the University of Michigan, both in North America, offered the full set of interventions comprising the core set for faculty, and no institutions were identified that offered the full core set for students \[[@CR22]\].Table 7Interventions included in the core sets for HPE students and faculty to counter caregiver discriminationCore set for studentsCore set for facultyPregnancy• Pregnancy/maternity and parental leave• Pregnancy/maternity and parental leave (paid)• Continuation and reentry policies that do not require pregnant students to terminate their education• Pregnancy/maternity leave replacement funding to hire temporary replacements for employees on pregnancy/maternity leave to ensure continuity of instructionPostpartum• Lactation breaks and spaces• Lactation breaks (paid) and spaces• Parental leave• Parental leave• Child care (daily and emergency)• Child care (daily and emergency)• Child care financial assistance (or at low cost)• Child care financial assistance (or at low cost)• Flexible training schedules, such as part-time schedules and reduced workloads• Flexible working hours\ • Flexible tenure ### Key implementation challenges for interventions to counter caregiver discrimination {#Sec14} Interventions to counter discrimination based on caregiver responsibilities also face implementation challenges, although these, too, can be met by strong HPE leadership commitment:Adverse consequences---or fear thereof---are a significant barrier associated with some interventions. For example, faculty who opt for reduced duty leave or flexible training programs may experience resentment from colleagues. Moreover, HPE faculty may not always take advantage of interventions for fear that others will perceive them as uncommitted or that their careers will be negatively affected.Work-life integration is a key concern for many current and prospective HPE faculty (both women and men) \[[@CR31]\]. Institutions with family-friendly policies may, therefore, have a competitive edge in recruitment. Indeed, outside of the HPE sector, the University of Washington law school has used its family-friendly environment as a student- and faculty-recruiting tool \[[@CR32]\], and the University of California and University of Michigan both highlight their family-friendly initiatives to faculty candidates.Families and communities may resist some of the changes required to address discrimination based on caregiver responsibilities, because the interventions challenge longstanding gender norms, expectations, and divisions of labor. Girls and women who go to school likely need a reduced workload at home, potentially adding to their families' workload. Communication of policy and education of faculty and students, as well as ongoing public support of faculty who use such flexible policies, is key to preventing adverse consequences. A complementary strategy is to proactively plan for pregnancy coverage and flexible scheduling. HPE planners must also anticipate the different levels of resistance that may arise in recruitment and retention efforts and deliberately mobilize communities around reducing women's and girls' housework, preventing early marriage and pregnancy and sharing responsibility for caregiving. This implies a long-term, multidimensional, and multisectoral strategy to keep girls in school from the primary through tertiary levels. This might include provision of reproductive health services (including family planning) through HPE institutions. ### Interventions that address multiple forms of gender discrimination and inequality {#Sec15} In addition to selecting the core sets of interventions targeting sexual harassment and caregiver discrimination, the reviewers identified gender centers and equal employment opportunity units as having significant gender-transformative potential. These are institutional structures that advocate for, coordinate, oversee, implement, and evaluate multilevel strategies. These entities generally work to:Develop gender equality, equal opportunity, or affirmative action policiesEngage in awareness raising and information sharingServe advocacy and accountability functionsConduct gender sensitization workshops or sexual harassment training for women and menConduct research and university assessmentsProvide financial assistance to female studentsOffer mentoring and faculty career and leadership programs to women The last two objectives are examples of special measures to counter systemic structural discrimination and promote substantive equality (see Table [2](#Tab2){ref-type="table"}). These special measures counter the discrimination that may occur when poor families allocate scarce financial resources to fund boys' education and to compensate for the career barriers faced by women in HPE institutions characterized by high concentrations of men in top faculty and administrative positions. Whereas equal employment opportunity units which aim to counter discrimination in employment and occupation \[[@CR33]\] are often backed by national equal opportunity laws, gender centers face the challenge of not necessarily being backed by law. In addition, implementation challenges may arise from funding or staffing constraints. Leaders in HPE reform should educate stakeholders and advocate for the need for resource allocations to fund special measures to counter systemic structural discrimination to achieve substantive equality. Research limitations {#Sec16} -------------------- The relative lack of descriptive contextual and evaluation data for the 51 interventions limited the expert reviewers' ability to determine the exact nature, feasibility, sustainability, or effectiveness of the various interventions and, therefore, constrained their capacity to make recommendations for specific contexts such as low-resource settings. Overall, more information was available for institutions in high-resource than low-resource settings. The research team invited a variety of published experts representing different sectors and countries to participate as reviewers, but most experts were unavailable. A larger expert review group (including stakeholders such as students and faculty), or one with a more diverse range of expertise, would likely strengthen future reviews. Discussion {#Sec17} ========== A comprehensive HPE reform agenda aiming to produce a robust and competent health workforce should consider core sets of interventions to counter gender discrimination and inequality---even when discrimination is not overtly recognized by perpetrators or victims. Female health workers already constitute a large proportion of many countries' health workforces (in both the professional and non-professionalized cadres), and there is a growing focus on the role of women in the health workforce in the emerging human resources for health agenda \[[@CR5]\]. Failure to address gender discrimination and inequality in HPE can jeopardize broader health workforce and health system reform. The review findings can serve as a basis for evidence-based decisionmaking in planning and implementing appropriate gender-transformative interventions. The review demonstrated that interventions can address some of the gender-related obstacles to entry, retention, career progression, and graduation in HPE, in particular those related to sexual harassment and caregiver discrimination. Although many HPE institutions may not have the resources to implement *all* the gender-transformative interventions identified by the reviewers, the "core sets" provide a basis for establishing priorities and taking steps to counter gender discrimination. Potential implementation challenges spanning financial, informational, and cultural barriers need to be considered, however. Financing considerations are particularly relevant when interventions entail restructuring physical resources and human resources arrangements (such as child care, lactation spaces, or reduced duties leave) or require human resources (e.g., education and awareness-raising activities or adequately staffed and trained sexual harassment committees). Merely offering interventions does not guarantee that interventions are used or that equal opportunities and gender equality in HPE will result. Ensuring that the institutional community and especially the intended beneficiaries are aware of, committed to, and actually use the interventions is equally important. However, because many of the interventions challenge social and institutional norms and cultural stereotypes, some community members (inside or outside the institution) may resist their implementation. Increasing the actual use of new policies or procedures requires information and advocacy, anticipating potential resistance, providing incentives, and ensuring that HPE leaders are accountable for work-life integration and the speedy and effective handling of sexual harassment so that institutional efforts actually prevent and end impunity for discriminatory organizational structures and behaviors. What are some implications for HPE reform? First, gender-transformational HPE reform will in some cases entail both institutional and instructional changes \[[@CR3]\] sometimes in combination, since some reforms to institutional governance may require, or can be enhanced by, instructional reforms. For example, the gendering of the health workforce involves the distinction between occupations, as well as the relations between occupations (for instance, between medicine, nursing, and allied health professions) \[[@CR34]\]. These processes of gender segregation or stratification could be mitigated by both equal opportunity policy (an institutional governance reform) and transprofessional education (an instructional reform). Institutional reform, such as introducing a new policy, could also be combined with professional instruction to raise students' and faculty awareness of gender discrimination as *unethical* professional conduct, as well as a human rights violation. For example, the problem of *quid pro quo* sexual harassment (i.e., sex in exchange for grades, see Table [3](#Tab3){ref-type="table"}) is a form of unethical professional conduct within an HPE, an abuse of power by faculty which impacts educational or occupational opportunity, which has both instructional and institutional consequences. Heads of HPE institutions can work with an equal opportunity unit to introduce a code of conduct to faculty and students and include the subject in a course in ethics or human rights \[[@CR35]\], thus playing a dynamic role in inculcating shared attitudes, values, and behaviors related to respect for the dignity and rights of clients, students, and colleagues. Instructors can be trained to model ethical one-on-one and team interactions and conducting fair assessments free from the taint of sexual harassment, stereotyping, and other forms of discrimination. Third, the review highlights the importance of a gender-relational perspective in HPE reform. Gender relations shape health systems---including HPE systems---through their effects on the occupational segregation and stratification among health providers, the conditions of work, and processes of regulation, supervision, and management of health labor forces \[[@CR36]\]. The theoretical foundations of the approach in this paper include gender-relational theory, social dominance theory, and the sociology of patriarchy \[[@CR34], [@CR37], [@CR38]\]. This perspective gives a central place to the patterned relations between women and men (and among women and among men) that constitute gender as a social structure \[[@CR39]\]. The structure of gender relations (including power relations and hierarchies) in a given institution is its "gender regime" \[[@CR34]\]. The institutions through which health care is delivered (e.g., hospitals, clinics, private practices) have well-defined gender regimes \[[@CR34]\], and this applies to HPE institutions. If so, heads of HPE schools would benefit from a gender analysis approach to shed light on the "gender regime" in their institutions and possible impacts on faculty and students at various points in the academic career. A practical first step would be for heads of schools, equal opportunity staff, or human resources managers to conduct a gender audit of its institutional "gender regime," including policies and practices in relation to pregnancy, family responsibilities and sexual harassment, instructional content, socialization processes, and the socioeconomic characteristics of its student body and faculty. Ending all forms of discrimination against all women and girls everywhere (i.e., SDG 5.1) does not speak to the multiple or intersecting axes of gender discrimination, bias, and marginalization (e.g.,, economic class, region, race, or caste) which likely impact present and future student bodies. Such analysis could result in changes in recruitment, admissions, and financing \[[@CR3], [@CR40], [@CR41]\]. Operationally, this is a "field waiting for an analytical breakthrough" (\[[@CR36], p. 9\]) and should be part of an HPE-learning agenda. Fourth, institutional governance reforms suggest the utility of gender centers or equal employment opportunity units to drive policy and accountability. This unit could coordinate the development and enforcement of supportive, evidence-based policies to promote equal opportunity, non-harassment, and social equality. As a part of policies such as flexible working arrangements and career paths that accommodate temporary breaks, HPE leaders and HR managers should embrace gender equality in social roles and promote the value of caring and work-life integration *for both women and men* in educational and employment systems \[[@CR3], [@CR42]\]. Finally, the review highlighted the striking lack of contextual and evaluation documentation for gender-transformative interventions in HPE and the need to invest in a stronger learning agenda in relation to sexual harassment, caregiver discrimination, and, in fact, in all forms of discrimination already documented in health employment systems. We recommend building a rigorous knowledge base as well as evaluating the feasibility, sustainability, and effectiveness of gender-transformative HPE reforms. Table [8](#Tab8){ref-type="table"} outlines suggested areas of focus for this learning agenda.Table 8A learning agenda for gender-transformative health professional education reformForm of discriminationSuggested elements of learning agendaSexual harassment• Elements and features of grievance procedures that effectively resolve sexual harassment cases and end impunity\ • Effectiveness of various (HPE) institutional and non-institutional actors in investigating sexual harassment\ • Role of extra-institutional legal advocacy and redress\ • Desirability of separate grievance procedures for students, faculty, and staff versus a single procedureCaregiver discrimination• Impact of basic bundle on student retention, performance, and graduation\ • Impact of basic bundle on faculty recruitment, retention, and development\ • Feasibility and effectiveness of offering free or low-cost family planning for men and women in the basic bundle\ • Effective community messaging related to the need to reduce women's and girls' family and domestic burdens\ • Factors that contribute to use of family-friendly services by both women and menCross-cutting• Approaches for inter- and transprofessional education to promote equality and teamwork and effectively eliminate gender stereotyping and stratification (e.g., siloes, hierarchical chains of command)\ • HPE curricula that transmit core professional values and competencies (including human rights, social justice, gender equality, professional ethics and conduct, respectful care, and critical inquiry) and foster health workers as change agents in inequitable systems\ • Test new and efficient gender audit methodology and information systems to support improved institutional governance policies and monitor progress of reforms, targeting the analysis of gender regimes and intersecting discriminations, including the collection of data for indicators of social position (e.g., gender, income, geographical region, race, caste, disability)\ • New approaches to teach human rights and ethical professional conduct in combination with HPE governance.\ • Coordinated efforts across universities to implement special/affirmative measures and enabling conditions such as work and learning design to take parenting and caregiving responsibilities into account for all students and faculty, universal flexible working/study arrangements.\ • Structural and resource requirements for and impact of equal opportunity units in HPE institutions Conclusions {#Sec18} =========== The paper's broad aim is to encourage HPE leaders---heads of schools, human resources administrators, donors, and other human resources for health (HRH) stakeholders---to make gender-transformative changes in the current way of doing business and commit themselves to countering gender discrimination and inequality in HPE. These affect the functioning, quality, outputs, and outcomes of health professional education. Gender-transformative HPE reforms that eliminate impunity for sexual harassment, promote and protect the educational and labor rights of students and faculty, and develop student and faculty understanding of professional ethics will require changes in mindsets, institutional cultures, and leadership capability and the reframing of health workers as change agents committed to gender equality in often inequitable systems. Greater focus on the application of the core sets of interventions identified in this paper and a strong learning agenda should be part of ongoing HPE reform efforts. **Competing interests** The authors declare that they have no competing interests. **Authors' contributions** CoN, CrN, and SPM were responsible for the conception and design, acquisition of data, and analysis and interpretation of data for the systematic review on which the article was based and for drafting and revising the manuscript and graphics. DF was involved in interpreting findings from the systematic review and critically revising the manuscript. All authors gave final approval of the version to be published. **Authors' information** Authors Crystal Ng and Sara Pacqué-Margolis are former employees of IntraHealth International. The research described in this paper was funded through the United States Agency for International Development's (USAID's) Capacity*Plus* Project, under Grant GPO-A-00-09-00006-00. The views expressed in this paper are solely those of the authors and do not necessarily reflect the views of USAID or the United States government. The authors are grateful to Mesrak Belatchew, formerly of IntraHealth International; Asha George at the Johns Hopkins Bloomberg School of Public Health; Rebecca Bailey, formerly of IntraHealth International; and Nonie Hamilton, Lois Schaefer, and Temitayo Ifafore of USAID. Special thanks to Claire Viadro for editorial support.

{ "pile_set_name": "PubMed Central" }

Introduction {#S0001} ============ Acne is a common, multifactorial, inflammatory skin disease affecting the pilosebaceous unit resulting from altered keratinization, androgen-induced increased sebum production, inflammation involving both innate and acquired immunity and microbial colonization by *Propionibacterium* (*P*.) *acnes*.[@CIT0001]--[@CIT0003] Genetic predisposition, anxiety and depression, hyperglycemic diet, sunlight or artificial light, are additional factors known to be involved in the pathogenesis of acne.[@CIT0004] Acne is clinically characterized by different clinical aspects and severity, spanning from mild forms, mainly characterized by retentional lesions (RTL) (open and closed comedones) with or without few inflammatory lesions (IFL) (papules/pustules), to moderate or severe forms, with multiple IFL that seriously impact on patient's quality of life.[@CIT0005],[@CIT0006] The treatment vary according to disease severity. For mild acne, topical pharmacological treatments are generally recommended and include the use of benzoyl peroxide (BP), retinoid, and antibiotics.[@CIT0005]--[@CIT0010] The clinical efficacy of these agents is widely demonstrated, and it is well known that their use may cause some adverse effects, such as hypersensitivity reaction, erythema, burning/stinging in case of BP/retinoids, and limitations for the possible development of resistant bacterial strains for antibiotics.[@CIT0005]--[@CIT0010] Recent evidence suggests that some cosmetic agents if correctly prescribed may improve the therapeutic outcome with the advantage of minimizing the side effects and enhance the efficacy of prescription drugs.[@CIT0011] The aim of this study was to assess the efficacy and tolerability of a new cosmetic regimen consisting in the daily use of an anti-inflammatory/corneolytic/sebum controlling fluid in the morning and of an anti-inflammatory/corneolytic cream at bedtime in the treatments of mild facial acne by clinical and instrumental evaluations for 8 weeks duration study. Materials and Methods {#S0002} ===================== Study Design {#S0002-S2001} ------------ This was a multicenter, prospective, observational clinical study. Setting and Study Period {#S0002-S2002} ------------------------ From October 2017 to May 2018 ninety-one adult patients (71F/20M) with mild acne were enrolled at the Dermatology Department of the University of Catania (Italy) and of University "Federico II" of Naples (Italy). Study duration was 8 weeks. The study was performed in accordance with the ethical principles originating from the Declaration of Helsinki 1996 and Good Clinical Practices. The protocol was approved by an internal review board. A written informed consent was obtained from each patient before study procedures were started, which included consent to use their images. Inclusion/Exclusion Criteria {#S0002-S2003} ---------------------------- Inclusion criteria were: adult patients with mild facial acne (open and closed comedones and up to 10 papules), a wash out period of at least 2 weeks for any topical acne treatments, and 1 month for oral antibiotics and 2 months for hormonal therapy and/or oral isotretinoin. Exclusion criteria were: presence of pustules or nodules lesions, trunk localization, severe underlying diseases, concurrent exposure to sunlight and/or artificial ultraviolet sources, pregnancy and breastfeeding. No other topical products or drugs were allowed, except for a mild cleanser, make-up and oil-free SPF 50+ sunscreens. Methodology {#S0002-S2004} ----------- Patients were instructed to apply a fluid containing Licochalcone A combined with Salicylic acid, and L-Carnitine in the morning and a cream based on Licochalcone A and Hydroxy-complex 10% at bedtime for 8 weeks. A mild, perfume-free cleaning gel principally based on Ampho-Tenside 6% was suggested for daily care. In order to reduce potential evaluator bias, all subjects were assessed by an investigator not directly involved in the study at baseline (T0), and at 4 weeks (T1), and 8 weeks (T2). Clinical and Instrumental Evaluation Criteria {#S0002-S2005} --------------------------------------------- Clinical and instrumental evaluation were carried out at baseline, and at 4 and 8 weeks. Acne severity was rated by clinical evaluation using: 1) *the Global Acne Grading System* (GAGS) score consisting in a subjective assessment of acne severity proposed by Doshi et al in 1997 that divides the face into six areas, forehead, nose, chin, right cheek, left cheek, chest and upper back and assigns a factor (from 1 to 3) to each area (2=forehead; 1=nose; 1=chin; 2= right cheek; 2= left cheek; 3=chest and upper back): the local score is calculated using the formula factor x grade (depending on the most severe lesion type; 0= no lesions; 1= comedones; 2=papules; 3=pustules; 4=nodules); the global score is the sum of all global scores: a score of 1--18 is considered mild; 19--30 moderate; 31--38 severe and \>39 very severe and by 2) lesions count, method based on count of the total number of retentional and/or inflammatory lesions.[@CIT0012],[@CIT0013] Instrumental assessment included: 1) measurement of sebum by Sebutape^TM^ strips (CuDerm Corp., Dallas, TX, USA) placed on the patient's forehead, nose, cheeks, and chin and subsequently checked against a black background of a score card; sebum spots are scored on a scale ranging from 1 to 5, where 1 indicates a dry skin without sebum, and 5 identifies a very oily skin, and 2) facial imaging performed by high-tech facial photography characterized by 15 mega pixel resolution and flash cross-polarized Light by Reveal Photo Imager (Canfield Scientific Inc., Parsippany, NJ, USA) at Naples site and by VISIA-CR™ imaging system (Canfield Scientific Inc., Fairfield, NJ, USA) at Catania site.[@CIT0014]--[@CIT0016] Additionally, evaluation of product tolerability by a self-administered questionnaire based on 5 parameters (erythema, scaling, dryness, stinging/burning and itch) from 0 to 3 (0 =none; 1=mild; 2=moderate; 3=severe) was carried out. Study Endpoints {#S0002-S2006} --------------- Primary endpoint was the clinical efficacy at week 4 and 8 evaluated by reduction of GAGS score and lesion count; secondary endpoint was the evaluation of tolerability and cosmetic acceptability at the end of the study. Statistical Analyses {#S0002-S2007} -------------------- The characteristics of the patients were reported as absolute number and percentage for the categorical variables and as mean and standard deviation (SD) for the continuous variables. The analysis of longitudinal data was performed using a mixed effect regression model, which considered the correlation between measurements carried out over time in the same subject. In the mixed effect regression models, time, age, gender and center were considered as covariates. Interaction between center and time was also tested. For time covariate Dunnett post-test was used for multiple comparisons vs baseline. A two-tailed P value \<0.05 was considered significant. All statistical analyses were performed using SAS version 9.4 (SAS, Inc., Cary, NC). Results {#S0003} ======= Ninety-one subjects (71F/20M; age range: 18--30 years; mean age 21.5±3.8 years; mean GAGS: 11.6±2.4) with mild acne were enrolled. Eighty-eight (70F/18M) completed the study whereas three subjects were lost to follow-up for personal reasons. At 4 weeks a statistically significant reduction from baseline of GAGS (mean from 11.6±2.4 to 10.3±3.0; p\<0.001) was observed. In addition, the mean total count of RTL and IFL was significantly reduced of 41% and 45%, respectively, from baseline (comedones: mean from 42.3±14.3 to 24.8±12.7; p\<0.001; papules: mean from 5.5±5.0 to 3.0±3.8; p\<0.001) along with a significant reduction of mean sebum of about 47% (mean from 2.13±0.67 to 1.13±0.38; p\<0.001) ([Table 1](#T0001){ref-type="table"}, [Figure 1](#F0001){ref-type="fig"}).Table 1Results from Clinical and Instrumental Evaluation of GAG Score, Total Lesion Count of Comedones/Papules from Baseline to 8 WeeksEvaluations (Mean Value)Baseline4 Weeks8 WeeksP ValueMeanDSMeanDSMeanDS**GAGS**Catania11.22.29.63.47.73.3\<0.001Naples12.02.610.92.68.72.4\<0.001Overall11.62.410.33.08.22.9\<0.001**Total Count Comedones**Catania39.311.920.813.612.610.9\<0.001Naples45.116.028.710.517.27.6\<0.001Overall42.314.324.812.715.09.6\<0.001**Total Lesions Papules**Catania4.23.32.32.41.32.0\<0.001Naples6.76.03.74.81.83.7\<0.001Overall5.55.03.03.81.63.0\<0.001**Sebutape**Naples2.130.671.130.381.030.37\<0.001 Figure 1Advanced digital photography of two patients affected by mild acne at baseline (**A,C**) and after 4 weeks of treatment (**B,D**). At 8 weeks a further statistically significant reduction from baseline of GAGS (8.2±2.9; p\<0.001) was also recorded; similarly the mean total count of RTL and IFL was significantly reduced of 64% and 71%, respectively, from baseline (comedones: mean from 42.3±14.3 to 15.0±9.6; papules: mean from 5.5±5.0 to 1.6±3.0; p\<0.001), along with an additional mean sebum reduction of about 52% (p\<0.001) ([Table 1](#T0001){ref-type="table"}, [Figure 2](#F0002){ref-type="fig"}). For all considered outcomes, Dunnett multiple comparison post-tests versus baseline showed a significant reduction (p\<0.001) of 4-weeks and 8-weeks values. The two centers showed similar results in the reduction of GAGS (interaction center\*time, p=0.655) and total count of comedones (interaction center\*time, p=0.192), while for total count of papules a significant interaction center\*time was found (p=0,007) showing a greater reduction in the Naples center in comparison to Catania center ([Table 1](#T0001){ref-type="table"}).Figure 2Advanced digital photography of two patients affected by mild acne at baseline (**A,C**) and after 8 weeks of treatment (**B,D**). Local side effects were documented in 3 cases only (severe erythema: 1 case; severe hitch: 2 cases) and product tolerability was rated as excellent in 90% of patients. Discussion {#S0004} ========== Cosmetic treatment for acne, if correctly prescribed and used, may improve outcomes in acne therapy.[@CIT0011] The topical cosmetic approach for acne treatment includes: hygiene and cleansing, ideally with a non-comedogenic, non acnegenic, non-irritating and non-allergenic cleanser; sebum controlling agents, generally based on the so-called "mattifyng" ingredients able to absorb sebum from skin surface, in case of hyperseborrea; corneolytics agents, particularly indicated for comedonic acne as able to induce a comedolytic effect, making comedones more superficial and at the same time may also facilitate skin absorption of other products; antimicrobial agents that have been proven to effectively reduce P.acnes growth both in vitro and in vivo clinical study; anti-inflammatory agents indicate to control the pivotal role of inflammatory events in the development of acne lesions; moisturizers product specifically designed for acne patients that improve adherence to therapy by alleviating the discomfort and feeling of dryness following topical and systemic retinoids or topical BP.[@CIT0011] The use of specific photoprotective agents and shaving products should also be encouraged. Finally, tailored camouflage make-up technique using non-comedogenic products may minimize the esthetic problem associated with acne and improve patients' self-esteem and adherence to treatment.[@CIT0011] The results of our study suggest that the tested products are effective in the treatment of mild acne achieving a good resolution in cure rates due to a synergic action of its agents concurrently able to control different factors involved in the pathogenesis of acne. In detail, licochalcone A is a natural flavonoid substance isolated from *Glycyrrhiza glabra* and *Glycyrrhizainflata* (licorice root) with anti-inflammatory and anti-microbial effects, as confirmed by inhibition in vitro and in vivo of pro-inflammatory cytokines (PGE2, LTB4, IL-6 and TNF-α),[@CIT0014]--[@CIT0015] as well as anti ROS-induced cell damage action by activating of the expression of antioxidant transcription factor (Nf-E2-related factor 2/Nrf2) and of detoxifying enzymes (Heme Oxygenase 1/HO-1, Glutamate-Cysteine Ligase Modifier subunit/GCLM).[@CIT0017]--[@CIT0019] Hydroxy Complex is a combination of three exfoliating agents (Glycolic, Salicylic and Polyhydroxy Acids) able to cause intercorneocyte cell detachment so as to induce a comedolytic effect.[@CIT0020]--[@CIT0022] Finally, L-Carnitine is a sebum controlling agent by β-oxidation of intracellular lipid content as demonstrated in vitro and in vivo clinical study.[@CIT0023] Conclusions {#S0005} =========== Our results suggest that the daily regimen-based Licochalcone A in combination with Salicylic acid/L-carnitine as fluid or with Hydroxy-Complex 10% as cream represent an interesting approach for treating mild acne that undoubtedly deserves attention. It may represent a significant adjunct to the therapeutic armamentarium for the management of acne patients without any significant side effects. We are aware that our study has some restrictions being a non-controlled trial but it was a multicenter study with a considerable number of patients enrolled. Further studies on larger series of adolescent acne patients are necessary to confirm our results. Disclosure {#S0006} ========== The study was supported financially by Beiersdorf SpA (Milano, Italy), who manufacture the product. Paolo Chiodini received an honorarium for a consultancy from Beiersdorf SpA (Milano, Italy). The other authors declare no other conflicts of interest in this work.

{ "pile_set_name": "PubMed Central" }

Introduction ============ Chronic recurrent multifocal osteomyelitis (CRMO) is a rare, noninfectious, inflammatory bone disease, which occurs mainly in childhood \[[@b1-dp0804a09]\]. We present a case of CRMO and palmoplantar psoriatic skin lesions in a 12-year-old girl. Case Presentation ================= A 12-year-old girl presented with recurrent erythematous palmoplantar plaques and pustules. She also complained about pain in her left ankle that started 7 months earlier. Previous magnetic resonance imaging (MRI) had consistently revealed a multifocal bone edema of the left foot. Symptoms of weakness, fever, and morning stiffness were absent. The family history was unremarkable. Physical examination revealed well-demarcated erythematous plaques with remnants of dried pustules in a palmoplantar distribution ([Figure 1](#f1-dp0804a09){ref-type="fig"}). The active range of motion of the left upper ankle joint was painfully decreased by 50%. Laboratory results showed a slight increase of inflammation parameters, including c-reactive protein level and erythrocyte sedimentation rate. Antinuclear antibody level, rheumatoid factor, HLA-B27, and Lyme disease testing were negative. Serial MRIs revealed fluctuating T2 hyperintensities and T1 hypointensities involving the left talus, calcaneus, and metatarsal bones with undulating discrete joint effusion in the left upper ankle joint ([Figure 2a--d](#f2-dp0804a09){ref-type="fig"}). Additionally, synovial thickening of the left talocalcaneonavicular joint was noticed. T1-weighted images after contrast application were acquired at several MR-measurements over the course of 1 year, mainly reflecting the edema seen as T2 hyperintensities. Based on these results, the diagnosis of CRMO accompanied by palmoplantar pustular psoriasis (PPPP) was made. Treatment with oral nonsteroidal anti-inflammatory drugs (NSAIDs) and topical mometasone furoate 0.1% cream was initiated. CRMO was first described by Giedion et al in 1972. It primarily occurs in the distal metaphyses of long tubular bones \[[@b1-dp0804a09]\]. The involvement of the calcaneus, as described herein, was rarely reported. PPPP is found in approximately 15% of CRMO patients. The pathophysiology of CRMO is not well understood. Recent studies of CRMO patients described a reduced production of interleukin (IL) 10 by monocytes. This impairment may result in an increased activation of the Nod-like receptor family pyrin domains containing protein 3 inflammasome (NLRP3) leading to an enhanced expression of IL-1 β, which has a role in osteoclast activation via receptor activator of nuclear factor kappa-B ligand (RANKL) stimulation. Bissonnette et al \[[@b2-dp0804a09]\] described high levels of IL-1β and IL-17A in patients with PPPP leading to a secondary chemokine production of keratinocytes with accumulation of neutrophils. Thus IL-1β seems to play a key role in both CRMO and PPPP. Conclusion ========== CRMO should be treated interdisciplinarily, and NSAIDs should be the medication of first choice. Skin lesions may be alleviated by topical steroids. Moreover, bisphosphonates, TNF antagonists, IL-1-inhibitors, sulfasalazine or methotrexate have been described as effective. Fortunately, our patient showed a complete remission 2 years after the onset of symptoms, which is also observed in 30% to 40% of reported cases. In summary, we describe a rare case of CRMO initially presenting as PPPP with joint pain. Dermatologists and pediatricians should be familiar with the association of CRMO and PPPP in children to lead the way to correct diagnosis and treatment. **Funding:** None. **Competing interests:** The authors have no conflicts of interest to disclose. ![Well-demarcated erythematous scaly plaques with remnants of dried pustules on the left sole of the patient. \[Copyright: ©2018 Epple et al.\]](dp0804a09g001){#f1-dp0804a09} ![Serial MRIs over the course of 12 months (a--d) with fat suppressed T2-weighted images showing the fluctuating hyperintensities of the left ankle involving the calcaneus, talus, and metatarsal bones with an undulating joint effusion in the left upper ankle joint. The date of each MRI is indicated in the top row of images. \[Copyright: ©2018 Epple et al.\]](dp0804a09g002){#f2-dp0804a09} [^1]: All authors have contributed significantly to this publication.

{ "pile_set_name": "PubMed Central" }